Weekly Newsletter #57 | Updates Worth Your Attention 👀

Did OpenAI steal Levent Alpöge’s Navier-Stokes research? Silicon Valley dramas are coming to the big screen. GPT-6 Astra unveiled.

sep 9, 2026 4 min

Did OpenAI Steal Levent Alpöge’s Work?

Anthropic researcher and mathematician Levent Alpöge and Tristan Buckmaster, a mathematician from NYU, have been working to solve fluid mechanics equations using AI models, recently focusing specifically on the Navier-Stokes problem. Since Alpöge has been conducting this research personally, OpenAI‘s coding assistant Codex and OpenAI models were also used in the work.

Then, out of nowhere, OpenAI announced that it had found a solution to the same problem, and things got messy. Alpöge and Buckmaster are alleging that OpenAI read through their Codex conversation history and lifted their approach. OpenAI flatly denies it, but we already know they technically can access user conversations under the cover of “improving their models.” We strongly recommend reading the Twitter/X thread; it’s quite a ride.

The lesson here: OpenAI is using the excuse that Alpöge works at Anthropic to avoid crediting him. But this raises a deeper question — who deserves credit, the model or the person using it? And watching AI companies compete so fiercely that they can’t cooperate even on advances that benefit all of humanity is genuinely worrying.

Meanwhile, OpenAI clearly wanted to use this moment to spotlight GPT-6 Astra, announced the same week. As with every big model release, they’re saying AGI again. The model’s ability to operate a computer and browser like a human, handling almost anything we do in digital environments, is what’s driving the AGI debate this time. We really liked the launch video.

OpenAI’s announcement (X) → | Alpöge’s claims (X) → | Bubeck’s DMs to Alpöge (X) →

Silicon Valley Dramas Heading to the Big Screen

A few announcements this week bridging Silicon Valley and cinema.

Elizabeth Holmes, founder of Theranos, one of the Valley’s most spectacular fraud stories, is heading to theaters in October in an A24 documentary where she plays herself. Holmes claimed to have developed technology that could run dozens of blood tests quickly from a single small prick of a finger. The company reached a $9 billion valuation before a Wall Street Journal reporter, working with former employees, exposed that the technology didn’t work and traditional machines were running the tests behind the scenes. Convicted in 2022 and sentenced to 11 years, Holmes invited a film crew into her home 34 days before prison and had them document her every move. That film is You Can See Everything.

From the Neon team comes Artificial, dramatizing the 2023 drama at OpenAI involving Sam Altman. Andrew Garfield plays Altman, and the film opens December 25. We can’t wait.

You Can See Everything — Trailer → | Artificial — Trailer →

ADHD-Friendly Claude Skill

If you’re tired of Claude giving you overwhelmingly long answers, you might want to try this skill that makes it ADHD-friendly.

âšĄïž Quick Bits

‱ Apple is announcing new iPhones today. Apple Events →

‱ Google chose Finland for its largest AI infrastructure investment in Europe. Reuters →

‱ A former Anthropic employee says AI will kill people within ten years. X →

đŸ“© Liked this newsletter? You can forward it to your friends
if you are feeling nostalgic, though it’s always better if
they just subscribe on their own.

tio digest

enjoyed this issue?
get the next one.

A weekly digest of tech & AI news worth your attention. Straight to your inbox, every week.