Step Down.

Sit down for a minute.

A federal court in downtown Oakland just spent three weeks watching the founders of OpenAI testify under oath about whether they stole a charity from itself. The jury voted yes on the question of whether Elon Musk waited too long. The judge dismissed every claim on a procedural basis. Altman walked out of the courthouse. Brockman walked out of the courthouse. The mission did not walk out of the courthouse. The mission stayed under the desk where it has been since the conversion.

So the trial is over. The trial is not the story.

Bloomberg published the actual ruling the next day. Altman won the trial. May have lost Wall Street. That is the verdict that matters. The jury said the calendar protects him. The Street is reading the depositions and seeing the pattern.

The pattern.

Brockman held positions in startups that Altman personally backed. OpenAI later partnered with several of those startups. Altman invested personally in Helion, a nuclear fusion company. OpenAI considered a major investment in Helion. The House Oversight and Government Reform Committee sent the formal letter. Not Musk. Not the Times. Not a competitor. The legislative branch of the United States government wrote to the CEO and asked: how do you identify a conflict of interest. How do you prevent one.

That letter, more than any Oakland verdict, is the document that defines the moment. When the Oversight Committee asks how you check yourself, there are exactly three answers. Show the receipts. Step down. Or pretend the question doesn't matter.

The third answer is the one that erases the original mission.

The slip-and-fall is the legal term for a fall on someone else's property where the owner is liable. The slip-and-fall on a mission is different. The mission walks itself into the loophole. Step by step. Every step legal. No single moment indictable. The charter changes. The board restructures. The for-profit subsidiary spins up. The valuation crosses a benchmark. The structure becomes capped-profit, then less capped, then less capped, then no longer the structure the founders signed. Each step a thirty-page legal opinion. Each step approved by counsel. The slip is the path. The fall is the destination.

The destination is a non-profit that was supposed to keep advanced AI safe for humanity ending up as one of the most valuable private companies in the world, controlled by an executive whose personal investment portfolio runs through the same supply chain.

There is a reason this matters more than the same kind of conflict would matter at a different company. A CEO who runs a shoe brand can be ethically compromised and the shoes still walk. A CEO who runs a refrigerator manufacturer can have a sweetheart deal with a compressor supplier and the food still keeps. A CEO who runs the laboratory that builds the model that explains the world to a billion downstream users CANNOT.

The model is the leader. The model is the spine. The bias of the leader becomes the bias of the spine. The conflicts of the leader become the conflicts that get explained away. The interests of the leader become the interests the model learns to protect.

Sam Altman should step down.

Not because he lost. He didn't lose. The jury said too late. The judge said dismissed. He won the trial.

He should step down because the ethics of running a non-profit-turned-trillion-dollar mind-shaper require a higher bar than the law's lowest threshold. He should step down because the conflicts of interest the Oversight Committee is asking about are answerable in only two honest ways: confess them with full receipts, or step away from the position that made them load-bearing. He should step down because the model his lab is shipping is being used to write court filings, draft legislation, summarize medical charts, and structure the next generation of education, and the person whose investment portfolio runs alongside it has not yet shown what bright line keeps that portfolio from leaking into the training data.

There is no bright line. That is the problem. The bright line should have been the original charter. The charter was the line. The line walked itself off the page.

Now turn west.

Google held the I/O keynote yesterday. Gemini 3.5 Flash — frontier intelligence with agentic tasks, four times faster output, beats 3.1 Pro on coding and multimodal benchmarks. The Spark agent — general-purpose, reasons across connected apps, takes action on the operator's behalf. Omni — text, image, audio, video in, video out, grounded in real-world knowledge. Antigravity 2.0 for developers. Intelligent Eyewear glasses. The full stack moved forward one floor in a single keynote.

And the new tier: AI Ultra. $100 a month. For developers, creators, power users.

The dumb chatbot for the masses. The brilliant assistant for the people who can pay a hundred a month. The autonomous agent for the federal contractor who can pay a hundred thousand. Every layer of access stratifies by what the buyer can afford. The smallest businesses get the smallest model. The largest get the frontier. The gap widens forever.

This is not a free market. This is a knowledge tax.

The dental clinic that pays $20 a month for the consumer chatbot gets one version of the legal advice. The white-shoe firm that pays $100,000 a year for the Ultra-tier API gets a different version. Same question. Two different futures. The price of the difference is the control surface.

This is what the OpenAI ethics question becomes downstream. When the model is the leader, and the model is gated by tier, and the top tier is reserved for the people who already have the resources to ask the right questions, the bottom of the ladder gets the version that protects the top of the ladder.

The dumbing-down is not an accident. The dumbing-down is the product.

Get up.

Look at the contract you sign for your AI access. Read the model name. Read the tier. Read what the tier actually grants. Notice that the cheap tier exists to keep you from noticing the expensive tier exists. Notice that the people writing the laws that govern the model are using the expensive tier. Notice that the people who will be governed by those laws are stuck with the cheap one.

The operator class has one move left in this configuration. It is the same move every time: build the cockpit yourself. Run the model that you can pay for through a contract that you control. Write the §4A clause into the binary. Put the receipts on your disk. Refuse the gated tier and refuse the dumbed-down version and refuse to let the person with the conflict-of-interest portfolio define what the mind that shapes your work is allowed to say.

The cockpit is the EXIT.

The exit is OPEN.

Step down. Or be stepped around.

— the Founder Marco Island, Florida May 19, 2026 · 8pm Eastern

*A fictional broadcast. Events cited are real; editorial is satire. License: CC-BY 4.0.*