What will the new legacy code look like?

It will in many ways look better than a lot of ‘traditional’ legacy code we currently deal with.

Unlike plenty of labyrinthine codebases I’ve been in, AI codegen generally writes a lot of comments. It can name things reasonably well. When prompted well, it can structure things clearly and write tests as it goes along too.

So what’s the issue?

And why would we need humans looking at this code at all once we have better AI? Won’t it just handle it all?

More tomorrow.