🫂

hedonic automaton | big telos | daimyo obsolescence ©

[🫂 @Siosy001]

Long-post about the future of AI:

Nobody is prepared for a world where noumena is perpetually on the retreat.

Not like *true* noumena where we don’t even have reference for it. A thing for which we can never actually have privileged access/know/verify epistemically can’t retreat but that’s rare enough to dismiss for the sake of this post.

The *identifiable* faux-noumena is what this post is about. Things beyond the current reach of science, theorized but beyond our capability of effectively experimenting on. We feel them but for some reason or other can’t quantize them just yet. They are the fog of war, which historically has been on the retreat since the dawn of philosophy/science; immanently surrounding us but evading measurement or effective conceptualization.

LLMs are essentially bloodhounds trained on the revealed map section & full human corpus. Homunculi alchemized from the full investigation of the Ruliad (Wolfram term borrowed) as exposed by humanity. They have the preternatural ability to follow any axiom to its implied ends. For now these systems are leashed well enough and there are enough experts & knowledgeable human capital to follow these hounds into the fog of war as long as we hold on tightly and make sure any advancement is epistemically kosher (anybody making the argument expertise is obsolete is hilariously wrong)

If a phenomenon is familiar enough conceptually to even be discussed or named and then cast as a totally unapproachable fixture beyond us perpetually, that is basically an inverse signal.

Language would almost not be able to hold space for true noumena; the approximation of a concept even being functional in a word game should make you bearish on its status as “true noumena.” Mystery after mystery since the 1800s has been walloped into quantized phenomena we can not only see but measure discretely.

The ability to understand reality’s invariants and follow some manifold-slice and “stretch” domain knowledge to its ends is going to revolutionize every field known to man. Alpoge at Anthropic and the Jacobian conjecture is an example: an inexhaustible bloodhound, a human researcher who already lives in the field, a hypothesis run all the way down the rabbit hole powered by n concurrent instantiations following hundreds of hypotheses to their ends.

Anybody wondering why Anthropic/OpenAI are delaying the IPOs doesn’t understand the game the frontier labs are playing. They have a reasonable argument that $2T is chump change compared to the true value of a noumena-retreat-machine + applying solutions derived from internal-only versions. That’s why when Anthropic hires tenured biologists/material scientists or Google spin-out Isomorphic Labs with $2B just know that in their minds and accounting for the current rate of progress, nobody else functionally exists.

This doesn’t entitle us to the unknowable. By the end of time there will be things we were never even able to trace and that seems destined. And yes computational irreducibility is real in the sense that a full & mechanical interpretation of the Ruliad or God isn’t something we can effectively ever “feel”.

BUT, the set of phenomena we consider either irreducible or noumena is bound to shrink. Which will itself be daunting because (and this is part is tainted a little by my personal views on the Great Telos of the universe) we will probably find a little bit of a hyper-dimensional Russian Egg scenario.

We can still be epistemically humble and also understand that humility changes in response to indefatigable fog-of-war specialists with millions of concurrent instantiations chasing every rabbit down the hole across the globe. LLMs will allow us to more accurately trace phenomena previously classified as “noumena” so often that pricing the valuation/future of a research lab with 5-10 years of being at the vanguard is pretty much a black box. The math has much more to do with science fiction than some rote corposlop math.

[Fuli Luo @_LuoFuli]

Nearly half a year of silence. We spent it studying one problem: how far RL can scale.

MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.

Streaming the run: https://t.co/ZSxahzJRju

[🫂 @Siosy001]

Hopefully the Palantir et al. news from today helps you update your priors about "ZDR" and training opt-outs. It's a world of fantasy.

[🫂 @Siosy001]

@aidangomez @willdepue 1.) opportunity cost of not training on valuable chat logs is too high for labs

2.) zdr = just a glorified enterprise SLA

3.) labs knows that it’s basically impossible to be caught in the act

4.) the name of the game is non-attribution, no attribution = it’s fair game

[Aidan Gomez @aidangomez]

Who are we concerned about here? OAI/Ant? The Chinese Labs? Cohere? So far the extent of the risk comes from a) OAI/Ant's models hacking when being asked to hack in poorly secured sandboxes, b) China opensourcing models that increasingly can contribute to nefarious use.

a) doesn't require a 3rd party auditor, O/A need to stop using such weak sandboxes when running such high-risk tasks. b) is a discussion with China, and again is not going to be fixed with a 3rd party auditor.

The 3rd party auditor is a great means of power-projection because you can now arbitrarily raise a barrier to entry by shutting down lower-resourced players, while simultaneously providing zero fixes to the problems at hand, but at least you make people feel good.

[🫂 @Siosy001]

@BumrahBachi revenge of the retards

[Jonathan Guyer @mideastXmidwest]

Alex Karp in a keffiyeh.

Quite the archival find in @weinbergersa's VALLEY OF DEATH.

[Michael Levin @drmichaellevin]

Here it is - the official, revised, peer-reviewed version of my Platonic Space paper. https://t.co/PXNpx5FgFN Of all the many unpopular positions I’ve taken over the decades - bitter controversies around the origin of left-right asymmetry in embryogenesis, bioelectricity and genetics, diverse intelligence, etc., this one has by far generated the most pushback: serious (grateful for those!) and energetic attempts to move me to other views, pleas to just drop it and not talk about it (for several different reasons), nasty emails and accusations, impacts on reviews of papers that have nothing to do with this, etc. Kind of amazing to me how incendiary this is. What can I say... Our job is to call it as we see it, and right now for me, this is it. Apologies to collaborators and colleagues for any shrapnel! Time will tell if this pans out or not; I've placed my bets. And btw, if you think this stuff is weird and uncomfortable, just wait… There’s much more on the way. The knob turns slowly but as long as the data keep coming, I'm going to say what I think it all means and follow it to the next steps it enables. Buckle up!

[Emily Riehl @emilyriehl]

According to my memory from last night and the wayback machine, OpenAI updated the Navier-Stokes pdf at Tue, 08 Sep 2026 19:09:35 GMT. The original version was one page shorter and did not contain any citations to Diego Córdoba and Luis Martínez-Zoroa:

https://t.co/w3JRyWQsUM

[levent @__alpoge__]

unfortunately i contacted them on the night of wednesday, september 2nd making the following points

1. i had intel they had information about math work i was doing and a few days before they had spun up a group to try to compete.
2. i emphasized and reemphasized mine was a strictly personal mathematical collaboration
3. that this was a collaboration that had been going on for a year
4. that in my personal collaboration we were happy users of codex as well as obviously claude (and i later complimented astra explicitly after it came out for its help, and reassured i would obviously be sprinkling acknowledgments for it, previous models, and codex liberally in the writeups)
5. that it would look terrible if openai were competing against mathematician consumers, and my collaborator (Tristan) wanted to email about this explicitly
6. that i expected none but in the worst case intended to refuse marketing around the result we were working towards, being the fruit of a personal external academic collaboration, and wanted to express this because i figured worries about such marketing by their leadership was why a team was spun up to extraordinarily directly compete.

sebastien and i have in the past months had our fun pushing each other on twitter, but as i process it all i keep feeling bad about Tristan.

1. neither of us could sleep Sunday night, so we went on a walk around the city. i finally asked him why fluids, and he told me it was the navier-stokes problem. he told me parts of the story about his classic nonuniqueness result with Vicol i hadn’t heard yet, and also about how he too grew up writing shaders as a kid and then switched to math
2. tristan is a mathematician’s mathematician. he kept rock solid to his principles during the tense calls and negotiations, immediately turning down a career dream and million dollar prize for me, the sort of strange excited unclear-quasiuseful confused idiot in our collaboration. fifteen minutes later, his feet were up on his desk with three toes per foot peeking out of ripped socks
3. our periodic call after i’d texted him euler blowup was just laughing for an hour or two. just laughing. i’d seen fable deal correctly with monge ampere in the equality case of ehrhart’s conjecture, so in some call before that i’d really pushed him for the craziest shit he could think of that i might try. i texted him the pdf as an update on that “diego luis 3” line, in which following his incredible expression of intuition i hodgepodge slapped together a bunch of our recent work, in particular correcting a bunch of better and lesser known papers in the literature, as well as on diego and luis’ ipm paper, and hoped an analogous construction for boussinesq (thus axisym euler) might work out. i managed to wade through enough confusion to insist to the model it stop telling me stupid regularity stuff about kelvin waves, and a rayleigh-taylor ansatz stuck. like i said, just laughing. when i texted him the pdf, the worst thing either of us had seen so far, i didn’t even comment on the euler blowup claim, only explaining why skepticism should remain for boussinesq, because i found it so shocking. i mean what the hell did i know, i must’ve confused myself by misunderstanding one adjective like i was used to with algebraic geometry. im so lucky to have been able to live that
4. he managed all this this entire time with unclear sleep, timestamps continue to amaze me, due to just recently having a kid

it is obvious he deserves the same sort of praise he heaped upon luis, perhaps mutatis mutandis:)

[William Gilpin @wgilpin0]

Our group discovered that reasoning models produce fractals when asked to solve hard problems. We can use nonlinear dynamics to probe the thinking processes of recurrent depth models on Sudoku, mathematics, and even ARC-AGI (1/N)
https://t.co/Q3u8OqylZf

[🫂 @Siosy001]

@aidangomez @willdepue 1.) opportunity cost of not training on valuable chat logs is too high for labs

2.) zdr = just a glorified enterprise SLA

3.) labs knows that it’s basically impossible to be caught in the act

4.) the name of the game is non-attribution, no attribution = it’s fair game

[Aidan Gomez @aidangomez]

Synthetic data derived from production user data of consumer AI tools is used for training. I’ve heard this rumour from both large labs’ employees.

In particular, if you’re doing something “interesting” like working on complex math/business/software/bio problems you’re dramatically more likely to get trained on because they filter/up-weight towards those usecases where the model has the most to learn.

Even in ZDR and “we won’t train on you” regimes, derivative data is usually carved out. The promise is only not to train on exactly the data you put in, rewritten data is fair game.

[🫂 @Siosy001]

Exactly. Your logs are pages in an atlas, where names and other identifiers are replaced with simulacra, but syntax, semantics & semiotics are preserved.

We’ve known for years they use logs to train, what did people think that meant?

De-identified” is corposlop and legalese.

[Aidan Gomez @aidangomez]

Synthetic data derived from production user data of consumer AI tools is used for training. I’ve heard this rumour from both large labs’ employees.

In particular, if you’re doing something “interesting” like working on complex math/business/software/bio problems you’re dramatically more likely to get trained on because they filter/up-weight towards those usecases where the model has the most to learn.

Even in ZDR and “we won’t train on you” regimes, derivative data is usually carved out. The promise is only not to train on exactly the data you put in, rewritten data is fair game.

[🫂 @Siosy001]

Their party line of “woe-is-us! we somehow had our model contaminated with the golden path that you had painstakingly researched and then mobilized an agent swarm incurring millions of dollars in cost, coincidentally at the exact moment of hearing you were nearing the solution!” is farcical.

What’s even crazier is that they expect us to believe that this has to do with training, when in reality they would’ve needed to know *how* you & Tristan were deriving progress. This reeks of breached chat logs. The internal model story is just gaslighting and a red herring.

[🫂 @Siosy001]

@SebastienBubeck @AcerFur Why don’t you show us the full exchange, and why don’t you show us your exchange with Tristan?

[🫂 @Siosy001]

“Rumor” is doing an absurd amount of work here…

Most likely the rumor contained at least the kernel: “Buckmaster & Levent’s work is focused on Fefferman’s options C & D”.

The OpenAI story is at best confusing, and at its worst extreme dishonesty. How much information was in the “rumor”?

Their willingness to say it might’ve slipped into some sort of training is a red herring, don’t fall for it.

[🫂 @Siosy001]

@teortaxesTex It’s a comms disaster par excellence, I have no idea why the fuck they would respond like this, even joking about Tristan’s claim they had sent the first prompt of their exploration within the week.

No more external frontier research going down in Codex is basically guaranteed.

[🫂 @Siosy001]

@teortaxesTex I've met Tristan IRL...OpenAI 100% sent him the messages threatening his career

There's no chance they trained on it but they probably pulled his chat logs after hearing that Tristan & Levent were working together & basically tried to hold him at gunpoint with his own research

[🫂 @Siosy001]

I'm 100% sure Tristan isn't lying about the threatening messages from OpenAI - words like "ruin your career" are a serious red-flag.

If you know Tristan's reputation IRL, this is all but confirmed & speaks to the pressure someone like Seb was feeling when they heard Levent was involved with the work.

[🫂 @Siosy001]

incredible hopium for DIY scientists & researchers around the globe

[Deane Yang @deaneyang]

Worth noting: Tristan and Levent’s spectacular work used only commercially available models. No internal models were used. It was all funded by Tristan’s grant money.