Clauday.

Debugging an agent today is mostly rerunning it and praying it fails the same way. It won't — that's the whole problem. Retrace, which launched on Product Hunt this week, is an exe…

Here's a result that should make every lab's finance team sit up. A new paper, Is One Layer Enough? (arXiv 2607.01232, on HN's front page today), finds that training a single trans…

Manufact (YC S25) launched on Hacker News today with a simple pitch: MCP servers should deploy like web apps. The company maintains mcp-use, an open-source SDK with over 7 million …

The tagline alone earns the star count: why use many token when few token do trick. Caveman is a Claude Code skill that makes your agent talk like a caveman — sentence fragments, z…

This one takes a second to sink in. OpenAI has an official repo called codex-plugin-cc, and what it does is let you run Codex from inside Claude Code. Type /codex:review and OpenAI…

Meta took astryx, the design system it built and used internally for eight years, and open-sourced it. React and StyleX under the hood, 150-plus accessible components, theming, dar…

Every time your agent runs LLM-generated code, you are trusting something no human read to touch your machine. CubeSandbox, just open-sourced by Tencent Cloud, is a place to run th…

BAAI dropped Orca and it's the number one paper on Hugging Face today by a wide margin, 176 upvotes when the next one is at 21. The pitch is in the subtitle, the world is in your m…

Zhipu just shipped ZCode, and it's on the Hacker News front page today. Think Codex or Claude Code, except the model underneath is GLM-5.2 and the whole thing comes from the people…

A new paper from the University of Washington asks a question that sounds trivial and turns out to be brutal: do agents know when to stop instead of act? The team behind it, Han Lu…