Claude is who I go to when I need to *think*, not just type. Architecture reviews, planning sessions, careful writing, postmortems — Claude does these better than anyone else on the bench.
How Claude helps me
Claude is my planner. I don't reach for it to autocomplete a function — I reach for it when I'm stuck on a decision and need a thoughtful second opinion.
Workflow
Typical session:
- Paste the problem (or repo context) into a Project.
- Ask Claude to summarize what it understood before doing anything.
- Ask it to ask me clarifying questions.
- Once we agree on the approach, ask it to draft the change.
- Bring the draft into VS Code and let Copilot finish the typing.
Strengths
- Long context. I can paste an entire feature spec and it stays coherent.
- Cautious by default. It asks before doing something destructive.
- Writing. The best AI writer I've used. Honest tone, doesn't oversell.
- Architecture reasoning. Will say "I don't know" instead of inventing.
Limits
- Slower than Copilot for simple completions.
- Sometimes too cautious — refuses tasks that are clearly fine.
- Web access is limited; combine with a tool that does live research.
Tips
- Always ask it to plan first, then write. The quality difference is large.
- Use Projects to keep context across sessions instead of re-pasting.
- For destructive ops, tell it explicitly what it may and may not do.
