Comment by viraptor

Apart from a few things already mentioned, I don't believe developers are really trying to engage with the LLM development. Some are just completely opposed to it ethically, some think it's not good enough and don't try, some don't care, some were not satisfied before. There's quite a few people using copilot, but that's the most basic and simple approach.

I don't personally know anyone trying to use more fancy tools like agents or ide-integrated helpers. They're not perfect by any means and you actually need to learn how to use them well, but the difference is massive. I've definitely saved some hours when developing smaller scope tools. It's not a time save that would drastically change my total productivity, but... it exists and it's going to increase in the future. And it requires upfront investment into the tooling and learning that few people seem to be interested in.

But even given current issues, how can you tell there hasn't been an improvement? How would you be able to tell across all the open source in the world?