How to Review AI-Written Code Without Trusting the AI (or Yourself)
The most useful stage in my workflow is the one built on trusting nothing — least of all the agent that just wrote the code. Four reviewers, one rule (a reviewer is worth how UNLIKE the builder it is), and the honest ledger nobody selling AI review will show you: the scariest findings are usually the wrong ones.
russalo@blog ~ % ls -t posts/ | grep "#claude"
#claude
2 published posts
How I Run Claude, Gemini, ChatGPT, and Copilot as One AI Crew
Two and a half years of how I actually use AI — Claude, Gemini, ChatGPT, and Copilot — run like a construction job, because the workflows genuinely translate. The models quietly lie to themselves (here's the exact mechanism), so I build everything around one rule: keep the truth outside the machine. The frame, the crew, and the playbook, with the deep-dives to come.
russalo@blog ~ % cd ..