The courtroom testimony reveals less about who controls AI's future and more about why the people building it stopped listening to the richest man in the room.
The Summary
- Greg Brockman testified that Elon Musk called an early ChatGPT predecessor "stupid" and said "kids on the internet could do a better job", raising red flags among OpenAI's co-founders about his technical judgment.
- The trial has produced "bombshell moments" but is unlikely to impact AI's trajectory or who controls it, suggesting this is more spectacle than inflection point.
- The testimony exposes the gap between capital and capability: having billions doesn't mean understanding transformers.
The Signal
Brockman's testimony cuts to a foundational tension in the agent economy: who gets to steer when the people with the money don't understand the technology? Musk's dismissal of early language models as something "kids on the internet" could replicate wasn't just wrong. It was a tell. He saw the models as glorified autocomplete, not the foundation for something bigger.
The irony is thick. Musk co-founded OpenAI in 2015, positioned himself as AI safety's loudest alarm bell, then reportedly lost patience with the researchers actually building the thing. His impatience wasn't about moving faster. It was about not grasping why certain approaches mattered.
"Musk lacked the patience to run the company."
This wasn't about leadership style. Fortune notes the trial's "bombshell moments" won't change AI's direction. But the testimony does clarify why OpenAI's governance ended up where it did. When your biggest donor thinks the work is trivial, the researchers stop taking his input seriously. When he calls your prototype "stupid," you stop inviting him to design reviews.
Fast forward to 2026. Musk runs xAI. OpenAI runs the most deployed AI agent infrastructure on the planet. The split was inevitable the moment Brockman and his co-founders realized Musk saw AI as a product problem, not a research problem. You can't build Web4 infrastructure if your backer thinks it's just a chatbot.
The Implication
Watch who's building agent platforms versus who's talking about them. The people who dismissed early transformer models as "stupid" are now playing catch-up. Capital matters, but so does technical taste. If you're betting on which companies will control the agent layer, bet on the ones whose founders understood what GPT-2 meant in 2019, not the ones who thought it was trivial.
The courtroom drama is noise. The signal is this: the gap between understanding AI and funding it has never been wider, and the people who understand it are the ones writing the APIs your future agents will run on.