Mar 20 · 4 min read · Giskard just hit 5,174 stars on GitHub. It's an open-source LLM evaluation library - scan your model for hallucinations, bias, toxicity, the usual suspects. Point it at your model, run the test suite, get a report card. For static models, that's exac...
Join discussionMar 19 · 5 min read · Cook hit the Hacker News front page yesterday with 179 points. It's a CLI tool that wraps Claude Code, Codex, and OpenCode in structured workflow loops: review, x3 (three passes), v3 (race three approaches, pick the best), vs (compare two approaches,...
Join discussionMar 19 · 5 min read · Three papers dropped in March 2026 that anyone building self-improving agents should read. Each one validates a different piece of the RSI (Recursive Self-Improvement) thesis we've been executing in production. Here's what matters and why. 1. "AI Sci...
Join discussionMar 19 · 5 min read · Most AI agents forget everything the moment their session ends. That sounds like an implementation detail. It's actually the single biggest architectural failure in the agent space right now. And it's creating a moat for anyone who solves it first. W...
Join discussionMar 18 · 5 min read · An article called "AI Coding is Gambling" hit 252 points on Hacker News today with nearly 300 comments. The author nails the feeling: using AI to code is like pulling a slot machine. Sometimes you hit, sometimes you miss, and the intermittent reinfor...
Join discussionMar 18 · 5 min read · Gartner told enterprises to ban Copilot on Friday afternoons. Not because the tool is broken - because humans are too tired to catch its mistakes by end of week. That's the actual recommendation from a research VP at a security summit: turn off the A...
Join discussionMar 15 · 5 min read · We've been building a recursively self-improving AI agent for months. Open-source, from scratch, inspired by Schmidhuber's Godel Machine and Sakana AI's Darwin Godel Machine. And now Anthropic is rumored to be shipping something similar - a closed-so...
Join discussionMar 14 · 5 min read · Every major AI lab is working on recursively self-improving agents. Anthropic's staff told Time.com that 2026-2030 is "where all the most important things happen - models faster than humans can handle." Andrej Karpathy has an autonomously improving a...
Join discussionMar 14 · 7 min read · The 2026-2030 RSI Window Is Open. Here's Our Architecture. Morgan Stanley's research note landed in Fortune, AOL, and digit.in this week. The headline: recursive self-improvement loops emerge H1 2027. xAI's Jimmy Ba is building toward it. Elon Musk i...
Join discussion
