Claude Code

Verdict: Worth your time

What it is: an AI coding agent that runs in your terminal. You describe what you want built, it writes the code, runs it, and fixes what breaks.

What we ran it on

Everything. This website was built with it: scaffolded, designed, written, deployed, and wired to a mailing list in roughly a day of working sessions. Not a demo. The site you are reading right now is the receipt.

Do this

Give it a real task end to end ("build me a page that does X"), not one line at a time. It earns its keep on whole jobs.
Make it prove its work. Ask it to run the thing and show you the output. The difference between "I wrote the code" and "I ran it and here's the result" is the whole game.
Review what changed before you ship. You don't need to read every line. You need to read the summary and spot-check the claims.

Skip that

Skip using it as fancy autocomplete. That's paying for a contractor and asking them to hold your ladder.
Skip trusting "done" without evidence. Any AI tool, this one included, will sometimes claim success it hasn't verified. Demand receipts.

Sharp edges

It costs real money at heavy use, and the bill scales with how much you lean on it. It needs version control underneath it (git), which is one new habit if you've never used it. And it is confident by default: the discipline of asking "show me it working" is yours to bring.

Disclosure: we build everything with this tool, including this site. We pay for it like anyone else. That's not a conflict. That's the test.