I appreciate your reply. My tone was a little dismissive; I'm currently deep deep in the trenches trying to unwind a tremendous amount of LLM slop in my team's codebase so I'm a little sensitive.

I use Claude every day. It is definitely impressive, but in my experience only marginally more impressive than ChatGPT was a few years ago. It hallucinates less and compiles more reliably, but still produces really poor designs. It really is an overconfident junior developer.

The real risk, and what I am seeing daily, is colleagues falling for the "if you aren't using Cursor you're going to be left behind" FUD. So they learn Cursor, discover that it's an easy way to close tickets without using your brain, and end up polluting the codebase with very questionable designs.

Oh, sorry to hear that you have to deal with that!

The way I'm getting a sense of the progress is using AI for what AI is currently good at, using my human brain to do the part AI is currently bad at, and comparing it to doing the same work without AI's help.

I feel like AI is pretty close to automating 60-80% of the work I would've had to do manually two years ago (as a full-stack web developer).

It doesn't mean that the remaining 20-40% will be automated very quickly, I'm just saying that I don't see the progress getting any slower.

GPT-4 was released almost exactly two years ago, so “a few years ago” means GPT-3.5.

And Claude 3.7 + Cursor agent is, for me, way more than “marginally more impressive” compared to GPT-3.5