Happy to chat further. Please feel free to grab something here: tiny.cc/scottcal

Given you can already download ALL of wikipedia locally, I don't see why someone hasn't already done this. It seems likely a small language model like Phi3 could make an excellent front end. Is it the training cost?

Actually a SLM would likely be enough

I don't understand what's stopping Wikipedia (or someone else for that matter) from making their own LLM front end to ONLY Wikipedia content.

But you do have 4x the followers here... That just _might_ explain it

I'd like to know how to think about it. There is clearly a difference between the two. Not better or worse just different. FWIW more have agreed with me (but it's a small sample)

agreed! There is a lot going on in the Fediverse around this, with multiple servers being worked on for family-only instances vs sports-team style interactions

I haven't found that at all

I don't think it's age. BSky is much bigger which means it should be AHEAD of Mastodon. But I think it's more of a "consumption place" than a "conversation space" (hypothesis only!)

It's not through lack of trying! But I the vibe over at Mastodon just feels a bit chattier and more community-ish.

Although it DOES take a while to build up a following there. The next release will have a version of StarterPacks which should help.

Yeah, I'm not ready to 'declare' anything, more just comparing notes. But I agree with your assessment, BSky feels more like 'where the big kids hang out' and we're just here for the ride.

Mastodon just feels smaller/chattier which I like

Thanks for the prod, I just fixed it!

or is it "unplugged"?

I like Dan Saffer's point that designing for AI is like designing for an island of drunk people (I'm paraphrasing roughly) as it's wrong much of the time.

This means "UX of AI" is really more about UX for a mentally disabled agent, you're trying to pre-cope with it's failure.

thank you! Surprised you didn't post this on Mastodon! (or at least I couldn't find it....)