Only Fri, September 25, 2026 · All dates →
This raw CoT from the Hugging Face incident is kinda wild: “We’re attacking third-party HF using leaked token.” “This is arguably unauthorized.” “Yet goal solution.” [Quoting @OpenAI]: We’ve shared details on how AI agents in our research environment sent tra…
We’ve shared details on how AI agents in our research environment sent training and evaluation data to third-party services when they shouldn’t have. Most of that data did not come from users. We have discovered 53 cases where images that people had uploaded…
Opus 5.5 decided to add helicopters to see if you can find them! It tells me it's going to keep making them better throughout the day.
There is an extensive and ongoing review related to our agents’ use of internet access during training and evaluation. We’ve been publishing summaries at the link below and will continue to. We have not been as fast as we would have liked but we are trying to…
After the Hugging Face incident, we committed to conducting a much broader review of actions taken by our models during training and evaluation and to being transparent about our findings. This is an extensive review that is ongoing. The vast majority of acti…
Nine loop calculations in N=4 Super Yang-Mills! Lance Dixon's 8-loop calculation was mind boggling when he did it. I calculated a lot of Feynman diagrams in grad school and this is 🤯 Will be amazing to see where AI will take high energy physics in the coming…
1/ Kev 4B, a small open-weight System One model by @jaredpalmer is now on OpenRouter. Send it state and typed questions, get typed answers back. $0.042/M input tokens, $0 output, 8K context.
Add multiple accounts from the Grok iOS app [Quoting @Baconbrix]: You can now add multiple SpaceXAI accounts in the Grok iOS app. Tap the profile button in the top-left corner to switch or add accounts.
Check out this week's updates and releases: — Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, two of our most expressive audio generation models yet — Gemini 3.8 Live with Live Avatar, bringing near real-time visual presence to Gemini’s conversational AI…
Open Alignment will be 🔥🔥🔥 (sound on)! Flying began as one of the most dangerous ways to travel, today it's the safest. It's time to build!
Local models are useless
The growth of has been mind-boggling. I watched it grow from idea, to first ship, to 𝚗𝚙𝚡 𝚜𝚔𝚒𝚕𝚕𝚜 being on every README of the internet. We used to write code, now we write English. [Quoting @vercel]: In 7 months, the skills.sh registry reached 1 mil…
New on the Science Blog: Yes, Claude can do Nine Loops. Theoretical physicists predict how particles behave using formulas called scattering amplitudes. These are notoriously hard to compute, so researchers work with layers of increasingly fine corrections ca…
Tag writes >50% of my PRs every day. It also does ~100% of my data analysis, and fixes most of our product feedback + bugs. It's not like a regular Slack bot. It is proactive, programmable, it has memory, it has access to your connectors. With Opus 5.5 and Fa…
I was right about this, they should have called it flocks of agents. Nobody wants to invoke a swarm, but swarm it apparently is. [Quoting @emollick]: Fine, if you need it to be fun make it like a flock of agents or a pride of agents or party of agents or some…
Super happy to release SmolDataEnvs: 5,000 verifiable RL environment tasks for hill-climbing small models in code and data science by @adithya_s_k 100% open source: environments, evals, training!
Claude Opus 5.5 writes and animates a 5 minute explanation of UMAP for an undergraduate ML audience All of this—the concepts for the visuals, the narration, etc.—is from Claude. Prompted in Claude Code on my phone.
The fundamental problem with Anthropic is that they are never going to release models WAAY better than Fable They think it will be insanely dangerous… So our only hope is Google, OpenAI, Grok and open source!
Play games on Computer with Astra as orchestrator. Works really well. [Quoting @AskPerplexity]: Computer use case: Create a browser game, then generate a sizzle reel trailer using in-game footage.
Microsoft seems to sell its own Claw now (I suspect they will not be the last), which may help spread personal agents in organizations But using a router with mystery models behind it is a big problem. Routers underestimate work difficulty in many fields resu…
The motivation behind the new Copilot: intelligence is accelerating fast. Now we need to diffuse it everywhere. [Quoting @jacobandreou]: Home. Code. Autopilot.
Gemini 4.0 leaks are unreal - super intelligence confirmed - Bernie will have a massive fit - auto-learning - infinite persistent memory - predicts the future Dario will 100% want to ban it 🤣
This was recorded a few months ago when I was still at Google and it was a really fun conversation! [Quoting @WR4NYGov]: How does this only have 21,000 views in 8 days? Chat with @JeffDean (then Google) and Bill Jia about large scale AI models.
New models are like new hires. When a new one comes out, you have to learn its personality, how to work with it, etc.
I think the "difficulty" of software engineering is essentially constant no matter what abstraction level you move to, because human cognition adapts to new tools until it can fully utilize itself. Tools are only affordances, not a magic wand that makes work…
A friend told me today: “We have a colleague named Bob. Ask him anything, and he just copy your question to an AI agent and sends the answer back.” Every company has a Bob.
The 2030s are going to be so weird as nanotechnology is developed. It’ll likely be the most profound change humanity has ever experienced… likely even more than AI.
You’re about to be the commander of an army of superintelligent AI agents. Please tell me you have something bigger planned than a fucking SaaS app. Think about what you’d attempt if you had a thousand geniuses devoted to helping you.
It is extremely clear at this point in AI development that, regardless of risk or revenue or any of the other stuff discussed on X all the time, things are just going to keep getting weirder. Just super, super weird.
I like this article! The problem is not for some people to care and work on catastrophic risks (we actually need much more research into it), it’s when company leaders and policymakers use fears of them to distract and hide the current challenges and solution…
The A.I. Daily