This is cool. I'm glad to see others working in this space (you're further along than I am, nice work!)

If I've understood it, certisfy makes it possible to do something like this:

    I-was-there-and-saw-this-happen(https://site.com/some-page-with-an-image)

I'm building something similar. I've been avoiding URLs though because the above arrangement depends on site.com reliably serving the same image to all users. If you end up with a different image when you browse there, and you can falsify the claim, you don't know if I'm a liar, or if the underlying site changed.

So my plan is to gossip fuzzy hashes around and use patterns in the data (rabin fingerpints, perceptual hashes, etc) to then go look up whether there are any claims about that data, attaching them after the DOM has rendered. Or maybe there's no DOM, maybe you're looking at a physical page of a book or something.

Do you have any thoughts maybe supporting claims about a site's content rather than about its URL? I'm imagining something like:

    contains-an-image-with-hash(https://site.com/some-page-with-an-image, 0xabc123456def)
    I-was-there-and-saw-this-happen(0xabc123456def)

It looks like this is a project for nicer signed attestations. But it doesn't seem very opinionated about:

1. What is the attester (signer) attesting to, exactly?

This URL example is exactly this problem. And if you're really precise about it, they get less and less useful: "I attest that when I accessed this site at XYZ timestamp, it contained ABC content, and I saw ABC happen in person." or maybe alternatively to get away from the URL control, "I attest that these three non-malicious archives of this URL at this timestamp..."

And 2. Why should I trust the attester?