taj

taj

[kache @yacineMTB]

behold. THE WORLDS FIRST SIX PENDULUM CARTPOLE SOLVE. Including a sponsor!

To solve this task, I built an environment to train an AI. This is what mechanize does, but for larger AIs. Apply! Salaries are up on their page

Thank you to mechanize for sponsoring!

[Andrei Bursuc @CVPR @abursuc]

State of things in 2026 we have even more language: “computer vision by asking nicely” #cvpr2026

[taj @expygmalion]

coding tools will become so minimalist that a user will never know what an ide is again

devs are getting cooked

[taj @expygmalion]

@adilmania Coin

[taj @expygmalion]

@adilmania Overindexed on the wrong side of the doin too.

[taj @expygmalion]

@GaryMarcus @elonmusk No, of course corse not! The startup is premature, you see, it didn’t exactly suckle from the teats of his genius and it probably failed to realize his vision.

[thebes @voooooogel]

opus 4.8 offers some structural pushback

[taj @expygmalion]

@pors @TrueAIHound

[taj @expygmalion]

Accelerate.

[taj @expygmalion]

@teortaxesTex The geodesic dome inspired some unrelated designs in power tech the day i was taught of its existence. Sounds counterintuitive but the architect is the final-end user of ai.

[taj @expygmalion]

@AISafetyMemes though they lack the qualifying backgrounds, you can actually blur together the abstraction layers of these two superficially conflated concepts to make the argument work in their favor

skill issue

[taj @expygmalion]

mchewww

[Y Combinator @ycombinator]

.@ApolloAtomics builds the most compact nuclear reactors with the highest uptime and a deployment time of less than 24 months.

Apollo took the pressurized water reactor technology that already powers 80% of the world’s nuclear plants and flipped one part, the steam generator, to make the plant an order of magnitude smaller without compromising power.

Congrats on the launch, @AssilHalimi & Drew!

https://t.co/5lGDpZhmQ5

[alphaXiv @askalphaxiv]

New paper from Yann LeCun!

"When Does LeJEPA Learn a World Model?"

This paper proves that under Gaussian latent dynamics, LeJEPA can recover the hidden state behind nonlinear observations up to rotation.

The intuition is that linear latent features are the most stable across nearby views, while nonlinear features decay faster, so the objective naturally selects the real world variables.

The key caveat is that this guarantee holds under specific assumptions, and Gaussian latents are the unique case that guarantees this.

[taj @expygmalion]

yep
i picture human progress as nothing short of a dream

[taj @expygmalion]

the thing i hate about this is that one always acquires the language of importance precisely when the moment to speak up has expired

[Valerio Capraro @ValerioCapraro]

The Pope is making exactly our point. LLMs “may imitate or even simulate, but they do not understand.”

This is the core epistemic fault line.

Most AI evaluation is still based on one assumption: if a system statistically approximates human behaviour, then it is close to human intelligence.

But approximation is not intelligence.
Simulation is not understanding.

LLMs can produce the right answer without knowing why it is right. They can simulate empathy without feeling. They can imitate judgment without responsibility. They can generate coherent explanations without having a world to which those explanations are accountable.

Stop confusing behavioural similarity with cognitive equivalence.

Human understanding is embodied, affective, relational, motivational, and normative. It is not just the production of plausible text.

*
Full paper in the first reply

[taj @expygmalion]

@MikushRab this kind of task needs a specialized neurosymbolic ai to give you decent results, you can build it today with claude itself if you know what to do exactly

[Michael Rabinovich @MikushRab]

Opus 4.8 just dropped and I ran it through our CAD tasks. 4.6 → 4.7 → 4.8 side by side.

The results are unexpected!

[taj @expygmalion]

@YiMaTweets yes and that's just the normative question, the eschatology of science forces the reinterpretation of the past 5-6 centuries (roughly the Scientific Revolution to today?) of scientific progress as strictly opinionated

[taj @expygmalion]

i just love how oai force-deploys a new tool/feat every time the competition is about to drop something not so dissimilar

the only thing that is both figuratively and literally "open source" is their optimization function

[OpenAI @OpenAI]

Windows users, this one’s for you.

Computer use now works on Windows, so Codex can take action on your Windows computer.

And with Windows support for Codex in the ChatGPT mobile app, you can start, review, and steer tasks on the go while work continues on your Windows machine.

An early experience, but we’re working on more ways to keep your work moving, wherever you are.

[taj @expygmalion]

RAAS
robot as a service
horrible

[The Humanoid Hub @TheHumanoidHub]

EDEN-1

A bimanual bot you can get for effectively $10/hour.