Articles
Industry news, technical articles, and product introductions
📚 Claude Tutorials Hub
40+ step-by-step Claude guides — prompt engineering, Claude Code, API, agents. Browse by topic →
Loading...
Industry news, technical articles, and product introductions
📚 Claude Tutorials Hub
40+ step-by-step Claude guides — prompt engineering, Claude Code, API, agents. Browse by topic →
Loading...
We ran an 'LLM proposes, code decides' auto-tuning loop on a daily WeChat newsletter for 40 days: 18 versions, 8 rollbacks. Three of those rollbacks overturned decisions a human had made, and one version whose knobs were identical to its parent still got rolled back. Using our real numbers, telling apart a 20% difference needs about 25,000 reads per version. At 60 reads a day, that is more than 400 days. This postmortem covers five reasons the flywheel never turned, and what actually moved the numbers.
DeepSeek-V4-Pro-0813 posted a Terminal-Bench score 0.1 behind Fable 5, then scored 53 on Artificial Analysis — one point above its own small model — before the announcement vanished that afternoon. The Hugging Face commit log tells a different story than 'the model is bad.'
Nineteen database corruptions in six months, all traced to a data race that had been sitting in SQLite for roughly 16 years. Here are 10 lessons from Tailscale's investigation into the WAL-Reset bug.