Cloudflare published the actual economics of AI code review this week: 131,000 reviews a month, $0.98 each, and an 85.7% cache hit rate. Swizec’s counterpoint landed the same week, arguing that most of what those reviewers catch is a deterministic linter rule you could run in seconds for nothing. Meanwhile Cursor rebuilt Git storage on write-ahead logs in S3 to get past the limits consensus-based Spokes ran into, and pgrust rewrote the Postgres query engine with batching, operator fusion and SIMD to make analytics 300x faster.
If you’re interested in sponsoring this newsletter, feel free to reach out via LinkedIn or simply reply to this email.
Newsworthy stories
Tutorials of the week
Videos of the week
Projects of the week
Bruno replaces Postman with an offline API client that saves every request as a plain-text file, so collections branch and are reviewed in Git like the rest of your code.
Munder Difflin runs your terminal coding CLIs as a self-coordinating team of agents on the subscriptions you already pay for.
secfiles collects wordlists and payload files for penetration testing and bug bounty work.
OpenViking is a self-evolving context database that unifies agent memory, RAG knowledge, and skills.
triagebot-action puts every new GitHub issue through reproduce, diagnose, and fix, then ships a preview release and waits for the reporter to confirm before opening the PR.
Agent Safehouse sandboxes coding agents on macOS so they only reach the files and integrations they actually need.
Meme of the week
Cloudflare pays $0.98 per review for AI to catch things a linter would catch for free. Which one is actually running in your pipeline today, and would you drop the other?