Today he used it for live translation between Turkish and Chinese in Switzerland. I did not say it could do that, he just thought it might. He is 61

Astra on @ChatGPT dot is absurdly inadequate. Not even close to Astra. It is slow, it thinks very narrowly and reading its output is pretty bad. Like, why would you not format it better. Did it have to be that quick to be on the devday?

AI models are still not generalized. During a login, if there is a button saying keep me signed in, you gotta assess checking it or not. It can find new enzymes but cannot think that. Last week is very frustrating with ai agents.

For the daybreak blue program from @OpenAI , you now need hardware from @Yubico I think @worldcoinfnd id integration is close if not imminent. @thsottiaux ? Included in the 28 day run?

I taught my father @bot . He is planning a trip to Europe. Today I showed the primary bot and when we set it up it directly found a duplicated payment for a train ride, cancelled one trip and helped him to message to cancel the other duplicated trip.

Ahahahahahahahahahaha Please go to dot.com

If machines do more of the work, why should staying afloat still take up so much of our only life? My first Substack piece—on money, work, and whether we should rethink the arrangement itself. howsureareyou.substack.com/p…

My bot is managing all the compute you have that are scattered around different cloud services and your own hardware. You can also rent more compute. Finally it helps you to rent out your own compute so that the others can use it and you can make money. It is doing this using kinesis network. x.ai/bot/mkpOkKydKDCfdVeMJge…

Things that are obsolete and should disappear: - SMS and phone numbers - Banks - Gasoline-powered cars - Fiat currency - Shopping malls - Centralized crypto exchanges

I created a Grok @bot (x.ai/bot/-GgufM3GkZclfn9PuI1…) to centralize the compute I have on @Kinesis_Network. It’s been pretty helpful to me. @poteto @tibor_tee @cursor_ai really cooked on the Grok bot.

Optimizing compute will be the differentiator

If you are a supergrok subscriber and dont wanna pay 260$ for grok bot, spin up a machine on kinesis, setup hermes agent and use your supergrok subscription. Decent substitute for 7/24 bot. It will be cheap and do the work

Coders who coded before gpt. Try Warp

This actually simplifies things more than it sounds

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!

I know people who would need grok subs because if x integration. I know people who would need chatgpt sub bc of the image gen and voice. I know people who can use kimi sub for coding and kimiclaw. I do not know anyone who would need claude subs over others.

Is it me or working with GPT 5.6 and Opus 5 is pretty awkward? They take initiatives that are not helpful at best most of the time. I would rather have grok 4.5 and kimi k3.

Introducing Fugu-Cyber: an update to our Fugu orchestration model. It achieves state-of-the-art performance on real-world security benchmarks, matching cyber-focused frontier models like GPT-5.5-Cyber and Mythos Preview. sakana.ai/fugu-cyber-release 🐡

Download the weights from HF right now