There's automation and you can do it manually if needed. For example I have a couple of emulators that pull every 24 hours from GitHub just in case nint tendo gets a little lawsuit heavy. I also have one offs from GitHub that pull down when I want.
You can also mirror a public repo from GitHub into a private repo so it does not gets indexed/ai trained.
I had the same issue. I had to turn off AI features and it started working again. Specifically it stopped working on a peice of legacy code that had hundreds of thousands of lines of code all in one file. No idea what version of vscode but it was fairly recent.
Yeah it's been a little less than a year since it stopped working. I saw the GH issues, looks like the Lemmy devs took the position the Peertube devs need to fix....something with activityhub. I hope they fix things. Lemmy development vs the actual instances seem a bit more fragmented than the rest of the fediverse.
I tried to take a look with my local Lemmy instance but it looks like the issue resides in how different Lemmy uses activityhub than other software. I don't claim to know why they made the change but they seem to be the odd one out.
On the plus side pixelfed seems to be popping off. And every time one of the big companies f up, we get a slow trickle of new users. Which is much more sustainable in the long run.
As a dev, there's still quite a bit ai can't do and will most likely not be able to do.
AI is good at solving old problems but it's not trained on anything new. Its good at boilerplate and templates, but not good at original material. If it gets tremendously better, and really does get to the point where it's better than we are at development, then the industry will shift into prompt engineering. But I can see a huge reduction of jobs.
When the worst case is to lose your job and never deal with the legacy code again.