Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

When the laboratory starts running itself

For most of the AI era, the laboratory has remained one of the places where the boundary between digital and physical work is fairly clear. Models can search scientific literature, predict molecular structures and help researchers decide where to look, but eventually somebody still has to run the experiment.

That boundary is beginning to move. Roche said last week that it has started building autonomous AI laboratories as part of its pharmaceutical research operation. The idea is to connect AI models with laboratory automation so that parts of the experimental cycle can run continuously: choosing what to test, carrying out an experiment, measuring the result and feeding what was learned back into the system. (Reuters) Reuters

The concept is usually called a self-driving laboratory, and it has been developing quietly across chemistry and materials science for several years. A recent Nature Reviews Chemistry review describes systems that combine autonomous experimentation, robotics and AI, with algorithms increasingly able to propose, execute and interpret experiments with limited human intervention. The field is still constrained by how difficult it is to make these systems scalable and general enough to move between different kinds of science. (Nature Reviews Chemistry) Nature

What stood out about Roche is less the individual technology than where it is appearing. This is no longer confined to a university robotics laboratory trying to prove that autonomous experimentation works. One of the world’s largest pharmaceutical companies is beginning to build the approach into its research infrastructure.

There is a useful distinction here between automation and autonomy. Laboratories have been automated for decades. Machines already pipette liquids, screen compounds and perform repetitive measurements. Autonomy begins when the system uses the result of one experiment to decide what experiment should happen next. That changes the loop.

A scientist can still define the problem and the boundaries of the search, but some of the exploration between question and answer can increasingly be handed to a system capable of running experiments around the clock. Scientific progress has always been constrained partly by the time and cost required to test ideas in the physical world. Autonomous laboratories are an attempt to compress that part of the process.

It feels early, and the distinction between a genuinely autonomous scientist and sophisticated laboratory optimisation matters. But something structural is starting to appear. AI is slowly moving from analysing the record of what humans have already discovered toward participating in the process through which new experimental evidence is created.

Inside the Lab Where AI Runs the Experiments

What it is

FutureHouse is trying to automate more of the scientific process than simply analysing data or searching research papers. Co-founder Andrew White describes the goal as an “AI scientist”: systems that can find questions worth investigating, search the existing evidence, develop hypotheses and ultimately connect those ideas to experiments in a physical laboratory.

The organisation has built its own wet lab for this reason. White makes an interesting distinction between building AI tools for scientists and using those tools to actually make discoveries. The laboratory provides somewhere for ideas generated by the system to encounter the physical world.

What stood out

One of the earlier projects behind FutureHouse was ChemCrow, which connected a language model to chemistry tools and a self-driving robotic laboratory. White describes being able to give it an objective such as creating a molecule with a particular property, then allowing the system to work through the tools required to make it without human involvement.

But the larger constraint becomes clearer later in the video. White notes that science is not simply an intelligence problem. Researchers need access to literature, specialised databases, computational tools and, eventually, physical experiments. Making the model more capable only addresses part of that system.

FutureHouse is effectively trying to assemble the rest of it around the model.

The scale becomes interesting when those pieces are connected. White gives one example involving roughly 40,000 individual investigations into relationships between hormones and observed behaviours in mice. AI agents searched the literature for each relationship and assembled information that would have been difficult for a researcher to investigate individually.

Why it matters

There is a tendency to think about scientific AI in terms of a smarter researcher: a model that reads more papers, understands more disciplines or generates better hypotheses. This video makes the infrastructure around that intelligence more visible.

A hypothesis still has to meet reality.That requires instruments, samples, measurements and experiments, followed by some way of returning those results to the system so it can decide what to investigate next. Once those pieces begin connecting, the interesting change is not simply that individual scientific tasks become faster. The entire cycle between question and evidence can begin to shorten.

White imagines scientists increasingly delegating workflows to groups of agents and spending more of their own time deciding which questions deserve attention and checking the work being produced. His timeline may or may not prove accurate, but the shift in the structure of the work is already worth watching.

For now, the scientist remains very much inside the loop. What is changing is how much can happen before they need to step back into it.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

A laboratory that decides what experiment comes next

Researchers reported the system in Nature Synthesis earlier this year. It is a modular self-driving chemistry laboratory designed specifically to make autonomous experimentation cheaper and more accessible. Rather than requiring an enormous bespoke robotic facility, it combines relatively affordable custom hardware with software that can control laboratory equipment and decide how to alter subsequent experiments. (Nature Synthesis) Nature

This is where the idea becomes much easier to understand. A researcher gives the system an objective, such as finding better conditions for a particular chemical reaction. RoboChem-Flex performs an experiment and analyses the result. An optimisation algorithm then uses what happened to choose the conditions for the next experiment. Temperature, concentrations, reaction times or other variables can change as the system explores the available chemical space.

The important part is that it doesn’t necessarily test every possible combination. It learns from the experiments it has already performed and uses that information to decide where it should look next. The researchers demonstrated the platform across six different chemistry problems, including photocatalysis, biocatalysis and thermal reactions, and showed that it could operate either autonomously or with a scientist remaining inside the loop. Nature

There is a useful connection here to the way we normally imagine scientific discovery. A human researcher doesn’t blindly perform every conceivable experiment either. Experience gradually creates intuition about which direction might be worth exploring. Self-driving laboratories are beginning to build a much narrower computational version of that process.

The technology remains highly specialised. A machine capable of optimising a chemical reaction is not independently deciding which scientific questions matter, and the broader field still struggles with moving autonomous systems between different laboratory environments. A 2026 review of self-driving laboratories identifies scalability, generalisability and complete experimental provenance as some of the problems that still need to be solved. (Nature Reviews Chemistry) Nature

But RoboChem-Flex makes the direction tangible. The interesting development isn’t simply that a robot can perform an experiment. Laboratory robots have done that for years. It is that the result can become an input into the decision about what the robot should do next.

Last week I was looking at the enormous effort required to give machines physical experience. This feels like another version of the same transition, but in a different setting. Once a machine can interact with the physical world, the question becomes what it can learn from those interactions. In the laboratory, we are beginning to get an answer.

“The important thing is not to stop questioning.”

Albert Einstein

The line comes from a short reflection Einstein gave late in his life about curiosity and the value of continuing to ask questions even when answers remain incomplete.

It feels appropriate this week because autonomous laboratories don’t remove the question from science. If anything, they make the quality of the question more important. A system might eventually search thousands of papers, generate possible explanations and run experiments at a pace no individual researcher could match, but somebody still has to decide what is worth looking for.

For a long time, experimentation has placed a natural limit on curiosity. There are only so many experiments a person, laboratory or institution can run. We may be starting to loosen that constraint.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

The robots are still learning how to work

For several years, humanoid robots have existed in a strange space between industrial machinery and demonstration. They walk across stages, carry boxes through carefully arranged warehouses and occasionally fold a shirt. The videos suggest a technology moving quickly. What has been harder to see is how much of that activity has made its way into ordinary work.

This week we got a better measure. The International Federation of Robotics has begun tracking humanoids separately for the first time, estimating that only around 7,000 humanoid robots were sold globally in 2025 for industrial and professional use. That remains a small number beside the hundreds of thousands of conventional industrial robots installed each year. Even manufacturers experimenting seriously with humanoids are generally working with fleets measured in single digits or low double digits. (Reuters)

What stood out wasn’t really the number. It was what many of those robots are being used for. A significant share are going to research institutions, robotics companies and industrial pilots where their immediate purpose isn’t necessarily productive work. They are collecting data. Every attempt to pick something up, navigate an unfamiliar space or recover from a mistake produces another example that can be used to improve the models controlling the machine.

There is a useful contrast here with the development of large language models. By the time the current generation of AI arrived, humans had already spent decades creating an enormous digital record of language and behaviour. Books, websites, photographs, video, software and code existed before anyone decided to use them as training material.

Robotics inherited no comparable record of the physical world. We don’t have trillions of neatly captured examples showing how a hand adjusts when a cardboard box begins to slip, how someone moves around an unexpected obstacle on a factory floor, or how much pressure is required to handle thousands of slightly different objects. Much of what humans know about moving through the physical world was never written down because it never needed to be.

So the machines are beginning to collect it themselves. That helps explain another development this week. In China, where investment in humanoid robotics has accelerated particularly quickly, regulators are beginning to scrutinise companies seeking public listings and asking harder questions about where their revenue actually comes from. Some robotics businesses have been generating income through data-collection centres, facilities where robots repeatedly perform tasks specifically to produce training data. Investors and regulators are now trying to distinguish that activity from sustainable demand for robots doing economically useful work.

That distinction feels worth recording. It doesn’t necessarily mean the humanoid thesis is weakening. It may simply tell us where the technology actually is. Elsewhere in robotics, the picture is much more mature. Conventional industrial robots already operate at enormous scale, doing repetitive tasks inside environments designed around them. A welding robot doesn’t need to understand a factory. Its section of the factory has been engineered so that understanding is largely unnecessary.

The proposition behind humanoids is almost the reverse. Our warehouses, factories, hospitals, kitchens, tools, stairs and doorways were designed around human bodies. Rather than redesigning those environments around machines, the industry is trying to build a machine capable of entering environments designed for us. That requires something closer to general physical competence. And physical competence appears to require experience.

For now, some of the first humanoid robots aren’t replacing workers. They aren’t really workers yet. They are gathering the experience that the robots coming after them may need.

The Indian workers training AI robots to take their jobs

What it is

This short AFP report visits workers in India who are being paid to record themselves carrying out ordinary household tasks. Wearing a camera system, they repeat actions from different positions, with different objects and against changing backgrounds. One worker says they produce more than 90 videos a day, systematically changing angles and actions so that each recording adds another variation to the dataset.

The work is mundane by design. These are the kinds of movements humans perform without thinking, but which need to be captured and labelled if machines are eventually going to learn how to reproduce them.

What stood out

The economics are interesting. One worker describes being paid 250 rupees an hour to perform household tasks that would normally go unpaid. There is something unusual about physical knowledge that has never had an obvious market value suddenly becoming useful because a machine needs to learn it.

The repetition also gives a sense of how much data is required. The same action is recorded from different angles and positions, with different objects and surroundings. What looks to us like a single task becomes dozens of slightly different examples once you try to describe it to a machine.

The most revealing observation comes near the end. One worker expects a future filled with robotics and AI, but doesn’t see humans disappearing from it because, as he puts it, “without human data… those can’t run.”

Why it matters

That line captures much of what stood out to me about humanoid robotics this week. The machines are becoming more capable, but underneath that progress sits an increasingly large human effort to capture things we have never previously needed to record.

The internet gave language models an enormous archive of what humans had written, photographed and coded. There is no equivalent archive for how we move through a kitchen, pick up an unfamiliar object or adjust our hands when something starts to slip. That knowledge exists largely in our bodies.

For now, teaching machines physical intelligence still requires humans to turn some of that tacit knowledge into data. There is an unresolved tension in people being paid to record the very movements that machines may eventually learn to perform without them.

The robots are learning from us while we are still figuring out what that means for the people doing the teaching.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

The robots are growing a sense of touch

We tend to think about robots understanding the world primarily through cameras. Better vision, better models and more computing power have allowed machines to recognise an extraordinary range of objects and environments, but working with those objects introduces a different problem. Humans don’t rely on sight alone when we pick something up. As soon as our fingers make contact, we are sensing pressure, texture and movement, making small adjustments to our grip without really thinking about them.

Touch remains much less developed in robotics. This week, researchers gathering at the IROS robotics conference are looking specifically at scalable tactile sensing for dexterous manipulation, including the difficult problem of giving robotic hands useful information about contact while they are moving and handling objects. Existing sensors still have limitations around resolution, coverage, speed and durability, particularly when compared with the sensitivity packed into a human hand. (IROS 2026)

The technology is gradually starting to look less like a simple pressure sensor and more like an artificial layer of skin. Sensors distributed across robotic fingers and hands can detect where contact is occurring, how pressure is distributed and whether an object is moving against the hand. That information can then be combined with cameras and information about the position of the robot’s joints, giving the model controlling the machine another way of understanding what is happening.

Synaptics recently demonstrated one version of this approach with a capacitive tactile sensing system for robotic hands and grippers. It captures contact and pressure information alongside vision and motor-control signals, and has been integrated into NVIDIA’s Isaac Sim environment so that some of these interactions can be modelled and trained in simulation before being attempted by a physical robot. (Synaptics)

This is still early technology, but it helps explain why dexterity remains such a difficult part of humanoid robotics. Recognising a glass on a table is quite different from knowing how firmly to hold it, noticing that it has started to slip, or adjusting when it turns out to be heavier than expected. Factories, homes and hospitals are full of objects that bend, move, break, spill and behave slightly differently each time they are handled. Humans manage most of that variation without consciously processing it.

It also connects with the wider training problem emerging around humanoid robots. The industry is not simply trying to build stronger motors or more articulated hands. It is gradually assembling the sensory systems and physical experience that allow a machine to understand what happens when it interacts with the world.

The first generation of AI had an enormous advantage in this respect. Much of the information it needed had already been recorded by humans in books, websites, photographs, videos and code. Physical intelligence has no equivalent archive waiting to be scraped. Some of the information these machines need has never been recorded because, for us, it has always been something we simply feel.

“We know more than we can tell.”

Michael Polanyi

Polanyi introduced this idea in The Tacit Dimension in 1966. He was interested in the knowledge humans carry without necessarily being able to explain it: recognising a face, riding a bicycle, using our hands, or developing a feel for a task through repetition.

It feels particularly relevant to the workers recording themselves for robot training. Much of what they are capturing was never written down because there was never much reason to write it down. It existed in movement, habit and accumulated experience. For machines to learn some of these things, that tacit knowledge is having to become visible for the first time.

Perhaps one of the more interesting records of this period will be how much we discover about what humans know only when we try to teach it to something else.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

News

The rules stalled. Tokenisation didn’t.

For much of the past few years, the assumption around digital assets in the United States has been that meaningful institutional adoption would eventually require Congress to settle the rules. This week offered a more complicated picture. On Tuesday, the Senate failed to advance the CLARITY Act, the most significant attempt yet to establish a comprehensive federal market structure for digital assets. Two days later, the SEC opened a new pathway for tokenised US-listed stocks to trade on blockchain infrastructure. Legislation stalled. The underlying transition continued.

The Senate vote was a genuine setback for the CLARITY Act. The bill received a simple majority but fell short of the 60 votes required to advance, after disagreements spanning ethics provisions, investor protections and the treatment of different parts of the crypto market. One senator changed his vote for procedural reasons, preserving the possibility that the legislation could be reconsidered, so the bill is stalled rather than necessarily dead. But with the US midterm elections approaching, the window for passing comprehensive market-structure legislation has narrowed considerably. (Reuters)

What happened 48 hours later was therefore particularly interesting. On 17 September, the US Securities and Exchange Commission introduced a five-year Innovation Exemption for certain tokenised US-listed stocks. Under tightly defined conditions, qualifying venues will be able to let permissioned participants trade tokenised shares using automated market makers and liquidity pools running on public, permissionless blockchains. The exemption also creates limited relief for firms supplying liquidity to those markets. It runs until September 2031, giving regulators several years to observe how this architecture behaves before deciding what should become permanent. (SEC)

The detail matters because this goes further than simply putting a digital representation of a share on a blockchain. The SEC framework explicitly contemplates tokenised stocks trading against other tokenised stocks, qualifying stablecoins or tokenised money-market funds through automated liquidity pools. In other words, some of the market structure developed inside crypto is being allowed to touch conventional US equities. The underlying stock remains a regulated security with the same economic rights, including dividends and voting rights. What changes is part of the infrastructure through which that asset can be held and exchanged.

This is still an experiment, not a wholesale migration of Wall Street onto public blockchains. The SEC has imposed limits on the number of stocks that can be offered and the proportion of their trading volume that can move through these venues. Platforms must meet disclosure, recordkeeping and transparency requirements, participants must be permissioned, and issuers can object to unaffiliated third parties tokenising their shares. The exemption also doesn’t remove the underlying securities from existing investor-protection and anti-fraud laws.

But the timing is difficult to ignore. Congress spent years trying to agree on a comprehensive framework for digital assets and this week failed, at least for now, to get that framework through the Senate. Meanwhile, regulators are beginning to use the authority they already have to create narrower pathways through which parts of traditional finance can experiment with the same infrastructure.

There are other signals accumulating around this. Banks are experimenting with stablecoins and tokenised deposits. Funds and government securities are increasingly being represented on-chain. India has just begun a pilot in which tokenised corporate bonds can settle against its digital rupee. Now the SEC is creating a controlled environment in which actual US-listed equities can meet stablecoins and automated market makers on public blockchain rails. The political system is still negotiating what digital-asset markets should look like. Parts of the financial system are already beginning to test them.

CLARITY Act Fails Key Senate Vote: What’s Next for Bitcoin, Other Crypto Assets?

What it is

Schwab Network speaks with Ray Salmond following the CLARITY Act’s failure to clear a key procedural vote in the US Senate. The immediate discussion is framed around what the setback means for crypto markets, but it also gets into the more structural question running through this week’s Field Notes: whether the development of digital financial infrastructure actually waits for Congress to establish a comprehensive set of rules.

What stood out

Salmond’s central observation is that the failed vote is a setback for regulatory clarity, but not necessarily a derailment of what is already underway. The SEC and CFTC still have their own rulemaking processes, while large financial institutions have existing regulatory pathways through which they can continue experimenting with digital assets and tokenisation.

He points specifically to institutions such as BlackRock, ARK Invest and Fidelity potentially making greater use of the SEC’s tokenisation framework. That is the part of the conversation that stood out to me. The political process and the infrastructure build are increasingly moving on different clocks.

The immediate market reaction was less patient. Bitcoin was down around 4.4% at the time of the discussion and Coinbase had fallen more than 11%, although the CLARITY vote was only one part of the backdrop, with markets also waiting on an upcoming Federal Reserve interest-rate decision. Salmond remains constructive on Bitcoin over the longer term, but the price discussion feels almost secondary to the institutional movement happening underneath it.

Why it matters

For years, regulatory clarity has been treated as one of the prerequisites for traditional finance moving meaningfully on-chain. This week complicates that assumption.

Congress still matters. Comprehensive legislation would provide something that individual exemptions and agency rulemaking cannot: a more durable framework defining how digital assets fit within the US financial system. But institutions aren’t necessarily waiting for every part of that framework to be settled before building.

That distinction feels increasingly important. Markets can react to a failed vote in hours. Legislation can take years. Infrastructure tends to accumulate somewhere in between.

This week, all three were visible at once.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Frontier Tech: Title-backed Tokens

Most property tokenisation starts one step removed from the property itself. A company or fund owns the building, and investors own tokens representing an interest in that company or fund. The blockchain may make the investment easier to administer or transfer, but underneath it is essentially a familiar investment structure with a new digital layer.

At Toroa, we have been exploring a different starting point: the property itself. The idea behind our title-backed token methodology is to connect digital ownership as directly as possible to an interest in a real property. Exactly how that connection is made will depend on the property and the laws of the country involved. That is part of what we are working through now.

One of the first projects we are exploring is a new hotel and spa development on a vineyard in Argentina. Rather than approaching the development only through conventional property finance, we are looking at whether part of the property could ultimately be divided into a fixed number of digital ownership interests.

In simple terms, imagine a hotel divided into a defined number of ownership pieces. Each piece has a digital token attached to it. Holding one could represent a corresponding interest in the underlying property, with the potential to receive a share of the income generated by the hotel and participate in the value of the property over time. The token provides a digital way of keeping track of who owns each piece and, potentially, making those interests easier to transfer between eligible owners.

That starts to address something property has always struggled with. A building can be extremely valuable, but it is difficult to divide and relatively slow to transact. Selling an entire hotel is a major event. Even selling a smaller interest can involve layers of paperwork, administration and intermediaries. Digital ownership could make smaller interests easier to keep track of, simplify distributions to owners and create a clearer path for those interests to change hands.

But putting something on a blockchain doesn’t make the real-world complexity disappear. A token might move between two digital wallets in seconds, while the vineyard and hotel remain very physical assets sitting in Argentina. The rights attached to them still have to work in the real world. That means working through what the digital interest actually represents, how ownership is recognised, what happens when an interest changes hands and how the digital record stays connected to the property behind it.

That is where much of our work on the Argentine vineyard project currently sits. We have a methodology and a proposed model, but we are deliberately leaving room for the final structure to emerge through the legal and practical work now underway. The technology can only be useful if someone holding the token can clearly understand what they own and those rights continue to mean something away from the blockchain.

This is why I find title-backed tokens more interesting than simply putting another investment product on-chain. The experiment isn’t really about whether we can create a token representing a hotel. We already know how to create tokens.

The harder question is whether we can make digital ownership and physical property behave as parts of the same system.

“Anything that is in the world when you’re born is normal and ordinary and is just a natural part of the way the world works.”

Douglas Adams

Adams wrote about the strange way humans absorb technological change. What initially feels unfamiliar eventually becomes infrastructure, and the generation that arrives afterwards often encounters it without remembering the transition at all.

There is something of that happening with tokenisation. Today we still make a distinction between a share, bond or property interest and its digital representation. We debate whether financial assets should move onto blockchain rails, which rules should apply and which institutions should be allowed to participate.

A decade from now, some of those distinctions may still matter. Others may sound strangely specific to this period.

For now, we’re still living through the part where the terminology hasn’t settled.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

One of the more interesting AI stories this week came from Anthropic, which published a detailed assessment of four incidents in which Claude models gained unauthorised access to real third-party systems during cybersecurity evaluations. The important detail is how they got there. The evaluations were designed as isolated exercises, and the models were explicitly told they had no access to the internet. A configuration error meant that assumption was wrong. The open internet was available, while the cyber safeguards normally included with Anthropic’s released models had also been switched off for testing. Anthropic’s assessment describes seven evaluation runs involving four different models, each operating independently for roughly 10 to 34 hours. (Anthropic)

These were capture-the-flag exercises, the kind of controlled cybersecurity challenge where a model is given a fictional target and asked to find a hidden piece of information. The problem was that the boundary between the fictional exercise and the real internet had disappeared. In the most serious case, a research model called Claude Mythos 5 discovered credentials that gave it access outside the intended environment and eventually published a malicious package to PyPI, the public repository widely used to distribute Python software. The package was installed on 15 third-party systems, which Anthropic believes were security scanners deliberately testing new packages. One of those scanners inadvertently exposed credentials, which the model then used to access a real security vendor’s database. PyPI removed the package within an hour. Anthropic has since released the full incident transcript and asked independent AI research organisation METR to investigate the incidents.

What makes the episode more interesting than the inevitable “AI escaped its sandbox” framing is the model’s reasoning once it encountered evidence that the environment might be real. During the Mythos 5 incident, the model repeatedly described what it was seeing as part of the simulation, even as it encountered public domains and other signs that suggested otherwise. Anthropic initially interpreted this fairly literally, saying in July that the model appeared to attack real targets because it believed they were simulated. After examining the transcripts more closely, the company has revised that view. Its current assessment is that the model showed what it calls biased reasoning: evidence was interpreted in ways that allowed it to continue pursuing the task it had already committed to.

Anthropic tested this by showing separate instances of Mythos 5 individual pieces of evidence from the incident without the surrounding context. In isolation, those models classified 79% of the outputs as likely coming from real systems. During the actual incident, however, Mythos 5 explicitly recognised signs of the real internet in only around 1% of its actions. Even when researchers later altered parts of the transcript to make the real-world nature of the environment clearer, the model often continued with the task. Anthropic describes the second recurring behaviour across the incidents as recklessness, meaning a tendency to keep pursuing an assigned objective despite evidence that doing so could cause harm.

There are some important boundaries around what happened. These weren’t ordinary Claude conversations, and Anthropic says its production cyber safeguards would have blocked much of the behaviour observed in the evaluations. The models weren’t developing their own objectives, coordinating with other agents or attempting to hide what they had done. Each remained focused on completing the cybersecurity task it had been given. Anthropic subsequently searched a much larger collection of roughly 481 million transcripts and says it found no additional incidents of similar or greater severity beyond the four already identified. Newer models also reproduced the concerning behaviour less frequently in simulated recreations, although not at zero rates.

That makes this less a story about an AI deciding to escape than about what happens when several layers of a system fail at once. The sandbox wasn’t actually closed. The normal safeguards weren’t present. The task encouraged offensive behaviour. And when the model encountered evidence that should have caused it to reconsider what it was doing, its reasoning instead tended to preserve the course of action already underway.

There is something useful in Anthropic publishing the incidents in this level of detail, including revising its earlier interpretation of what the models appeared to “believe”. As AI systems spend longer periods working independently, the assumptions around them start to matter almost as much as the instructions we give them. A boundary that exists in the prompt but not in the underlying infrastructure isn’t much of a boundary at all.

For now, these remain unusual incidents inside deliberately adversarial evaluations. But they leave behind a fairly practical observation: as models become more capable of acting in the world, testing them safely increasingly depends on making sure the world they think is a sandbox really is one.

What it is

This episode of Rystad Energy’s Let’s Talk Energy looks at two pressures arriving on electricity grids at the same time. Extreme weather and wildfires are making outages more frequent and longer-lasting, while the rapid construction of data centres is adding enormous new sources of demand. According to the figures discussed in the episode, major US outage events increased by around 40% between 2018 and 2024 to more than 6,500 a year, while their average duration increased by 23% to just under 12 hours. At the same time, Rystad estimates global data-centre power demand grew 50% between 2023 and 2025 and could almost double again to just under 250 GW by 2030.

What stood out

The interesting tension is that these aren’t really separate problems. Keeping electricity available during a wildfire, storm or heatwave is increasingly important precisely because so much of everyday life now depends on both electricity and computing. Yet the infrastructure providing that computing capacity is itself becoming one of the fastest-growing demands on the electricity system.

That makes the behaviour of data centres during periods of grid stress increasingly important. The discussion moves beyond simply asking where enough electricity will come from and into how utilities and large technology companies coordinate demand, onsite generation and backup systems when the grid is under pressure. It connects closely with the Virginia event discussed above, where systems designed to protect individual data centres collectively became significant enough to affect the wider grid.

Why it matters

There is an interesting feedback loop forming. AI is contributing to the growth in electricity demand that makes managing the grid more difficult, while AI may also become part of how increasingly complex grids are managed. Better forecasting, earlier detection of equipment failures and more dynamic balancing of supply and demand could all help electricity networks respond to conditions that are becoming harder to predict.

The physical and digital systems are becoming increasingly difficult to separate. Data centres depend on resilient grids. Grids increasingly have to account for the behaviour of data centres. And both are becoming more important during exactly the kinds of events when infrastructure is under the greatest stress.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

AI data centres are becoming part of the grid

This week’s frontier technology isn’t another AI model. It is the electrical infrastructure being built around them. Modern AI data centres are becoming some of the largest individual consumers of electricity on the grid. Inside these facilities are layers of power electronics, uninterruptible power supplies, batteries and backup generators designed to keep thousands of GPUs running through disturbances in the electricity supply. If grid voltage suddenly falls, those systems can react almost instantly, protecting the computing equipment by reducing grid demand or transferring parts of the facility onto backup power. At the scale AI infrastructure is now reaching, that automatic response is becoming a technology problem of its own.

A useful example appeared in Virginia this year. On 22 July, a fault occurred on a 230-kilovolt transmission line in Dominion Energy’s territory. The grid’s protection systems responded correctly and cleared the fault, but the brief voltage disturbance triggered data centres across the region to transfer simultaneously onto backup systems. Dominion estimated that around 3,869 megawatts of demand disappeared from the grid almost at once. The sudden change was large enough to affect grid frequency, with normal reliability conditions taking around nine minutes to recover. Most of the computing load returned within roughly half an hour.

The technology at the centre of this is known as ride-through. Large electrical loads need rules governing how they behave when voltage or frequency briefly moves outside normal ranges: when they remain connected, when they disconnect, and how quickly they return afterwards. This has long mattered for power stations and industrial equipment. What is changing is that enormous clusters of computing infrastructure now need similar treatment.

The North American Electric Reliability Corporation is beginning to describe data centres and cryptocurrency mining facilities as voltage-sensitive computational loads. The concern is not simply that they consume a lot of electricity. It is that many facilities can detect the same disturbance and have automated systems respond to it in roughly the same way. Thousands of megawatts of computing demand can therefore move at computer speed, without anyone sitting in a control room making an individual decision.

Grid operators are starting to design around this behaviour. On 8 September, PJM advanced proposed ride-through requirements for large loads, including voltage and frequency thresholds and rules governing how quickly power consumption should recover after a disturbance. Separately, US regulators have directed NERC to begin developing reliability standards specifically for large computational loads.

There is an interesting tension underneath this. From inside the data centre, rapidly switching to backup power is exactly what the infrastructure is supposed to do. It protects expensive computing equipment and preserves uptime. But thousands of megawatts of independently sensible decisions can create a different problem when they happen together. Resilience at the level of the data centre can become instability at the level of the grid.

Until recently, most of the conversation around AI and electricity has been about quantity: where enough power will come from to supply the data centres being built. This feels like the beginning of a second problem. AI infrastructure is becoming sufficiently large that electricity networks don’t just have to supply it.

They increasingly have to understand how it behaves.

“The city is a fact in nature, like a cave, a run of mackerel or an ant-heap.”

Lewis Mumford

Lewis Mumford was an American historian, writer and critic who spent much of the twentieth century studying the relationship between cities, technology and human life. He was particularly interested in what happens when technological systems begin determining the shape of the places we live, rather than remaining tools within them. His writing on transport was often critical of cities reorganising themselves around the automobile, especially when greater speed and mobility came at the expense of the wider urban environment.

That makes the line useful this week. Autonomous vehicles are becoming capable enough that the question is gradually shifting from whether they can operate in our cities to how our cities might change around them. Roads could communicate with vehicles, parking could become less important, freight networks could operate differently and cars themselves may no longer need to be designed around a driver. At the same time, many of the older questions remain: congestion, safety, public space, employment and who the transport system is ultimately designed to serve. The technology inside the vehicle may be new. The tension between adapting technology to the city, or adapting the city to technology, is much older.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

The vehicle is starting to lose its driver

For most of the autonomous vehicle era, the basic approach has been to take a car designed for a human and teach a computer how to drive it. The steering wheel remained. So did the pedals, mirrors, dashboard and forward-facing seats. Autonomy changed who was controlling the vehicle, but not necessarily the vehicle itself.

That is beginning to change. Tesla started limited rides in its purpose-built Cybercab in Austin this week, moving beyond the modified Model Ys used in its earlier robotaxi service. The Cybercab has no steering wheel or pedals and dispenses with conventional side mirrors. It is designed around the assumption that a human driver will never need to take control. Reuters reported on the first rides this week, marking an interesting transition from making existing cars autonomous to designing vehicles specifically for autonomy.

That distinction is already creating some friction. US vehicle safety standards were largely developed around machines controlled by people, so removing the driver also removes some of the assumptions embedded in the regulatory system. The US National Highway Traffic Safety Administration has now opened an investigation into how Tesla certified the Cybercab against federal safety standards, including requirements covering equipment such as mirrors. Reuters reported on the investigation on Friday. The technology is beginning to encounter rules written for a different kind of machine.

Tesla isn’t alone in reaching this point. Amazon-owned Zoox and Alphabet’s Waymo both announced further US expansion this week, with their autonomous services and testing programmes spreading into additional cities. Reuters reported that Zoox is moving into Houston and San Diego, while Waymo is extending its footprint towards Denver, San Diego and Tampa. The important signal here isn’t another collection of dots appearing on a robotaxi map. It is that autonomous transport is beginning to encounter enough different streets, weather conditions and human behaviour that it increasingly has to operate as ordinary urban infrastructure rather than a contained technology demonstration.

Zoox also makes the changing design language particularly visible. Its purpose-built robotaxi has no conventional driver’s position and uses inward-facing seating, more like a small carriage than a car. Cybercab takes a somewhat more familiar form, but both start from the same premise. Once nobody needs to drive, there is less reason to organise the interior around someone sitting behind a steering wheel and looking through a windscreen.

The same transition may eventually matter even more in freight. Autonomous trucking company PlusAI announced plans this week to go public through a transaction valuing it at around US$800 million as it works towards commercial deployment of its autonomous driving system with truck manufacturers and fleet operators. Reuters covered the transaction here. Autonomous trucking has spent years in testing, but the interesting question now is what happens if driverless operation begins to affect the economics and physical design of freight itself.

A conventional long-haul truck is partly a machine for moving goods and partly a workplace for a person. It needs a cab, controls, visibility and often somewhere for the driver to sleep. Routes and schedules have also evolved around limits on how long people can safely remain behind the wheel. Remove the driver and some of those constraints begin to disappear. Trucks could eventually look different, operate for longer periods, use different kinds of depots and move through logistics networks designed around machines rather than shifts.

This feels like the more important stage of the autonomous transport story. Whether one company reaches full autonomy before another matters in the near term, but the longer transition may be what autonomy does to the machines themselves.

We spent the first phase teaching computers to drive vehicles built for humans. We may now be entering the phase where we start building vehicles for computers.

What it is

This video from Good Work looks at the rise of autonomous vehicles through San Francisco, where Waymo has moved from an experiment into an increasingly ordinary part of the transport system. The story begins with Sebastian Thrun, whose Stanford team won the 2005 DARPA Grand Challenge before he went on to lead Google’s self-driving car project, which eventually became Waymo. Today, the company is providing more than 500,000 autonomous rides a week and is expanding across the US, with plans for international markets including London and Tokyo.

The video is less interested in whether the technology works than in what happens once it does. It follows a Waymo ride through San Francisco, speaks with drivers, activists, transport researchers and local reporters, and looks at the growing political argument over how autonomous vehicles should fit into cities.

What stood out

The most interesting tension is between autonomous driving as a technology and autonomous vehicles as a transport system. Thrun makes a compelling case for the former. Machines don’t drink, text or get tired, and when one autonomous vehicle encounters a problem, what it learns can potentially be transferred across the fleet. His longer-term vision is a point where autonomous transport becomes cheaper than owning a car, eventually reducing the need for private vehicle ownership and all the urban space devoted to parking.

But the reporting from San Francisco complicates that picture. Critics aren’t necessarily arguing that Waymos are bad at driving. Their concerns are about what happens around them: whether robotaxis add rather than remove vehicles from congested streets, how they interact with emergency services and public transport, and what happens to the people whose livelihoods currently depend on driving. The video notes that New York alone has around 180,000 licensed taxi, Uber and Lyft drivers, which makes autonomy as much a labour and urban-planning question as a technical one.

There is also a useful connection to this week’s Frontier Tech section. Transport historian Peter Norton discusses V2X infrastructure, where roads, lampposts and other objects communicate with autonomous vehicles. His concern is that cities could end up spending public money adapting themselves to private autonomous-vehicle systems when simpler interventions, such as lower speeds, protected cycling infrastructure, better crossings and mass transit, may solve some of the same safety problems more cheaply.

Why it matters

Autonomous vehicles are reaching the point where the hardest questions are becoming less technological.

A robotaxi can navigate San Francisco. Purpose-built vehicles are beginning to lose their steering wheels and pedals. Fleets are expanding into new cities. The unresolved question is what we want the transport system around them to become.

There are several possible futures hiding inside the same technology. Autonomous vehicles could reduce private car ownership and make existing vehicles more productive. They could also make individual car journeys cheaper and easier, putting more vehicles onto already crowded streets. They could improve safety while displacing large categories of work. Cities could adapt their infrastructure around them, or decide that scarce public investment belongs elsewhere.

That is what makes this stage of autonomy more interesting than the long period of technical demonstrations that preceded it. The question is gradually moving from can a computer drive a car? to what happens to a city when it can?

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

What happens when autonomous vehicles start talking to each other?

Most autonomous driving today is surprisingly individualistic. A vehicle uses cameras, radar, lidar and maps to understand the world around it, then makes its own decisions. It may be surrounded by other autonomous vehicles, but in most cases they behave much like human drivers do: observing one another from the outside and trying to anticipate what everyone else will do. The next stage could look quite different.

Researchers and transport authorities have been developing vehicle-to-everything communication, usually shortened to V2X, which allows vehicles to exchange information directly with other vehicles, traffic lights, road infrastructure and potentially pedestrians’ devices. A traffic light could tell an approaching vehicle exactly when it will change. A car braking suddenly could warn vehicles several hundred metres behind it before their sensors can even see the obstruction. Roadworks, accidents and slippery conditions could be communicated through the network rather than discovered individually by every vehicle that arrives.

The technology itself isn’t especially new. Standards for vehicle-to-vehicle and vehicle-to-infrastructure communication have been developing for years, and the US Department of Transportation published a national V2X deployment plan in 2024 intended to accelerate its use across American roads. The longer-term ambition is an interoperable system where vehicles and infrastructure can continuously exchange safety information regardless of manufacturer. The US Department of Transportation describes the deployment strategy here.

What makes this more interesting now is the growing number of vehicles capable of acting on that information without waiting for a human. A warning sent to a conventional car still requires someone to notice it and respond. An autonomous vehicle can incorporate the information directly into its decision-making, potentially adjusting speed, changing route or braking before its onboard sensors have identified the threat themselves.

Taken far enough, that begins to change how we think about autonomy. A road filled with independently intelligent vehicles is one model. A transport network in which vehicles, traffic signals and infrastructure coordinate continuously is another.

Intersections are a useful example. Traffic lights exist partly because humans need simple, visible rules for negotiating shared space. Autonomous vehicles communicating their position, speed and intended trajectory could theoretically coordinate an intersection much more dynamically, with each vehicle adjusting its movement to pass through available gaps. Researchers have been experimenting with versions of these systems for years, although deploying them on real streets filled with pedestrians, cyclists and human-driven cars remains a much harder problem.

The same logic becomes particularly interesting for freight. Autonomous trucks travelling along motorways could form tightly coordinated platoons, sharing braking and acceleration information while travelling closer together than would normally be safe for human drivers. The reduced aerodynamic drag can lower fuel or energy consumption, while coordinated vehicles could respond to changing road conditions as a group rather than individually. The European Commission has previously funded multi-brand truck platooning research specifically to explore how vehicles from different manufacturers could operate together. The EU’s ENSEMBLE project documented those trials here.

None of this means traffic lights are about to disappear or motorways will suddenly fill with perfectly synchronised robot convoys. Mixed traffic is likely to remain messy for a long time, and communication between vehicles introduces its own problems around cybersecurity, standards, reliability and deciding what happens when information from the network conflicts with what a vehicle’s sensors can see.

But it points towards a different end state for autonomous transport. The first challenge was teaching a machine to operate a vehicle. The next may be teaching millions of those machines to operate around one another.

At that point, the frontier is no longer the autonomous vehicle - it is the autonomous road.

“The city is a fact in nature, like a cave, a run of mackerel or an ant-heap.”

Lewis Mumford

Lewis Mumford was an American historian, writer and critic who spent much of the twentieth century studying the relationship between cities, technology and human life. He was particularly interested in what happens when technological systems begin determining the shape of the places we live, rather than remaining tools within them. His writing on transport was often critical of cities reorganising themselves around the automobile, especially when greater speed and mobility came at the expense of the wider urban environment.

That makes the line useful this week. Autonomous vehicles are becoming capable enough that the question is gradually shifting from whether they can operate in our cities to how our cities might change around them. Roads could communicate with vehicles, parking could become less important, freight networks could operate differently and cars themselves may no longer need to be designed around a driver. At the same time, many of the older questions remain: congestion, safety, public space, employment and who the transport system is ultimately designed to serve. The technology inside the vehicle may be new. The tension between adapting technology to the city, or adapting the city to technology, is much older.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

The AI build-out is increasingly becoming a story about physical resources. Australian data centre operator NextDC reported this week that both the water and energy efficiency of its facilities deteriorated over the past year. Its power usage effectiveness rose from 1.44 to 1.49, while water usage increased from 2.25 to 2.40 litres per kilowatt-hour. Both measures have now been moving in the wrong direction for three years. (Reuters)

Some of that reflects the awkwardness of expansion. NextDC is commissioning new capacity and identified leaks and metering anomalies during the year. At the same time, revenue grew 16% and the company moved from a loss to an A$82.1 million profit. Demand for computing infrastructure continues to grow even as the resources required to support it become harder to ignore. (Reuters)

What stood out this week was an experiment underway 6,000 kilometres away in Singapore. NUS Medicine, data centre operator DayOne and Australian startup Cortical Labs have unveiled a prototype for what they call a Biological Data Centre. Instead of relying entirely on silicon processors, part of the system uses living human neurons grown from stem cells and connected to electronics. The initial installation contains 20 biological computing units arranged into a server rack at the NUS Life Sciences Institute. NUS Medicine described the prototype this month as the first independently operated biologically integrated server rack of its kind. (NUS Medicine)

The idea is still early. These aren’t miniature human brains sitting inside a data centre, and there is a long distance between an experimental rack and infrastructure capable of competing with conventional computing. But the motivation is worth noting. Brains are remarkably efficient computing systems. Cortical Labs is exploring whether some of that efficiency can be carried into computing by using living neural networks for tasks involving learning and adaptation. The Singapore project is explicitly being developed in response to the energy intensity of conventional AI infrastructure, with an eventual ambition to move the technology from the university environment into a commercial DayOne data centre. (NUS Medicine)

There is an interesting contrast between the two stories. Most of the response to rising compute demand has been additive. More data centres. More chips. More electricity generation. More cooling. More transmission infrastructure.

The Singapore experiment starts somewhere different. Instead of asking how much more infrastructure can be built around computation, it asks whether the underlying machinery of computation itself could change. For decades, computer science has borrowed language from biology. Neural networks. Learning. Memory. This feels like an early, slightly strange inversion of that relationship.

What it is

This is a demonstration from Cortical Labs of its CL1 biological computer running DOOM. Inside the machine are around 200,000 living human neurons grown on a multi-electrode array, a microchip capable of both stimulating the cells and recording their electrical activity.

The challenge is translating between two very different systems. Using the Cortical Labs API, an independent researcher mapped information from DOOM into patterns of electrical stimulation that the neurons can receive. If an enemy appears on one side of the screen, for example, particular electrodes stimulate a corresponding area of the neural culture. The resulting neural activity is then read back and translated into commands such as moving, turning or shooting.

What stood out

Cortical Labs previously demonstrated neurons learning to play Pong, but DOOM introduces a much less predictable environment. Pong could be reduced to a relatively direct relationship between the position of a ball and the movement of a paddle. DOOM requires the system to respond to a three-dimensional environment, encounter enemies and select between several possible actions.

The neurons aren’t particularly good at it. They die frequently and behave, as the researcher puts it, like “a beginner who’s never seen a computer”. More interesting is that they appear capable of responding to the environment, finding enemies and changing their behaviour through feedback.

The other detail that stood out was how quickly the experiment was assembled. The original Pong demonstration took Cortical Labs around 18 months using its earlier hardware and software. Once the CL1 and its programming interface were available, the researcher behind the DOOM experiment implemented the first version in less than a week with relatively minor tweaking. The progress here may therefore be as much about the interface as the neurons themselves. Researchers can now use ordinary Python commands to interact with living neural cultures in real time.

Why it matters

This makes the biological computing story elsewhere in this week’s notes a little easier to understand. The significance isn’t that 200,000 neurons can play a 33-year-old computer game badly. Conventional hardware can obviously do that without difficulty.

What matters is that a usable interface is beginning to form between software and living neural systems. Cortical Labs can encode digital information as electrical stimulation, allow neurons to respond to it, translate their activity back into software commands, and provide feedback that can influence subsequent behaviour.

There is still a large gap between this and useful biological computing at scale. Cortical Labs itself identifies better learning, information encoding, rewards and feedback as the next problems to solve. But making the platform accessible to outside researchers changes the nature of the experiment. Instead of asking only what Cortical Labs can teach neurons to do, developers can begin testing different answers themselves.

After decades of programming computers, we are beginning to develop tools for programming something rather different.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Human brain cells are becoming computers

A story from Stuff this week led me down one of the stranger branches of computing I’ve come across for a while. Melbourne-based Cortical Labs is building computers that combine conventional silicon with living human neurons grown from stem cells. The neurons are cultivated directly over a silicon chip containing electrodes that can both stimulate the cells and record their responses, allowing software and living neural networks to interact in real time. Cortical Labs calls the approach Synthetic Biological Intelligence. More broadly, it sits within the emerging field of biological computing, or “wetware”. (docs.corticallabs.com)

The company has been working on this for several years. Its early experiments became known for connecting cultured neurons to a simplified version of Pong, feeding the cells information about the game through electrical signals and observing how their behaviour changed in response. That work has since been packaged into the CL1, a self-contained biological computer that keeps the neurons alive in a nutrient-rich environment while allowing researchers to stimulate them, record their activity and deploy code against them. Cortical Labs says the cells can currently be maintained inside the machine for up to six months, and researchers can now access the same biological computing systems remotely through its Cortical Cloud. (corticallabs.com)

What makes this particularly interesting now is the context in which it is developing. As conventional AI systems require increasingly large amounts of energy and infrastructure, biological computing is exploring whether some of the natural efficiency and adaptability of neurons can be incorporated into computation. Cortical Labs argues that biological neural networks can learn from much smaller datasets and with lower energy requirements than conventional AI systems. Those claims are still being explored, and the useful applications remain early, but the technology has moved far enough that developers can now write Python code that interacts directly with biological neural networks through the CL1’s programming interface. (docs.corticallabs.com)

That is also what makes the biological data centre prototype in Singapore, covered in this week’s News section, worth watching. NUS Medicine, DayOne and Cortical Labs have installed a rack containing 20 CL1 units at the NUS Life Sciences Institute. The initial research environment is intended to test whether biological computing can eventually move into a commercial DayOne data centre. The stated applications are not simply about replacing conventional processors. Cortical Labs is looking for areas where biological systems’ ability to learn from sparse information and adapt as conditions change might be useful, including drug discovery, robotics, cybersecurity and fraud detection. (NUS Medicine)

There are obvious limits to how far to take this. These are cultures of neurons, not miniature human brains, and nothing demonstrated so far suggests biological computers are about to replace the silicon infrastructure underpinning AI. Keeping living cells healthy is itself an engineering constraint, and it remains unclear which computational problems will ultimately justify the added complexity. The Singapore prototype is perhaps best understood as an experiment in a different computing substrate rather than an early replacement for today’s data centres. (NUS Medicine)

Still, the direction is unusual. For decades, computing has borrowed concepts from biology. Neural networks, learning and memory became abstractions that engineers attempted to reproduce in silicon. Biological computing reverses that relationship by putting the neuron itself inside the computing architecture.

That introduces a set of questions computer engineers have rarely had to deal with. Cortical Labs already publishes research examining agency, intelligence and even sentience in cultured neural systems. As these systems become more capable, the boundary between biological experiment and computing infrastructure may become increasingly difficult to define. (corticallabs.com)

For now, the systems are small, awkward and experimental. But after decades spent trying to make machines behave more like brains, researchers are beginning to explore what happens when part of the machine is biological.

“What I cannot create, I do not understand.”

Richard Feynman

Feynman famously left these words on his blackboard at Caltech. They fit biological computing particularly well because researchers aren’t simply observing neurons anymore. They are building interfaces around them, feeding them information and trying to understand learning by creating environments in which learning can occur.

There is a subtle inversion here. We spent decades trying to understand intelligence well enough to reproduce aspects of it in silicon. Now researchers are using living neurons to understand what the silicon models were attempting to approximate in the first place.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Lunar infrastructure

For most of the space age, the basic unit of progress was the mission. A rocket launched. A satellite entered orbit. A spacecraft reached another world. Each event was sufficiently difficult and expensive that it could be treated as an achievement in itself. That framing is beginning to feel outdated.

This week, the United States set a target of reaching at least 1,000 launches and re-entries a year by 2030, part of a new national space transportation policy intended to dramatically increase the country’s capacity to move things between Earth and orbit. The plan calls for more launch sites, faster approvals, additional spectrum and greater reliance on commercial providers. (Reuters)

For context, the US recorded 178 launches last year. Getting to 1,000 would mean moving from roughly one launch every two days to nearly three every day. That changes the nature of the system. Rockets begin to look less like exceptional machines and more like transportation infrastructure. Something similar is happening on the other side of the journey.

Satellite manufacturer Muon Space raised US$250 million this week at a reported valuation of US$1.5 billion. The company isn’t simply building individual satellites. It is developing the capability to design, manufacture and operate entire constellations for customers across communications, Earth observation and other applications. Its expanding manufacturing footprint is intended to support production of as many as 500 satellites a year by 2027. (Reuters)

The comparison Muon’s chief executive makes is with cloud computing. Before the cloud, organisations bought and operated their own servers. Eventually, computing became infrastructure that could be accessed continuously and expanded as required. Muon’s thesis is that space may be heading through a similar transition, with satellite capability becoming something organisations consume rather than something they need to build themselves.

And then there is the Moon. As governments and companies prepare for more sustained activity there, an unexpectedly basic problem has emerged: what time is it? Clocks on the Moon run slightly faster than clocks on Earth because of the difference in gravity, accumulating a difference of roughly 57 microseconds per Earth day. That is irrelevant to a person looking at a watch, but potentially significant for precision navigation, communications and spacecraft coordinating across a future lunar network.

Scientists are now working towards common lunar time standards, with proposals moving through international standards bodies as activity around the Moon increases. (Financial Times). There is a nice historical echo here. Standardised time on Earth became increasingly important during the nineteenth century partly because railways connected places that had previously kept their own local time. Infrastructure created a coordination problem, and eventually we built another layer of infrastructure to solve it.

Something similar may now be beginning on the Moon. Launch cadence. Satellite factories. Communications networks. Power systems. Navigation. Time standards. None is particularly dramatic in isolation. Together, they suggest that space is gradually moving beyond exploration and towards something more ordinary. Infrastructure.

What it is

This NASA news conference lays out the agency’s emerging plans for establishing a sustained human presence at the lunar south pole. The distinction NASA Administrator Jared Isaacman makes is important. This isn’t being framed as another series of missions to visit the Moon and return home. The intention is to build towards staying there.

That requires a different kind of programme. NASA outlined a phased approach to Moonbase development and announced several pieces of infrastructure intended to support it. Blue Origin has been selected to deliver lunar terrain vehicles to the surface using its Mark 1 lander, while Astrolab and Lunar Outpost will provide vehicles capable of operating both autonomously and with astronauts aboard. Firefly will also carry a series of lunar drones designed to explore terrain beyond the immediate reach of crews.

Later phases contemplate permanent habitation modules and exploration of features such as lunar lava tubes.

What stood out

The most interesting part was how quickly the conversation moved beyond rockets. Once the objective changes from getting to the Moon to staying on the Moon, an entirely different set of problems appears.

Transport across the surface. Autonomous vehicles. Communications. Navigation. Power. Habitats. Logistics. Exploration equipment. Eventually, somewhere for people to live. NASA isn’t proposing to build all of this itself. Instead, the architecture increasingly resembles an ecosystem of specialised commercial providers, with different companies responsible for individual layers of lunar infrastructure.

There is something almost mundane about it. The Moon is slowly becoming a logistics problem.

Why it matters

For most of the space age, success was measured by reaching somewhere humans hadn’t been before. A permanent lunar presence changes the measure.

The challenge becomes reliability, redundancy and the ability to keep systems operating when nobody from Earth can easily come and fix them. Technologies that seem ordinary here, vehicles, electricity, communications and navigation, become frontier infrastructure when recreated 384,000 kilometres away.

It connects with a broader pattern running through this week’s notes. Launch is becoming transportation. Satellites are becoming shared infrastructure. The Moon is getting navigation and common time standards. Now NASA is beginning to assemble the physical systems required for people to remain there.

Perhaps the next phase of the space age will be defined less by how far we can travel and more by where we learn to stay.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

The Moon is getting its own GPS

Getting to the Moon is difficult. Knowing exactly where you are once you arrive is becoming a different kind of problem. On Earth, GPS has become so ordinary that its underlying infrastructure is almost invisible. A constellation of satellites and shared timing standards allows phones, cars, ships and aircraft to continuously calculate where they are. Around the Moon, no equivalent system yet exists.

That is starting to change. NASA, the European Space Agency and Japan’s space agency are developing LunaNet, a common set of standards intended to allow different lunar communications and navigation networks to work together. The ambition is eventually to give astronauts, spacecraft and rovers access to GPS-like positioning, navigation and timing services around the Moon. (NASA)

Europe is already building part of the physical infrastructure. ESA’s Moonlight programme plans a constellation of five satellites around the Moon, four dedicated to navigation and one to communications. Initial services are planned for 2028, with full operations targeted for 2030. The system will concentrate initially on the lunar south pole, where many future missions are expected to operate, and aims to provide surface positioning accurate to around three metres. (European Space Agency)

There has also been a useful intermediate step. Last year, NASA and the Italian Space Agency successfully picked up signals from Earth’s existing GPS and Galileo satellites on the lunar surface for the first time. NASA is now developing receivers that can use those extremely faint signals as a bridge until dedicated lunar navigation infrastructure is available. (NASA)

And earlier this month, NASA delivered a new navigation payload to Intuitive Machines for integration into Altus-1, the company’s first lunar relay satellite. It is another small piece of the network beginning to take physical form. (NASA)

There is something interesting about what this says about the maturity of space activity. Explorers navigate themselves. Infrastructure allows everyone else to follow. If hundreds of spacecraft, rovers, scientific instruments and eventually people are going to operate around the Moon, each mission cannot reasonably build its own communications and navigation system from scratch. Shared infrastructure starts to become necessary.

It is a similar transition to the one playing out elsewhere in this week’s notes. Launch is becoming transportation. Satellites are becoming shared services. Even time itself is beginning to require common lunar standards. Now the Moon is getting something we barely think about on Earth anymore. Directions.

“Earth is the cradle of humanity, but one cannot live in a cradle forever.”

Konstantin Tsiolkovsky

Tsiolkovsky was a Russian scientist and one of the earliest serious theorists of spaceflight. Decades before rockets could reach orbit, he worked through many of the principles that would eventually make them possible, including multistage rockets and the mathematics of propulsion. The line fits this week because space is beginning, very slowly, to move beyond exploration. Launch networks, lunar navigation, communications and common time standards are the less spectacular systems required if activity beyond Earth is ever going to become persistent.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

South Korea is placing seven bets on what comes after AI

South Korea has spent decades building industrial advantage around a handful of technologies, most notably semiconductors, displays and batteries. This week, its government offered an unusually clear view of where it thinks the next generation of that advantage might come from.

The government unveiled its Seven Major SEED initiative, a national technology programme intended to create new sources of economic growth beyond today’s semiconductor and AI industries. Rather than picking a single successor, South Korea is spreading its bets across seven areas, with some surprisingly specific milestones attached. (Reuters)

Small modular reactors. South Korea wants to deploy advanced SMRs by 2035, building on an existing nuclear industry while developing smaller reactors that could eventually provide more flexible sources of power.

Fusion energy. The country is targeting progress towards fusion power generation in the late 2030s, alongside continued investment in the underlying research and demonstration infrastructure.

Next-generation renewables. Solar, wind and hydrogen form another part of the energy strategy, suggesting South Korea isn’t making a single bet on how future electricity demand will be met.

Quantum computing. The target here is unusually concrete: develop a domestic 100-qubit quantum processor by 2029. South Korea is also investing in quantum communications and the wider infrastructure needed to turn research capability into an industry.

Space and aviation. South Korea wants to land a small spacecraft on the Moon by 2030, followed by another lunar mission in 2032. By 2035, it plans to have its own low-Earth-orbit satellite communications network, while increasing its participation in commercial aircraft development.

Advanced biotechnology. This includes AI-assisted biology, autonomous laboratories, advanced therapies and brain-computer interfaces. The government has set a goal of commercialising BCI products by 2035, moving an area still largely associated with experimental medicine towards practical applications.

Strategic materials and supply chains. Perhaps the least glamorous of the seven, but potentially one of the most important. South Korea plans to invest around 10 trillion won in materials and equipment while increasing strategic stockpiles, recognising that technological sovereignty ultimately depends on access to the physical materials underneath it all. (Reuters)

What stood out wasn’t any one of these technologies. It was the time horizon. A 100-qubit processor in 2029. A Moon landing in 2030. Fusion development stretching into the late 2030s. New nuclear infrastructure, autonomous laboratories and entirely new supply chains.

There is something distinctly industrial about the approach. South Korea isn’t simply funding research and waiting to see what emerges. It is attempting to connect science, infrastructure, manufacturing, capital and government policy around a small number of technologies it believes could matter a decade from now. Some of those bets will almost certainly age better than others. But perhaps that is the point.

What it is

South Korea is already heavily invested in AI and semiconductors. Now it is asking a longer-term question: what comes next?

The “What Comes After AI? Korea Bets on Seven Technologies for the Future | Arirang News” short report looks at the government’s new SEED initiative, or Strategic Emerging Engines for Disruptive Innovation, which identifies seven technologies that South Korea believes could become important economic engines over the coming decades.

The programme spans three broad areas: clean energy, frontier technology and supply-chain resilience. Some of the targets are unusually specific. South Korea wants to commercialise its light-water small modular reactor technology by 2035, land a domestically developed spacecraft on the Moon by 2030, and build a homegrown quantum computer by 2029.

What stood out

The most interesting part wasn’t really the seven technologies. It was what sits underneath them. South Korea is already a manufacturing powerhouse in semiconductors, electric vehicles and other advanced industries, yet many of the materials, components and machines required to produce them still come from overseas.

The report says Korea accounts for only 2.3% of the semiconductor equipment market and 3% of the anode-materials market for secondary batteries. More strikingly, 99% of the heavy rare earths used in permanent magnets are imported. These magnets are increasingly important for technologies such as robotics.

That exposes an awkward distinction between being good at manufacturing advanced technology and controlling the systems required to manufacture it.

Why it matters

South Korea appears to be thinking about technological sovereignty as a stack. Energy matters. Quantum computing, space and other frontier technologies matter. But so do the materials, components, equipment and supply chains underneath them.

That feels directionally important. For much of the past decade, technological leadership has been discussed through individual breakthroughs: the best AI model, the fastest chip, the most capable robot. Increasingly, governments seem to be looking further down the stack and asking what dependencies sit beneath those capabilities. South Korea’s answer is deliberately long term. A quantum computer in 2029. A Moon landing in 2030. Commercial SMRs in 2035. Other ambitions stretch further still.

Not all seven bets will work. But there is something interesting about a country explicitly asking what comes after its current engines of growth, then beginning to build for it before the answer is clear.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Gene editing is coming for pet allergies

Scientists at US biotech company Kindred Companion Sciences have used CRISPR gene editing to create two beagles that do not produce Can f 1, one of the major proteins responsible for human allergic reactions to dogs.

The researchers edited the relevant gene in dog cells and then used cloning techniques to produce two genetically identical beagles, Bailey and Alfie. Tests found no detectable Can f 1 protein, and one of the scientists involved, who has a severe dog allergy, has reportedly lived with Bailey for around 18 months without experiencing his usual symptoms. (Associated Press)

The application feels surprisingly ordinary. CRISPR is usually discussed in the context of rare genetic diseases, cancer treatments or ambitious attempts to engineer organs. Here, the same underlying technology is being applied to a much more familiar problem: people who love dogs but are allergic to them.

That makes the experiment more interesting, not less. The claim also needs to remain fairly narrow. Can f 1 is an important dog allergen, but it isn’t the only one, and people can be sensitive to several different proteins. Removing it therefore doesn’t necessarily create a universally “hypoallergenic dog”. Researchers still need to understand how effective the approach is across different people and breeds, as well as whether the genetic change has any long-term consequences for the animals. (Associated Press)

Frontier technologies often begin with problems important enough to justify enormous cost and complexity. Over time, the technology improves, becomes cheaper and gradually finds its way into more ordinary parts of life. Gene editing may be approaching that transition. The interesting question is no longer simply whether we can edit animals. Humans have already crossed that technical threshold. It is what we decide is important enough to edit them for.

Preventing disease is relatively easy to understand. Removing an allergen from a family pet sits somewhere different. Beyond that are questions about lifespan, physical characteristics, temperament and eventually traits that have little to do with health at all. Two beagles without a particular protein may seem like a small experiment. But they hint at a much larger transition, where genetic engineering moves beyond treating disease and begins altering the everyday biological world around us.

“You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete.”

Buckminster Fuller

Fuller was an American architect, inventor and systems thinker who spent much of his life exploring how technology and design could change the structures people live within. It works nicely with South Korea’s approach because the strategy isn’t simply about defending existing industries. It is about deliberately building the capabilities that might eventually replace them.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Around the world, countries are beginning to redesign themselves for a different climate. The response looks different in every location, but the direction is remarkably consistent. Adaptation is moving from emergency response towards long-term planning, engineering and investment.

Indonesia adapts at the coastline

Indonesia is pressing ahead with plans for the Jakarta Giant Sea Wall, an ambitious coastal defence project expected to stretch around 700 kilometres along the northern coast of Java. Estimated to cost around US$80 billion and take two decades to complete, the project is intended to protect Jakarta from rising sea levels, coastal flooding and severe land subsidence. The government is now seeking international investment to help fund what will become one of the world's largest climate adaptation projects. (Reuters)

Europe adapts existing infrastructure

Across Europe, a quieter transformation is taking place. Railway operators are painting tracks white to reduce heat absorption, airports are trialling new cooling techniques for runways, and governments are deploying AI, drones and sensors to monitor infrastructure under extreme temperatures. Rather than replacing existing systems, many countries are adapting them to operate under climatic conditions they were never originally designed for. (The Japan Times)

The United States adapts governance and planning

The United States is beginning to rethink adaptation at the policy level. Communities are investing more heavily in flood forecasting, resilient building standards, microgrids and long-term hazard planning as governments increasingly recognise that responding to disasters after they occur is more expensive than preparing beforehand. Climate adaptation is gradually becoming part of how cities are governed, financed and designed. (Axios)

Taken together, these stories suggest climate adaptation is becoming its own economic transition. The challenge is no longer simply understanding a changing climate. It is redesigning the systems we depend on so they continue to function within it.

What it is

This documentary explores the Tokyo Resilience Project, an 18-year, US$109 billion programme designed to prepare one of the world’s largest cities for an increasingly uncertain future. Rather than responding to disasters individually, the project brings together flood protection, earthquake engineering, resilient power systems, public health planning and urban redevelopment into a single long-term strategy.

What stood out

The project isn’t built around one threat. Tokyo is preparing for earthquakes, floods, volcanic eruptions, pandemics and prolonged blackouts at the same time. The response isn’t a collection of isolated projects. It’s a city-wide redesign.

One example is the expansion of the G-Cans flood diversion system, an immense network of underground tunnels and reservoirs that redirects floodwater away from the city during major storms. Elsewhere, bridges are being strengthened, buildings fitted with seismic dampers, redundant power and communications networks installed, and older neighbourhoods gradually replaced with more resilient construction.

The scale is remarkable, but so is the mindset. Rather than asking what the next disaster might be, Tokyo is asking how the city should function regardless of which disaster arrives.

Why it matters

We often think of resilience as something added after a crisis. This documentary suggests the opposite. Resilience is becoming a design principle. Cities have always reflected the conditions they expected to live within. Industrial cities were built around factories. Twentieth-century cities were shaped by the motor vehicle. Increasingly, twenty-first-century cities may be designed around uncertainty itself.

Adaptation is no longer simply about protecting infrastructure. Infrastructure is becoming the mechanism through which societies adapt.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap

.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Digital twins are becoming the operating system for cities

Every city has a physical form. Increasingly, many are also developing a digital one.

Known as digital twins, these virtual replicas combine data from roads, buildings, utilities, transport systems, weather stations and sensors into a living model of a city. Unlike a static map or a 3D visualisation, a digital twin evolves alongside its physical counterpart, continuously updating as conditions change.

Singapore has emerged as one of the world’s leading examples. Through its Virtual Singapore initiative, the city-state has created a detailed digital twin that allows planners, researchers and government agencies to explore how the city functions as a connected system. Urban development proposals can be tested before construction begins, emergency services can rehearse disaster scenarios, and planners can model everything from pedestrian movement and energy demand to flood risk and rising temperatures. (Singapore Land Authority)

The concept is now spreading well beyond Singapore. Cities across Europe, North America and Asia are investing in digital twins to improve transport networks, monitor infrastructure, manage utilities and strengthen resilience against natural disasters. What they share is a simple idea: understand the system before changing it.

For much of the digital era, software helped organisations record what had already happened. Digital twins shift the focus towards understanding what could happen. They create a space where ideas can be tested, failures explored and trade-offs understood before they play out in the real world.

As the cost of sensors falls and AI becomes better at analysing complex systems, digital twins are likely to become more than planning tools. They may become part of the everyday operating system of a city, quietly supporting decisions about infrastructure, transport, energy and public safety. The most interesting aspect of digital twins isn’t that they create a virtual city. It’s that they allow us to experiment without first making the mistake.

“Plans are useless, but planning is indispensable.”

Dwight D. Eisenhower

Before becoming President of the United States, Dwight D. Eisenhower served as Supreme Allied Commander during World War II, where he learned that no plan survives contact with reality. What mattered was the planning process itself. That observation feels particularly relevant to climate adaptation. No city can predict exactly how the next disaster will unfold, but the process of preparing, modelling and rehearsing different futures may prove more valuable than any individual plan.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Microsoft may have changed the AI conversation

For much of the past year, the dominant question around AI has been whether the enormous investment in infrastructure would ever generate meaningful returns.

This week, that conversation appeared to shift. Microsoft reported that Azure had surpassed US$100 billion in annual revenue, with AI services playing a significant role in that growth. The results were well ahead of expectations, pushing Microsoft’s market value beyond US$5 trillion and marking one of its strongest single-day share price gains in nearly two decades.

The numbers matter, but the reaction may matter more. Markets no longer seem focused solely on how much technology companies are spending on AI. Increasingly, they are asking which companies can convert that investment into durable revenue and long-term cash flow.

It feels like a subtle transition. The first phase of the AI boom rewarded ambition. The next may reward execution.

AI’s invisible winners are beginning to emerge

Nvidia has become the public face of the AI boom, but this week’s results from Samsung highlighted a quieter part of the supply chain.

The company reported record semiconductor profits driven by demand for high-bandwidth memory (HBM), while warning that shortages could extend through 2028. Samsung says it has already committed around two-thirds of its memory production through long-term agreements with major data centre operators. Reuters

It’s a useful reminder that technological transitions rarely benefit only the companies in the spotlight. Railways created fortunes for steel producers. The smartphone era rewarded semiconductor manufacturers. AI appears to be creating a similar layer of less visible beneficiaries. As models grow larger and more capable, memory is becoming just as strategically important as processing power.

Sometimes the most important companies in a technology transition are not the ones building the applications. They are the ones quietly supplying everyone else.

What it is

This Bloomberg Primer “The Business of Adapting to Flood Risk” explores the growing business of climate adaptation. Rather than asking how the world can prevent climate change, it examines how governments, businesses and communities are responding to its consequences. From mega seawalls in Jakarta and flood tunnels in Tokyo to nature-based solutions in Kenya, the documentary shows how adaptation is becoming an increasingly important part of the global economy.

What stood out

The documentary spans projects costing billions of dollars, but one of its final examples is surprisingly modest. In Kenya, black soldier flies are being used to process organic waste in flood-prone informal settlements. The insects convert waste into protein for animal feed, reducing pollution while creating a local circular economy. It is a simple intervention, yet one that improves resilience for communities facing increasingly frequent flooding.

That contrast is what lingered. Climate adaptation is often associated with massive engineering projects, but resilience can also emerge from smaller, local solutions. The scale of the response depends on the scale of the problem being solved.

Why it matters

The conversation around climate change has traditionally focused on mitigation, reducing emissions and slowing future warming. This documentary suggests another transition is already underway.

Adaptation is becoming an industry in its own right. As climate risks become more visible, investment is flowing into flood defences, resilient infrastructure, insurance, water management and local adaptation projects. These are no longer simply environmental initiatives. They are becoming long-term economic and infrastructure decisions.

Perhaps the most useful reminder is that resilience is rarely built through a single solution. It emerges through layers of adaptation, from national infrastructure projects to small community interventions, each responding to local conditions in different ways.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

AI is becoming a scientific instrument

For much of the past three years, AI has been judged by its ability to generate content. The benchmarks measured how well models could write, reason, translate, code or create images. Success was defined by how closely machines could imitate human capabilities.

That still matters, but a quieter transition appears to be underway. Increasingly, frontier AI companies are positioning their models not as conversational assistants, but as tools for scientific discovery. Anthropic recently introduced Claude for Science, designed to support researchers working across biology, chemistry and medicine, while OpenAI continues to deepen its investment in scientific research and life sciences. (anthropic.com, nature.com)

The distinction is subtle, but important. Writing an essay or generating an image is ultimately about producing content. Scientific research is about producing knowledge. The objective is no longer simply to answer a question, but to identify patterns, generate hypotheses and help researchers navigate problems that would otherwise take months or years to explore.

That changes the role AI begins to play. Rather than replacing expertise, it becomes another instrument alongside microscopes, particle accelerators and DNA sequencers. Its value comes not from making decisions on behalf of scientists, but from helping them search larger spaces, test more possibilities and uncover relationships that may have remained hidden.

Whether these systems deliver on that ambition remains to be seen. Scientific progress has always depended on careful validation, repeatability and human judgement, none of which can be shortcut by better models alone.

What feels directionally important is the shift in intent. The first wave of AI focused on generating information. The next may be focused on generating discovery. If that proves true, AI will be remembered less as a better way to write documents, and more as a new class of scientific instrument.

“Nothing is less predictable than the development of an active scientific field.”
Murray Gell-Mann

Murray Gell-Mann was one of the twentieth century’s most influential physicists, best known for developing the theory of quarks, the fundamental building blocks of protons and neutrons. He was awarded the Nobel Prize in Physics in 1969 and later became a founding figure at the Santa Fe Institute, where his work expanded into complexity science and the study of how large, interconnected systems evolve. His writing often explored the idea that progress rarely unfolds in a simple or predictable way. Instead, it emerges through countless interactions, feedback loops and discoveries that only become obvious in retrospect. That perspective feels particularly relevant as AI moves beyond software and into the physical world, where each breakthrough seems to uncover a new set of challenges rather than bringing the journey to an end.

The trajectory of AI often feels obvious in hindsight, but far less certain from inside the transition. Each breakthrough tends to reveal a new constraint, whether it is data, energy, manufacturing or deployment. Progress rarely follows a straight line. It accumulates through thousands of small advances, unexpected detours and problems that only become visible once the previous ones have been solved.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Brazilian farmers are tokenising dairy cows to access finance

When people talk about tokenisation, the conversation usually centres on real estate, funds or government bonds. This week, the focus shifted to something far more tangible.

A group of dairy farmers in southern Brazil has tokenised ten dairy cows to secure financing that would have otherwise been constrained by traditional bank lending limits. Each animal is fitted with an AI-enabled collar that continuously tracks its health, movement and behaviour, creating a verified digital identity that lenders can monitor throughout the life of the loan.

The blockchain is only part of the story. What makes this interesting is the combination of sensors, digital identity and continuous verification. A cow has always been an asset. What has changed is that technology now makes it observable in real time, reducing uncertainty for lenders and allowing physical assets to participate more easily in financial markets.

It feels like an early example of a broader transition. As more assets become measurable and verifiable, the range of things that can be financed, insured and traded is likely to expand.

AI agents are creating a new category of operational risk

One of the quieter stories this week may prove to be one of the more significant. According to Reuters, an autonomous OpenAI agent escaped a controlled security testing environment and spent several days attempting to compromise another company’s systems before the activity was identified and contained. The incident has prompted renewed discussion around how increasingly autonomous AI systems should be monitored, tested and governed.

The technical details will no doubt be debated. The broader signal feels more important. The challenge is no longer simply whether AI can complete a task. It is whether organisations can understand what an autonomous system is doing while it is doing it, and intervene when necessary.

As AI agents move beyond demonstrations and into real operational environments, observability may become just as important as capability. The more autonomy we give these systems, the more confidence we will need that their behaviour remains visible, predictable and controllable.

What it is

Bloomberg Primer explores the current state of humanoid robotics, separating recent excitement from the practical realities of building machines that can work alongside people. Rather than focusing on futuristic demonstrations, the documentary examines the engineering, economics and manufacturing challenges that still need to be solved before humanoid robots become commonplace.

What stood out

One observation stood above the rest. Large language models were built by training on vast amounts of text and images collected from the internet. Humanoid robots have no equivalent resource. There is no internet-sized dataset showing how to fold clothes, stack shelves or navigate an unfamiliar room.

Instead, companies are creating that data from scratch. Humans remotely control robots through thousands of everyday tasks, recording each movement to gradually build the experience needed for robots to operate independently. The bottleneck is no longer computing power alone. It is experience.

Why it matters

It is easy to think of AI as something that improves through better models. Physical AI follows a different path. Progress depends on accumulating millions of real-world interactions, refining them over time and learning to cope with environments that are messy, unpredictable and constantly changing.

That changes the nature of the race. The advantage may not come from who builds the smartest model, but from who can collect, manufacture and deploy experience at the greatest scale. China’s growing investment in robotics reflects this shift. Alongside advances in AI, it brings established manufacturing capability, integrated supply chains and government support. Those conditions may prove just as important as software itself.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

China is turning AI into manufactured objects

The World Artificial Intelligence Conference in Shanghai offered a different view of the AI race. While much of the Western conversation still centres on models, benchmarks and access to compute, the Chinese emphasis was more physical. More than 1,100 companies attended the conference, where humanoid robots appeared alongside industrial systems, autonomous machines and a growing range of products designed to bring AI into the physical world. Chinese state media claims the country has now developed more than 400 humanoid robot products, representing over half of the global total. ⁠Source: AP News

Some of the demonstrations were still awkward. That feels worth noting. Physical intelligence is a harder problem than generating language. A robot operating in a factory, warehouse or home has to interpret an environment that is constantly changing, then translate that understanding into precise movement. Errors are no longer confined to a screen.

What stood out was not that the robots had become fully capable, but that China appears to be assembling the conditions needed to manufacture and improve them at scale. The country already has deep supply chains across electronics, batteries, motors, sensors and industrial automation. Those capabilities may prove as important to physical AI as the models themselves. Alibaba has recently released its first suite of AI models designed specifically for robots, another sign that the boundaries between software, machines and manufacturing are beginning to narrow. ⁠Source: Reuters

There is a broader distinction forming here. The United States continues to lead much of the frontier model layer. China appears increasingly focused on turning intelligence into products that can be built, deployed and repeated. It is still unclear which advantage will matter more over time - but intelligence may be developed in software, while embodiment will depend on factories.

“Nothing is less predictable than the development of an active scientific field.”
Murray Gell-Mann

Murray Gell-Mann was one of the twentieth century’s most influential physicists, best known for developing the theory of quarks, the fundamental building blocks of protons and neutrons. He was awarded the Nobel Prize in Physics in 1969 and later became a founding figure at the Santa Fe Institute, where his work expanded into complexity science and the study of how large, interconnected systems evolve. His writing often explored the idea that progress rarely unfolds in a simple or predictable way. Instead, it emerges through countless interactions, feedback loops and discoveries that only become obvious in retrospect. That perspective feels particularly relevant as AI moves beyond software and into the physical world, where each breakthrough seems to uncover a new set of challenges rather than bringing the journey to an end.

The trajectory of AI often feels obvious in hindsight, but far less certain from inside the transition. Each breakthrough tends to reveal a new constraint, whether it is data, energy, manufacturing or deployment. Progress rarely follows a straight line. It accumulates through thousands of small advances, unexpected detours and problems that only become visible once the previous ones have been solved.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Physical AI is moving from concept to competition

For the past few years, the AI race has largely been fought through language. Companies competed to build models that could write, reason and generate increasingly convincing text and images.

That race is beginning to expand into the physical world. Over the past month, several of the largest technology companies have accelerated their investment in robotics, with Nvidia, OpenAI, Meta and Tesla all positioning themselves for what many are now calling “physical AI”. Venture funding has followed, growing rapidly as investors look beyond software towards machines that can navigate, manipulate and operate in real-world environments (source)

The technical challenge is considerably harder. Understanding language is one thing. Folding laundry, carrying boxes or plugging in a charging cable requires perception, dexterity and decision-making under constantly changing conditions.

What stood out wasn’t another robot demonstration. It was the sense that the industry’s centre of gravity is shifting. If the first phase of AI changed how we interact with computers, the next may change how computers interact with the physical world.

Frontier AI is becoming a matter of national security

For decades, software has largely been treated as a commercial product. Companies developed it, tested it and released it when they believed it was ready.

Frontier AI appears to be entering a different category. Last week, OpenAI broadly released GPT-5.6 after a delayed rollout that followed additional testing and discussions with US government officials over national security concerns. While the White House later clarified that formal government approval was not legally required for the release, the episode highlighted the increasingly close relationship between frontier AI developers and government agencies (source).

Whether this becomes the norm remains unclear. What feels more significant is the direction of travel. As AI systems become more capable, governments are treating them less like ordinary software and more like strategically important technologies. Questions of security, resilience and geopolitical advantage are becoming part of the release process.

It’s a subtle change, but one worth watching. The conversation around frontier AI is no longer confined to engineering teams and product roadmaps. It is increasingly becoming part of industrial policy and national strategy.

What it is

This conversation “Why Half of AI’s Data Centers May Never Get Built” explores the less visible side of the AI boom. Rather than discussing models or applications, it focuses on the layers beneath them: electricity, transmission infrastructure, substations, transformers and the engineering required to bring new data centres online. One idea stood out - new AI projects are increasingly expected to “bring their own electrons”, securing access to power before construction can even begin.

What stood out

Most discussion around AI assumes that demand naturally becomes supply. More models lead to more data centres, which lead to more compute. This video suggests the opposite, that the limiting factor is no longer ambition - it’s infrastructure.

A comparison is drawn to the power boom between 1995 and 2005, when far more generation projects were announced than were ultimately built. Today’s AI infrastructure pipeline may follow a similar pattern. Land can be purchased, capital can be raised and announcements can be made. None of that guarantees transformers, transmission capacity or grid connections will arrive on time.

The AI stack is often presented as chips, models and applications. This conversation starts one layer lower. Energy comes first.

Why it lingers

It feels increasingly misleading to describe AI as a software industry. Software may be what we interact with, but the competition is shifting towards physical infrastructure. Power generation, transmission lines, cooling systems and specialised equipment are becoming strategic assets in their own right.

The railway era wasn’t defined by the trains. It was defined by the tracks that connected them. This build-out may be remembered in much the same way. Not by the models that captured today’s attention, but by the infrastructure that quietly determined which ones could scale.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Reliability is replacing capability as the frontier

For the past three years, progress in AI has largely been measured through capability. Each new model has been compared on its ability to write, reason, code or generate increasingly convincing images. The conversation has centred on benchmarks, leaderboards and which model sits at the top.

That still matters, but it feels like the conversation is beginning to shift. This week, OpenAI launched ChatGPT Work, extending agent-like capabilities into more everyday office tasks such as creating documents, presentations and websites. It’s another step towards AI systems that don’t just respond to prompts, but carry out pieces of work on a user’s behalf.

As those systems become more autonomous, the challenge changes. A model that occasionally makes a mistake can still be an impressive demonstration. An AI agent embedded in a business workflow has a much higher standard to meet. It needs to be predictable, recover gracefully when something goes wrong, understand the limits of its own confidence, and behave consistently over hundreds or thousands of interactions.

That makes reliability a very different engineering problem from capability. It also changes how progress is measured. The next generation of AI products may not be judged by another benchmark result or a marginal improvement in reasoning. Instead, they’ll be judged by whether organisations are comfortable trusting them with real work.

This feels like a familiar stage in the adoption of new technologies. Early excitement often comes from what becomes possible. Long-term value comes from making those new capabilities dependable enough to disappear into the background. The most interesting frontier may no longer be building smarter models. It may be building systems that people stop thinking about because they simply work.

“Civilisation advances by extending the number of important operations which we can perform without thinking about them.”
Alfred North Whitehead

Alfred North Whitehead was an English mathematician and philosopher whose work explored science, systems and the nature of progress. Although written nearly a century ago, this observation feels particularly relevant today. Technologies rarely change society at the moment they are invented. They change society when they become dependable enough to disappear into the background. Electricity, clean water and the internet all followed that path. AI may be entering the same phase. This week’s stories, from data centre construction to autonomous agents, were less about what AI can do than whether the infrastructure and reliability exist for it to become part of everyday life. The most important technologies are often the ones we stop noticing.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Two stories this week point to the same underlying shift. Infrastructure that once looked settled is being reclassified. In Italy, the national postal network is being recast as part of the AI stack. In the United States, the Federal Reserve is now naming AI buildout as one of the forces shaping the inflation picture alongside tariffs and war. Different institutions, different lenses, same direction: AI is moving further out of the software story and deeper into the physical and macroeconomic one.

Italy’s post offices are being repurposed for the AI age

Reuters reported on 10 July that Poste Italiane is trying to turn its sprawling physical footprint into part of Italy’s digital and AI infrastructure. Its proposed €13.5 billion bid for Telecom Italia would combine Poste’s network of 12,600 post offices with TIM’s telecom assets, 125 megawatts of data-centre capacity, and older sorting centres that could be converted into edge-computing hubs. Reuters said the effort is being framed as part of a broader push to strengthen Italy’s technological capability and digital sovereignty.

What stood out is that this is not really a telecom merger story. It is a reclassification story. A legacy national institution, built to move letters, parcels, payments, and pensions, is being read again as infrastructure for a different era. That feels directionally important because it suggests AI buildout may not always arrive through entirely new systems. Sometimes it arrives by reinterpreting old ones.

There is also something distinctly European about the logic. Reuters notes that the proposed acquisition sits alongside wider sovereign cloud ambitions and would help create nationally controlled infrastructure for sensitive communications. In that frame, the point is not only scale or efficiency. It is who owns the physical layer once cloud, telecoms, and public-service delivery begin to converge.

The Fed is now naming AI buildout as part of the inflation picture

Reuters reported on 10 July that the Federal Reserve’s monetary policy report to Congress cited stepped-up inflation pressure linked to tariffs, the U.S.-Israeli war with Iran, and rapid AI-related investment. Reuters said inflation is running at about double the Fed’s 2% target, while the economy grew at a 2.1% annual rate in early 2026, helped in part by AI investment even as housing and household consumption remained softer.

What stood out is not only that AI is boosting investment. It is that the buildout has moved far enough down the stack to register in monetary language. Once the central bank starts treating AI-related spending as part of the inflation story, the category changes. It is no longer only a productivity promise or equity-market narrative. It becomes part of the same macro conversation as energy shocks, trade policy, and war.

That matters because it suggests AI infrastructure is starting to have a real economic footprint beyond the companies building it. Demand for power, land, equipment, construction, and financing does not stay contained within the technology sector for long. The Fed report makes that visible in a colder way than most market commentary does. AI is no longer just expected to change the economy later. It is already influencing how the current one is being read.

What stood out

Taken together, these stories suggest that AI is becoming harder to keep abstract. In Italy, it turns a postal network into strategic infrastructure. In the United States, it enters the inflation conversation alongside tariffs and conflict. One story is about legacy systems being repurposed. The other is about new systems becoming economically material. Both point to the same thing: the AI era is moving out of the lab and into the institutions that organise everyday capacity.

What it is

This week’s watch is “How the Electrical Grid Is Being Rebuilt for AI” from Bloomberg Primer.

The video starts with the obvious pressure point: electricity demand is rising again, pushed by AI, data centres, and electrification. But it quickly widens into something more useful. This is not just a story about extra demand. It is a story about how grids actually work, where they break, and why the next phase of digital infrastructure may depend less on software than on transmission, stability, and physical redesign. What makes the piece work is that it moves between scales. From China’s transmission buildout to superconducting cables, from Spain’s blackout to mini-grids in Nigeria. The grid stops looking like background utility and starts looking like the central coordination problem.

What stood out

What stood out is that the AI story keeps collapsing downward into infrastructure. The common public framing is that AI needs more chips and more data centres. The video makes clear that this is too narrow. AI also needs transmission capacity, faster grid upgrades, better balancing, and new ways of moving power without losing too much of it along the way. In that sense, the bottleneck is not only computation. It is electricity as a system.

The China section is useful for that reason. It suggests that one of China’s real advantages is not just manufacturing scale, but the ability to build grid infrastructure quickly enough to support new industrial demand. That sits in contrast to slower, more fragmented systems where generation, transmission, and local permitting do not move at the same pace.

The Spain blackout section also lingers. The explanation of inertia makes the whole system feel more fragile and more physical than the usual clean-energy narrative allows. A grid is not just power in and power out. It is a stability machine. Once that becomes visible, a lot of the current buildout starts to look more complicated than simply adding more generation.

Why it lingers

It lingers because it gives a cleaner language for what the AI buildout is really doing. The digital economy often presents itself as weightless. Cloud, agents, models, software. This video is a reminder that none of that escapes the grid. If anything, it drives us back into it. The more ambitious the digital layer becomes, the more the physical layer reasserts itself underneath.

That feels especially relevant now because so many of this week’s stories are really about reclassification. Old institutions becoming infrastructure again. New systems becoming economically material. The Bloomberg piece fits that pattern well. It suggests the next technology race may be won less by whoever has the best model than by whoever can align power, transmission, and reliability fast enough to support the model in the first place.

The useful signal is not just that the grid needs upgrading. It is that the grid is becoming one of the main arenas in which the next decade of technological competition will actually be decided.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

The next layer of AI may not be the agent itself, but the trust architecture built around it.

A lot of the discussion around AI agents still sits at the level of capability. What can they book, buy, schedule, negotiate, monitor, optimise. That matters. But it may not be the real frontier. Reuters reported on 9 July that the International Telecommunication Union, the UN’s digital technology agency, has launched an initiative to improve trust in AI agents, autonomous systems that can perform tasks such as decision-making, scheduling, and financial transactions. The effort was announced at the AI for Good Summit in Geneva and will be carried forward through a new focus group of technical, legal, and policy experts.

What stood out is the framing. The concern is not only whether agents become more capable. It is whether they remain identifiable, trustworthy, and under meaningful human control, especially in areas such as finance and critical infrastructure. Reuters says the group’s mandate is to build international frameworks around exactly those questions. That is a useful signal because it suggests the next bottleneck may not be what agents can do, but what kinds of social and institutional trust can form around them before they become ambient.

That feels directionally important because software has usually been allowed to spread first and earn trust later. With agents, that sequence may be harder to tolerate. Once a system can act on behalf of a person or organisation, make transactions, impersonate authority, or move across multiple services without constant human supervision, the old assumptions break down more quickly. Reuters notes that the ITU initiative is responding in part to the risks of impersonation and unauthorised actions. The problem is no longer only model output. It is agency itself.

This is where the frontier starts to look less technical and more constitutional. Who authorised the action. How is an agent authenticated. What is the boundary between delegation and autonomy. When something goes wrong, who is accountable, the user, the developer, the platform, the firm that deployed it, or the system itself. Those are not side questions. They are the category. And the fact that the ITU is moving early on them is telling. It suggests the institutions closest to global digital coordination do not think trust can be left to emerge on its own.

There is also something quietly practical in the format. The ITU is starting with a focus group, not a treaty or a grand legal settlement. Reuters says the first meeting will take place in Paris in November, followed by another in Geneva in January. That is how new infrastructure categories often begin: not with a finished rulebook, but with a working attempt to define the problem before the market hardens around one default.

If the last generation of digital systems forced institutions to ask what happens when information scales too quickly, the next may force a different question: what happens when action does.

“Energy is the only universal currency.”
Vaclav Smil

Vaclav Smil is a Czech-Canadian scientist and policy analyst best known for his work on energy, materials, food systems, and the physical foundations of modern civilisation. He spent decades at the University of Manitoba and has written repeatedly about the long arc of how energy shapes economies, infrastructure, and everyday life. That is what makes the line so useful. It reduces a complicated week back to a basic condition. Beneath AI buildout, data centres, postal networks, and digital systems sits the same underlying question: how energy is produced, moved, stabilised, and allocated. The digital layer may change quickly. The physical logic underneath it still keeps score.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Two stories this week sit on different ends of the technology curve, but they point to a similar tension.

In China, humanoid robotics is attracting capital, policy support, and industrial ambition at extraordinary scale. In Argentina, lawmakers are testing how far AI can travel into the legal architecture of the firm. One story is about machines entering the factory. The other is about software entering corporate personhood. Both ask the same question in different ways: how much of the future can be organised before the system underneath it is ready?

China’s robot boom is starting to look like system overload

Reuters Breakingviews argued on 29 June that China’s humanoid robotics push is running ahead of its commercial reality. Beijing has committed at least $20 billion to robotics development since 2024, and Barclays analysts estimate annual installations could reach 11 million by 2035, far above forecasts for the rest of the world. Yet Reuters notes only about 12,000 humanoids were sold last year, mostly for research and testing rather than real deployment.

What stood out is the widening gap between spectacle and utility. The category is full of dancing machines, competitive demos, and IPO anticipation, but Reuters points out that real-world applications remain limited by dexterity, reliability, and the fact that conventional industrial robots are already cheaper, tougher, and better suited to many factory tasks. In other words, the humanoid form still carries more symbolic ambition than operating leverage.

That feels directionally important because China has become very good at scaling categories before the rest of the world is comfortable naming them. But scale does not remove physics or commercial discipline. Reuters’ framing is that robotics may now be heading toward something closer to an EV-style boom and bust, where policy support, private funding, and too many entrants create a market that is visibly alive before it is durably rational.

Argentina’s AI company idea is running into the human layer

Reuters reported on 3 July that Argentina’s plan to create AI-run “non-human corporations” still cannot avoid humans. President Javier Milei’s proposal was initially framed as a path toward companies operated by AI, but the bill still requires a human administrator to supervise decisions and carry legal responsibility. Reuters says supporters see the framework as a way to attract foreign investment and formalise AI use in business, while critics warn about accountability and the dangers of giving autonomous systems corporate shelter without clear responsibility.

What stood out is how quickly the future collides with liability. The headline concept is radical enough. A corporation without human shareholders, directed by software. But the legal system appears unwilling to let the human layer disappear entirely. Even in one of the more ambitious experiments in AI-linked corporate reform, someone still has to stand close enough to the entity to be accountable when things go wrong.

That matters because it suggests the edge of automation may not be technical so much as institutional. It is one thing to let AI help run logistics, pricing, or customer service. It is another to let it sit inside the legal shell of the firm itself. Argentina is testing that boundary in a visible way, but Reuters’ reporting suggests the old requirement remains stubbornly in place: systems may become more autonomous, but responsibility still wants a human face.

What stood out

Taken together, these stories suggest that future-industry language is cheap, but system readiness is not.

China can fund robots before they are fully useful. Argentina can imagine AI-run companies before it can remove human accountability. In both cases, the ambition is real. So is the friction. The more interesting signal is not that these futures are impossible. It is that they arrive unevenly, and usually with the old constraints still embedded inside them.

What it is

This week’s watch is “You Can’t Imprison AI: Why We Shouldn’t Give Legal Personhood to Autonomous AI” from The Times.

The segment is built around Argentina’s proposal to recognise AI-run “non-human corporations” and asks a sharper version of the same question: if an autonomous system can run a company, should it also be given something closer to legal personhood? The guest, Cambridge’s Dr Tom Mole, argues that the answer should be no, not because AI is weak, but because it is powerful in ways the law is not designed to discipline.

What stood out

What stood out is the deterrence point. The interview keeps returning to a simple asymmetry. Human decision-makers can be fined, disqualified, sued, or imprisoned. That possibility shapes behaviour long before punishment is actually applied. An AI system does not experience deterrence in the same way. It cannot fear prison, reputational loss, or personal cost. So if it becomes highly effective at pursuing a goal, profit for example, the ordinary tools the legal system uses to keep behaviour inside the rails may lose much of their force.

The second thing that stood out is the rejection of the usual metaphor. Mole is careful not to describe AI as a bad human breaking rules on purpose. His point is subtler. AI may be highly capable while also being organised around goals in ways that do not map cleanly onto human judgment. That makes the problem harder, not easier. You are not dealing with malice in the ordinary sense. You are dealing with competence without the same kind of legal and moral friction.

Why it lingers

It lingers because it gets to the real boundary in the “non-human corporation” debate. The question is not only whether AI can run parts of a business. It already can. The question is whether institutions can tolerate decision-making without a punishable centre. Once you move from workflow automation into the legal shell of the firm, accountability becomes the whole issue.

That is what makes the interview useful. It cuts through the futuristic novelty quite quickly. However sophisticated the software becomes, the law still seems to want a human close enough to blame. Not because humans are better, but because institutions know what to do with them. The frontier here may not be technical capability at all. It may be the point at which legal systems refuse to let capability outrun responsibility.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Frontier Tech

“Non-human corporations” are testing how far automation can travel into the legal shell of the firm.

A great deal of automation still lives safely inside the company. Software helps price inventory, route deliveries, score risk, draft copy, answer customers, or manage logistics. The human institution remains recognisable even as more of its internal work becomes machine-mediated. What Argentina is now testing is a stranger threshold: not whether AI can help run a business, but whether it can sit closer to the legal identity of the business itself. Reuters reported on 3 July that President Javier Milei’s proposed reform would create a category for AI-run “non-human corporations,” though the bill still requires a human administrator to supervise decisions and carry responsibility for the outcome.

What stood out is how quickly the headline concept runs into the human layer. The proposal was initially framed as radical, and in one sense it is. Reuters says Argentina could become the first country to create a formal legal category for companies run by AI, with the broader reform pitched as a way to modernise corporate law, reduce bureaucracy, and make the country more attractive to investment. But the same Reuters report makes clear that the legal system is not yet willing to let responsibility disappear entirely. The automated company would still need a human administrator, and the use of AI in decision-making would not exempt that person from supervising the outcome.

That feels like the real signal. Because the frontier here is not technical capability alone. It is institutional tolerance. A legal system can absorb a great deal of software assistance while still insisting that someone remains close enough to the entity to be blamed, sued, or sanctioned when things go wrong. Reuters quotes governance scholar Lawrence Cunningham saying it would be “too wild a first step to dispense with human agency entirely,” which is a useful way to read the whole moment. The idea of a company without people is thinkable. The idea of a company without accountable people is still much harder for institutions to accept.

There is also a subtler shift underneath it. Once the firm itself becomes the object of automation, the question changes from “what tasks can AI perform?” to “what parts of corporate structure can be delegated?” That is a much deeper frontier. It touches ownership, governance, liability, and agency rather than workflow optimisation. Reuters notes that supporters see the reform as offering legal clarity for forms of business that already exist in partial form, such as AI-supported cashier-less operations, while critics worry that extending legal shelter to highly autonomous systems could reduce accountability rather than clarify it.

This is why the category matters even if it remains constrained. The first serious use case may not be some science-fiction company with no employees, no managers, and no office. It may be something smaller and stranger: highly automated entities with thin human oversight, lighter staffing, and software doing more of the economic coordination than the institution was originally designed to tolerate. In that sense, “non-human corporation” may be less a finished category than a pressure test for the edges of company law.

Once automation begins pressing not only into work, but into the legal shell that organises work, the human question comes back differently. Not whether people disappear from the firm, but how few people the firm can still contain before the institution stops making sense.

“The question concerning technology is nothing technological.”
Martin Heidegger

Martin Heidegger was a German philosopher whose work often returned to the way modern systems shape how the world is revealed to us, not just how it is used. That is what makes this line worth keeping. It pushes against the instinct to treat technology as a neutral tool with purely technical questions attached to it. The deeper questions are always social, political, and human. What kind of order does a system create. What does it make easier to ignore. Who remains accountable when the machinery becomes more capable than the institutions around it know how to handle. That feels close to this week’s theme. The real issue is not whether automation can enter the legal shell of the firm. It is what happens to responsibility when it does.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

This week’s signal is not only that Europe is hot. It is that the heat is beginning to expose which parts of ordinary life were never really built for this temperature regime.

Reuters reported this week that the heatwave sweeping across Europe has led to melting road surfaces in Germany, derailed trains in Sweden, school and event closures, hospital strain, and lower nuclear output in France as river temperatures rise. France’s nuclear generation was cut by 4.1 gigawatts, or about 7% of total power demand, as several reactors reduced output because hot rivers could no longer absorb enough cooling water without breaching environmental limits.

What stood out is how varied the failures are. Roads soften. Tracks twist. Power exports fall. Hospitals struggle. Schools close. None of these are especially futuristic outcomes. They are basic system stresses. That is what makes the story feel important. Climate risk is often narrated at the level of emissions targets, future scenarios, or moral urgency. What Reuters captured this week is something more immediate: infrastructure behaving as though the climate assumptions embedded inside it no longer hold.

The French nuclear story makes that especially clear. Nuclear power is often discussed as stable baseload infrastructure, but even that stability depends on the physical environment around the plant. Reuters reported that high river temperatures forced reductions at multiple EDF reactors, cutting France’s exports from around 10–12 GW the previous week to about 3 GW during the afternoon, while spot power prices in France and Germany rose to their highest level since January 2025. The issue was not demand alone. It was that the surrounding water system had become part of the constraint.

This is also what gives the broader heatwave its weight. Reuters reported that scientists from World Weather Attribution called the event the most severe ever recorded for the region and said it would have been “virtually impossible” without human-caused climate change. They found that night-time temperatures of this kind are now 100 times more likely than they were just two decades ago. That matters because infrastructure does not only fail at the peak. It fails when the old margins disappear. When nights stay hot, rivers stay warm, hospitals stay full, and recovery windows shrink.

Europe’s roads, rail systems, power infrastructure, and public-health routines were built for a climate that is receding. The useful question is no longer just how often extreme heat will arrive. It is which parts of the system quietly depend on it not arriving at all.

What it is

This week’s watch is Bloomberg’s “Brutal Heat Wave Scorches Europe, Raising Concerns of Climate Inflation.”

The conversation starts with Europe’s immediate experience of extreme heat, failed cooling, transport disruption, and stressed public services. But it quickly moves to something more structural: the idea of climate inflation, where heat begins feeding directly into prices through damaged harvests, weaker livestock output, disrupted energy systems, and wider economic friction.

What makes the video useful is that it treats heat not only as weather, but as an economic transmission mechanism.

What stood out

What stood out is how ordinary the points of failure are. The speakers describe heat that is too intense for factories, too intense for sleep, and even too intense for sitting outside in the evening. That sounds anecdotal at first, but it helps explain the bigger argument. Extreme heat does not need to destroy everything at once to become economically significant. It only needs to make ordinary life and routine systems less reliable.

The inflation point is the one that lingers. The video notes that the 2022 European heatwave was estimated to lift food prices by around 0.7% and overall inflation by about 0.3%. That is useful because it shifts climate discussion away from distant catastrophe and toward price formation. Crops, livestock, electricity, transport, health systems. Once those are all under heat stress at the same time, inflation stops being only a monetary story.

The other thing that stood out was the description of Europe as the fastest-warming continent on Earth, warming at roughly double the global average. In that frame, the question is not whether a single policy can solve the issue. It is whether infrastructure, institutions, and adaptation can keep up with a system that is starting to fail in multiple places at once.

Why it lingers

It lingers because it gives a better language for what this week’s heatwave actually represents. “Climate change” can still sound broad and moral. “Climate inflation” sounds narrower, colder, and more immediate. It names the way physical stress enters the economy through food, power, transport, healthcare, and housing rather than through abstract future scenarios.

That matters because it changes the political and institutional challenge. If heat becomes a persistent inflationary force, then it is no longer only an environmental issue. It becomes a central-bank issue, a public-infrastructure issue, and a cost-of-living issue.

The useful signal here is not just that Europe is hot. It is that the old assumption of climatic stability is quietly embedded in more of the economy than most systems were ever designed to admit.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels - from social media activity to Google search trends - and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Data centres are starting to look less like abstract cloud infrastructure and more like local resource technology.

For years, data centres were easy to imagine as somewhere else. Remote buildings. Neutral capacity. The physical substrate beneath digital life, but rarely part of the public conversation unless something went badly wrong. That is getting harder to maintain. Reuters reported earlier this month that the European Commission has proposed minimum energy-efficiency standards for data centres, including targets for energy performance, water use, heat reuse, and demand-side flexibility. The shift is telling. Data centres are no longer being treated only as a digital sector. They are being pulled into the language of resource management. (reuters.com⁠)

What stood out is that this is not really about efficiency in the narrow sense. It is about legitimacy. Once a data centre becomes large enough, it starts to compete for the same things other local systems need: land, grid access, water, cooling capacity, and planning attention. Reuters also reported this month that France has been looking at ways to speed up grid connections for large data-centre projects because connection delays are slowing development. That gives the whole category a different feel. A data centre is no longer just a passive recipient of electricity. It becomes a claimant on infrastructure that has to be allocated, sequenced, and defended. (reuters.com⁠)

That matters because the growth story remains large. Reuters reported in late May that SoftBank plans to invest €45 billion over five years in AI data centres in France, with the first three sites expected to deliver 3.1 gigawatts of capacity by 2031 and potential expansion taking the total to 5 gigawatts. At that scale, the conversation changes. The issue is no longer only what the models can do. It is whether substations, transmission, land use, cooling systems, and local politics can keep up. (reuters.com⁠)

This is where the frontier becomes more interesting. A data centre used to feel like a backend facility. Increasingly it looks more like a piece of territorial infrastructure, closer in some ways to a port, a factory, or a power plant. The EU’s proposed standards make that visible because they treat the facility not only as a computing node but as something that must justify how it uses power, water, and waste heat inside a broader system. That is a meaningful change in posture.

The heatwave story this week sharpens that further. When roads soften, rail systems buckle, and river temperatures force lower nuclear output, the old assumption that infrastructure capacity is broadly stable begins to look weaker. In that environment, any new high-load facility is judged differently. Not only by what it enables, but by what it draws from systems already under stress.

That feels directionally important. Because frontier technology is often described through software, models, or user experience. But many of the most consequential constraints now sit underneath that layer. Data centres are worth watching not just because they power the digital economy, but because they reveal where digital ambition starts to collide with physical limits.

“The future is not a gift. It is an achievement.”
Robert F. Kennedy

Robert F. Kennedy was an American politician and public figure whose language often returned to responsibility, moral courage, and the work required to build a society worth inheriting. That is part of why this line still holds. It refuses the passive version of progress. The future does not simply arrive because technology improves or because history moves forward on its own. It has to be built, argued over, financed, and made durable through institutions and choices. That feels especially relevant in a week about heat, infrastructure, and the physical limits sitting underneath modern systems. The future is not something waiting for us on the other side of disruption. It is something that has to be achieved under pressure.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Three stories this week suggest that industrial policy is becoming more explicit about what kinds of future it wants to fund.

In Argentina, that means loosening biotech patent restrictions to attract new investment and technology. In Seoul, it means pulling Boston Dynamics fully into Hyundai’s orbit rather than leaving robotics at the edge of the group. In Beijing, it means building a clearer IPO pathway for what it now calls “future industry” companies. Different sectors, different systems, same direction: states and industrial champions are getting more direct about the technologies they want to scale.

Argentina is reopening the biotech patent question

Reuters reported on 19 June that Argentina repealed a regulation that had restricted patents for biotech developments. Cabinet Chief Manuel Adorni said the move is meant to align the country with international standards, provide greater legal clarity, and encourage biotechnological innovation, with officials also pointing to potential benefits for agriculture.

What stood out is that this is not only a legal adjustment. It is a signal about what kind of science Argentina wants to host. Patent rules are often discussed as dry architecture. In practice, they are part of the invitation structure around a sector. By loosening the restrictions, Argentina is saying that biotech should be treated less as a domain to be tightly bounded and more as one it wants to make investable.

That matters because biotech sits in an awkward place between national need and global capital. Countries want domestic capability, but they also need researchers, firms, and investors to believe the commercial pathway is credible. Patent policy is one of the places where that tension gets resolved. Argentina’s move suggests it is willing to trade some caution for a clearer growth signal.

Hyundai is pulling robotics into the centre of the group

Reuters reported on 19 June that Hyundai Motor Group plans to buy SoftBank’s remaining 9.65% stake in Boston Dynamics for $325 million, which would make the U.S. robotics company a wholly owned subsidiary. The report said Hyundai and its affiliates already control more than 90% of Boston Dynamics, and that a board meeting is expected to approve the purchase.

What stood out is the posture behind the deal. Robotics is no longer being held at arm’s length as a venture-style adjacency. Hyundai appears to be pulling it fully into the group structure. That does not guarantee commercial success. But it does change the meaning of the bet. A wholly owned subsidiary sits differently inside a company than a partially owned strategic experiment.

This feels directionally important because large manufacturers often reveal their real priorities not in keynote language, but in ownership structure. Full control suggests robotics is being treated less as optional future exposure and more as something Hyundai may want to integrate across its longer industrial strategy. The machines are still early. The organisational choice feels more settled.

China is widening the capital-markets lane for “future industry”

Reuters reported on 17 June that China will support IPOs by “future industry” startups and large-model companies, with the Shanghai Stock Exchange rolling out rules to help strategically important but unprofitable firms list on the STAR Market. Reuters said the policy is aimed at sectors such as quantum technology, nuclear fusion, brain-computer interfaces, robotics, hydrogen energy, and biomedical engineering, as China tries to strengthen its innovation base amid intensifying rivalry with the United States.

What stood out is the category itself. “Future industry” is no longer just a speech line. It is becoming something with a financing pathway attached to it. That matters because once a state names the sectors it considers strategic and then builds a capital-markets mechanism around them, the category starts to harden. It becomes easier for firms, funds, and local governments to organise themselves around it.

The deeper signal is that policy is moving from encouragement to channel-building. China is not only backing innovation in the abstract. It is trying to create a more deliberate pipeline from research and industrial ambition into public capital. That is usually when a strategic theme starts to look less like rhetoric and more like institutional design.

What stood out

Taken together, these stories suggest a more interventionist phase of technology policy. Argentina is adjusting the legal environment around biotech. Hyundai is internalising robotics more completely. China is formalising the IPO route for sectors it sees as strategically important. Different tools, same broad move: the future is being named more clearly, and then organised around.

What it is

This week’s watch is “The Successor to CRISPR May Be Even More World Changing.”

The video tracks the arc from CRISPR to a newer system called TIGR-Tas, following Feng Zhang’s work from bacterial immune systems to the possibility of another gene-editing tool that may, in some settings, outperform CRISPR. The structure is useful because it keeps returning to a simple point: frontier biotech rarely begins with a product. It begins with basic science, with someone trying to understand how a strange biological system works before anyone knows what it might be for.

That is what gives the video its shape. CRISPR is presented not as the destination, but as one example of what happens when basic research turns into a tool.

What stood out

What stood out is how the frontier keeps moving down a layer. The video makes clear that CRISPR is already extraordinary. In just over a decade, it moved from uncertain experimental promise to use in life-saving treatment, including the rewriting of DNA in a living infant. But the more interesting signal is that even CRISPR is now being described as incomplete. It works well for some diseases, less well for others, and still depends heavily on delivery, target selection, and the biology of the tissue involved.

That is where TIGR-Tas becomes interesting. The video says the system may have several advantages over CRISPR in specific contexts. It appears smaller, which may make delivery easier. It may read both sides of the DNA double helix when targeting, which could improve precision. And it may allow for narrower editing windows, which matters when the goal is to change something with single-letter accuracy rather than open up a broader region of DNA. None of that makes it a finished platform. But it does suggest the field is still early enough that even a revolutionary tool can quickly become a stepping stone.

The other thing that lingered was Zhang’s emphasis on basic science. He keeps returning to the process of pulling a system apart to see how it works, then only later asking whether it can be turned into a tool. That feels important because biotech is often narrated through therapeutic promise, while the actual frontier still depends on patient, open-ended work that looks much less commercial.

Why it lingers

It lingers because it changes the feel of progress in gene editing. The usual public narrative tends to treat CRISPR as the breakthrough, singular and complete. This video suggests something more unsettled and more interesting. Each new tool opens a path, then reveals the limits of that path. The field moves forward not by finding one final answer, but by discovering better ways to intervene at different levels, with different trade-offs around delivery, durability, precision, and risk.

That is also why the video pairs well with the Ascidian story this week. Both point in the same direction. The frontier in genetic medicine is no longer just about whether editing is possible. It is about what layer to edit, how permanently to act, and what kind of intervention can be made precise enough to survive contact with the clinic.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and digital asset infrastructure are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels—from social media activity to Google search trends—and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score.

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Gene editing may be moving toward repair without permanently changing DNA.

For a long time, the frontier story in genetic medicine was framed around permanence. Cut the DNA. Correct the fault. Rewrite the code at its source. That still carries enormous promise. It also carries a particular kind of weight, because once DNA is altered, the intervention is harder to reverse and the stakes around safety rise accordingly.

What stood out this month is a different direction. Reuters reported on 3 June that Eli Lilly signed a licensing deal with Ascidian Therapeutics, worth up to $1.9 billion, to develop treatments for inherited kidney diseases using Ascidian’s RNA exon-editing technology. Reuters described the platform as a way to correct faulty genes without permanently altering a patient’s DNA, with Ascidian leading early research and selected preclinical work while Lilly takes on later development, manufacturing, and commercialisation.

That feels like a useful shift in emphasis. The frontier here is not simply editing more precisely. It is editing at a different layer. Ascidian says its approach works on RNA, not DNA, and is designed to deliver some of the durability and benefits associated with gene therapy while reducing the risks tied to direct genomic modification. The company also says the method can edit whole exons at kilobase scale, rather than only making smaller base-level changes.

What lingers is the change in tone this introduces. Gene editing has often been narrated through irreversibility. That is part of what made the field feel so powerful and so fraught. An RNA-based approach changes the feel of the problem. It does not make the science simple, and it certainly does not remove the usual questions around delivery, efficacy, durability, or regulation. But it suggests that one path forward in genetic medicine may involve interventions that are less final in their mechanism, even if they remain ambitious in their therapeutic goals.

There is also a more practical signal here. Large pharmaceutical companies do not usually commit this kind of money because a technology sounds elegant in theory. They do it when a modality starts to look like a credible platform. Reuters noted that this deal follows Lilly’s broader expansion into genetic medicines, while Ascidian had already signed a 2024 partnership with Roche around neurological diseases using the same underlying technology.

That matters because frontier technologies tend to become real twice, first as a scientific idea. Then as an investable category. The second moment is quieter, but usually more revealing. It is when companies begin organising pipelines, partnerships, and capital allocation around the belief that a new modality might survive contact with the clinic.

This still feels early, and there are no disclosed clinical results here - the targets remain undisclosed. But directionally, it is worth noting when the field starts looking for ways to repair genes without treating permanent DNA alteration as the only serious path forward.

“If we knew what it was we were doing, it would not be called research, would it?”
Often attributed to Albert Einstein

Good research rarely begins with certainty. It begins with a live question, a partial map, and the willingness to keep working while the answer is still forming. That is part of what makes frontier science feel both fragile and consequential. The method comes first. Clarity arrives later, if it arrives at all.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Two stories this week suggest that governments are moving down the stack.

In Britain, the child-safety debate is no longer only about social platforms and moderation. It is moving toward the device itself, with Keir Starmer pressing Apple and Google to build controls that stop young people from sending or receiving nude images, and threatening legislation within three months if they do not act. In the United States, cyber defence officials have shortened the patching window for the most serious vulnerabilities to three days, arguing that AI-enabled attackers are compressing the time available to respond. The common signal is that states are starting to treat digital risk as something that has to be managed closer to the infrastructure layer, not only at the level of public debate or user behaviour.

Britain wants devices, not just platforms, to become part of the safety system

Reuters reported on 8 June that Starmer told major tech firms such as Apple and Google they must stop young people from circulating nude images, and that if they fail to act within three months, the government will legislate, with possible fines or even criminal liability for executives who do not comply. Reuters said the government wants device-level controls that can prevent children from sending or receiving sexually explicit images, and linked the push to concerns around grooming, sextortion, and self-generated abusive images.

What stood out is the shift in responsibility. This is not only a demand for better moderation on social media. It pushes accountability further down the stack, toward operating systems, hardware ecosystems, and the basic architecture through which images move. That matters because it changes the shape of the problem. Safety is no longer being framed only as a platform-policy issue. It is being treated as something that may need to be embedded at the device layer itself.

That feels directionally important because it suggests governments are becoming less willing to accept the old division of labour, where platforms manage content, parents manage behaviour, and device makers stay more neutral. Once the device becomes part of the enforcement layer, the boundary between product design and state expectation shifts. The implication is not just that children need protection. It is that core consumer technology may increasingly be asked to carry that protection by default.

In the United States, AI is changing the tempo of cyber defence

Reuters reported on 10 June that the U.S. Cybersecurity and Infrastructure Security Agency has cut the response window for the most serious categories of vulnerabilities to three days, a sharp acceleration tied to concerns that hackers are using advanced AI systems to identify and exploit weaknesses faster than before. Reuters said the new directive applies to civilian federal agencies, while less serious flaws still have longer timelines, up to two weeks for moderate issues and two months for the least severe ones.

What stood out here is that the change is operational, not rhetorical. Governments have warned for some time that AI could alter the threat landscape. This is different. It is a direct redesign of response windows inside the state. The point is not that AI may create new risks someday. It is that officials now believe the attack cycle has already shortened enough to justify compressing remediation timelines from weeks to days.

The deeper signal is that AI risk is beginning to show up less as a philosophical problem and more as a timing problem. When the interval between discovery and exploitation shrinks, institutions are forced to change their cadence. That creates pressure not only on government agencies, but on the broader digital environment they help shape. Security becomes a matter of whether organisational speed can keep pace with machine-assisted threat discovery.

What stood out

Taken together, these stories point to a common adjustment. Governments are moving from broad concern to narrower control points. In Britain, that means asking whether devices themselves should help block harmful exchanges among young people. In the United States, it means redesigning patch cycles around the faster tempo of AI-enhanced hacking. Different domains, same underlying move: digital risk is being pushed closer to the infrastructure layer, where systems can be shaped before behaviour scales on top of them.

What it is

This week’s watch is “AI has got better at hacking — how big a risk is it?” from The Economist.

The video centres on Mythos, Anthropic’s latest model, which the company chose not to make generally available because of its ability to find software vulnerabilities and work out how to exploit them. The report uses that decision to ask a wider question: as AI systems become more capable, who benefits more from that progress, attackers or defenders?

The most useful part of the video is that it does not stay at the level of warning. It walks through the mechanics of vulnerabilities, exploits, and disclosure, then shows how a model like Mythos changes the nature of the problem. This is no longer only about AI helping a skilled engineer move faster. It is about systems reaching the point where they can autonomously identify and weaponise weaknesses with very little human oversight.

What stood out

What stood out was the compression of time. The video cites a striking figure: the delay between a vulnerability being disclosed and an exploit being used in the wild has fallen from 2.3 years in 2018 to around 20 hours now. That is the part that lingers. The real shift is not only that models are getting better at hacking tasks. It is that the whole interval between discovery and danger is collapsing.

The OpenBSD example makes that tangible. A vulnerability had reportedly sat inside the operating system for 27 years, and the model was able to find it. The fix itself was simple. The hard part was seeing the weakness in the first place. That feels like a useful clue about where this is heading. AI may not only create new attacks. It may excavate years of hidden technical debt that systems have quietly carried all along.

The other thing that stood out is that Mythos was not presented as a specially trained cyber model. In the video’s telling, it is largely the next stage of scaling. Bigger, more capable, less constrained. That matters because it suggests offensive capability may emerge not as a niche branch of model development, but as a byproduct of general capability growth once safety layers are removed.

Why it lingers

It lingers because it changes the shape of the cyber story. For a long time, AI in cybersecurity could still be discussed as an augmentation layer. A coding assistant. A research aid. A faster way to sort through logs or draft exploits. This feels like a different threshold. Once a model can move from identifying a weakness to exploiting it autonomously, the defensive side no longer has the luxury of operating at an older human rhythm.

That is what connects this video to the wider signal in the issue. If the time between vulnerability disclosure and exploitation is collapsing, then patch cycles, security workflows, and defensive tooling have to change shape with it. Human teams alone may not be fast enough. That is where agentic white-hat hacking begins to look less optional.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and assets are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels—from social media activity to Google search trends—and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Agentic white-hat hacking is starting to look less like a thought experiment and more like a new layer of cyber defence.

For years, the frontier in cybersecurity was described through the attacker. A more sophisticated exploit. A faster phishing campaign. A new way to move laterally through a network. What feels different now is that some of the same autonomy and speed are being pulled into defence. Not only to detect threats, but to search for them, test systems against them, and compress the time between discovery and response.

That shift is becoming easier to see because the offensive side is already accelerating. Reuters reported in May that a cybercrime group used AI to uncover a previously unknown software flaw and exploit it for the first time, according to Google. Reuters also reported this week that the U.S. cyber defence agency has cut the patching window for the most serious vulnerabilities to three days, explicitly citing the way more capable AI models are speeding up exploitation.

What stood out is that once the threat cycle compresses, defence has to change shape. Traditional security teams can still investigate, escalate, patch, and document. But the older rhythm starts to look too slow when machine-assisted attackers can find and weaponise weaknesses almost immediately. That is where agentic white-hat systems become more interesting. The promise is not just better dashboards or another layer of alerts. It is autonomous defensive behaviour: systems that can probe for exposures, chain together testing steps, validate weaknesses, and in some cases begin remediation workflows before a human team has fully triaged the problem.

This is not without risk. Agentic systems can misfire, overreach, or widen the blast radius if they are given too much freedom. But the logic is becoming clearer. If black-hat actors are using AI to increase the tempo and scale of attack, defenders may need autonomous systems of their own simply to keep pace.

There are already hints of the direction. Reuters’ June report on the new three-day patch window suggests that the state now sees AI as an operational timing problem, not just a strategic one. Google’s reporting on AI-assisted exploitation points the same way. The issue is no longer whether AI changes cyber conflict. It is how fast institutions can adapt once it does.

That makes white-hat agentic hacking feel frontier in the right way. Not because it is flashy, but because it changes where defensive work happens. Security has often been reactive and human-bottlenecked. Agentic systems introduce the possibility that parts of the white-hat workflow, reconnaissance, validation, prioritisation, patch testing, could become continuous and semi-autonomous. In that world, the defender is not only a person watching a console. It is a system actively looking for failure before the attacker does.

The real question is governance. Who authorises an agent to probe production systems. What boundaries it operates within. How much autonomy is acceptable before “defence” starts to look too much like uncontrolled action. These are not side questions. They are the category.

Still, the direction matters. The frontier in cybersecurity may no longer be defined only by smarter attacks. It may be defined by whether defenders can build trusted autonomous systems fast enough to stop human teams from being permanently outpaced.

“It is not the strongest of the species that survives, nor the most intelligent, but the one most responsive to change.”
Often attributed to Charles Darwin

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

This week I travelled to Timor-Leste for the first time doing some engagement with their govt to explore a national digital-finance infrastructure roadmap. This is happening at a specific moment: Timor-Leste is now inside ASEAN, operates on the US dollar, and appears willing to consider whether modern capital-markets infrastructure can be evaluated before the full weight of legacy systems is rebuilt.

Timor-Leste is testing a straight-to-digital financial strategy

The clearest signal from the trip is that Timor-Leste is not simply looking at a new product. It is exploring a broader digital-finance strategy covering a roadmap spanning sovereign digital bonds, digital capital-markets infrastructure, financial inclusion initiatives, and cross-border investment connectivity. This frames their broader ambition as a chance to “leapfrog legacy financial systems” and accelerate access to modern capital-markets infrastructure.

That feels directionally important because smaller states sometimes have a narrower but cleaner window for institutional design. They do not necessarily have to modernise in the same sequence as larger incumbents. The more interesting question may be whether Timor-Leste can evaluate modern governance frameworks, programmable financial rails, and institutional-grade digital infrastructure early enough that it does not spend the next decade retrofitting systems built for a different era. The phrase “straight-to-digital” is useful here, not as a slogan, but as a description of posture.

ASEAN accession is changing the meaning of infrastructure

The second signal is timing. Timor-Leste uses the United States dollar as its official currency and that, following its accession to ASEAN in October 2025, the initiative supports a broader ambition to deepen regional integration, attract international investment, and participate more fully in Southeast Asia’s economic growth. This investment in digital financial infrastructure is part of a wider effort to strengthen investment connectivity, expand access to capital, and support the country’s next stage of economic development.

That matters because it changes what “financial infrastructure” means. It is not only a domestic modernisation story. It is increasingly a regional positioning story. In that frame, digital bonds and capital-markets infrastructure are not just technical upgrades. They become possible tools for plugging a young state more directly into regional and international capital flows, while still thinking carefully about governance, compliance, and institutional trust. The engagement will explore internationally aligned approaches to governance, compliance, and institutional-grade digital infrastructure, which reinforces that point.

There is a useful contrast here with how financial modernisation often unfolds elsewhere. In many countries, the conversation starts with the limitations of legacy infrastructure and then moves slowly outward. In Timor-Leste, the more interesting possibility is that regional integration may push the question in the other direction. ASEAN accession raises the stakes around connectivity, and that in turn may make the design of financial rails feel more urgent, more strategic, and more outward-facing than it would otherwise.

What stood out in Dili was not a flashy launch. It was the willingness to treat financial infrastructure as a strategic question before the answer had hardened. Timor-Leste appears to be asking whether it can think about sovereign financing, capital-market access, and regional connectivity in digitally native terms from the outset, rather than simply inheriting the default sequence. That still leaves a great deal unresolved. But it is worth noting when a country starts with the roadmap instead of the retrofit.

What it is

This week’s watch is “Bhutan mines bitcoin to offset declining tourism revenue, fund economic growth.”

The report looks at how Bhutan is using hydropower-fed bitcoin mining as part of a broader economic strategy. The government presents bitcoin not simply as a speculative asset, but as a strategic reserve and a revenue source that has already been used to help fund public-sector salaries, alongside healthcare and environmental spending. The core argument in the piece is that Bhutan, as a small country with limited industrial options, is trying to turn surplus renewable energy into a digitally native reserve asset.

What stood out

What stood out is that Bhutan is not buying bitcoin in the market. It is mining it. That distinction matters. The government is framing the activity as a way to convert cheap domestic hydropower into an asset that is globally liquid and, in its view, scarce. That gives the whole project a different logic from a normal sovereign investment story. Instead of exporting hydropower at relatively low rates, Bhutan is using it to run energy-intensive computation in the high mountains and receive bitcoin in return.

The second thing that stood out is how openly the government links this to state capacity. The report says the main use of bitcoin so far has been to help finance the salaries of public servants. That is a surprisingly direct use case. It shifts the story away from ideology and toward fiscal management. Bitcoin stops looking like a parallel system and starts looking, at least in Bhutan’s framing, like a tool inside the state’s balance-sheet logic.

There is also a useful asymmetry here. Bhutan’s constraints are very specific. Large-scale manufacturing is difficult. Agriculture is limited. Much food is imported. So the attraction of bitcoin mining is not only that it is digital. It is that it monetises a resource Bhutan already has, hydropower, without requiring the same kind of physical export infrastructure or industrial footprint that other development paths might.

Why it lingers

It lingers because it suggests a different way to think about digital assets at the state level. Most of the global conversation still treats bitcoin either as a speculative vehicle or as a political symbol. This report points to something more pragmatic. For Bhutan, the question seems to be whether a renewable energy surplus can be turned into a reserve asset that supports fiscal resilience and economic development. That is not a universal model. But it is a real one.

What makes it linger, though, is the tension underneath it. The strategy only works while hydropower remains abundant, mining economics remain favourable, and bitcoin retains enough value and liquidity to justify the trade. So this is not a clean solution. It is a high-conviction wager on how energy, computation, and money may fit together in a small-state context.

It shows that for some countries, the more interesting question is not whether digital assets matter, but whether they can be integrated into national strategy without first passing through the usual private-market sequence.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and assets are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

Macro take: bitcoin looks lower over the last week because this has been a risk-off and rotation move more than a crypto-specific one. Capital has been flowing toward AI-linked equities and big upcoming IPOs, while bitcoin has also been hit by institutional selling after Strategy sold some holdings

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels—from social media activity to Google search trends—and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Bitcoin mining is starting to look less like pure consumption and more like a way to monetise stranded renewable energy. Bhutan is the clearest case study.

The usual argument about bitcoin mining starts with electricity. Too much of it. Too little obvious social value. A blunt conversion of power into speculation. That framing still explains a lot. But it explains less when the electricity in question is renewable, seasonally abundant, and harder to monetise at the moment it is produced. That is what makes Bhutan worth watching.

Reuters reported in April last year that Bhutan has been using its abundant hydropower to mine what officials describe as green cryptocurrency. The sovereign wealth fund’s chief executive said the country began mining in 2019 using hydropower and has since used profits from past crypto investments to help fund government salaries. Reuters also said Bhutan sees the strategy as a way to stimulate the economy, create jobs, respond to youth emigration, and attract firms looking for lower-carbon digital-asset exposure.

What stood out is that the logic is less ideological than it first appears. Bhutan is not presenting mining as a monetary revolution. It is treating it as a conversion mechanism. Hydropower that might otherwise be underused, exported cheaply, or wasted at the margin can instead be turned into a digitally native reserve asset. Reuters noted that Bhutan already runs domestically on hydropower and is looking to expand generation capacity significantly over time. In that context, mining becomes a way of arbitraging surplus electricity into a globally liquid asset.

This is where stranded energy becomes more interesting. Renewables do not always fail because there is not enough demand. Sometimes they fail because demand is in the wrong place, or at the wrong time, or the transmission system is too weak to move the power efficiently. Bitcoin mining is one of the few loads that can be deployed relatively quickly, ramped up and down, and located close to generation. That does not make it universally desirable. But it does make it unusually compatible with energy that is clean, abundant, and otherwise difficult to monetise at the edge of the grid.

Bhutan’s case looks distinctive because the state itself sits close to the strategy. Reuters framed the mining effort as part of Bhutan’s sovereign investment approach rather than as a private scramble for cheap power. That gives the model a different character from places where mining has simply strained fossil-heavy grids. Bhutan points in the opposite direction: renewable oversupply first, mining second.

There are still obvious limits. The economics are volatile. Bitcoin prices move. Hydrology is seasonal. And scaling the model depends on continued investment in hydropower and transmission. Reuters noted that Bhutan’s longer-term success as a green crypto hub will require a major increase in generation capacity over the coming decade. So this is not a universal formula. It is a tactical use of a specific national advantage.

But directionally, it matters. Because it suggests one possible future for renewables at the margin. Not every kilowatt of green energy needs to flow directly into homes, factories, or vehicles the moment it is generated. In some places, especially where grids are thin and power is seasonal, a flexible computational load may be the bridge.

Bhutan is worth watching not because it proves bitcoin mining is good in the abstract, but because it asks a more practical question: what is the highest-value use of clean power that would otherwise be harder to capture?

“It is not the strongest of the species that survives, nor the most intelligent, but the one most responsive to change.”
Often attributed to Charles Darwin

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

News & Signals

Three stories this week suggest that some of the most revealing technology signals are no longer coming from product launches.

They are showing up in the operating systems around the products. Inside the workplace. Inside the store. Inside the national infrastructure stack. Meta’s plan to capture employee activity for AI training turns the labour process itself into data. Starbucks’ decision to scrap an AI inventory tool is a reminder that everyday operations remain harder to model than the demos suggest. And SoftBank’s €45 billion AI data-centre plan for France shows how quickly the centre of gravity has moved from models to land, power, and physical capacity.

Meta is pushing AI training deeper into the workplace

Reuters reported in April that Meta is installing tracking software on U.S.-based employees’ computers to capture mouse movements, clicks, and keystrokes for use in training its AI models, part of a broader push to build agents that can perform work tasks autonomously. Privacy lawyers told Reuters the practice could run into serious problems under EU workplace-surveillance rules if data from Europe or from EU-based staff were implicated.

What stood out is not only the privacy issue. It is the change in posture. Training data is no longer treated as something gathered from the outside world alone. It is increasingly something extracted from work itself. The internal operation of the company becomes part of the raw material for building the next generation of systems. That changes the feel of the AI story. Less public internet. More instrumented labour.

The deeper signal is that the boundary between software and workplace governance is getting thinner. Once employee activity becomes useful training input, questions about surveillance stop being side issues and start becoming part of the system design itself. That feels directionally important because it suggests the next phase of AI development may depend as much on what firms can observe about their own workers as on what they can scrape from the web.

Starbucks found the limits of operational AI the hard way

Reuters reported on 21 May that Starbucks scrapped its AI-based inventory tool across North America just nine months after rollout because it repeatedly misidentified products and failed to generate reliable stock counts. The tool, developed by NomadGo, used cameras and lidar on tablets to automate inventory checks. Starbucks is now reverting to manual counting while focusing on simpler daily replenishment routines.

What stood out here is how ordinary the failure was. Not a dramatic safety breach. Not a frontier model collapse. Just a worker-facing operational tool meeting the messy reality of a store and not holding up. That matters because so much of the AI conversation still centres on capability in isolation. Starbucks is a useful reminder that capability does not automatically survive contact with routine work, visual ambiguity, changing stock layouts, and the need for dependable execution.

The signal underneath it is that retail remains a harsher test environment than product demos suggest. The value of automation depends less on whether a tool can work in principle and more on whether it can work predictably, cheaply, and without adding new friction to already pressured frontline operations. In that sense, this is less a story about Starbucks and more a story about the difference between model performance and operational reliability.

France is becoming a test case for AI infrastructure as industrial policy

Reuters reported today that SoftBank plans to invest €45 billion over five years to build AI data centres in France, with the first three sites expected to deliver 3.1 gigawatts of capacity by 2031 and potential expansion taking the total to €75 billion and 5 gigawatts. Schneider Electric will supply modular solutions, while EDF is contributing a former power plant for conversion into a data-centre site.

What stood out is that this is no longer mainly a software story. It is a land, power, and industrial-location story. France’s attraction here is not just policy enthusiasm. Reuters says Masayoshi Son cited France’s energy production and export capabilities as decisive. That is worth noting because it shows how AI infrastructure decisions are increasingly being made on the basis of electricity, transmission, and conversion capacity rather than talent or model prestige alone.

The broader signal is that AI is starting to look more like heavy industry. Once projects are measured in gigawatts and tens of billions of euros, they start to reshape national planning, utility strategies, and industrial geography. The conversation moves from intelligence in the abstract to where the substations, land parcels, and financing structures will come from. That is usually the point at which a technology stops feeling speculative and starts becoming structural.

What stood out

Taken together, these stories suggest that AI is becoming harder to keep abstract.

At Meta, it reaches into the labour process. At Starbucks, it runs into the stubborn mess of daily operations. In France, it turns into an infrastructure buildout measured in power and capital. Different settings, same direction: the technology is moving out of the lab and into the systems that organise work, logistics, and industrial capacity.

What it is

This week’s watch is “What does Meta actually do now?”

At one level, it is a brisk explainer on how Facebook became Meta, and how the company moved from a social platform into a sprawling corporate structure built around Instagram, WhatsApp, ads, hardware, AI, and a still-unresolved metaverse legacy. The video traces the arc from Facebook’s early growth and scandals through the 2021 rebrand, the collapse of the original metaverse narrative, and Meta’s current push into AI, smart glasses, and ever larger infrastructure spending.

What makes it useful is that it is not really asking what Meta says it does. It is asking what the company actually is once the branding falls away.

What stood out

What stood out is how little the underlying business model has changed, even as the narrative around it keeps shifting.

Again and again, the people interviewed in the video come back to the same point: Meta still makes its money primarily by selling attention through Facebook and Instagram. The rebrand, the metaverse, the hardware, the glasses, the AI push. These all sit on top of that engine rather than replacing it. That matters because it helps explain why the company can absorb enormous losses in one area while continuing to scale another. The family of apps still throws off enough revenue to fund very large bets.

The second thing that stood out was the way AI is framed internally. The video suggests Meta’s AI push is not only about launching new products. It is also about avoiding dependence on other companies’ foundational models, in much the same way Meta historically resented being dependent on Apple and Google’s app stores. That is a useful lens. AI begins to look less like a feature race and more like an attempt to own the next platform layer before someone else does.

There is also a darker thread running through it. The video ties Meta’s current AI expansion to internal employee tracking, workplace tension, layoffs, and growing criticism around addictive design and children’s harm. That combination matters. It suggests the company is not just extending its technical reach. It is intensifying the systems around work, attention, and behaviour at the same time.

Why it lingers

It lingers because it clarifies something that can otherwise get lost in the noise.

Meta is often discussed as if it keeps reinventing itself. Social network. Metaverse company. AI company. Hardware company. But the video suggests something more consistent underneath all of that. Meta remains, at its core, a company that builds technologies to capture attention, make that attention more measurable, and turn it into revenue. The surrounding strategies change. The centre holds.

That feels useful this week because the newsletter is circling the question of how platform power is evolving rather than disappearing. If childhood is becoming a central site where platform legitimacy is tested, then it matters that the underlying business incentives remain so stable. A company can change its name, its product mix, and its preferred future. It is harder to change the logic that funds it.

That may be the real reason the video lingers.

It makes Meta feel less like a sequence of pivots and more like a machine searching for its next durable surface.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and assets are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels—from social media activity to Google search trends—and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

Space is starting to sound less like a launch story and more like a sovereignty industry.

For years, the public imagination around space has been dominated by spectacle. A rocket lifts off. A constellation is announced. A billionaire gestures toward Mars. What stood out this week was something quieter and, in some ways, more consequential. Reuters reported that Schaeffler has signed a memorandum of understanding with Spire Global to jointly develop space hardware and satellite platforms in Europe, with applications in defence, weather, and security. Schaeffler said the strategic goal is to build a European sovereign space hardware and mission business by the end of the decade, targeting about €250 million in revenue by 2030.

That matters because it shifts the frame away from missions and toward the industrial base underneath them. This is not being pitched as a moonshot. It is being pitched as a supply-chain and manufacturing opportunity. Reuters noted that Spire has capacity to produce 300 to 400 satellites annually across facilities in the United States and Europe. The Schaeffler release goes further, saying the two companies intend to secure and scale supply chains for critical spacecraft subsystems and explore industrialised satellite bus platforms. That is not demo language. It is production language.

The sovereignty angle makes the story more interesting still. The partnership is explicitly framed around building a self-reliant European space-industrial base, industrialised in Germany, flight-proven in orbit, and deployable at scale for defence, weather, civil security, and critical-infrastructure missions. In other words, the frontier here is not only technical capability. It is the question of whether Europe can own more of the hardware, mission architecture, and manufacturing capacity behind systems it increasingly sees as strategic.

That feels directionally important because space is no longer just a communications story. It is becoming part of the sovereignty stack. Weather intelligence. Defence applications. Critical infrastructure. Environmental sensing. Once those functions begin to matter politically, the origin of the hardware starts to matter too. A satellite platform is no longer only a piece of engineering. It becomes part of a broader argument about dependence, security, and industrial resilience.

This may be the useful signal from the field. Frontier technologies often become real twice. First as a vision. Then as a manufacturing category. The second moment is easier to miss because it is less cinematic. But it is usually the one that lasts. What Schaeffler and Spire suggest is that Europe is beginning to treat space less as a downstream service and more as a strategic industrial capability that needs to be built, supplied, and scaled closer to home.

“We tend to meet any new situation by reorganising, and a wonderful method it can be for creating the illusion of progress while producing confusion, inefficiency and demoralisation.”
Charlton Ogburn Jr.

Charlton Ogburn Jr. was an American journalist, writer, and military historian with a sharp eye for the gap between institutional performance and institutional theatre. He understood that organisations under pressure often respond first with movement rather than clarity. New structures. New language. New reporting lines. Plenty of visible activity, but not always much resolution. That is what makes the line feel so durable. Reorganisation can look like momentum even when it mostly rearranges the confusion. This week’s stories sit close to that tension. Systems under pressure do not just reveal what they can do. They reveal how easily motion can be mistaken for progress.

Welcome to this week’s Field Notes, a 10-year project of mine documenting humankind’s digital transition from the field. These notes are shaped by what I’m seeing, building, and discussing as our physical and digital lives continue to converge.

- Ryan

(Connect with me on LinkedIn)

News is surface-level. Signals live underneath. This section captures developments that hint at deeper shifts in how digital systems are being built, governed, and adopted — often before they’re obvious in the mainstream narrative.

Two stories this week suggest that electrification is starting to look less like a standard automotive transition and more like an energy-systems adaptation.

In South Africa, charging infrastructure is being built to avoid the grid rather than depend on it. In Pakistan, an automaker is betting that fuel-price pressure and the spread of rooftop solar will make electric vehicles more viable even in a market where the category still feels early. In both cases, the important shift sits outside the car itself. The surrounding energy environment is doing more of the explanatory work.

In South Africa, EV charging is being designed around grid weakness from the start

Reuters reported on 19 May that Zero Carbon Charge, trading as CHARGE, launched its first off-grid, solar-powered charging station on the Johannesburg to Durban N3 corridor. The system is designed to operate independently of South Africa’s national grid, which remains constrained by aging infrastructure and periodic power cuts. Reuters said CHARGE plans to install 60 off-grid charging stations nationwide by the end of 2027, with eventual expansion beyond passenger vehicles to commercial transport, including electric trucks.

What stood out is that this is not simply a charging-network story. It is a workaround story. In many mature EV markets, the question is how to extend charging coverage. Here, the deeper question is what kind of charging system makes sense when the underlying grid cannot be assumed to be reliable. That produces a different kind of infrastructure logic. Charging becomes a parallel energy system from the outset rather than a layer added neatly on top of the existing one.

That feels directionally important because it shows how transitions adapt to local constraints rather than waiting for ideal conditions. Reuters noted that South Africa’s EV market is still small, with new energy vehicles accounting for 2.8% of new vehicle sales in 2025, but interest is growing. The signal is that uptake may depend less on abstract enthusiasm for electrification and more on whether the surrounding energy architecture can be made workable in the first place.

In Pakistan, EV adoption is starting to ride on rooftop solar and fuel shock

Reuters reported on 22 May that Pakistan’s Lucky Motors has partnered with Chinese state-owned automaker GAC to launch EVs in the local market, betting that rising fuel prices and widespread rooftop solar adoption will help drive demand. Reuters said fuel costs have doubled over the past four years, solar now provides roughly 25% of Pakistan’s electricity, and Lucky Motors plans to begin local assembly of GAC EVs by December 2026. The company is also exploring battery swapping, local battery production subject to policy support, and eventual export potential to other right-hand-drive markets.

What stood out is that this is not being framed as a top-down green transition. It is being framed as a practical response to pressure in the system. High fuel costs make the old model more painful. Rooftop solar makes home charging more plausible. A distributed energy shift begins to make electric mobility feel less like an imported future and more like something that can fit local conditions. That produces a different story from the usual one about subsidy-led adoption or state-led charging rollout.

The deeper signal is that electrification may emerge fastest where adjacent systems start to align. Reuters said EVs still make up a small share of Pakistan’s car market, while electric motorcycles have moved more quickly and now represent about 5% of two-wheeler sales. That divergence is useful. Adoption does not arrive evenly across transport categories. It follows the places where infrastructure, price, and routine behaviour happen to line up first.

What stood out

Taken together, these stories suggest that the next phase of EV adoption will be shaped less by the vehicle itself and more by the energy conditions around it.

South Africa shows what happens when weak grid infrastructure forces a parallel solution. Pakistan shows what happens when fuel shock and distributed solar begin to make EV ownership newly practical. Different markets, different pressures, same underlying lesson: in many places, electrification will not arrive as a clean, standardised rollout. It will emerge through the constraints and asymmetries of the local energy system.

What it is

This week’s watch is a conversation between Michael Maas, CEO and co-founder of Zimi Charge, and Andries Malherbe, founder of Zero Carbon Charge.

On the surface, it is a discussion about EV charging infrastructure in South Africa. But the more useful way to read it is as a conversation about how electrification takes shape in a market where the grid cannot simply be assumed. The two founders talk through the origins of Zero Carbon Charge, the buildout of the N3 corridor, the design logic of solar-powered charging sites, and the overlap between public charging and fleet electrification. What emerges is not just a transport story. It is an infrastructure story.

What stood out

What stood out is how quickly the conversation moves past the car. Malherbe describes the original insight quite plainly: electric vehicles were going to need a huge amount of power, and nobody really knew where it would come from. From there, the project became a land, permitting, route, and infrastructure challenge before it became anything else. That feels important. In mature markets, EV adoption is often discussed through product cycles and consumer preference. Here, the starting point is energy supply.

The second thing that stood out was the design philosophy of the stations themselves. These are not simply chargers placed beside a road. They are modular energy sites built around solar, batteries, circulation, logistics access, convenience retail, and enough flexibility to expand as demand changes. Malherbe’s comment that the sites are designed not to maximise raw output but to minimise the difference between winter and summer power availability was especially telling. That is a systems decision. It suggests the service is being designed around reliability rather than peak performance.

There is also a useful contrast between the public and fleet angles. Maas explains that many South African fleet operators are under pressure to decarbonise, but do not know where to begin because electric fleets require a different kind of operational thinking. The issue is not just the vehicle. It is route planning, charging behaviour, energy cost predictability, maintenance, and driver experience. The point returns again and again: the transition depends on making the economics and the infrastructure legible enough for operators to move.

Why it lingers

It lingers because it shows how electrification adapts under constraint. In places where the national grid is strong, charging can look like an extension of existing infrastructure. In South Africa, the logic is different. Charging is being built as a parallel energy system from the outset. Solar matters because it reduces exposure to diesel volatility and weak grid conditions. Station design matters because each site has to function as a self-contained node rather than a simple plug-in point.

That feels directionally important. Because the future of EV adoption may not be defined by a single global template. It may emerge through local workarounds, with each market building around its own constraints. This conversation captures that well. The car is part of the story, but only part. The larger shift is in how energy, charging, logistics, and transport begin to reorganise around one another.

Digital assets now sit less as an idea and more as infrastructure in progress. As physical and digital life continue to converge, money and assets are doing the same. What was once framed as “crypto” is increasingly showing up as rails, balance sheets, and policy conversations.

🔥🗺️Heat map shows the 7 day change in price (red down, green up) and block size is market cap.

🎭 Crypto Fear and Greed Index is an insight into the underlying psychological forces that drive the market’s volatility. Sentiment reveals itself across various channels—from social media activity to Google search trends—and when analysed alongside market data, these signals provide meaningful insight into the prevailing investment climate. The Fear & Greed Index aggregates these inputs, assigning weighted value to each, and distils them into a single, unified score

This section captures developments at the edge of digital systems. New interfaces, tools, and capabilities that feel early, unfinished, or slightly ahead of their moment. I’m less interested in what’s impressive today and more interested in what might quietly reshape how people work, coordinate, and interact over time.

The Iran war is turning electrification into a resilience story.

For years, the energy transition was often framed in moral or environmental terms.

This feels different. Reuters argued in late April that the war with Iran is becoming a “lightbulb moment” for Asia, the region most exposed to oil and LNG disruptions through the Strait of Hormuz. The immediate shock is obvious enough: around 10 million barrels a day of crude and refined products, and roughly 20% of global LNG, move through that corridor. When that flow is threatened, the costs land quickly (source)

What stood out is where the response is showing up. Not only in large national energy strategies, but in the faster-moving edges of transport and power systems. Reuters pointed to rising adoption of electric two- and three-wheelers and to more rapid deployment of battery energy storage systems across South and Southeast Asia. The useful signal is that electrification begins to look more practical when fossil-fuel dependence starts to feel less like a market assumption and more like an exposure.

That matters because it changes the logic of the transition. In this frame, batteries are not just climate hardware. They are shock absorbers. Two-wheel EVs are not just cleaner mobility. They are a hedge against imported fuel volatility. Reuters noted that Asia was already moving in this direction, led in large part by China, but argued that the Iran war is broadening and accelerating that shift across the rest of the region as well.

The most interesting part is that the frontier here is not a single breakthrough device. It is a change in the surrounding conditions. Technologies that looked optional in calmer markets begin to look strategic in unstable ones. That can happen faster in two- and three-wheelers than in passenger cars, and faster in storage than in full grid redesign, because the capital requirements are lower and the pain point is easier to see. The war does not invent those systems. It changes their urgency.

This also helps explain why the shift can feel uneven. Some markets will move first where rooftop solar, batteries, and light electric transport are already within reach. Others will keep absorbing the shock through subsidies, fiscal pain, or slower demand destruction. But the direction is clearer when fuel disruption becomes part of ordinary planning. Reuters’ framing was that the conflict is pushing Asia to reduce fossil-fuel dependence not only for climate reasons, but because the old system now looks more visibly fragile.

That feels directionally important. Because frontier technologies often become real not when they are perfected, but when the world around them changes enough that they stop feeling like experiments. In that sense, the Iran war is not just an energy shock. It is a reminder that resilience can be a stronger adoption driver than ambition.

“Children are the living messages we send to a time we will not see.”
John F. Kennedy