simdpl

ML Eng.

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Gabriel Lattanzio @GabLattanzio]

A ce moment du forum, partie du public commençait à dire que les sept milliards pour l'Ukraine étaient un scandale, et Jordan Bardella disait vaguement soutenir l'Ukraine mais faisait le détail de tous les façons par lesquelles ils ne les aurait pas soutenus.

L'occasion de répondre qu'il y a des soutiens qui font penser à de l'hostilité, que la guerre des douanes précède la guerre tout court, que l'hostilité de la Russie pour la France n'était pas à prouver, et qu'être français, ce n'est pas détourner le regard face à un envahisseur.

[Kimi.ai @Kimi_Moonshot]

🚀 Hello, Kimi K2 Thinking!
The Open-Source Thinking Agent Model is here.

🔹 SOTA on HLE (44.9%) and BrowseComp (60.2%)
🔹 Executes up to 200 – 300 sequential tool calls without human interference
🔹 Excels in reasoning, agentic search, and coding
🔹 256K context window

Built as a thinking agent, K2 Thinking marks our latest efforts in test-time scaling — scaling both thinking tokens and tool-calling turns.

K2 Thinking is now live on https://t.co/YutVbwktG0 in chat mode, with full agentic mode coming soon. It is also accessible via API.

🔌 API is live: https://t.co/EOZkbOwCN4
🔗 Tech blog: https://t.co/n7xxaszqzF
🔗 Weights & code: https://t.co/4ukcXB0iP6

[Andrej Karpathy @karpathy]

I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter.

The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language person) is whether pixels are better inputs to LLMs than text. Whether text tokens are wasteful and just terrible, at the input.

Maybe it makes more sense that all inputs to LLMs should only ever be images. Even if you happen to have pure text input, maybe you'd prefer to render it and then feed that in:
- more information compression (see paper) => shorter context windows, more efficiency
- significantly more general information stream => not just text, but e.g. bold text, colored text, arbitrary images.
- input can now be processed with bidirectional attention easily and as default, not autoregressive attention - a lot more powerful.
- delete the tokenizer (at the input)!! I already ranted about how much I dislike the tokenizer. Tokenizers are ugly, separate, not end-to-end stage. It "imports" all the ugliness of Unicode, byte encodings, it inherits a lot of historical baggage, security/jailbreak risk (e.g. continuation bytes). It makes two characters that look identical to the eye look as two completely different tokens internally in the network. A smiling emoji looks like a weird token, not an... actual smiling face, pixels and all, and all the transfer learning that brings along. The tokenizer must go.

OCR is just one of many useful vision -> text tasks. And text -> text tasks can be made to be vision ->text tasks. Not vice versa.

So many the User message is images, but the decoder (the Assistant response) remains text. It's a lot less obvious how to output pixels realistically... or if you'd want to.

Now I have to also fight the urge to side quest an image-input-only version of nanochat...

[Jackson Atkins @JacksonAtkinsX]

My brain broke when I read this paper.

A tiny 7 Million parameter model just beat DeepSeek-R1, Gemini 2.5 pro, and o3-mini at reasoning on both ARG-AGI 1 and ARC-AGI 2.

It's called Tiny Recursive Model (TRM) from Samsung.

How can a model 10,000x smaller be smarter?

Here's how it works:

1. Draft an Initial Answer: Unlike an LLM that writes word-by-word, TRM first generates a quick, complete "draft" of the solution. Think of this as its first rough guess.

2. Create a "Scratchpad": It then creates a separate space for its internal thoughts, a latent reasoning "scratchpad." This is where the real magic happens.

3. Intensely Self-Critique: The model enters an intense inner loop. It compares its draft answer to the original problem and refines its reasoning on the scratchpad over and over (6 times in a row), asking itself, "Does my logic hold up? Where are the errors?"

4. Revise the Answer: After this focused "thinking," it uses the improved logic from its scratchpad to create a brand new, much better draft of the final answer.

5. Repeat until Confident: The entire process, draft, think, revise, is repeated up to 16 times. Each cycle pushes the model closer to a correct, logically sound solution.

Why this matters:

Business Leaders: This is what algorithmic advantage looks like. While competitors are paying massive inference costs for brute-force scale, a smarter, more efficient model can deliver superior performance for a tiny fraction of the cost.

Researchers: This is a major validation for neuro-symbolic ideas. The model's ability to recursively "think" before "acting" demonstrates that architecture, not just scale, can be a primary driver of reasoning ability.

Practitioners: SOTA reasoning is no longer gated behind billion-dollar GPU clusters. This paper provides a highly efficient, parameter-light blueprint for building specialized reasoners that can run anywhere.

This isn't just scaling down; it's a completely different, more deliberate way of solving problems.

[Keith Sakata, MD @KeithSakata]

I’m a psychiatrist.

In 2025, I’ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I’m seeing the same pattern.

Here’s what “AI psychosis” looks like, and why it’s spreading fast: 🧵

[OpenAI @OpenAI]

Want to see our open models in action?

Watch how gpt-oss builds a video game—using tools step-by-step within chain-of-thought reasoning 👾🍓

[Ethan Hays @ethanhays]

Cloudflare CEO @eastdakota is having the most honest conversations I've come across about the current & future of content creation

"6 months ago, 75% of queries to Google get answered on Google. Which means if you're an original content creator, your content is getting summarized & sold (they still put ads there), but you don't get that traffic.

And that's the good news. It used to be that for every 2 pages G scraped, you would expect 1 visitor. 6 months ago that deteriorated to 6 pages scraped to get 1 visitor.

Today the traffic ratio is: for every 18 pages Google scrapes, you get 1 visitor. What changed? AI Overviews

If the business model of the web has been search, fundamentally, for the last 35 years. You get value by subscriptions, ads, or fame. All 3 of those things are going away, and they are going away fast.

And that's STILL the good news. What's the ration for OpenAI? 6 months ago it was 250:1. Today it's 1,500:1. What's changed? People trust the AI more, so they're not reading original content.

People aren't following the footnotes. So if you believe the business model of original content creation is driving people to that content... I just have a really bad story for you.

The future of the web is going to be people reading the summaries of content, not the original. What I'm worried about is - if you can't sell subscriptions or monetize ads or get the ego boost from people reading your stuff, why anyone is going to create content?"

[Matt Dancho (Business Science) @mdancho84]

🚨 BREAKING: Microsoft launches a free Python library that converts ANY document to Markdown

Introducing Markitdown. Let me explain. 🧵

[undefined behavior @undebeha]

i've been trying to understand the transformer architecture for ages but it never clicked. so recently i started trying to learn its history and that has helped a LOT. to try to cement what i learned imma write up a essay.

Undefined Behavior presents: The Evolution of Attention

[Mistral AI @MistralAI]

Introducing Mistral Small 3.1.
Multimodal, Apache 2.0, outperforms Gemma 3 and GPT 4o-mini.
https://t.co/BHLAAaKZ9w

[Hugh Knight V @Hughbear54]

@natemcgrady It's only there when you're observing the towel. This is what it looks like when no one is looking

[Pavel Durov @durov]

China’s success in AI stems from its Soviet-style education system, which fosters strong competition — unlike Western schools that hide grades to protect feelings. If the US doesn’t reform its education system, it risks ceding tech leadership to China. https://t.co/VmO6hRvkJX

[DeepSeek @deepseek_ai]

📜 License Update!

🔄 DeepSeek-R1 is now MIT licensed for clear open access
🔓 Open for the community to leverage model weights & outputs
🛠️ API outputs can now be used for fine-tuning & distillation

🐋 3/n

[Jack Morris @jxmnop]

watched the Attention Is All You Need re-release on IMAX 70mm.. so epic

[Hypatie @Hypatiealex]

Les profs alertent sur leurs conditions de travail, leur traitement, leur salaire inférieur à éch. égal aux autre cat A. de la FP
Les gens ➝ gneu change de métier si t'es pas content, forme toi et pars !
les profs : ok ⤵️
https://t.co/sYfh3ZH1OT

[Massimo @Rainmaker1973]

Dutch journalist demonstrates real-time Al facial recognition glasses.
https://t.co/z32bs4ozjr

[Mistral AI @MistralAI]

We also released Pixtral Large, a new SOTA vision model.

https://t.co/cE7Drzasvv