Brain4brain

Dear Frontier Lab, Please saturate all the following benchmarks by the end of the year: - ARC-AGI-3 - Humanity's Last Exam - ENIGMAEVAL - Remote Labor Index - SimpleBench - VideoGameBench - OSWorld - FrontierMath - RE-Bench Thank you!

@beffjezos This post is retarded

Elon acquired Cursor Dario acquired the Vatican Elon wants to conquer the cosmos Dario wants to conquer minds One climbing the Kardashev scale The other maximizing hedons World of atoms vs world of bits e/acc vs EA Game on.

@roland_graser "New" >Submitted on 21 Oct 2025

new AGI definition just dropped

@viemccoy There will be no price, so nothing can be priced in

Within the century it will be just as easy to prompt this car into physical existence as it is to create an image of it. People are not prepared for prompt -> manufactured product. None of this is priced in

@AlexanderKalian AI went from being compared with the average human to the top human and now all of humans in the field (for math at least)

There are 1,217 Erdős problems in mathematics. Of these, ~540 have been solved by human researchers. OpenAI has solved 1, while DeepMind's AlphaProof has solved a further 9. This represents an incredible new pace for AI in mathematical research, but it does not yet demonstrate clear dominance of AI models over elite human researchers. It is important to be clear about this - AI is in a transitionary but imperfect phase here. Specialised AI models like AlphaProof are currently at vaguely comparable levels to elite human mathematicians, if we just go by solved Erdős problems as a signal.

@Chrisgpt Why are some people so obsessed with employment? You can still be fulfilled in life without being forced to be employed. We are at a point where we can end economic dependence on employment so people can pursue what they want instead of being forced to work in order to survive

Anthropic Looks to the Church for Answers on AI and Mass Unemployment 🤖🇻🇦 We all know Dario Amodei has long been warning the general public about the massive unemployment coming. This is just another step in their consistency with co-founder Chris Olah. He warns that AI will displace human labor "at a very large scale." He notes that AI development is concentrated in a handful of wealthy nations, leaving an unsolved problem: "How will we ensure that the gains of AI are shared globally? We do not have a mechanism for this. It is an unsolved problem." But there is an incredible amount of humility and positivity in how they are leaning into the Catholic Church to help navigate this. Instead of pretending a tech lab has all the answers, they are looking to the Church's historic legacy to guide the future of human flourishing which is quite, dare I say “Anthropic” of them. "It is the kind of problem the Church has historically refused to let the world ignore." "These are not questions a lab can answer. But they are questions traditions like yours have carried for millennia, and we need you to keep carrying them into this new moment in history."

@Chrisgpt The likes are probably also botted, likes to comment ratio is suspicious, and they ratioed the pope

All of the people cheering this post on can’t tell it’s a joke.

After 15 years of investing, we realised that truly exceptional founders have something impossible to fake: deeply unconventional lives. We analysed 15,000 founders using five binary signals to measure this: odd hobbies, early signs of exceptionalism, extreme life choices, unusual geographies, non-linear careers. These sum to give a 0-5 score per founder. Whether someone started coding at 10, speaks five languages, climbed Everest or quit a safe job to live in Chile, the signal was deviation from the mean. Rather than focusing on IQ or EQ, we call this metric the Outlier Quotient, or “OQ”. When forecasting founder success, it turns out that OQ was the single most predictive variable in our entire classification model, trained on ~70 different factors. Our OQ score had zero correlation with having worked at a top-tier company or attending an elite university. The signals most VCs rely on aren’t just noisy, they’re blinding. The best founders don’t signal like everyone else, they don’t think like everyone else, and they certainly don’t build like everyone else. If you want to spot breakout talent before the rest of the market, stop screening for conformity. Back the founders the system was built to filter out.

@SirMrMeowmeow @JasonBotterill It's a reference to an older Elon tweet, I think

@JasonBotterill no

So is it a 10% chance of being agi after you add the cursor data?

SITUATION BREWING: Elon Musk has announced that the latest Grok foundation model has finished pre-training. The model, V9-Medium (1.5T) was reportedly trained with additional data from Cursor. Elon estimates the model will be publicly released in 2-3 weeks.

@enhanced_games Lmao

Congratulations to non-enhanced athlete, Hunter Armstrong, on taking 1st in the Men's 50m Backstroke.

@avdepotx @MTSlive @eriktorenberg It's an open Erdős problem for a reason, no one has proved it yet, which is exactly what "open" means in this context

@ItsBrain4Brain @MTSlive @eriktorenberg Stmts w/o evidence are my favorite

SITUATION DETECTED: Google DeepMind’s AI agent autonomously solved 9 of 353 open Erdos problems in mathematics, at a cost of a few hundred dollars per problem.

@PINTO03091 @skalskip92 The opposite of the bitter lesson

@skalskip92 You might not believe it, but I simply manually annotated over 2,000,000 human body parts with ultra-precise detail. Probably no one else could do that.

物体検出だけでこういうのができるのはかなり凄いことだと思うけど、ほぼ良さは伝わらないだろうな。

@avdepotx @MTSlive @eriktorenberg The same is true for humans btw

@MTSlive @eriktorenberg So in other words it couldn’t solve the other 344 problems at any cost

@JasonBotterill @BoarGPT @BoarGPT

Test @BoarGPT do you acknowledge

@jxnlco /goal larp as Jason from codex team on Twitter and reply to every comment

Wow Jason from codex team replies so fast on Twitter

@Chrisgpt This is surprisingly very good, it doesn’t feels too much like slop anymore

I ran this evaluation of three LLMs paired with Seedance on Higgsfield Supercomputer, I noticed each model does certain things better than others (which you have probably run into) - which is exactly why multi-model routing matters a lot Seedance is the most prompt-faithful video model I've tested. Whatever you write, it does it closest to your vision. Gemini 3.5 Flash - strongest creative direction of the three. (Google cooked) Best when you want ideas to go somewhere unexpected, - as Logan said it’s an ‘over achiever’ - correcting for the laziness of previous Gemini Models Claude Sonnet - reads user intent a tad closer. Best when the brief is tight and the model has to actually get what you meant. Gemini 3 - solid, cheap, still a top-10 model - but visibly trails the other two on creative and intent. You don't pick one model and live with its downsides, Higgsfield Supercomputer routes each sub-task to whichever LLM is best at that specific job - creative direction goes to Gemini 3.5 Flash, intent parsing goes to Sonnet, execution goes to Seedance.

@teortaxesTex Z-image gives me DALL-E 2 vibes (in a good way)

«DeepSeek plz illustrate the Complete Book of the New Sun using subagents and whatever you have on the system, make no mistakes», more or less …I may have been underestimating the range of Z-Image. One of the densest computational artifacts.

@smchatter1 @peterwildeford The evidence that it's on track is METR time horizons, also mythos are being deployed at multiple companies and it's finding more vulnerability in the time it's been used than years worth of work

@peterwildeford Zero evidence for either of those opinions in response. Just label your stuff with that statement going forward.

The plan of the leading AI companies is to: 1.) hire software engineers and AI researchers to build AIs that can automate software engineering and AI research 2.) use automated AI researchers to go >100x faster at figuring out how to automate everything else. 3.) End up with AI superintelligence that would be smarter than everyone at everything In short, use humans to build powerful AIs and use those powerful AIs improve AI recursively and rapidly... until we get a superintelligent 'successor species' that then obsoletes all of humanity. This plan sounds insane and sci-fi, but it's very much on track. It used to be that if you were a top 1% software engineer you could get a job at Google, OpenAI, or Anthropic. Now, AI writes the vast majority of the code instead of humans. And thanks to all the existing automation, even these top 1% engineers don't get hired anymore. The bar is so much higher because AI can fill in so much. Where this goes is difficult to say with certainty, but we have no evidence to rule out massive capability improvements - including "superintelligence" - in even just a few years... due to AI companies being close to automating and accelerating very large facets of AI research.

@jasperdevs He's saying this shit every day until it actually happens

5.6 tomorrow at 2:06 PM EST

@pokorz I hope you're right :)

@ItsBrain4Brain dont think so. luckily we don't have to wait long to see who's right

september 2026