Automated website feedback with browser use + LLMs

I am building a tool that gives automated qualitative feedback on websites. This is the early and embarrassing MVP: https://vibetest-seven.vercel.app/product

You provide your URL and an LLM browses your site and writes up feedback. Currently working on increasing the quality of the feedback. Trying to start with a narrower set of tests that give what I think is good feedback, then increase from there.

If a tool like this analyzed your website, what would you actually want it to tell you? What feedback would be most useful?

I want this but an LLM.

OpenwebUI and you can run quantized or low end models (Llama 3 4b or gemma 4b) on a 4-6gb graphic card.

It's a game changer to run local (no usage caps for a weekend blitz project)

I played with gemma-3-4b-it-qat recently using a mid-tier graphics card and a few things stood out to me:

1. It was very fast, between 35 and 70 tokens per second, with initial response in under 200ms. That kind of speed is a feature.

2. It was very useful. I had a brainstorming session with it that was both fluid and fruitful

3. I can't wrap my head around so much knowledge being contained in about 3GB of data. It seems to know something about everything. Imperfect, but very useful.

From the maker(s) of mythicspoiler.com

I used QUERY_ALL_PACKAGES among other things for my app Limit Buddy (https://www.limitbuddy.com). It would be impossible to make the app without it. But for more normal use cases there's no reason to have it.

Apple has a much more robust solution privacy wise with their ScreenTime API but it makes an app like Limit Buddy much harder to build.

I would be much more interested in a Goodreads alternative that focuses on articles/substacks/newsletters etc.

There isn't a good way of sorting through all the stuff out there and I feel like I am missing a bunch of content worth reading as a result.

Out of curiosity how did you initially populate your site?

That's a really good idea. The only trouble would be keeping the astroturfers away, it's much easier for articles - a company I used to work for would send out company-wide requests to upvote their articles on HN.

But if someone could make this work in a way that it's like a "if you liked this you might like" version of HN/lobste.rs it could be really useful.

Cool interactive, also +1 for being from my hometown

I worked there until 10 years ago and it was quite rewarding. (In fact this interactive uses some of the "lab" framework code I worked on.)

The founder was inspired by Concord (Massachusetts)'s history of progressive educational experimentation, going back to the transcendentalists (Emerson, Thoreau).

They're based in Concord but they also have a Bay Area office, after merging with the nonprofit org that built CODAP (https://codap.concord.org/)