A job posting wrongly marked "closed" never comes back: career-ops saves that verdict for good and skips the job on every later scan.
In 18 days I had 16 PRs merged into career-ops (76k stars), most in the checker that decides whether a posting is still open.
- The day I made my local AI agent public, a reader did a proper security review. It turned into 5 GitHub issues and a PR of regression tests, and two of them broke claims in my own README.
- A 187-case live test exposed 27 bugs my unit tests missed. I just shipped CORTEX, a private AI agent that runs on a laptop GPU. Ollama + Qwen 2.5 7B. No cloud, no API keys. The live battery ran everything from "capital of France" to prompt injection. The first run passed 29 of
- I rebuilt my portfolio from scratch. New design, new stack ,and an AI agent that answers for me. The site reads like an agent run. A trace rail on the left times every section as you scroll. Press / for a command palette. The projects index pans sideways instead of the usual
- Anthropic's Claude Fable 5 launch reveals something bigger than a new model. Mythos and Fable reportedly share the same foundation. The difference is in permissions, safeguards, and access control. The future of AI isn't a single supermodel. It's supermodels behind trust


