What changed my workflow

The productivity gains are real. Not in the way the demos suggest, AI doesn't write features while you watch, but in the mundane: boilerplate, test scaffolding, documentation drafts, converting a data shape from one format to another.

Tasks I used to procrastinate because they were tedious but not technically interesting now take minutes. That friction reduction compounds over a week.

Where it actively costs you time

Confident wrongness is the failure mode. An AI assistant that said "I'm not sure" would be easy to use correctly. One that produces plausible-looking but subtly incorrect code is dangerous, because it passes the visual scan.

I've caught: outdated API usage that was correct in an older version, off-by-one errors in generated loops, security issues in generated authentication code, and race conditions in async logic. All looked right at a glance.

The skill that matters

The engineers who benefit most from AI assistants are the ones who can evaluate output quickly. They use it as a drafting tool and apply the same scrutiny they'd apply to code from a junior engineer, maybe more.

The engineers who struggle are the ones who treat it as an oracle. The tool amplifies whatever judgment you bring to it.

The honest summary

Useful. Not transformative. Requires maintenance of the skills it's nominally replacing. Worth adopting with clear eyes about what it does and doesn't do well.