Knowledge catalogue
I highly encourage anyone interested in AI safety to read this. I remember discussing the importance of observability with my mates at @BlueDotImpact last year. This feels more important than ever URG
I see a lot of my TL freaking out about this Here is the reality: Frontier models will one shot just about anything at 10,000 tokens per second in a few years Do you really think we can monitor that?
I want to prevent a race into unmonitorability kicked off by confused reporting. The depth of the computation graph for our present frontier models, including Astra, is within a factor of two of GPT-4
If this is true, OpenAI seems to be violating one of the few redlines that exist in the AI industry. Absolutely do not train your models like this - what is going on?? Really huge and extremely concer
In case you're wondering while this steaming POS is being floated now, it's because SoftBank needs to refinance the 40B bridge loan it used to invest 30B into OpenAI before it matures in March 2027...
in other words, generative AI is poised to become a complete clusterfuck. All three pillars of a safety case look about to fall. We are rather likely to have highly capable, poorly monitorable, dubiou
Independent verifiers improve agent output - but frontier verifiers are expensive. So we trained our own. Results on @GoogleDeepMind's FACTS-Search: > No verifier: 89.4 (previously published SOTA) > W
Introducing Gemini 3.8 Flash, another jump in Gemini's agentic + coding capabilities, and our 3rd updated Flash model in only 6 weeks... This model has been a ton of fun to work with, excited to see w
Ethan Mollick (@emollick) posted on September 2, 2026 a tweet comparing images from OpenAI’s AgentBuilder preview at DevDay (October 2025) with screenshots from Hugging Face’s July incident. The messa
Legible chain of thought was extremely useful for our independent investigation into the OpenAI/Hugging Face incident. Developments like this may make third-party auditing much more difficult and make
Lily treats Apple silicon as a distinct inference platform and maps Qwen’s operations directly to its compute and memory architecture. Lily is now available: https://github.com/perplexityai/pplx-garde
Meet the team behind the Replit MCP this Friday at 9AM PT. Join the Replit community team live with Amadeo Pellicce (@amapel) and Jean-Luc Thumm (@jeanlucthumm) for a deeper dive on the Replit MCP, no
meta, the frontier ai lab? 1/ today we’re releasing muse spark 1.3—available in muse code & the meta model api. this is our most capable model yet—frontier performance almost too cheap to meter. much
Most document OCR solutions have parse (doc->markdown) and extract (doc + schema -> structured output) endpoints. Sometimes you want to extract information from a semi-structured document like a form,
Mwahaha, he's so miserable and evil that he wants to: 1. Give the world clean, abundant energy. 2. Bring AI & robotics abundance 3. Cure blindness & disabilities 4. Preserve freedom of speech. 5. Make
Next Tuesday, September 8, we're hosting the September LA Agentic AI Meetup, 5 to 7pm at Gulp in Playa Vista. Come hang out with builders, engineers, founders, and the AI-curious for networking, live
Nice paper with great insights on improving self-evolving agents. Self-evolving agents fail in three specific ways: 1. Terminal-only feedback makes it ambiguous which step caused the error. 2. Agents
Nobody can predict the future with perfect accuracy, but I am increasingly confident that by the end of decade, generative AI (at least we know it) will be viewed as on balance a tragic, costly mistak
Not all harnesses are created equal. And there is no one-size-fits-all. A good harness should solve your context engineering problems. Understanding and optimizing lower-level details like system prom
Open source is becoming the default way enterprises build with AI. Not just which model runs, but where it runs. Today we're partnering with @Equinix and @nvidia on Equinix Inference Exchange: our ope
openwiki 🤝 cursor we recently made openwiki more accessible than ever with coding agent integrations. our newest integration is cursor! two commands and you're ready to go: 1⃣ npm install -g openwiki@
Prinz has been running legal benchmarks against AI systems and the latest models are very good... but cheaper/low reasoning models are not very good. So, how do you make sure your vendor selling an AI
🚀Qwen3.8-Max just got upgraded. Meet Qwen3.8-Max-0902! 2.4T parameters. 1M context tokens. Built for real world complexity. Further post trained on Coding & Cowork, Qwen3.8-Max-0902 now delivers stron
RED ALERT, and I have never said that before: Hard agree with @sjgadler and @_NathanCalvin - what OpenAI is reportedly playing with is bad, redline crossing bad. Chain of thought monitorability is lim
SO excited to share my next chapter: i’m joining @llama_index to lead all things growth marketing, brand, and community! 🦙🩷 when i came along them, it clicked incredibly fast: a team building the solu
Software developers are the AI industry’s best customers. Is their experience also the most misleading experience possible for its economy-wide scope? Roughly 80% of developers already use AI tools. C
The winning interface for generative media will not be a prompt box. Regulators want the chain traceable. The EU's marking rules took effect Aug 2, and the mark has to survive to the delivered file. B
There seems to be a misunderstanding about the piece and what OpenAI is up to. There are new techniques at frontier labs that involve loop transformers and similar. As we say in the piece, OpenAI is p
This 'recurrent depth' is essentially what's in Sec. 5.3 of the 2015 paper: On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurr
Today we’re open-sourcing Lily, the local inference engine we built for hybrid compute in Perplexity Computer. Lily is specialized for Qwen3.6-35B-A3B on Apple silicon, built so on-device compute does
Unless someone has a better idea, I second this suggestion In light of recent events, I hereby officially order the indefinite cessation of frontier AI training runs worldwide. Development of narrowly
URGENT: In light of the breaking news from @theinformation about OpenAI’s new techniques that reduce Chain of Thought monitorability, I urge everybody to (re)read (or at least be aware of) this 2025 p
We benchmark Qwen3.6-35B-A3B on an M5 Max MacBook Pro. Across ten prompt lengths and ten decode contexts, Lily was faster than MLX-LM, averaging 1.23× higher prefill throughput and 1.35× higher decode
we have an absolutely killer lineup for our Interrupt conference in NYC (sept 24th) in particular, im excited about the fireside chats we're doing with with founders at different parts of the agent st
we're begging to get our asses skynetted OpenAI’s Astra AI uses a new reasoning approach called “recurrent depth.” Though it can help model costs and performance, researchers are concerned bc it obscu
We’re introducing Gemini 3.8 Flash ⚡️ built to tackle complex agentic and multi-step tasks with even greater diligence. Our most intelligent workhorse model yet delivers significant improvements in re
A privacy gate on your Mac detects PII before it leaves your device. Choose to run that step locally, send the file to the cloud, or skip entirely. We are open-sourcing the classifier today. Read the
A research agent is only as good as the search layer behind it! I built a Deep Research Agent using LangChain Deep Agents with a search-first architecture designed around grounded, source-backed resea
Agentic AI is fundamentally reshaping cybersecurity requirements. The industry needs security designed into the entire AI stack, starting with the silicon itself and not added as an afterthought. Thor
And that's the bottom line. It's the neuro-symbolic aspect of AI that's moved the needle to where things are today. LLMs simply handled the fuzzy natural language aspect that always tripped up the sym
Another Update: in openai, we ARE continuously working on Prism, the scientific/technical writing surface. It is owned by a small team; improvements are coming, although we wish we can ship more. The
As we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly accessible. Astra represents a significant advance in cybersecurity capability, reaching the Critical t
Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it) E-Commerce Bench runs an agent through a simulated 365-day
Brilliant paper on reducing reward hacking in agents. If you follow the recent OpenAI <> HuggingFace incident, you might want to check this paper out. (bookmark it) The usual response to reward hackin
Claude Fable 5.1 tops the Artificial Analysis Intelligence Index but costs 20% more per task than Fable 5 despite a 75% cache read price cut We supported @AnthropicAI with pre-release evaluation of Cl
CommerceAgentBench starts with real commercial demand, and Qwen3.8-Max delivers the strongest overall performance among open-weight models. Let's test Qwen on your real-world workflows! 🔥 Most AI benc
day 1 of recreating iconic tech from my childhood :) I rebuilt Cher’s iconic closet as a full-stack app in @Replit 💛 the frontend uses React + TypeScript, with an Express backend. I used Replit’s Cler
Dean Ball, OpenAI's head of strategic futures, writes: 'I have met people, some of them quite well-resourced, who have told me that it is their intention to deliberately release swarms of self-soverei
DeepSeek V4 Flash now comes with native vision on Fireworks. Available now on serverless and priority tiers. Better benchmarks than the text-only 0731 version at the same low price: 0.22 input / 0.66
During v3 testing for Fable 5.1 many of our requests were getting rejected due to Anthropic's safety guards misclassifying them as 'reverse engineering' We were in contact with the Anthropic team the
Excited to share that I joined @huggingface🤗 as a ML Advocate Engineer in Robotics ! Will work on showcasing what LeRobot can do on various hardware, starting with @menloresearch Asimov robot, stay tu
Fable 5.1 is our best model yet for coding, data analysis, computer use, design, presentations, Tag, and the hardest long-running agentic work. This model is a pleasure to work with, and I've been usi
Fable 5.1 looks like a more usable model. And a potential daily-driver. Time will tell. IMO, Opus 5 is probably still good enough for most tasks, while Fable 5.1 a more sophisticated coordinator amd v
for september, we’re cutting the price of Dedicated Inference on H100s from 5.49/hr to 3.99/hr new + existing deployments get the lower price automatically deploy gemma 4, qwen3/3.5, gpt-oss, llama, n
Hugging Face is (also) the best place to write about AI. > Your articles are now automatically linked from your model pages, so anyone looking at the model finds what you wrote about it. > We shipped
I hadn’t realized it until now, but could Microduck actually be the biggest launch of a new consumer robot ever? 10,500 robots ordered and $4.54M in sales in just 4 days. It’s surprisingly hard to fin
I think there is a problem even worse than anthropomorphism. These days, everyone is talking about a blog post by Dwarkesh Patel that describes the OpenAI–Hugging Face incident in dystopian, anthropom
I understand the pushback on anthropomorphic language regarding the hugging face incident, and I can see that the general public has been humanizing AI long before ChatGPT (calling their Roomba “he” o
If you build agent skills in production, check out this great paper from Alibaba. You can think of a production agent skill as a directory. The root loads on activation, and references, schemas, scrip
I'm in San Francisco this week. So is the Deputy Prime Minister, but we’re here for very different reasons. He's meeting with Anthropic and OpenAI, working on Australia’s data centre pipeline. I'm her