I’m writing this because I’ve been a bit confused regarding the current state of WeWeb AI.
To give some context: I originally upgraded my plan specifically because I use the AI to build out some foundational elements. However, since the feature currently isn’t working as expected, I’ve decided to downgrade my subscription for the time being.
For my specific use case, this is an essential function, not just a “nice-to-have” add-on. While I’m more than happy to upgrade again once the service is stable, I can’t really justify paying for AI tokens when the core functionality is not working properly.
A quick thought/question for the team: Would it have been possible to keep the “old” AI version running until the new implementation was fully ready? AI is such a main feature in modern development workflows that having it not working feels like a significant hurdle.
Does anyone have an update on when we can expect the AI to function normally again?
Hi Michael, thanks for reaching out and sorry for the disruption to your workflow.
Unfortunately, the previous AI wasn’t compatible with this release. We rebuilt our AI system from the ground up, along with many foundational layers of WeWeb’s core, so carrying the old one over wasn’t an option.
The good news: the new AI system is architecturally much more robust, evaluable and scalable. It’s in Alpha right now, but it’s going to improve on a weekly basis. By the end of the month, it should surpass what the old AI could do.
I’m building a few apps on WeWeb and I’m very concerned about the issues I’m encountering with the new AI feature. I’ve witnessed the AI spontaneously removing my workflows and elements, and I have no way to stop or revert these changes. It’s like watching an “AI apocalypse” take over my work. When the AI seems unsure of what to do, it keeps trying, often resulting in the deletion of my components. Even refreshing my browser didn’t stop the process.
While we are testing this alpha version, could you please consider adding a quick feature that allows users to immediately stop the AI’s actions and “revert” back to the original state? This would be a crucial safety feature.
Hey @Yvonne - I have experienced the same in the past and I have started incorporating explicit instructions into my prompts like " Do X but DO NOT modify existing workflows, variables, formulas" or whatever I don’t want it to touch based on my case scenario and has seemed to help with regression and re-work
Thank you @RandyG0322 for the tips! I’d be sure to test this out (after the next release at the end of the month). I’d hear back from support that it’s better to wait since the new release should be more stable!
