Generation quality seems very off today?

[reuben-flyweelco (Reuben_from_Flyweel)]

Is anyone else having major issues getting the AI to follow instructions today? It is significantly worse today for some reason

[pawlean (Pauline P. Narvas)]

Hi, @reuben-flyweelco! Welcome to the Vercel Community

Sorry to hear that this has been your experience. Feel free to drop any chat link examples, or prompts so that we can share this with the v0 team.

Appreciate the patience as we continually improve the platform!

[pawlean (Pauline P. Narvas)]

Also wanted to cross-post a reply from another thread

Best Practices for Messaging Input Tokens

I also wanted to follow up with more tips after we shared this with the v0 team: You can set a custom instruction in v0 to output only code when building applications. This helps reduce token usage by skipping explanations about code changes if those aren’t needed. We’re also introducing V0-smol soon—a lighter-weight option for simpler tasks that don’t require high-level intelligence, making those tasks more cost-effective. Additionally, smaller projects and modular file structures tend to us…