GPT-6 Astra. AGI?

[AerisVahnEphelia]

1 minute ago, Carl-Richard said:

Leiv Eiriksson discovered America before Columbus. Checkmate buddy.

No one discovered America. It was already inhabited.

[Schizophonia]

15 hours ago, LordFall said:

Avec du jambon bien sûr, et même des champignons! Moi, j'adore. 

Yuck 

oeufs à la coque is the only good way to eat eggs 

34 minutes ago, AerisVahnEphelia said:

Racist colonialist perspective.
French has a pass.
It founded America.
Leo fix it write (except french because we founded the united states)

If Leo were so enlightened he would abandon that stupid peasant language and force everyone to learn French.

Edited yesterday at 05:16 PM by Schizophonia

[Lucasxp64]

That's not AGI, it's expected improvements. We knew it would get better, we already know that AI systems can do spatial reasoning in video  generators, anybody not seeing it coming is silly. But it didn't change fundamentally how LLMs operate, it still has the same issues, and it doesn't need true AGI to disrupt society and work at all. Human technology is always moving towards efficiency in all areas, people just get to operate in higher-level abstractions. A 3D graphics designer crying about it is like a bricklayer crying about pre-made concrete slabs. Now you just work with concrete slabs, you still need people and capital to put together the slabs.

We might get AGI, but this isn't it yet, and incremental improvements will be extremely disruptive, but it won't be a cataclysmic replacement. 

[Lucasxp64]

On 06/09/2026 at 8:14 PM, Leo Gura said:

You can't trust any of these models because these companies spend billions of dollars to game the benchmarks and make their models better and better at faking intelligence.

These companies don't care about science or intelligence or anything genuine. They are just doing anything that works. And by works, I mean, fools people into thinking it is intelligent.

Can you train a child to fake reasoning? Yes. Does that mean he can actually reason? No.

If you give me 5 years and a billion dollars, I can train a group of kids to be fake geniuses. They will fool most of you. But they are not actual geniuses and they cannot do the work of even one genius.

You will say, "So what? It's good enough." No, it's not good enough when it comes to doing the work that needs to be done by a genius. Or even an average man. You are assuming that this fakery is enough to do good work, but it isn't. This is like assuming that fool's gold can do the work of actual gold just because they both look yellow.

You can take lead and paint it yellow, but that doesn't mean it can do what gold does. Once you start to really work it, it will not behave like gold.

I completely agree with Leo here. This is not some kind of cataclysmic improvement, just expected improvements. Nothing fundamentally changed about their architecture, I won't be holding my breath for "AGI"

If AGI it happens it will be like a nuclear war anyways, no amount of hiding under the table will save anybody if you're within the blast radius.

Edited yesterday at 06:22 PM by Lucasxp64

[oOo]

9 hours ago, oOo said:

 

Consider your AGI question now solved @Joshe et al. 

See below.

 

 

 

 

More formally:

 

 

[Lucasxp64]

5 hours ago, Schizophonia said:

If Leo were so enlightened he would abandon that stupid peasant language and force everyone to learn French.

D'accord. Bah oauis, la suprématie de la langue française est un fait. English is an abhorrent Germanic folk peasant language, romance languages are successors of the great roman empire.

[Leo Gura]

@Schizophonia

[Natasha Tori Maru]

@AerisVahnEphelia @Carl-Richard @Schizophonia Any more out of you three and I'll be delighted to slap you upside with warning points! 

Absolute UPSTARTS! 

Edited 20 hours ago by Natasha Tori Maru

[Carl-Richard]

17 minutes ago, Natasha Tori Maru said:

@AerisVahnEphelia @Carl-Richard @Schizophonia Any more out of you three and I'll be delighted to slap you upside with warning points! 

Absolute UPSTARTS! 

You can't threaten me with a slap 😏😏 Sry 🙏😌

Edited 20 hours ago by Carl-Richard

[Natasha Tori Maru]

Just now, Carl-Richard said:

You can't threaten me with a slap 😏😏

The zen sting of rebuke! 

[Jirh]

On 08/09/2026 at 0:10 AM, Schizophonia said:

Omelette du fromage 

I know this one. From Dexter 😆

[Lucasxp64]

On 08/09/2026 at 3:32 AM, Natasha Tori Maru said:

Excuse me, boys.

Naughty. Naughty boys.

Motherfucking TROUBLE TROUBLE TROUBLE ON THE INTERWEBS.

See below markup in green.

hErE cOmES tHe FuN POlIcE  

 

@CARDOZZO 🤣 Bora encher o saco deles falando português ahhahaah

[Lucasxp64]

 

OPENAI PROBABLY STOLE THE WORK OF A MATHEMATICIAN WORKING ON A VERY IMPORTANT UNSOLVED MATHEMATICAL PROBLEM BECAUSE IT GOT LEAKED INSIDE OF CHATGPT, AND THE AI MODEL GOT TRAINED ON IT.

OpenAI claims that their unreleased state of the art model "did solve it by itself". That will be very controversial.

Edited 6 hours ago by Lucasxp64

[AerisVahnEphelia]

50 minutes ago, Lucasxp64 said:

 

OPENAI PROBABLY STOLE THE WORK OF A MATHEMATICIAN WORKING ON A VERY IMPORTANT UNSOLVED MATHEMATICAL PROBLEM BECAUSE IT GOT LEAKED INSIDE OF CHATGPT, AND THE AI MODEL GOT TRAINED ON IT.

OpenAI claims that their unreleased state of the art model "did solve it by itself". That will be very controversial.

A lot of crazy people constantly claim that “their work” is being stolen, even before AI. They see themselves as geniuses far too much.

Edited 5 hours ago by AerisVahnEphelia

[Schizophonia]

15 hours ago, Lucasxp64 said:

D'accord. Bah oauis, la suprématie de la langue française est un fait. English is an abhorrent Germanic folk peasant language, romance languages are successors of the great roman empire.

Et ouai 

7 hours ago, Jirh said:

I know this one. From Dexter 😆

very good serie btw

15 hours ago, Leo Gura said:

@Schizophonia

What an ungrateful person

15 hours ago, Natasha Tori Maru said:

@AerisVahnEphelia @Carl-Richard @Schizophonia Any more out of you three and I'll be delighted to slap you upside with warning points! 

Absolute UPSTARTS! 

D’accord on arrête 🧐

[Lucasxp64]

5 hours ago, AerisVahnEphelia said:

A lot of crazy people constantly claim that “their work” is being stolen, even before AI. They see themselves as geniuses far too much.

What actually happened was that those researchers were working on it, they were actually using AI to help them solve the boring parts, and they had published publicly their partial results so far, once they seemed to be close enough at solving the problem using a specific method that they theorized

We have a hypothesis that is still contested that: OpenAI did outspend them on AI compute to arrive at the answer faster, with professional mathematics team inside of OpenAI guiding the model.

LLMs have not spontaneously evolved into autonomous mathematical deities. They have not changed fundamentally. What changed is the engineering wrapper placed around them. By combining an LLM with a formal verification engine (like Lean) and scaling it across 10,000 coordinating agents doing brute-force tree search, OpenAI created an industrial-scale syntax-checker and inequality-solver. It is a massive, automated trial-and-error machine guided by linguistic heuristics, not a conscious "genius."

When an LLM produces 300 billion tokens, billions of those tokens are discarded trial-and-error search waste dead-end paths. Humans do not need to generate billions of external tokens because our brain performs massive subconscious filtering before a single symbol is ever written. The advantage of LLMs is they have an external way of proving the correctness of their slop, they can outpace humans, meanwhile when humans collaborate we spend the majority of our time in communication and checking each other.

If a single human consciousness could experience 88 hours as 3,000 subjective years generating and internally verifying 300 billion tokens of self-consistent thought, they would not just solve Navier–Stokes. They would likely restructure theoretical physics, solve the remaining Millennium problems, and synthesize entire branches of philosophy.

My research alongside Gemini 3.8 Flash/Gemini 3.1 Pro and fact-checked with Claude Sonnet 5, and other videos from specialists that I watched as context:

Quote

# The Buckmaster–Alpöge–OpenAI Navier-Stokes Dispute (Updated, Fact-Checked)

*Updated September 9, 2026, to correct and add context following independent verification.*

### The Mathematical Reality and Epistemic Divide

The core event is a priority dispute over finite-time singularity formation (blowup) in fluid dynamics. Building on the established program of Diego Córdoba and Luis Martínez-Zoroa, Tristan Buckmaster (NYU) and Levent Alpöge (Anthropic-affiliated) pushed a specific iterative cascade past the regularity barrier, achieving blowup with *smooth* external forcing.

There is a clear epistemic divide in what has been proven:

- **The Verified Results:** Buckmaster and Alpöge published proofs for the 3D incompressible Euler equations, the 2D Boussinesq system, and the 2D Incompressible Porous Medium (IPM) equation, on September 7–8, 2026. These are fully formalized and verified in Lean 4, with the code posted to a public repository. Terence Tao independently reviewed the work and called it a "remarkable achievement," saying he sees no fundamental obstacle to extending the approach to full Navier–Stokes.
- **OpenAI's Claim:** On September 8, 2026, OpenAI announced that an internal, unreleased model — run over 88 hours with roughly 10,000 coordinating agents launched September 1 — had produced a ~100-page proof of finite-time blowup for the **forced** 3D Navier–Stokes equations, satisfying Statements C and D of Charles Fefferman's official Clay Millennium Prize formulation. OpenAI has not released the proof for independent verification, and it has not publicly confirmed whether it was Lean-formalized. OpenAI has said it will not pursue the $1M prize, since the forced version doesn't meet the Clay criteria.
- **The Physical Caveat:** Both the Buckmaster/Alpöge results and OpenAI's claim rely on an external forcing function (f ≠ 0). This does not resolve the unforced (f = 0) physical behavior of autonomous fluids, which is what the actual Clay Millennium Prize requires and remains unsolved.

### The Institutional Conflict and Authorship Dispute

OpenAI has confirmed it launched its 10,000-agent compute sprint on September 1, 2026, after rumors circulated that Anthropic-affiliated researchers were close to resolving one or more Millennium Problems.

Buckmaster's own statement (posted alongside his preprints) describes contact from OpenAI beginning September 3, meetings on September 6 with an OpenAI mathematician and Sébastien Bubeck, and two proposals: either a joint announcement the next day, or Buckmaster alone writing up the Navier–Stokes result while crediting an internal OpenAI model. He writes that Bubeck twice raised removing Alpöge from authorship, citing Alpöge's employment at Anthropic, and that when Buckmaster said he would go public, he was told "Why would you ruin your career?" and, when he pushed back, "If you don't want me to be nice, then I don't have to be nice."

**Important correction:** OpenAI's Sébastien Bubeck has publicly denied these allegations, calling them "false and inflammatory." This is a live, contested dispute, not a settled account — and it should be reported as such. Buckmaster himself is notably careful in his own statement not to make firm accusations: he writes "I am not accusing anyone of anything. I am stating what I was told, when, and what was proposed to me," and he explicitly says he has not seen OpenAI's proof and does not know whether his and Alpöge's data was used.

### Hypothesis A: Telemetry-Assisted Preemption (Unconfirmed)

This hypothesis posits that OpenAI's rapid success may have been aided by exposure to Buckmaster and Alpöge's own work, since the pair used OpenAI's Codex extensively for months. Per Buckmaster's account, he asked directly whether the model had been trained on or had access to their Codex sessions; he was told the model did not look up user data, but on the training question, "I did not get an answer." OpenAI has not since clarified this publicly. This remains speculative — a real open question, not a confirmed mechanism.

### Hypothesis B: The Existence-Proof Catalyst

This hypothesis holds that OpenAI's result required no data exposure at all: the Córdoba–Martínez-Zoroa program was already public, and once rumors confirmed that smoothing the forcing term was achievable, a massive compute sprint could plausibly out-search human mathematicians on a now-known-tractable target. Nothing in current reporting favors Hypothesis A over B, or vice versa — both remain live possibilities pending release of OpenAI's proof.

### Additional Context Left Out of the Original Summary

- **A third, independent result exists:** a team led by Anima Anandkumar at Caltech released a separate zero-viscosity blowup result on September 7, using a physics-informed neural network rather than a general-purpose LLM — complicating any narrative that frames this purely as an OpenAI-vs-Anthropic race.
- **Clay Institute's response was measured, not triumphant:** its president, Martin Bridson, called it "an exciting day" while noting the prize criteria (unforced Navier–Stokes) remain unmet.
- **OpenAI has not released its proof publicly**, so independent audit of the claimed result is not yet possible.

### The Structural Fallout

Regardless of which hypothesis is true, this event is a live case study in how commercial AI labs can move at compute-driven speed on rumors of academic progress, and in the tension between traditional academic priority norms and corporate incentives. It also illustrates a shift in mathematical epistemology, where machine-verified proof (e.g., Lean) is increasingly treated as more load-bearing than institutional trust or informal claims, given that AI-assisted output can now be generated faster than it can be peer-reviewed.

***

### Sources

- [Tristan Buckmaster's original Mastodon announcement](https://mastodon.social/@tristanbuckmaster/117233413705701198)
- [Buckmaster's full statement (PDF)](https://cims.nyu.edu/~tristanb/statement.pdf)
- [Euler blowup preprint (PDF)](https://cims.nyu.edu/~tristanb/euler.pdf)
- [IPM blowup preprint (PDF)](https://cims.nyu.edu/~tristanb/ipm.pdf)
- [Boussinesq blowup preprint (PDF)](https://cims.nyu.edu/~tristanb/boussinesq.pdf)
- [Lean formalization repository (GitHub)](https://github.com/tristanbuckmaster/fluid_lean)
- [Terence Tao, "Finite time blowup with smooth forcing term..." — What's New](https://terrytao.wordpress.com/2026/09/07/finite-time-blowup-with-smooth-forcing-term-for-the-incompressible-porous-medium-boussinesq-and-incompressible-euler-equations/)
- [Unite.AI: "Buckmaster and Alpöge Post AI Fluid Blowup Proofs, Dispute OpenAI Contact"](https://www.unite.ai/buckmaster-and-alpoge-post-ai-fluid-blowup-proofs-dispute-openai-contact/)
- [Forkast: "OpenAI's 10,000-Agent Navier-Stokes Claim Solves the Wrong Problem — and the Right One Has a Provenance Controversy"](https://forkast.news/openais-10000-agent-navier-stokes-claim-solves-the-wrong-problem-and-the-right-one-has-a-provenance-controversy/)
- [Neoteric: "10,000 Agents in 88 Hours Prove a 90-Year-Old Navier-Stokes Problem"](https://www.neoteric.no/blog/openai-navier-stokes-10000-agents-88h)
- [Science.org: "How an AI math breakthrough ignited a controversy"](https://www.science.org/content/article/how-ai-math-breakthrough-ignited-controversy)
- [Silicon Reckoner: "If you don't want me to be nice, then I don't have to be nice"](https://siliconreckoner.substack.com/p/if-you-dont-want-me-to-be-nice-then)
- [Clay Mathematics Institute: Navier-Stokes Millennium Prize Problem](https://www.claymath.org/millennium/navier-stokes-equation/)

SOURCE OF THEIR OFFICIAL STATEMENTS:
https://mathstodon.xyz/@tristanbuckmaster@mastodon.social/117233413735526010

 

 

 

Edited 11 minutes ago by Lucasxp64

[AerisVahnEphelia]

18 hours ago, Leo Gura said:

@Schizophonia

Never ever relate to Germans. Peter Thiel, Trump, and Hitler are Germans. There is no coincidence: their culture creates dysfunction. The biggest mistake the French made was repeatedly trying to subjugate Germany instead of finding a lasting solution.

Hitler was not an inexplicable monster detached from German society, he was german culture and his fruits.
Trump is proving it again & again.

How many more people on earth will need to suffer ?

I wouldn't be surprised when AGI will find out that all the dysfunction in America are due to germans immigrants.
 

[[NO, I DON'T HAVE AN ALSATIAN BIAS]]

Edited 2 hours ago by AerisVahnEphelia

[Joshe]

Thanks @oOo