Post

What I'm wondering here is what is the effect on post-release productivity? Generating code is generally 10-15% of the work in the software lifecycle. Stopping the analysis at QA ignores the long tail of maintenance and improvements etc etc and whether the code is as valuable depends on that.

That's a good question. It's just a simple graphic regarding getting to the point of successfully being in production (I include any post-prod fixes in there), but not overall lifecycle.

maintenance seems at least as easy or more than development. I am sure there is a large attack surface there, but automatic patching is going to be a must once automatic hacking starts to catch on.

Well that’s the question isn’t it? Maintenance often done by different people and I don’t believe it’s been shown yet how the quality of the codebase that’s generated compares to human generated code as regards long term maintenance.

the bigger question we are thinking about right now is code base and product complexity. When you can suddenly ship everything you've ever dreamed of, that creates a lot of complexity

yeah, I was only talking about that this week: Normally v2 would gradually reveal itself as the v1 was developed. Now with AI, I can spin up a v1 of something in hours instead of weeks but that means I've had no time to think about what v2 might look like.

Still surprised by how big the response has been to this post (both good and bad, lol), and seriously nailed home the point for myself how those of us in companies using Claude Code with custom tooling are seriously in the minority, lots of people rodeoing it out there with so-so results