LLMs are also a kind of extreme data compression.
Non-payment data on bitcoin maybe have some use (outside monkey jpegs), but blockspace is super scarse and expensive.
Could LLMs be used to reliably interpret super tiny data "bursts" stored forever as something much more complex/useful?
Just thinking out loud here.
nostr:npub1m2mvvpjugwdehtaskrcl7ksvdqnnhnjur9v6g9v266nss504q7mqvlr8p9
found in my bookmarks:
Yep, I'd be surprised if it didn't work for text too.
"almost deterministic prompts"?
Cought you there, you didn't read it ))
It's actually much better in text compression.
but to your idea about sharing things via btc blockchain ...
you can store a hint for any program. how it interprets is up to that program. that peogram is not immutable like blockchain. so ... why/how?
Saved it for later. Now that I read it, it doesn't really change anything (but it's a good confirmation it's working well).
The usecase is still unknown (to me). I kinda like "arbitrary data on blockchain" as some kind of proof or a certificate. But there is maybe some other usecase for uncensorable content with timestamp that doesn't make (financial) sense without extreme compression. And I don't mean collectables :).
Yes I know opentimestamps exist.
Again: this is just thinking out loud in case someone else might get inspired.
the network is very conservative. and that's good. it allows you to save shit there (for a price) but all other programs outside of that network can interpret the data in ANY way. it's outside of (btc) consensus. lightning is not an exception. any l2.
