A year of Claude Code and Cursor: the honest scorecard
I've used Claude Code and Cursor every day for about a year. Here's the honest scorecard, not the hype.
Where AI genuinely saved me time
- Boilerplate and glue code — the stuff you've written a hundred times. Gone in seconds.
- Refactors across many files — renaming, restructuring, moving patterns consistently through a codebase.
- First-draft tests — a solid starting point for unit and E2E specs that I then tighten.
- Getting oriented in an unfamiliar codebase fast — "explain how this flows" beats reading top-to-bottom.
Where it still falls down
- Architecture — it'll confidently pick the wrong abstraction and never tell you it's unsure.
- Subtle bugs — it writes code that looks right and quietly fails on the edge case you didn't mention.
- Your product — it doesn't know your users, your constraints, or last quarter's decision about why you don't do it that way.
The teams winning with AI aren't the ones who trust it most, they're the ones who review it hardest. I standardized these tools across a team and delivery got ~30% faster, but only because we treated every AI output like a junior's pull request: fast, useful, and always reviewed before it shipped.
Your turn: where has AI-assisted coding burned you, or genuinely saved you? What's the one task you'd never hand it?