Here’s another shout-out to Claude Code.

Three months ago we started switching our primary AI-enabled development environment from Windsurf to Claude Code.

We mostly run Claude Code outside the IDE — it feels more natural when it’s working through large tasks.

Since then, our throughput on Jira tickets has roughly tripled.

That includes the usual day-to-day work, but also some large refactoring tasks we’d simply avoided before — the unrewarding, difficult, risky kind. Adding Claude Code is the first time I’ve seen a tooling change whose productivity gain is so clearly visible in the Jira burn-down chart.

And we haven’t seen any drop in code quality. If anything, the opposite: unit test coverage for new code is significantly higher, and our overall coverage has finally crossed 50%.

Typically Claude Code takes a whole feature from a Jira ticket through to completion. My main job is making sure the ticket is detailed enough — relevant technical design aspects, clear test instructions. Claude Code then writes the code, compiles it, and runs the unit tests until everything passes.

The real productivity shift isn’t the coding speed. It’s doing the design and specification up front, instead of — as many developers do — figuring it out while coding. I’ve been preaching the value of upfront design for a very, very long time. Finally, with tools like Claude Code, everyone can see how it really pays off.

Claude Code is genuinely fast at writing code, analyzing existing code, adapting established patterns, and spotting inconsistencies. It asks when it needs more detail. But the key to success is starting with a well-specified ticket.

Next step: getting the rest of the team developing entire features from a single prompt off a Jira ticket. With a productivity gain this consistent, our backlog will soon be pretty empty — and we can start thinking about the new requirements we were once afraid to tackle.