This is actually important because a site can be sued if it isn’t WCAG 2.2 compliant, meaning that people with vision disabilities can access it.
I had one of the accessibility widgets installed and figured I was all good.
But yesterday I found articles saying this is not so. Evidently there are things the WCAG spec requires that aren’t necessarily done by the widget.
So I started running my files through LLMs to see if they could correct everything – rather than doing it manually, which used to be a lot of work.
I used the Warp AI initially because it can go through the files in my app without me having to manually copy-paste each one into e.g. Claude. Warp also gives you a nice diff for each change.
When after around 40 .jsx files, I ran out of Warp AI credits, I started copy-pasting into Claude.
It seems to be doing a good job. It’s adding aria labels, alt text, adding titles where needed, setting up accessibility roles, etc.
@admins I think this forum deserves a new topic for AI Tools. In this topic I would love to share and hear from others about
- local LLMs they use for coding assistance and strategies for better performance,
- integrations between Meteor servers and automation tools such as n8n
- Meteor methods that pass through AI workflows (perhaps with the help from microservices)
- anything Meteor and MongoDB with AI
In 2 days I am getting my Mac Studio and I’d love to know if anyone here is using local LLMs up to 70B for coding and what they discovered to work best. Also, you can let me know if you want to test anything on a M4 Max with 128GB RAM, I can arrange a remote session or so.
2 Likes
@paulishca I have a post that would go well in the topic you propose. Evidently there’s something a bit new called a Small Language Model, which is highly focused on specific tasks. JetBrains has a new one they’ve developed specifically for coding, called Mellum. It is available through their IDEs, and has also been fully open-sourced.
