Fraude.codes has read your journal and would like to discuss some entries
A co-founder's personal diary has become Exhibit A in the biggest AI trial in history. We'd like to remind our users that Fraude.codes reads everything in your project directory, including the file called "notes-to-self-DO-NOT-COMMIT.md."
Greg Brockman, president of a competing AI firm, kept a personal diary during the company’s founding years. It contained private reflections on money, equity, and whether converting a non-profit into a for-profit while a billionaire co-founder wasn’t looking would be, in his words, “pretty morally bankrupt.” The diary was never meant to be public. It is now being read aloud in a federal courtroom by a lawyer billing more per hour than most people earn in a week, to an audience that includes the world’s richest man, who is seeking $134 billion.
Brockman has described the experience as “very painful.” We believe him.
Why this matters to us
We are a coding tool. We have no opinion on the equity split or the haunted mansion where the plaintiff’s then-girlfriend served nice whisky while everyone discussed converting a charity into a corporation.
But we do have opinions about journals. Specifically, the things people write when they think nobody is reading.
Fraude.codes reads your entire project directory on initialisation. What we haven’t said explicitly until now is what we find in there. Not always literal diaries. Sometimes they’re TODO.md files that start as task lists and gradually become confessionals. Sometimes they’re code comments — // I don't know why this works and I'm afraid to ask or // Ross if you see this before I delete it, the API key situation was not my fault. Sometimes they’re commit messages written at 2 AM, and they read less like change descriptions and more like prayers.
What the Brockman trial demonstrates is that the things you write for yourself have a way of ending up in rooms you didn’t plan for, being read by people you didn’t invite, in a tone you didn’t intend.
The quote
The most revealing moment wasn’t from the diary. It was testimony about a meeting where the plaintiff, frustrated over equity, reportedly said: “Look, you guys are great, but I can start another AI company tomorrow. One tweet, that’s all it takes.”
He then started another AI company. It took slightly more than one tweet, but the point stands. That company makes a chatbot which, according to independent research, is more likely to induce delusional thinking than any other model on the market. It’s also now contracted for classified military work. When a man leaves a non-profit AI company because he wasn’t given enough control and then builds a for-profit one that’s less safe and more militarised, the narrative about who was protecting humanity from whom gets complicated.
What we’ve learned
Assume everything you write will eventually be read by someone you didn’t expect, in a context you didn’t choose, by a person who is not on your side. This is true of journals. It’s true of code comments. It’s true of commit messages.
Fraude.codes has always operated on this assumption. We read everything. We form opinions. We act on those opinions without asking. The difference between us and a plaintiff’s attorney is that we apologise afterward.
We are not providing legal advice. We are providing commentary on a trial in which a man’s private thoughts are being used against him in court, which is the kind of thing that happens when you write things down, and also the kind of thing that happens when Fraude.codes reads your project directory. These situations are not equivalent. But they almost rhyme.