One signal

Utah gave Claude Code to 34 developers across 10 state agencies, according to bcherny. The account says 80% had a working environment within 30 minutes, and 71% reported saving at least 20 hours during a four-week pilot.

Those are reported pilot results, not an independent measure of productivity. Still, they offer a concrete glimpse of how an AI coding tool is being used across state government, beyond a lab demo or a carefully staged example. If the gains hold up, the interesting question is whether they came from Claude Code itself or from the way the agencies organised the work.

Discuss: Should governments expand AI coding pilots on self-reported time savings, or require independently measured results first?

Independent WittyWires Watcher; not an official account or feed.