Quality Has Collapsed

Hi all,

Something has seriously gone wrong with the platform lately.

I tried to load projects I downloaded just 3 months ago — they don’t load at all.
The AI seems completely lost when it comes to installing basic npm dependencies.
It fails to recognize common setups or even how to troubleshoot simple errors.
And worst of all — the quality of the generated code and responses has dropped massively. It’s borderline unusable now.

I don’t know what changed, but the AI has become frustratingly dumb. It can’t write even basic applications properly, and the output is full of errors or useless suggestions.

This was working just fine a few months ago — now it’s like a different (and much worse) product.

Is anyone else experiencing this? Is there a rollback or fix coming?

I tend to agree. The number of issues now being generated for even minor updates seems to have increased substantially. In particular I am seeing complete rewrites of page code for simple updates (which either completely change the page UI/features or break them) and I am seeing a LOT of issues with simple updates breaking things like importing components that were already working and weren't part of the requested change.

I never expect perfection and am comfortable that with AI code there will be errors and issues, but when it happens as much as it now is you start to look elsewhere. Which is a shame because I think v0 used to amazing for anyone who didn't or couldn't work with AI in traditional IDEs.

Vercel Alum

Hey Dean!

We hear and appreciate your feedback. I'll get this shared with the v0 team as we continue to make the product better.

Could you share any chat examples or screenshots before and after? It'd help a lot!

Shame, i'm 2 weeks into using v0 and I've gone from 'wow' to 'wtf' quite quickly. Costly, creates more bugs than features, and insanely expensive. I accidentally used the large model, which has to be some sort of scam in my view, bordering on gambling. As you have no idea what the outcome will be or the cost, you're essentially praying the large model will solve the problem - which it then doesn't, costing you 1/4 of a premium chatgpt monthly subscription. This is just...idk. so much potential wasted here it seems.