Hey, AI Enthusiast! Let's talk tools
What's the closest thing to a "Kevin AI" currently or what needs to be built?
Is it a RAG system we can integrate into into existing AI with MCP? (or does it require a finetuning a model and A2A?)
AI tools are getting better all the time. But IMO, The current state of AI tools fails with CSS by suggesting...
1) outdated css advice
2) using bloated javascript where modern css will do
Does a good tool already exist, or do we need to have a bot watch all Kevin's videos and RAG them?
Traditional Frameworks have migration and upgrade recommendations. Is there an equivalent repo of examples on "old" vs "new" CSS that an AI can use?
5 Replies
I love the idea of a bot with knowledge from all of Kevin's videos! You should definitely give it a go and share with the community
To stay on top of CSS for an LLM you're on the right track with RAG, it would be more economic than training new models on all the latest CSS as a generic LLM can simply look up what new APIs are available through the spec, check https://caniuse.com data, suggest polyfills, progressive enhancement strategies etc.
I can see a good business opportunity here for a CSS helper, especially when models get better at reasoning
Honestly, I'm really hoping someone else is developing the tool. I just don't want to miss it. Could I technically make the product, probably. But time is more my enemy than my technical skills.
I think that using up to date AI to assist in writing modern CSS would actually be a huge boon to the CSS community, as I think it would lessen the reliance on JavaScript frameworks like React. Especially if the javascript signals proposal gets approved.
I'm not sure how tool for writing css would lessen reliance on react
Dont the framework solve different problems than css
In my experience, the worst part of AI web development is the CSS and the best part is the JS.
Correction: AI is even worse at regular expressions than at CSS.
CSS, in my opinion, isn't "solvable" in the same way other languages are, where encapuslation is so important. The global nature, and interaction between everything, along with the visual element of it, the accessibility, etc. is incredibly complex.