Do you change your model or reasoning effort often?
How’d you decide?
Principal Software Engineer at Cloudflare. Fixer. I wrote that JavaScript Fatigue post 10 years ago. Sorry bout that.
- AI loves to writing tests way too early. I like tests that fortify the implementation after careful review. > “Don’t update or write tests until I’ve approved both the implementation and behavior. Then, use red-green-refactor against main that this PR resolves.”
- Protip - write the PR title and description yourself. It’s like taking a quiz to see if you’re understanding matches what the code actually does.
- Now that code generation is so fast, people are debating about the time to review the code. What about making QA fast with 🤖 too? Before/After screenshots & videos. Affected URLs & cohorts. Auto-updated product docs. Evergreen smoke tests. Perf deltas.
- I use this thing more than my trackpad. Dictate. Send. Switch. Scroll. Repeat.Oh no. Too many options. 😳



