Current AI builder makes a good amount of errors when implementing small changes (updating workflows/formulas) that the previous AI within the formulas or javascript didn’t make.
This new claude coding model has significant coding improvement that I would pay to have implemented if it meant that I could more easily take designs from figma to live within weweb.
Thank you for reading.
It’s been like 10 hours since release xd
Lmaoo very true the space is moving insanely fast. I guess a better way to frame the question:
are there plans to add ability for users to choose which coding agent/model being used within weweb (aka hit a different api endpoint)?
similar to cursor functionality*
Hi @imb
Thanks for the suggestion. It is not currently planned, and we may never implement it. The primary reason is that there are multiple agents available, and WeWeb must select the most suitable one for each task. We continuously test the latest models to ensure optimal performance. If you encounter any errors with AI generation, please don’t hesitate to reach out, so our team can investigate further.
I think we should use DeepSeek more because it is open source which will save more cost of weweb. Importantly, we can choose to update the system whenever we want without much impact because there is no change of API. However, relying on SaaS agent is not sustainable at all. When the service provider updates the API, changes the system, we have to update continuously. If they charge more, the token cost will be more expensive. Or if one day the service provider stops doing business, what will you do? All the trained data will be lost.
Thank you for the breakdown and congrats on the AI launch. Y’all are doing great work here.
WeWeb work environment has for sure got the best environment for customization and tweaking so excited for yall to perfect the AI.
we are doing tests now and working with our providers to enable enough tokens flowing for this new model, we hope to upgrade to it this week or the next, we’ll keep you posted!
Hey @imb! So, after extensive testing, we’ve decided to delay integrating Claude 3.7 with WeWeb AI for now. While powerful, it consistently produces oversized outputs with accuracy issues.
However, @Alexis discovered a promising approach for using it specifically for custom component generation. Initial results are exceptional, and we’re conducting additional testing this week. If all goes well, we plan to roll out 3.7 for component creation next week.
yall have been absolutely killing it with the updates. thank you!