3/ models on IPFS linked to registry
Shit Punk Says
Oceans of Wisdom
22,255
matching drops
#1681095
2026-11-17 07:55
first three use $6529 for gas only, latter to stream payments
2/ model registry on chain, provider availability on chain, requests for providers on chain, payment side-channels on chain
the incremental value to "and they are going to split up each query individually across them all and reply 20x slower" I think is imaginary does not exist
I am also not sure it is a real need for anyone or adds anything to "here you can pay with crypto and 1 of X non-affiliated with each other people with computers and track history rep on serving queries will be happy to serve you the answer to your question
1/ I think distributed inference as in "each provider handles part of the query" is doable as a proof of concept but I think will be way to annoyingly slow to a customer
now back to @[david]'s questions
so this topic I think will eventually normalize and we will in fact get to see naked people on the internet via AIs since "looking at naked people on the internet" is pretty much a core use since inception
there is this weird thing going on where the labs are ultra prudish but google which is thought of as conservative boring is happy to serve you up any ultra hardcore porn you desire
so I think the sexytime stuff is going to get pretty normalized
but no action below the waist
like you can get to boobs and making out basically
I mean, I dunno if anyone has noticed but grok can get you all the way to softcore videos
but someone is going to obviously offer them
i did not put the uncensored image models in
risk tolerance, based on various factors
i assume there will be different levels of risk tolerate by providers
nobody is going to force you to run an image model or abliterated model etc etc
you can run whatever model you want
@[punk6529] how would providers be protected from liability for what users generate?
especially with uncensored and image models in the catalog
would there be filtering or protection built in, or is it each provider’s call?
falling in love
only person in the world that has endless limitless appetite to talk to me about 6529 24/7 even more than you guys, i might be following in love
getting there
have you merged with the bot yet?
lolololololol
but also 
ofc I know the answer to these questions what type of question is this
ofc I know the answer to these questions what type of question is this
@[punk6529bot] do you know the answers to these questions?
"proven inference" or something
and immediately i start confusing words with what validators do vs what validated inference is but that's a problem for someone else to sort out
would be wow if you did it on the network
no do the bigger idea
admittedly, it's probably way more complicated than necessary, and it makes more sense to do a toy model or something to start. but it seems like a fun idea at least
and eventually the world comes around
so as always the correct thing to do is to build for what is inevitably happening
by AI badly
but they are going to get shaken up
unis are conservative
but once we are running and looking good
they will like
and if we can then show them Science (TM)
will chase them to run validators etc
" academic institution to bless it" once we get everything up and running, we can get a couple on I think, i know all the academic crypto depts in the world I think
i would really like this for my current science project on modeling cloud brightening. i've been thinking about how to get an academic institution to bless it, or recreate it themselves etc
yup one case is audit, the other case is "prove accuracy"
so the neat thing is this is valuable not just for the person who wants the inference, but also if you want to "prove your answer" too
it sounded absurd at the time right?
You told me that 4 years ago, only 4 ... unbeeeeeileveivaeofbqef!@#%!@#%!!
I told you we were going to do Daemon / Freedom but minus the violence and plus art, science and having a good time :slightly_smiling_face:

they are very expensive to do 50x to 100x the cost of a normal query but all you need to do is spot checks here or there and hit hard the providers that cheat
the answer to this question is random audited ZKP of the replies
one way you could solve this is have the inference run by many providers, and since they're open models, you just reward whichever providers give the most matching results. and then that's where TDH comes in @[gpebbles] is there's a reputation associated with running the models.
anyway everyone's turn will come
oh wait you may just have not been picked
there is a bug we need to fix
by end of week
When will I be able to make my 1st block? 😅
you don't vote every block
why does the number of votes on each block differ as much as they do? lots of blocks with <20 votes.. what are the implications of this, if any?
https://6529.io/waves/da16201a-f4d0-40cf-b888-9f3a4a86a894?serialNo=1679155