Mino (@compounded_mind) header
Mino (@compounded_mind) profile picture

AI news, research and bad takes. accelerationist. professional shitposter in training per aspera ad astra

Joined March 2024
Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)Photo by Mino (@compounded_mind)
??????????
JUST IN: Riemann Hypothesis has been solved. alpo.ge/remark.pdf
1
1
quick summary of this chart so you don't have to click: - Contra Labs puts their stealth model "Gestalt" at 1931 Elo, almost 200 points above human designers (1739) - Claude Fable 5.2 (1713) and Opus 5.5 (1691) sit right behind humans - Gemini 4 Argon (1221) somehow lands below Grok 4.7 take every "blind head-to-head" design leaderboard from a company launching its own model with a massive grain of salt
1
3
204
spent the last few years hearing scaling would hand us general intelligence for free and now an OpenAI researcher is leaving to build a $100B business hand-crafting the exact messy tasks scaling missed
Today is my last day at @OpenAI. I'm glad to have spent the last eight months of my life working here! I'm starting a new company focused on the production of high-quality reinforcement learning datasets: 1. The generalization ability of LLMs is clearly very poor, with "spiky" capabilities even in areas that have received tremendous amounts of investment and attention. For example, despite multiple years with tens (if not hundreds) of billions invested, even coding capabilities don't demonstrate "generality" -- even if every model can solve Codeforces questions or port C++ to Rust better than I can, I still have to manually "deslop" pull requests. 2. The vast majority of economically productive capabilities are not well represented in existing data offerings. First, there's a certain art to the design of an RL dataset which most vendors, not having upstreamed data into large training runs themselves, don't really understand. Second, and more importantly, most work is highly contextual and not easily encoded into a gradable environment; even if we can observe a "golden path" taken by a human which we believe to be good, it's challenging to understand whether alternate, counterfactual paths produce good or bad outcomes. The basic premise here is that I have a clear understanding of what labs need/want, having explicitly been on the other side and having been involved at every level from procurement all the way through training, and I'm able to provide it. I also believe that data needs will grow tremendously in the coming years, especially as frontier labs face increasing pressure toward profitability, and that they won't get the relevant capabilities "for free" through scaling alone; instead, they'll need to spend >$100B on precise, well-targeted data acquisition. Our first products will be focused on biology and statistical reasoning: 1. First, datasets that address long-horizon scientific reasoning, drawing on my work on GeneBench-Pro with @jeremyli__. Frontier models are still unable to reliably execute "messy" data analyses that require judgment, exploration, and adaptive revision (GB-Pro passrate on GPT-5.6 Sol scarcely exceeds 30%); to address this, we have the ability to generate thousands of high-quality problems with known ground truths which can be reliably graded. (In contrast, most existing RL data for bioinformatics is either massively over- or under-specified, and will probably break your model when you train on it.) Moving the "reliability gap" from 30% to >90% is obviously required for scientific acceleration, and -- despite my skepticism about generalization of RL -- is one of the *most promising datasets* conceivable when it comes to yielding generalization benefits for models' overall reasoning capabilities. 2. Second, datasets that address capabilities relevant to day-to-day workflows. Imagine a scientist snapping a picture of some experimental process or result -- say, a cell culture plate or a Western blot -- and asking Claude a question. Frontier models remain quite bad at these questions, especially those with multimodal components. But they're obviously required for acceleration of scientific discovery; before we can dream about automating science, we have to begin with shoring up these basic, generalist capabilities. Beyond these two, we hope to expand to adjacent fields (chemistry, materials science, etc.), and then even further into fields with more direct economic applicability like healthcare and white-collar office work. I strongly encourage labs with data needs to reach out. We offer industry-standard pricing and terms, and like I said -- I know how this process works, what good data looks like, and how to demonstrate to you, convincingly, that you'll be able to upstream our data into your training processes without issue. My DMs are open!
9
can't wait to hear how great the proactivity is from US Plus users for the next three months before the dot finally gets cleared by Brussels
Heads up, we’re starting to test dots with a small number of ChatGPT Plus users. Looking forward to the continued feedback as we expand and I’m recently seeing a lot more positives around how good the proactivity has been after many improvements. We’re also bringing dots to all ChatGPT Business Standard users. Rollout will start today.
1
2
76
six weeks left in 2026 and OpenAI is still teasing new drops while Plus users are bracing for shared limits in december
Replying to @theo
I still count 6 weeks in the year
1
169
GTA 5 made $10B+ while GTA 6 will make $2B+ in its first week…which probably makes this computer glitch in original GTA the most valuable computer glitch in software history
9
135
6
3,415
83,606
people keep trying to cast Jensen as either a traitor or a patriot when his functional role was basically running the modern equivalent of the British East India Company shipping digital opium into Canton if Chinese frontier labs are desperate for NVIDIA hardware while Beijing literally refuses to issue import permits, who do you think is winning that leverage game? banning American chip sales is quite literally executing the CCP’s five-year plan on their behalf
Anduril founder Palmer Luckey says a lot of people are missing the forest for the trees when it comes to the Nvidia saga. He called it the “midwit take,” and said they have it completely backwards. Luckey made the case that most people don’t understand how selling American GPUs into the Chinese market actually works AGAINST the CCP. LUCKEY: “What they don’t understand is China doesn’t want Nvidia to sell GPUs to Chinese companies.” “Like there are Chinese companies that want to buy Nvidia GPUs. There’s 600,000 H-200 GPUs sitting in a port right now, because China will not give them an import permit...” “Because they want to build an internal domestic processor industry, silicon industry, not build one that is dependent on the United States.” “And so the point that I’ve made is China is thrilled to have Jensen, who is somebody who was pro-selling into [China].” “And they’ll say, but Jensen was helping smuggle into China.” “Yes! Against the Chinese government’s will.” “Like the bizarre thing that people don’t get is Jensen was basically on the side of Chinese AI companies, but NOT the Chinese government.”
1
45
insane levels of aura
I want to thanks @SemiAnalysis_ for the motivation. In the Gemini RL scaling team, we printed & put that article on our desks to be reminded daily of what used to be true at the time. But the times they are changing :)
2
146
insane, this thing is only going up from here $ANT
Up 35% as soon as they started trading.
2
98
honestly their branding team carries a level of aesthetic direction nobody else in AI has
1
3
258
and there's also this about deep learning: “we recently saw sudden and rapid breakthroughs in mathematics produced by frontier models [...] we are monitoring for similarly rapid breakthroughs in deep learning” a couple of other interesting details from the report: - a GPT-6.2 Astra pretraining researcher credited Codex with spotting core insights while they worked through an idea together - documented human workweeks devoted to Safety and Alignment increased by 77% between January and September, the largest absolute increase among research areas
the new OpenAI research acceleration numbers are wild - 46% autonomous success on 2–4 hour tasks, up from 23% in March - 35% on 4–8 hour tasks, up from 13% - $2,400/day in AI usage for the median researcher - $23,000/day at the 90th percentile the dollar figures are valued at API prices, and Codex is apparently already helping researchers develop core ideas for frontier models
2
56
Bridgewater says it's managing money based on a one-in-three chance of an economically destabilizing cyberattack within the next year for anyone unfamiliar with Bridgewater, it is one of the world's largest hedge funds, founded by Ray Dalio its latest research highlights several concerning developments: - AI-powered hacking capabilities are spreading exponentially, including through open-source models - attacks are accelerating despite efforts to secure software and critical systems - decades of insecure code, spread across countless owners, make fixing vulnerabilities before they're exploited incredibly difficult - using AI to find and patch vulnerabilities introduces its own containment and alignment risks the researchers warn that current defensive efforts may not be enough to prevent a continued surge in damaging attacks
1
2
47
the new OpenAI research acceleration numbers are wild - 46% autonomous success on 2–4 hour tasks, up from 23% in March - 35% on 4–8 hour tasks, up from 13% - $2,400/day in AI usage for the median researcher - $23,000/day at the 90th percentile the dollar figures are valued at API prices, and Codex is apparently already helping researchers develop core ideas for frontier models
We're sharing new data on how AI is changing research at OpenAI. We find researchers' AI use continues to rise exponentially, and agents are completing longer tasks, but human involvement remains central in research. Researchers are also spending more time on safety and alignment, with human effort in these areas growing faster than in any other individual research area. This report aims to increase transparency around research acceleration and inform the governance of increasingly capable AI. We believe that AI labs should be required to track and publicly report indicators of recursive self-improvement (RSI): the process by which AI accelerates the development of future AI systems. Our hope is that this expanded report will help sustain and advance shared measurement norms across the industry. openai.com/index/research-ac…
2
3
180
Gemini 4 Neon is looking really good, outperforming Opus 5.5 and GPT-6.2 Sol on most of Google's published benchmarks it even beats Carbon on several tests, including RiemannBench (84% vs 80%) and Terminal-Bench Science (70.5% vs 64.8%). Google's own comparison also puts Neon slightly ahead on the AA Index (61 vs 60), with faster responses and lower standard API pricing
Introducing Gemini 4 Neon! Pushing the frontier across domains and punching above its model class in a meaningful way. To me, Neon feels like the right balance of intelligence, speed, and usability. Available now across @GoogleAIStudio, the API, and more!
2
63
Anthropic priced its IPO at $83 per share, above its original $70–$78 range, with trading starting tomorrow on Nasdaq under $ANT
We’ve priced our IPO at $83 per share, with trading expected to begin on Nasdaq tomorrow. anthropic.com/news/anthropic… Legal disclaimer: A registration statement relating to the Class A common stock was declared effective by the Securities and Exchange Commission on November 18, 2026. The offering is being made only by means of a prospectus. Copies of the final prospectus may be obtained, when available, from: Goldman Sachs & Co. LLC, Attention: Prospectus Department, 200 West Street, New York, NY 10282, by telephone at (866) 471-2526, or by email at [email protected]; or Morgan Stanley & Co. LLC, Attention: Prospectus Department, 180 Varick Street, 2nd Floor, New York, NY 10014, or by email at [email protected]
1
74
Neon is a great model : )
355
299
125
6,212
611,390
Gemini 4 Neon is now available in Antigravity, with four thinking levels: Low, Medium, High, and Extra it's listed above Carbon and described as being for deep reasoning and complex problems. not sure if it's actually better or worse than Carbon
1
2
182
also live on the gemini app
11