A newsletter covering several AI and engineering topics: LinearB's APEX framework for measuring AI impact on software delivery (beyond DORA/SPACE), GitHub Copilot's controversial PR advertisement incident, Anthropic's accidental leak of Claude Code's source revealing zero tests and a 3000-line function, Shopify's 75x inference cost reduction using fine-tuned Qwen 3 models, managing open source projects flooded with AI-generated PRs, and a new JavaScript library called Pretext that bypasses CSS for text rendering.
Table of contents
1. What happens when a PR becomes a billboard?2. Surviving the whiplash of “dead” AI trends3. Zero tests, 500k leaked lines, and virtual terminal pets4. Swapping giant models for highly tuned smaller ones5. Faster coding doesn’t mean faster delivery6. No more CSS in your browser?7. Managing open source in the age of agentic pull requestsSort: