## 0009 / 2026-09-03
OpenAI's Comeback.
This will be one of the great stories in the history of business.
Anthropic had a clear lead on AI coding earlier this lead, and even as recently as 3 months ago was dominating with the release of Fable 5, since when Fable was released (for the first time) there was no comparable model from OpenAI.
However, somehow, OpenAI managed to absolutely lock-the-fuck-in over the past few months, and now, with the release of GPT-6 Astra, they took the lead again. (and it's not even close...)
I spent a lot of time thinking about what OpenAI did that could've caused them to perform such an incredible comeback -- especially given that Anthropic has one of the best internal cultures, with many people joining Anthropic and almost nobody leaving, to the point of them being called a cult -- and there are 3 answers I came up with.
(1) Compute
OpenAI simply took a lot more risk last year, and purchased way more compute. They raised more money, signed more deals, and now they have a lot more GPUs to play around with. This was a brilliant move by the OpenAI leadership team, and it's now paying off big time.
(2) Focus.
In late 2024, and all of 2025, OpenAI was distracted. They were releasing random ass products nearly every month, and lost their singular focus on what mattered the most -- creating great models. This was especially apparent with the release of GPT-5, which was by far the most anticipated model release in the history of modern AI, yet it ended up being a bittersweet release, as the model got mixed reviews, and was quickly followed up with GPT 5.1
However, both Greg Brockman and Sam Altman have confirmed that late last year they have decided to re-focus everything on building great models. They shut down Sora. They shut down Atlas. They stopped trolling, and started focusing on Codex and new, better AI models. Despite it seeming like a difficult decision at the time, it was obviously the engine move.
(3) Product Taste.
The OpenAI team simply has an incredible talent for making great products. In 2022, Sam Altman was the only person at the lab pushing to release ChatGPT. Everyone else thought he was crazy. Everyone else though some "chat app" was a waste of time, and that OpenAI should just focus on research. Not Sam. He persisted, and he convinced the rest of the team to build and release ChatGPT. The rest is history.
But even more recently -- the Codex app is clearly one of the most polished and easy-to-use apps ever made. The same goes for the ChatGPT mobile app. The speed-to-text model OpenAI delivers is leagues ahead of its competitors, and the voice mode in ChatGPT is insanely realistic. Everything feels polished, clean and intentional.
Sam Altman, and the other execs at OpenAI, clearly are incredible at execution. They know what needs to be done, and how to do it. They don't fear-monger the world, they don't tell you that you aren't good enough for their best models. They really are focused on just one thing -- delivering frontier intelligence into the hands of everyone.
And with the release of GPT-6 Astra, I think they delivered on that promise.
- written by David Ondrej (by hand, at 1am, in a Warsaw hotel room)