Truflation shows inflation is 1.36% today. Real-time inflation has been crashing since tariffs were announced. Tariffs are deflationary, not inflationary. The Fed should be cutting right now.

Super Dario x.com/dnak0v/status/…

It looks like o3 can reason across long context better than any other model, including 2.5 pro! I went over one of this benchmark’s example questions a few weeks ago and it seems to be testing for the real deal, not just basic recall.

Using o3-high as architect and gpt-4.1 as editor produced a new SOTA of 83% on the aider polyglot coding benchmark. It also substantially reduced costs, compared to o3-high alone. Use this powerful model combo like this: aider --model o3 --architect aider.chat/docs/leaderboa…

Might just be me, but it feels like we’ve completed erased profoundly autistic people from society. A diagnosis that used to imply a lifetime of tragic disability—severe verbal challenges, struggles with emotional regulation, inability to care for oneself—is now lumped in on a

And we have our first fully green row on MathArena - o4-mini-high completely solves AIME 2025 II, marking the benchmark officially saturated!

OK, that it cold regenerate this maze, not just see the path, is insanely impressive, they have ≈solved multimodality

Don't mean to be mean but I don't think saying what they did on this podcast helps their startup cause

I've been having one of the most productive workdays of the year with o3 and apparently OpenAI is NOT happy about this Sam Altman come get your boy

Amazing to see the community excitement on Codex CLI! Still early days: we're adding support for MCP, local/different provider models, and a native plugin system. In the immediate term, also fixing rate limit issues that some people have reported. Keep the feedback coming!

So excited gdb is leading this Gonna kickass

Nearly perfect context o3 is genuinely a next-generation model, only 2.5-Pro is competitive But again, I believe in Neo-GDM velocity

I see what you mean. This is normal.

Weekend Twitter Summary - o3 is literally AGI manna from heaven - o3 has yet to create code that compiles - Gemini though! - Anthropic who?

DeepMind's AI is beginning to generate knowledge that goes beyond the human horizon. With the new “Streams” system, it no longer learns from us – but from the world itself. No human data sets. No predefined categories. The machine observes, experiments, abstracts. This is more

Scaling Instruction-Tuned LLMs to Million-Token Contexts via Hierarchical Synthetic Data Generation "we introduce a novel post-training synthetic data generation strategy designed to efficiently extend the context window of LLMs while preserving their general task performance.

O3 model is extremely bad for coding and developing projects with more than 1k LOCs Insane levels of hallucination and very bad instruction following Where gemini 2.5 pro will 0-shot 1k + LOCs o3 will struggle and write a draft 300 LOC script that will crash.

Rich Sutton just published his most important essay on AI since The Bitter Lesson: "Welcome to the Era of Experience" Sutton and his advisee Silver argue that the “era of human data,” dominated by supervised pre‑training and RL‑from‑human‑feedback, has hit diminishing returns;

If you interview at meta with Yann and you're llmpilled does he blackball you Or do you just have to stay in the closet Asking for a friend

Ted Mabrey tough position for you with 5k followers to his 2M, and his dishonest style of debate is very popular right now as people succumb to partisanship. but if it means fewer people at your company with as much contempt for our nation as paul, it will be a good thing.