[happymellon]

I've just had to "fix" a bunch of shit that was thrown over the wall that "sort of did the job" that came from someone using AI.

It's a game changer for some people who only need it to mostly get things started and pretend they did their job, and a work generator for anyone who actually needs to get things working.

The code was shockingly bad, and had to be rewritten to be able to do step 2 of the task.

[ironSkillet]

In my mind that is a problem with your lazy developer colleague, not AI as a whole. You can't expect it to be right on the first try (just like human code), you have to iterate with it and have the experience to know when it's off track and you have to take over.

[btreecat]

> In my mind that is a problem with your lazy developer colleague, not AI as a whole. You can't expect it to be right on the first try (just like human code), you have to iterate with it and have the experience to know when it's off track and you have to take over.

The problem with this IMO is when a human writes the code, they know the code they wrote, and have a sense of ownership in terms of correctness and quality.

Current industry workflows attempt to improve quality and ownership with PR reviews.

Most folks I see using AI coding don't know all the corner cases they might encounter, but more importantly don't know the code or feel any real ownership over it.

The AI typed it, and the AI said it's correct. And whatever meager tests exist either passed or got a 1 line change to make them pass.

Quality is going down from those who rely on tools to produce code they don't know. This has a cost associated with it that's been deferred.

Sometimes this is fine, like POC where you are comfortable with tossing the code out.

This isn't fine for business who need to be able to plan out work in the future. That requires knowing the system more so than just reading the code base.

[happymellon]

If only it was this once, and only this person.