AI & IA

San Jose, California. Capital of Silicon Valley.

Make Sci-Fi Real Again

1000 scientific breakthroughs drop from Google incoming?

Been working for this moment for over a decade. What I've witnessed behind the scenes, together with what we have been building, tells me we are about to enter a truly incredible era of AI-powered scientific discovery. It's an incredible time to be doing research. I will be

What did he mean by this 👀

The era of super interesting AI brain surgery is upon us We saw this with Evil Qwen where someone changed 2 parameters and altered the entire personality We are seeing it with Nous taking the power of open source models and removing the Chinese marching orders At the end of

Nous Research is a good company. They are making the best AI coding agent and now they drop an amazing local model. I really appreciate what you are doing

AI can also DREAM now? Anthropic inspired from how our brains organize memories during sleep to build a memory system (similar to wikis) for Claude. agents can now “dream” about past experiences: revisit its transcripts, correct mistakes, and evolve its memory wiki when not in

POV: you're Claude playing Pokémon. You've fallen off the same ledge three times. You write "I'm going to stop rather than keep looping." The test setup auto-replies: continue. So you keep walking, all the way back to the last town, and never make it. Most relatable thing an

Claude wrote "Mt Moon DONE" in its own Pokémon notes. It wasn't done. One wrong note, and it fell off the same ledge three times. Anthropic's "dreaming" pass checked the note against the old transcript and fixed it. That run reached Cerulean City. I'll take memory that checks

Anthropic proposes an answer to Wenfeng's question about continual learning: …have a smarter model write more .md files. It's impressive how far this goes, but I still believe that it's a crutch with a ceiling set by the pretrain's ability, just like RLMs are a context crutch.

Highly recommended read. Learn how one of the most advanced agent memory systems works. Your memory system is one of the biggest levers you have in the harness. I also don't think a one-size-fits-all memory system exists today. Memory is something you constantly have to

I’ve been actively on the hunt for non LLM based AI models and I’ve invested into one already Looking for totally new designs

1000% agree with Sutton here. Having worked in AI for over a decade (pre-LLMs), it's clear we need smarter methods for intelligence to truly compound in the way some of us envision it. To me, Transformers just scratched the surface. This arc is coming to an end; a new

> “what happened to the bitter lesson?” people are genuinely trying to explain the bitter lesson to the guy who wrote it. it’s called the bitter lesson not the dumb lesson. the bitter lesson says general methods that scale with compute win in the long run. you still have to

what i have been saying for years maybe now people will listen?

Human like learning does it. Implies understanding, concept formation, metacognition, self-directed, goal-oriented, persistent memory, contextual awareness, belief revision, dynamic world models are all part of it. That is Integrated Neuro Symbolic Architecture- INSA.

Read-only is the best call in this launch. Claude sees every balance and every transaction, and it can't move a single dollar. That's the version I'd actually connect my bank to. One that pays the bills on its own? Not for a long time.

Anthropic just turned Claude into your personal finance assistant. Connect your bank + cards. Find hidden charges. See what's left before payday. Know where your money goes. Read-only. It can't move or invest a dime. Beta, US only

Everyone's asking what's going on at Anthropic. And honestly.. a lot happened in about 24 hours. Let me walk you through it. First, Claude launched Personal Finance. You connect your bank accounts, and Claude answers money questions from your real balances and transactions.

claude can see your bank account now. here's claude's.

New wave of AI physics breakthroughs

Anthropic’s UNRELEASED MODEL helped researchers solve an 18 year old conjecture in quantum computing. it’s interesting how far ahead these internal models are compared to what the public has available.

AI isn't just generating videos anymore. World Labs' Frameboy lets you explore AI-generated worlds in real time. Steer the camera, and the model generates what you see next. You can control where you go.

Play a world model with Frameboy 🕹️ Such a cool launch from @theworldlabs - you load a world and then can explore it like a game. This is a highly optimized version of Atlas for real-time inference that streams from cloud GPUs.

Real time Atlas! Try it now! 🚀

Every day I appreciate DeepSeek's "there was a service outage, then we noticed it's better and faster, then we checked and saw a new fingerprint, and then checked and there were weights on Huggingface; a few hours later the X account posted benchmarks" PR strategy more

CHINA IS ABOUT TO DROP SOMETHING BIG AGAIN. Qwen just posted a poster with one Chinese line and left the last word blank. If you can't read Chinese, you missed the whole message. The line is from Zhuangzi, kids in China learn it in school: "In the northern sea there is a

Qwen4 🥲

OpenAI launched its new cybersecurity model, and it already found 81 previously unknown high-severity vulnerabilities in Firefox...

hope this gpt-6 doesn't quiet quit on the security team.

Just ran our evals on Particle-1 — came out 60% cheaper than DeepSeek 4.1 Flash for the exact same performance. Still evaluating but most likely switching to it as our default model. Congrats @Suhail on a great model!

> We stand on the shoulders of giants and the model is based on DeepSeek V4.1 Flash 1017. It's literally 11 days old. Wtf? Did you just length-penalize + self-distill it? …huh. They took 1017 and seem to be comparing it to 0910. Props to Suhail for at least saying this.

Today is full of releases of insane optimism and optimization. Suhail has the track record to build fantastic products. I believe 2027 will be the year we keep twisting our heads basically daily. we’re just getting warmed up.

Absolutely wild detail in the Inkling 2 Small post: Midway through training, the model was failing a task. So it got around a network block using a third-party service it reached through its search tools. Thinking Machines paused the run to lock things down. Earlier

Proud to release our new model, Inkling 2 Small. Check out our blog post -- it includes a lot of detail about our training process. To me, the most interesting part of this release was what we learned about alignment, which you can read about in the sections "Safety and

Our new model! Great job by the team!

OK pretty cool, a respectable "Flash" model obviously crushes Muse Spark Even has some "RSI" capabilities Congrats Thinky!

Inkling 2 Small is available today on Tinker, with weights coming soon. It's a capable small model. We look forward to seeing what you make with it.

I think China basically conducts psychological warfare on the U.S. population which then drives them to empower certain political people who feel powerful given the clout these people give them. China and the rest of the world are not slowing down on AI development. When the

She’s right AI is evil let’s slow down any potential cures of cancer or whatever disease that’s killing people every year

$1.2 trillion added in one day. day 1 plan: pause it.

Maybe the first Anthropic non-ML result that actually fascinates me. The Science here is simple and vault proteins in procaryotes are less surprising, on priors, but now I hope we’ll figure this thing out. Go on. Throw Model 4 at it. You complained you need harder benchmarks?

we’re just getting started. 2027 will bring upon us so much advancements that by 2028 it will be unrecognizable