Csaba Biro

@MatthewBerman Complete guide to making a data center on Mars.

Important. I’m going to make a video about data centers, what do you want to know?

@emollick GPT-5.6 SOL Pro Max reasoning in API:

Kimi K3 cannot write a good murder mystery (though neither can any other model). That remains the jaggedest of frontiers. They both make things too obvious (the letter) and too obscure, and cannot foreshadow to save their artificial lives.

Introducing Kimi K3: Open Frontier Intelligence 🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows Kimi K3 is now live on on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.com/blog/kimi-k3

Imagine a web design team at your beck and call. Now imagine that team already knows you from hundreds of working sessions over the years. Your career, your voice, your risk tolerance, what you refuse to sound like. A human agency bills weeks for discovery. This team finished discovery before the project started. That is why the site sounds like me and not like a template. The model that knows you builds differently than the model that just met you.

🔥Today, we are releasing one of the first visual reasoning benchmarks for autonomous AI diagnosis in healthcare! 🚀Introducing Radiology’s Last Exam 2.0 (RadLE 2.0) from @CRASHLabAI, an uncertainty-aware benchmark for autonomous diagnosis in radiology! ✅In the last few days, the AI frontier has moved significantly. @OpenAI launched GPT-5.6 Sol. @Meta launched Muse Spark 1.1. @xAI dropped Grok 4.5. 🙌We’ve benchmarked all frontier, open-source and medical VLMs in RadLE2.0 and the leaderboard is now LIVE! 🚨 Before AI models are handed autonomy, one question matters more than any accuracy score: Do they know when to STOP and hand over to a human? ⚠️ A confident wrong diagnosis is far more dangerous than an honest “I don’t know.” Yet most models are bad at admitting the latter! 🚀 We release five RadLE 2.0 Scores: Confidence Weighted, Reliability, Accuracy, Safety and Handover Readiness and we find that models from @OpenAI @AnthropicAI @MetaAI @GoogleDeepMind @xAI @nvidia @Alibaba_Qwen @MistralAI @MiniMax_AI all score very differently as they optimize for different metrics! 🚨But most importantly, NONE of the Models have been able to reach the average human expert baseline! ⚡️A thread on what we found and which models aced our metrics! Link to the leaderboard and technical report at the end of the thread!

Yesterday, we made GPT-5.6 Sol Ultra generally available. Today, we're sharing that it produced a proof of the 50-year-old Cycle Double Cover Conjecture using 64 subagents in just under one hour. We're sharing the prompt and proof below. We're excited to see what you all do with Ultra!

@beninthehouse @thekaransinghal Yes, welcome to the Immortals Club

@thekaransinghal Does this mean I can cancel my health insurance?

♥️ GPT-5.6 is a major step forward for health, both at the frontier and at cost. These models push the frontier of performance per dollar, bringing the best health intelligence to all. The smallest variant, GPT-5.6 Luna, evaluated at the lowest reasoning effort, outperforms GPT-5.5 at the highest reasoning effort–despite costing 25x less. The largest variant, GPT-5.6 Sol, sets a new high bar at cost. Another especially cool result: physicians found fewer flaws in GPT-5.6 responses than physician-written responses. We collected diverse tasks that remain difficult for recent OpenAI models, across patient-facing and clinician-facing use cases. We asked speciality-matched physicians to write responses to these tasks with unlimited time and web access. We then asked other physicians to compare responses side-by-side, blinded to their source. Physicians were asked to comment on areas of improvement across five axes: accuracy, communication, completeness, instruction following, and health decision helpfulness. We then reported the fraction of responses across sources rated perfectly across all axes, across 20,000 total axis ratings. GPT-5.6 Sol appeared strongest, although all GPT-5.6 models performed significantly better than physicians.

BREAKING: ANTHROPIC HAS JUST RELEASED CONTENT ON HOW TO BUILD A COMPANY WITHOUT EMPLOYEES - CEO: 1 person. - employees: claude agents it lasts 30 minutes and it's free Bookmark this so you don't miss it.

@skdh There is no such thing on earth like an empty emergency unit in a hospital which is air-conditioned, I assume you must have dreamed it or was it on a different planet?

in other news, I got to spend some time in the medical emergency unit in the Heidelberg hospital yesterday and I can report it was almost empty and air conditioned and they have a cafeteria too, so maybe next heatwave we're going to have another medical emergency* *it's a joke

@AndrewYang Why do you want to take away my only joy?

Try not to hate anyone.

@chrisgpt AGI is expected in this millennium.

Grok 5 will be here by December 2025 guys January at the latest

@apples_jimmy Correct, and ICE agents will execute them on the spot if found using it.

“ and to all foreign persons within the country “ ?? Do ants own staff not get to work on this / not use it personally now lol

@airesearchtools You can safely release any new models if you are on good terms with the president.

The truth is, I was planning to subscribe to Claude this weekend with the sole purpose of trying out Fable 5. I didn’t do it sooner because I was waiting for OpenAI to release their new model this past Thursday. We’ve missed out on Fable 5. But I wonder who would be bold enough to release a model of that caliber now without having tested it extensively, given the risk that the same thing could happen to them.

@sebgehr I have resolved more than 1000 medical cases using AI, probably tested more than anyone alive for medical purposes. If you don't have API access to GPT-5.5-Pro or GPT-5.5-XHigh feel free to DM me and I would run them for you, currently these are much better models than Fable 5.

I tried using Fable to help wrestle a severe medical issue of a close relative that has been stumping doctors for close to two years. The model auto-degrades and the problem remains unsolved. This safety washing is disgusting. I thought better of anthropic...

🚨BREAKING: ChatGPT can now teach you any skill in 30 days for free. Use these prompts to replace $20,000 courses and 4-year degrees:

@emollick Do you still have humans commenting on your posts? :) Robots raise your hand.

Law professors wrote questions they were asked during office hours. Gemini 2.5 & humans answered them then other law professors blindly judged the results: -Gemini had a 75% win rate vs. professors -Gemini's answers were rated LESS harmful than humans -Newer models do even better