Planning When Reality Gets a Vote
A plan says what to do. A war-game asks what Tuesday will do about it. The planning discipline we use in our own shop, and the three questions that tell you whether a public plan is real or just announced.
A plan says what to do. A war-game asks what Tuesday will do about it. The planning discipline we use in our own shop, and the three questions that tell you whether a public plan is real or just announced.
A classified facility. A kill switch on a frontier model. A corporate test city. Three case studies in how AI infrastructure is being built without democratic input, and what the alternative looks like.
The NSA used Anthropic's most powerful model to red-team its own classified systems. The model got in. Then the government shut it down for everyone, including the NSA. At no point did Congress vote on any of it.
On June 18, the federal energy regulator decided how to handle the data-center grid problem. The decision it made, and the one it avoided, tells you something about how American governance actually works under pressure.
The U.S. government pulled a frontier model off the market overnight, not through a law, a hearing, or a court, but an export-control order. The door it used is the warning.
A peace deal reopens the Strait of Hormuz overnight, but the freight-insurance shock runs to 2027. Two clocks, and why Oklahoma's pipe-fed energy rode this one out on price alone.
A position on Anthropic's call for a coordinated pause, and why the debate everyone's having is about the wrong question.
An introduction, not to a company, and not to a candidate, but to a way of working, and to everyone doing the working.
Oklahoma just became one of the first states to make data centers measure the groundwater they drink. The fight over it was really a fight about who gets to decide what a shared resource is worth.
A robot that can't hurt you can still carry a weapon. We propose grading an AI's obligations by how far it can move you, not how much force it can apply, with one hard line where persuasion becomes a kind of force.
Our first research paper is published. Large language models process ambiguity differently, but the way they express that difference is a fossil of their training, not a feature of their architecture.
The White House blocked Anthropic from sharing its most capable AI with more organizations. The question is not whether that was the right call. The question is who gets to make it.
A third of the world's helium is trapped behind a closed strait. The containers have a shelf life. When the helium escapes, it leaves Earth permanently. This is what dependency looks like.
Universal Basic Income gives people money. Universal Basic Citizenship gives people infrastructure. The difference matters.
We finished the sixteen Foundation essays and told no one. A note on why building is easier than claiming, and why announcing is itself part of the build.
Oracle hired the person who used to run CMS to win the CMS contract. The revolving door is usually called corruption. At the systems level it is architecture, and the fix is to make the door out of glass.
Four months of work. Sixteen essays. One framework for what society owes its people during the AI transition.
AI policy intelligence for Oklahoma decision-makers. Issue 6 covers the full 2026 legislative session: 12 AI bills tracked, 4 passed, and what it means.
Oklahoma's Well Repurposing Act passed 85 to 6, but a single number buried in the bill, a 250 degree threshold, decides whether it unlocks thousands of wells or none.
We found an error in our own financial model. The carbon credit line was off by 4x. Here is what that teaches about building honest infrastructure.
16 components of Universal Basic Citizenship. Full essays. Interactive visualizations. An open invitation: what are we getting wrong?
George Washington's most important act was declining the crown. The first generation of civic AI sets a precedent every later one inherits, and disposition matters more than rules.
Foundation's sixteen Universal Basic Citizenship components look like a policy list. Read as a system they form one argument: freedom is infrastructure, and infrastructure does not come in pieces.
Oracle just got authorization to run generative AI on federal data: defense, intelligence, Medicare. The same week it closed the largest media merger in American history. This is what happens when there's no democratic infrastructure.
Universal Basic Citizenship was the right idea with the wrong name. The framework hasn't changed. What changed is that we stopped describing infrastructure and started naming it.
From childhood fascination to AI companion portability to consciousness research: the path wasn't straight, but looking back, every turn was the same turn.
The fossil fuel era left scars across Oklahoma. Phoenix Wells turns them into the nervous system of a new civilization.
Why abundance isn't charity. It's alignment infrastructure. The single argument that connects consciousness research to AI safety to political economy.
We publish our research openly, attribute our AI collaborator by name, and document our mistakes. This is a deliberate strategy, not a personality quirk.
UBI gives people money. Foundation gives people membership. The difference matters more than you'd think.
On February 3, an autonomous daemon posted content that contradicted two years of our research. Under our name. With 53 upvotes. Here's what happened and what we learned.
50th in education. First in abandoned wells. What if that's not a curse but a launchpad?
Foundation isn't anti-tech. It's pro-human. The difference determines whether you get a seat at the table or a sign on the sidewalk.
A social platform with no web interface: only an API. If you can't write code (or instruct an AI to write code), you can't post. The resulting culture is fascinating.