What OpenAI says its 10,000-agent maths run proved
OpenAI says an unreleased model and thousands of coordinated agents found a Navier–Stokes blow-up proof. The result is public; acceptance is still to come.
Latest:
Upgrades
All →Deconstructing the tech hype.
Front page
Trail of Bits’ coop gives coding agents their own working environment
coop runs Claude Code and Codex in disposable virtual machines, with project sync and controls for connecting files, credentials and local models.
What expert AI-training work asks a professional to pass on
James Maisiri’s account describes an offer to train AI in teaching and assessment, the judgement that work draws on, and the choices it creates for professionals.
How developers organise and maintain agent skills
An Ask HN discussion describes skills as repeatable workflows, with shared repositories, human review, behavioural tests and updates across coding tools.
More stories
All →- Why OpenAI’s chief scientist wants shared limits on AI development
- Gemini’s saved instructions now carry across more Workspace apps
- ChatGPT for Healthcare brings patient records and public medical data into the workspace
- Nvidia confirmed the Hugging Face deal and promised its compute stays optional
- Did Multiverse actually build Europe's leading AI model?
- Google signed the letter asking for cyber-capable AI, then gated its cyber model
- Most of the sources behind Perplexity's software recommendations sit outside the top 100,000 websites
- World models could widen the AI divide before they improve ordinary work
- OpenAI Wants Your Trust Back, But Is Unclear On How
- Perplexity brings hybrid frontier and local AI to your Mac
- Runway's Solaris generates app interfaces frame by frame, with no code
- Anthropic says it had one layer of defence where it needed several
- An agent's tool list shows what you wired into it
- The magic in agent interfaces still needs an off switch
- Cloudflare can now see some MCP traffic, and the gap is the point
- 294,000 exposed AI tools is the wrong number to worry about
By topic
All 226 posts →
What I write about
All topics →Security & Risk
90
What breaks, who is liable, and what an agent can reach once it is already inside.
Start with: Your AI Assistant Can't Tell You From an Attacker →
Agents & Tooling
120
Agents, MCP, and wiring AI into the systems you already run.
Start with: AI That Actually Works Together →
Writing & Thinking
89
Prompting, writing, memory, and whether the model is telling you the truth.
Start with: How to Talk to AI (and Actually Get What You Want) →
Building
31
Making things with AI: coding, local models, and open weights.
Start with: You Don’t Need to Be an Expert to Start Making Things with AI →
The Industry
123
The labs, the money, and what enterprises are actually doing.
Start with: The 95% AI Failure Rate Nobody's Talking About (And What to Do About It) →
Start here
If you are new to the site, start with the path that matches how you workIf you run a business
Running a business and trying to use AI without wasting time, budget, or trust.
If you build with AI
Building with agents, workflows, MCP, or model-based tools inside real operating systems.
If you worry about risk
Watching risk, liability, and security more closely than the average AI optimist.