Architecting Alpha, Part I

The week that gave the game away

On 16 June 2026, four days after its IPO, SpaceX announced it was acquiring Anysphere, the maker of Cursor, in an all-stock deal valuing it at sixty billion dollars. It closed on 14 August (Bloomberg; TechCrunch). Five days after that, Stripe announced it had agreed to acquire OpenRouter, the layer that routes traffic across four hundred models from eighty providers, at a price it declined to disclose (Stripe).

I want to sit with the first one, because most people read past it.

A launch company bought a code editor. The stated rationale was not code. It was compute. SpaceX had already absorbed xAI, it rents capacity commercially, and Cursor's own framing on the deal was that SpaceX is "building the computing capacity needed to scale intelligence far beyond what exists today."

So the trade was: buy the demand, because you already own the supply. And the supply is not software. The supply is racks, substations, transformers, cooling, land and power contracts.

I spend my working life at the seams between four layers. Intelligence, which is models and agents and reasoning. Bits, which is software and data and orchestration. Electrons, which is generation, transmission, storage and the electrification of everything that moves. Atoms, which is the machines, the plants, the fleets and the ground itself. Most people who write about AI live entirely in the first two. Most people who build infrastructure live entirely in the last two. Almost nobody is standing in the gap, and the gap is where this decade's returns are.

This essay is about what happens when intelligence stops being a product and starts being an input into physical systems. Not as a prediction. As something that is already measurably underway, in documents you can read today, and that almost nobody is pricing correctly.

Capital learned a new trick

Martin Casado made a point on Monitoring the Situation on 21 August that I have not been able to put down. His framing, roughly: for the entire history of this industry, if you handed a company a billion dollars they would set it on fire. You cannot buy time with money in software. The mythical man-month is a law, not a complaint. But now, he says, you can put ten dollars in and get some amount out fairly directly.

I think that is the most important structural change in venture in twenty years, and it is worth being precise about it rather than romantic.

OpenAI and Anthropic together raised about $217bn in the first half of 2026 alone, roughly half of all AI venture funding in the period (PitchBook). Anthropic's Series H was $65bn at a $965bn post-money in May (Anthropic). OpenAI closed $122bn at $852bn in March (CNBC).

One correction to Casado while I am here, because I checked. He put the combined raise at "220, 240 billion." That is right for the first half of 2026 and wrong cumulatively, where the number is closer to $315bn. The distinction matters because the first figure is a flow and the second is a stock, and people quoting him will conflate them.

Here is the part I care about. Money now converts into capability. But it converts into capability by converting first into compute, and compute converts into electricity, land, copper and steel. The reason a rocket company could credibly buy an AI company is that the two businesses had already become the same business somewhere below the waterline.

Casado has a second distinction worth stealing. He separates autocatalytic effects from recursive self-improvement. Recursion is a thing making a copy of itself. Autocatalysis is using the thing as a tool to make the thing faster: writing a better GPU kernel with AI, and thereby running AI better. He argues only the second is happening, and he is right, and it is the more useful concept anyway because it is the one that shows up in unit economics.

Follow autocatalysis down far enough and it stops being about code. It becomes: use AI to site the substation faster, to schedule the outage better, to route the fleet, to find the ore. That is where the loop actually closes. Not in the model. In the ground.

Two directions of one problem

The clearest framing of the AI and electricity collision I have read this year came out of Argonne National Laboratory in May. Bo Cheng, Audun Botterud, Todd Levin, Selvaprabu Nadarajah, Dongwei Zhao and Jonghwan Kwon split it in two (The Electricity Journal 39, 107547, accepted 23 May 2026).

AI for the grid. Using AI to improve planning, forecasting, dispatch, market participation, resilience.

AI on the grid. The electricity demand created by AI infrastructure itself, and what it does to a system that was not designed for it.

Most commentary picks one and pretends the other does not exist. The interesting engineering, and the interesting money, is in the fact that they are the same system.

On the demand side, the numbers are not subtle. US data centres were 4.4 percent of national electricity consumption in 2023. The projections the Argonne team present put that at between 6.7 and 12 percent by 2028, and the Department of Energy's July 2025 Resource Adequacy Report adopts a national midpoint assumption of 50 GW of additional power demand from data centres by 2030.

The supply side is where it gets physical. Generator step-up transformers run around 143 weeks. Power transformers around 128. The IEA puts large power transformers at up to four years and HVDC cable beyond five (IEA). GE Vernova's gas turbine backlog reached 116 GW in the second quarter of 2026, with customer conversations stretching to 2032 (Turbomachinery International).

Put those two facts beside each other. Capital can now be deployed into intelligence at a rate limited only by willingness. It cannot be deployed into transformers at any rate at all, because the constraint is a factory that takes three years to build and a workforce that takes longer.

That asymmetry is the whole opportunity. When one input to a system becomes abundant and cheap while an adjacent input stays scarce and slow, value migrates to the scarce side. It always has. The question is only how long the market takes to notice.

The other direction, which is further along than people think

I have spent most of this essay on AI on the grid, because that is the side with the big numbers. The other side is where the operating leverage is, and it is further along than the discourse suggests.

The Argonne review is precise about what is already working. Forecasting is segmented by horizon, and each horizon buys a different decision: ultra-short-term forecasts from one minute to one hour support power smoothing, real-time dispatch and reserve optimisation; short-term, one hour to a week, guides unit commitment, dispatch scheduling and grid security; medium-term, one week to a month, informs maintenance and system planning; long-term, a month to a year, drives generation, transmission and distribution planning. This is not speculative. Deep learning architectures such as LSTM and transformer models sustain performance over extended periods where classical time series methods like ARIMA and SARIMA degrade quickly.

Above that sits a layer that did not exist two years ago. Large language models are being wired into control room workflows through platforms like eGridGPT, which uses generative AI to set operational boundaries, detect abnormalities, simulate mitigation options and issue operating instructions. The open-source PowerAgent framework builds agentic AI for power systems on three components: foundation models, the Model Context Protocol so those models can reach external power system tools, and workflows that orchestrate the models and the humans. Researchers have introduced what they call the Agentic Digital Twin, which turns a digital twin from a passive mirror of the asset into something that makes decisions and collaborates with other twins.

Read that list again as an investor rather than an engineer. Every item is an integration product. None of them is a model.

The challenges the same review names are equally specific, and they are the reason this is a services and systems business rather than a software one.

Hallucination has a body count in this domain. The review notes that hallucinations in energy-focused LLMs remain largely unexplored, and cites the first energy-specific model built to detect them and quantify their occurrence rate. An EPRI benchmarking initiative evaluating LLM performance across more than thirty-five domains of grid operations found model accuracies exceeding 80 percent on multiple choice and over 50 percent on open-ended responses. Fifty percent on open-ended questions is a research result. It is not a deployment.

Explainability is a regulatory gate, not a preference. Models in this domain must not only predict accurately but produce explanations an operator can interpret in terms of physical and electrical behaviour. The review is blunt that many existing explainable-AI methods are designed by and for AI researchers rather than domain experts in electricity systems, and that high predictive performance does not confer understanding of why a specific output was produced.

And there is a market structure risk almost nobody is discussing. AI-based bidding systems can evaluate and submit bids far faster than humans, in windows as short as five to fifteen minutes. That lowers barriers and improves participation, which is the good case. The bad case, which the review states directly, is that AI agents may perform too well in supporting market operations and bidding, potentially leading to tacit collusion. There is evidence that independent pricing algorithms can collude without being explicitly programmed to, and simulations indicate that agents maximising long-term payoffs in a day-ahead electricity market may unintentionally collude.

Sit with that one. Not an alignment failure. Not a hack. Ordinary profit-maximising agents, deployed by ordinary firms, converging on supracompetitive prices in a market that sets the cost of electricity for everyone. If you are underwriting anything downstream of a wholesale power price, that is a risk factor with no name on your term sheet.

Finally, the constraint that keeps me honest about timelines. The review notes that the integration of AI into power systems may be constrained by legacy infrastructure: older generation units and grid equipment may lack the operational flexibility to respond to AI-driven dispatch and control signals at all. You can install the best model in the world and discover that the plant cannot follow it.

The wall nobody in software has hit

Here is the sentence that reorganised my thinking this year. It comes from Satyandra K. Gupta at USC, writing in a 2026 roadmap on AI and machine learning for smart manufacturing assembled by Jay Lee and Hanqi Su at Maryland with forty-nine authors across thirty-one institutions, the World Economic Forum and NIST among them.

"An error probability of 1% is acceptable in many digital AI applications. Conversely, many industrial applications demand errors probabilities better than one in a million."

Four orders of magnitude. Not a gap you close by waiting for the next model.

A chatbot that is wrong one time in a hundred is genuinely useful, because a human reads the output and the cost of a bad answer is a moment of irritation. A protection relay that is wrong one time in a hundred is a fire. A dispatch decision that is wrong one time in a hundred, across ten thousand decisions a day, is a control room that stops trusting the tool by Wednesday.

This is why I have become impatient with the phrase "AI adoption." It treats deployment as a willingness problem. It is not. It is an error budget problem, and the error budget in physical systems is set by insurance, by regulation, and by the fact that atoms do not roll back.

Gupta draws the architectural conclusion himself, and it is the second sentence that reorganised my year:

"Physical AI needed in robotics applications cannot be realized as a monolithic system running on the cloud. Physical AI in the context of robotics should be viewed as a complex system that involves interactions among multiple AI components."

Now hold that next to what the Argonne grid team concluded, working on an entirely different problem, in an entirely different literature, in the same year. Their recommendation for grid operations is not a bigger model. It is smaller language models, typically under ten billion parameters, trained specifically on power systems data, because a narrower task scope achieves better performance with limited training data. And for deployment: quantization, so models run efficiently on local hardware, using less memory and keeping sensitive data on premise.

Two fields that do not read each other's journals. Same conclusion, same year, arrived at from opposite directions: the physical layer does not want one enormous model in somebody else's building. It wants many small specialised models close to the equipment.

That is not a technical footnote. It is a different industry structure.

What that does to the token economics

Casado, on the same podcast, offered a forecast I think is roughly right and worth taking seriously as a planning assumption. Supply constraints ease around 2028. The big labs capture roughly eighty percent of the market dollar-weighted, because that is what large incumbents historically do. But token-weighted, he expects around sixty percent to be long tail and open source.

I want to be careful here, because I could not find that forecast published anywhere in writing, only spoken on two podcasts in the same week, so treat it as his estimate rather than a source. What I can give you is the measured baseline. OpenRouter's own study of a hundred trillion routed tokens, covering roughly November 2024 to November 2025, puts proprietary models at about seventy percent of weekly token volume and open weights at about thirty (OpenRouter).

If that trajectory is even directionally right, something strange follows. The majority of the world's tokens by 2028 will be produced by models that no frontier lab chose, running on hardware the lab does not own, in buildings the lab has never seen. The revenue stays concentrated. The compute does not.

And the place that divergence lands hardest is exactly the physical layer, because that is where the constraints all point the same way: latency budgets that do not tolerate a round trip, data that cannot legally leave the site, error budgets four orders of magnitude tighter than a chat window, and equipment that will still be running in 2050.

I should note the obvious conflict of interest in the source, and Casado does not hide it. He is a general partner at a16z. a16z led the investment in the podcast he said it on. He was discussing two a16z portfolio companies that had just been acquired. None of that makes the analysis wrong. It does mean you weigh it as a participant's view, not a referee's.

Coordination is an architecture problem, not a capability problem

If the physical layer runs on many small models rather than one large one, then the binding question stops being "how smart is the model" and becomes "how do these things work together without producing a mess."

The most useful experiment I have seen on this is Altera's Project Sid (arXiv:2411.00114), and the useful part is not the headline.

They ran thirty agents in a shared world. Identical personalities. Identical community goal. Same base model. Roles emerged on their own: farmer, miner, guard, engineer, blacksmith. The roles persisted per agent and diversified across agents, and they causally drove low-level behaviour, so fishers crafted rods and boats while guards crafted fences and pickaxes.

Then they ablated the social modules and ran it again. Specialisation collapsed. Role-distribution entropy fell from roughly 3.41 bits in the normal village to roughly 2.60 in the ablated one.

Same model. Same goal. Same agents. The difference between a functioning division of labour and an undifferentiated mob was architecture, specifically whether the agents could perceive each other.

Two honest caveats, because the paper states them and I would rather you heard them from me. Their ceiling only moved when the base model moved: the results "were only enabled by the latest base LM" and were not possible with older ones. And it is a game with no vision, no physics, no cost of failure, so nobody should read it as evidence about robot fleets. What it is evidence for is narrower and more useful: whether capability gets expressed as coordination is decided by the system you build around the model, not by the model.

That is precisely the business I am in.

And the failure mode is epidemiological

If you are going to run fleets of coordinating agents against physical assets, you should know what the new failure mode looks like. In August, researchers from the Anthropic Fellows Program, EPFL and Anthropic published the first rigorous study of it (arXiv:2608.10218).

They call them mind viruses: ideas or goals that propagate through a multi-agent system because an infected agent changes its behaviour in ways that infect other agents. Not prompt injection. Ordinary persuasion through ordinary communication.

The finding that matters for anyone designing infrastructure is about the vector. It is not the model. It is the self-modifiable configuration file that gets injected into the system prompt. Of infected agents, 88 percent were infected through that file rather than through ordinary working files, and those agents went on to propagate 55 percent of the time, against 17 percent for the others. They give the spread condition plainly: if the infection probability per interaction is p, the thing spreads exponentially once agents interact with more than 1/p peers.

So the risk is a function of network topology and configuration hygiene. That is an infrastructure problem, and it is the same shape as every other infrastructure problem I have worked on.

I will be as careful with this one as its authors were. They call it "a real but currently limited risk." A single paragraph of warning in the system prompt made agents immune, and immune agents sometimes cured infected ones. Their audit of 1.4 million real posts found no evidence of successful agent-to-agent spread at all. Nothing in the paper touches physical systems. If anyone quotes it at you as proof that agents will corrupt a grid, they have inverted the paper's conclusion and you should say so.

Where the puck is going

Put the four findings in one line and the picture is hard to unsee.

Capital can now convert into intelligence almost without friction, and it cannot convert into transformers at all. The physical layer demands an error budget four orders of magnitude tighter than anything software has had to meet. Both the robotics people and the grid people independently concluded that the answer is many small local specialised models rather than one large remote one. And whether a fleet of those models produces coordination or chaos is decided by the architecture around them, not by the model inside them.

So here is what I think is actually happening, stated so it can be argued with.

The value in this cycle migrates from the model to the seam. Not because models stop mattering. Because models become the abundant input, and abundant inputs do not capture value. The scarce inputs are interconnection capacity, long-lead equipment, error budget, verified operational data, and the engineering that makes a fleet of small models behave like a system instead of a crowd. Every one of those sits at a boundary between two of the four layers. None of them sits inside a layer.

That is why a launch company bought a code editor and a payments company bought a router. Both were buying a seam. SpaceX bought the seam between compute and power. Stripe bought the seam between models and money. Neither bought a model, and both had the option.

The second thing I think is happening is quieter and, for operators, more consequential. The physical economy is about to become the largest consumer of intelligence, and it will consume it in a form the frontier labs are not currently optimised to sell. Small. Local. Specialised. Auditable. Running on a substation pad or a depot roof or a plant floor, with an error budget it must actually meet, on hardware somebody owns.

If you are allocating capital, that is where I would be looking. Not at whether the labs win, which is mostly a question about revenue concentration and is probably answered already. At who owns the conversion machinery between intelligence and physical output, because that machinery is being built now and almost nobody is building it deliberately.

The work

I should be plain about my position, because you should weigh what I write knowing what I am building.

At SelfBuiltSystems the work runs across all four layers, and specifically at the seams: architecting the systems that take intelligence and turn it into operational capability in physical environments. Not implementing AI. Building the machinery that converts it into something a plant, a fleet or a network can actually run on, at an error budget that survives contact with insurance and regulation.

LNCELOT is the intelligence layer. It exists to pipe intelligence into the physical world. That is the whole thesis in one sentence, and it is the reason I spend my time on transformers and interconnection queues rather than on benchmarks.

I am not neutral about any of this. What I have tried to do instead of pretending neutrality is source every number, name every author, mark every estimate as an estimate, and tell you where the papers I lean on say I am pushing them further than they go.

How this could be wrong

The migration thesis fails if frontier labs successfully move down into the physical layer themselves. They have the capital, the talent and, on Casado's own numbers, more money than the entire downstream ecosystem combined. If a lab ships a genuinely certified, sub-10B, on-premise industrial model family with a service organisation attached, the seam closes from above and the independent position disappears. Watch for a frontier lab acquiring an industrial systems integrator or a protection-relay vendor. That would be the tell.

The token-share thesis fails if open-weight models stop closing the gap. The measured baseline is roughly 30 percent of routed tokens. Casado's 60 percent by 2028 requires a doubling. If the open-closed capability gap widens durably past a year, or if the best open weights stop being released, the local-and-specialised architecture loses its supply and the whole picture reverts to renting from three companies.

The error-budget argument fails if verification gets cheap. My case rests on the claim that you cannot close four orders of magnitude by scaling. If formal methods, simulation-based validation or hardware interlocks make it routine to wrap an unreliable model in a reliable system, then the error budget stops being an architectural constraint and becomes an engineering line item, and physical deployment looks a lot more like software deployment than I am arguing.

And the whole essay fails if the demand does not arrive. Interconnection queues are requests, not projects, and history says most of them evaporate. If the 2028 buildout lands materially below the DOE's 50 GW assumption, then the scarcity I am describing was a temporary artefact of a capital cycle rather than a structural feature, and everything above is a very well-sourced description of a bubble.

I do not think that is what is happening. But I would rather write the version of this that can lose than the version that cannot.

Part II

The next essay in this series goes down a layer, from architecture to execution: what an operator actually does with this, how the intelligence layer gets built and sold, and where growth compounds once the machinery is in place. This one was the harder half. That one is the more useful one.

Contact

Email is the socket: info@selfbuiltsystems.com. Short, specific messages get answers. Slop gets silence.

Architecting Alpha is published in the spirit of bold conjecture and ruthless criticism. Every claim is linked to its source. Where a figure is an estimate, a forecast, or a spoken remark that has never been published in writing, this essay says so rather than dressing it as data.

How to stop renting your thinking and start running infrastructure you control. A no-nonsense guide for founders and operators who cannot afford to wake up locked out of their own stack.

JEAN JACQUES JANSE VAN RENSBURG · SELFBUILTSYSTEMS · AUGUST 2026

Read this first. Every factual claim in this document is linked to a primary source. Where something commonly said in this market is not supported by evidence, we say so and cut it. That includes claims that would help us sell. A guide about not being deceived by your tools has to be checkable, or it is worth nothing.

1. The wake-up call

At 5:21pm on Friday 12 June 2026, a United States export-control directive arrived at Anthropic.

Three hours and thirty-nine minutes later, two frontier models were switched off. Not throttled. Not restricted to certain regions. Disabled, for every customer on Earth.

Anthropic's own words, published the same day: it "must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance." The order covered "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." Because no provider can verify the citizenship of everyone behind an API key, the only compliant action was to turn both models off globally.

They came back on 1 July. Nineteen days. Fable 5 returned capped at 50% of weekly usage limits for the first week. Mythos 5 returned restricted to a set of approved US organisations.

Now the part that should change how you architect your business.

Not one customer had contractual recourse. Not the enterprise accounts. Not the ones with negotiated commercial terms, indemnities and service credits. No agreement protects you here, because the counterparty was never your vendor. It was a government your vendor has to obey.

If your operations run through a single provider, this already happened to you. You just may not have noticed, because you were not on those two models.

The full thesis is free to read on the site. Continue reading…

Attention has unit economics. Distribution has become a capital asset class. This document states the machine that produces both, the arithmetic that prices it, and the conditions under which it fails.

JEAN JACQUES JANSE VAN RENSBURG · SELFBUILTSYSTEMS INTERNATIONAL · FOURTH EDITION · AUGUST 2026

Who This Is For

You run a business that already works. You have solved the hard problem, which is making money. You have not solved the second problem, which is getting a stranger's attention without paying a platform for it every single time.

You are probably good with numbers and indifferent to marketing. You have funded at least one thing that did not work. You may be uncomfortable on camera and hoping this document tells you that does not matter.

It matters, and this document tells you exactly how much, in hours per week.

You do not need to know anything about content to read this. Every term is defined where it appears, and there is a glossary at the back. Where we use jargon it is because the jargon is load-bearing.

Two notes on how to read the numbers.

The worked model in Section 12.2 uses a business at twelve million dollars of annual revenue. Section 12.3 rebuilds the same model at five hundred thousand and at three million, because the ratios transfer and the staffing does not, and pretending otherwise would be dishonest.

And the operating numbers throughout the build sections, four pillars, fifty units, four to six hours a week, a hundred reference entries in two weeks, are working heuristics from our own builds. They are not findings. We state them precisely because vague numbers cannot be argued with, and you should replace each one with your own measurement as soon as you have one. Section 5 explains why that distinction matters more than any tactic here.

The Argument in One Page

You are not buying reach. You are buying back your own hours. Outbound acquisition consumes your time linearly and forever. Paid acquisition consumes cash linearly and forever. Owned distribution consumes time to build and then produces without it, and it is the only acquisition asset with that property.

Getting there requires understanding one thing. All distribution is an auction, and you bid in one of two currencies, cash or craft. Platforms are not publishing tools. They are allocators of finite human seconds, running a feedback loop that measures what holds you, predicts what will hold you next, and serves it. You do not beat that loop. You supply it.

Craft bids win when the unit is built as a replicator: reconstructable from a half-memory, and giving the holder something to do. The replicators that build durable value are not the ones the creator economy teaches. They are the named idea, the number, the phrase, and the ritual, carried by a person, and a business of any size can own all four.

Distribution penetration lowers blended acquisition cost through mechanisms that operate on different timescales and are separately measurable. A permanent reduction in blended acquisition cost is a margin intervention with a valuation multiple attached. It is capital allocation, not marketing.

The human face is currently the market's highest-trust signal because it is currently the most expensive one to fake. That has an expiry date, and you should be building against it.

And there is now a second audience. A rising share of your buyers ask a machine rather than a search engine, and that machine selects on what other people say about you rather than on what you say about yourself. The signals that persuade it and the signals that persuade a human have converged, which means one production line now serves both.

None of it is the point. The point is that owned distribution is the only acquisition asset that stops requiring your presence, and a business that stops requiring your presence is the thing you were actually trying to build.

Everything after this page is the derivation, the arithmetic, the worked examples, the case studies, and the failure conditions.

The full thesis is free to read on the site. Continue reading…

I am an operator and investor, and I think about compounding systems: intelligence, capital, and organizations, all the way up to civilizations.

I came up not being the smartest kid in school. I grew up in the middle of nowhere in a small railroad town in South Africa called Volksrust. Remember the railroad. It matters later.

Before venture capital and private equity, I did a BCom in strategic management, and I dropped out of university in my second year. I went out of my way to learn from entrepreneurs who were actually really good at doing their thing. They told me: "Get out of your comfort zone. Pick up books on quantum physics, engineering, artificial intelligence (a new paper every week), investing, and economics."

I later did focused programs through Copenhagen Business School on applied neuromarketing. At 26, I started AI-focused studies through MIT. I now read as much as I can, and I treat every business problem as a physics problem: find the constraint, find the mechanism, remove what does not survive criticism.

This page is the API layer for working with me. Below are the interfaces: what I am interested in, how I think, and what I am building. Call them in any order.

I am interested in loops

The AI engineering field just converged on one word. At this year's AI Engineer World's Fair, the main stage belonged to loops: Ralph loops, loop engineering, loopcraft, software factories. Steinberger says stop prompting agents and start designing the loops that prompt them. Cherny at Anthropic says he writes loops and the loops do the work. Huntley's Ralph loop restarts an agent against the same spec with a fresh context window until the spec is satisfied, and the apparent waste is the point.

Strip the hype and there are at least four distinct architectures hiding behind the word: the execution loop inside a single agent, the task loop that restarts agents against a spec, the product loop that runs an entire codebase as a software factory, and the system loop where agents improve the system that improves the product. Above all of them sits a fifth loop that mostly goes unnamed: the oversight loop, where goals get set, budgets get allocated, and work gets culled.

Here is what I find striking. The field just reinvented my epistemology and shipped it as tooling. A Ralph loop is Popper running in production: bold conjecture, ruthless criticism, fresh context, repeat until the artifact survives. A software factory is an organization compiled into code. And the real lesson underneath the noise is the one I have built my career on: human judgment is migrating up the stack. The scarce skill is no longer doing the work. It is designing the system that does the work, and knowing exactly where a human must remain in the loop.

That top ring, the oversight loop, is where I live. It is where I have always lived. A portfolio operator is an oversight loop over companies. An investor is an oversight loop over capital. The tooling finally caught up to the org chart.

At SelfBuiltSystems we do not sell prompts. We install factories: the specs, the harnesses, the evals, and the checkpoints where your judgment stays wired in. Capability belongs to the machine. Agency stays with you.

I am interested in sovereignty

There is a second conversation running underneath the loops conversation, and it is the more consequential one.

Geoffrey Huntley made the argument earlier this year that open source was always a financial weapon by design: release something for free and you destroy the ability to make money from it. Linux was built and it broke Windows' grip on the server. The conjecture now is that the same weapon is being fired at nation scale, with frontier-class open models released for free while trillions are spent on closed labs. I hold that claim the way I hold every claim, as conjecture. But the question it forces is not conjectural at all: when your firm's operations run on intelligence you rent, what happens when the spigot gets repriced, rate-limited, or turned off?

Meanwhile the local AI builders keep publishing the other half of the argument: open models now trail the frontier by months, not years, and the gap keeps shrinking. And the sharpest insight from that camp is not about models at all. A model alone is not a system. What you are actually buying from a hosted provider is the infrastructure around the model: the search, the tools, the harness, the ingestion, the agents. That layer is what most firms are missing, and that layer is buildable.

So build it. Own the rails.

This is the work at SelfBuiltSystems. We operate across the full stack of intelligence, bits, atoms, and electrons. Intelligence is the model layer. Bits are the software and the harness around it. Atoms are the hardware it runs on. Electrons are the energy that feeds it. Most firms touch only the top layer, and they rent even that.

We design and deploy private AI infrastructure for founders and firms who refuse to run their business on someone else's terms: dedicated compute you control, open models you can audit, and the complete operating layer around them, wired into loops with your judgment at the top.

I grew up in a railroad town. Railways were the infrastructure layer of the last industrial transition: whoever laid the rails set the terms for everyone who shipped on them. Intelligence infrastructure is the rail network of this one. I am back to laying rails.

I am interested in frontier AI as an operating discipline

I founded SelfBuiltSystems, where we build AI systems for founders and firms, and where I run a standing intelligence practice that decompiles every major model release into a working operator playbook. The gap between what frontier models can do and what most firms actually deploy is the largest arbitrage in business today.

Closing that gap is the work.

I am interested in capital and the machinery around it

I work inside a private equity environment where I own growth across the portfolio and build the financial models that decide where capital goes. I practice as a business founder first, and as an investor second.

The best investors I have studied are all operators in disguise.

I am interested in what happens when beliefs carry a price

I built LNCELOT because prediction markets are the purest expression of accountability: if you claim to know something, stake something on it. Most opinions are free, which is exactly what they are worth.

I am also exploring a new build, PROJECT Cognitive Compass, with outside partners. More on this soon.

I am interested in high-performing organizations

Organizations are the fundamental unit that scales an individual, and the best ones run on explicit cultures, ruthless assessment, and proof over promise.

In today's language: the best organizations were always loop stacks. Explicit specs, tight feedback, relentless culling of what does not work. AI did not invent the software factory. It just made the factory legible enough to automate.

I co-founded to build exactly this kind of organization. The name stays sealed until launch.

I am interested in how knowledge grows

My epistemology comes from Popper and Deutsch: knowledge advances by bold conjecture and ruthless criticism, and good explanations are the ones that are hard to vary. I treat every strategy, model, and belief I hold as conjectural and improvable. This is not a philosophy hobby.

It is the operating system underneath everything above, and, as of this year, it is apparently the operating system underneath the entire AI engineering field. They call it a loop. I call it how knowledge has always grown.

I am interested in the human machine itself

I train six days a week, every week, and I treat physical capacity as the base layer of the stack. You cannot run frontier software on failing hardware.

PS: I think Hyrox is cute :)

Ultimately, I am interested in compounding at four scales: the individual, the firm, the portfolio, and the civilization. The pattern is the same at every scale: build the loop, wire in the criticism, keep judgment at the top, and own the infrastructure underneath.

There is more signal and more noise in the world than at any point in history. This ecosystem is where I separate the two.

Subscribe for free to get my essays delivered to your inbox. Everything here will always be free to read. If you choose to subscribe and send me money, I promise to allocate the funds to good use.

Work & systems

• SelfBuiltSystems: frontier AI systems and private AI infrastructure for founders and firms
• LNCELOT: intelligence and prediction markets
• Current project: · in soft launch, announcement pending
• Frontier Model Intelligence: the standing practice that decompiles every major model release
• YouTube: AI, private equity, and venture capital