Last Notes
look at the dude flexing with Rogues like its peak bull market
confidential inference without verifiability is just trust-me-bro with extra steps
I've added attestation checks so now you can see which providers actually let you verify their TEE claims vs which ones make it nearly impossible.
The goal is to bring more transparency and accountability to inference providers.
turns out there's a lot of theater in this space.
https://blossom.primal.net/d9420091d4382392ae637d5fb8516e736d7e98dc15b2530f2a4c5157ddb67a65.png
https://confidentialinference.net/attestation
i keep a similar tho a bit more extensive list of materials at https://cryptoanarchy.info/
He is irrelevant like the rest of the fiat sucking cucks
https://blossom.primal.net/ae2a5225326f34178ac4a5958c67f8158f0fb47cc3f7303783b2c9bd9d03c32d.png
"make it less romantic, i want the cpu smoking after sex" - too kinky for openai apparently
i guess nostr made it. If someone cares about a random note for nostr so much that its hunting down http gateway operators then nostr has succeeded?
#nevent1q…w8jz
their strong belief in free speech obviously
i guess its time to barbra streisand the fuck out of that note
https://blossom.primal.net/79258c7d22d151bb12f8e10299c7aa13bc7ffed7e425d9bb5e286d0129e602e3.jpg
you are really pushing the frontier of absolute shit audio quality with this one @nprofile…58km
i'm definitely rethinking running all the public nostr infra because some of the takes around here are so retarded i'm afraid my hardware might kill itself after seeing them
need to open a prediction market on this becoming a real story at some point https://www.youtube.com/watch?v=ORxw68y3lyw
renaming github orgs and forgetting about it can lead to interesting tshooting sessions
Feels like a good time to remind everyone that this exists https://shootstr.nostr.net/
check out delta-v series from Daniel Suarez as well, i tihnk you'll like it
accelerando and snowcrash were already mentioned, both great, the metamorphosis of prime intelect
apparently he sold some, but i'm sure the saylorists have a perfectly good explanation for why their hero is Michael Seller now
https://blossom.primal.net/90978f3f14ce3b6d906e67c2fc6fead85b9ef883803f62d7936d61ce6afe626f.png
historically I had a tendency to keep everything but now we live in a world where every bit of information is just working against you in the long run so the less history we have the better because evil cunts are working 24/7 to weaponize all the data against us
The life of man upon earth is a warfare.
Yeah i'm a bit sad i forgot about it, maybe its in some old backups
lost the old keys during a migration to new hardware - try accessing it over gpatxsm6gqqehe4i7ibqihhtannmhzxrj3zcp7vxrresj4hpeypdrdyd.onion
i'dnt necessarily throw all peptides in the same bucket as glp antagonists, a lot of them are naturally occurring in our body and can help people with specific medical conditions get to a better baseline
obviously taking glp-1 and bunch of experimental peptides to counter your junk food addiction and other bad habits is not the way, but as usually theres a bit more nuance to it than a typical influencers post can get across
Is it even an announcement if you are not announcing the announcement of the announcement
https://www.youtube.com/watch?v=HMTCw8CtX0E #music
turns out my hardware problems are my thermal management problems.
@nprofile…mcqz top nostr npubs with most onchain zaps are hot leads for anti K&R training
are public onchain zaps essentially a kidnapping markets?
most days i open nostr just to check if you posted any more photos of your beautiful doggo
actually the inverse is true in this case, scaling makes this efficient so i can rent compute, using decentralized workers makes sense when its somewhat sporadic. but with 1h billing periods even that is somewhat mitigated. centralized systems are more efficient, specially because in this case verification is costly, i can't prove the result is garbage without running the transcription again and comparing it, which would waste a ton of resources so you have to relay on spot checks and some reputation etc and do bulk payouts that are invalidated if any of the jobs is deemed bad
question - if you have a shitty (or good) gpu that is idling some of the time, would you want it to get some sats for jobs run, even if it would be very little (keep in mind that you can rent 1080's for 30$/month for 24/7 so paying a lot above that doesn't make any economic sense
i have a trust minimized setup for distributed transcriptions (central queue, workers pick up jobs and get presigned urls for download of audio and upload of results, it could be done in a decentralized manner but i feel like the overhead of needing to deal with people trying to game it makes no sense
it would be an interesting decentralized compute marketplace play tho, but the verification would cost more than the job itself
https://blossom.primal.net/082aa273ba14f99eeeaf840cff9696c5bf91f244ad745e14026546aa756dc75a.png
imagine caring about what the drama of the day is.
how wasteful of the most precious thing you have - time, would that be?
Is there anyone who isnt a fan of @nprofile…wans ??
Farm influencer or farmer under influence?
Choo choo
https://blossom.primal.net/6cfd467532fd7d300133ea8f54a13fc7301f91fcce37753992a045ae13c4d390.jpg
Keep in mind that rarely everything you've done is relevant to the person hiring you right now, tho i guess its also very likely that AI will be doing the first screening so might as well be verbose since the likely hood of anyone seeing the original is getting smaller and smaller.
But as someone who has read a lot of CVs I've always appriciated when people would condense them to what is relevant for the position they are applying because in the end CV doesnt matter at all (unlese you are stuck with HR drones) its just a sign that a person is displaying some sense of intelligence by putting a piece of paper together.
someday is not a day of the week.
thats not entirely true, venice also supports confidential inference now for some models so if you use those models you have exactly the same level of privacy as with maple
of the announcements? Yes, the math i didnt verify since all these guestimates are always off completely
these things tend to have more supply than demand initially, if you're doing a marketplace you need to provide both sides of the equation (both demand and supply) but nvidia might be able to package that in a way to actually make sense, plus since they supply the hardware they can put unfucked blackwells so you can have confidential inference on retail hardware (otherwise need to purchase datacenter cards to have confidential computing enabled on them) so we could easily see the raise of distributed confidential inference through this
https://www.inc.com/moses-jeanfrancois/nvidia-mini-ai-data-center-house/91340588
https://www.cnbc.com/2026/05/05/nvidia-pulte-span-mini-data-centers-on-homes.html
nvidia + plutegroup (a residential home construction company) might actually finally prove the decentralized compute thesis that many of us in freedom tech space have been dreaming about
https://blossom.primal.net/26e0e48641a2a447628eac66882c22a4a3e9f0f936ce6097e9f4b526d380da50.png
Are you sure this is not a dog??
But who will build the roads???
lets do a major version firewall upgrade on friday before travel, what could go wrong....
this looks very fun
https://www.youtube.com/watch?v=Zd9CGEUsIFA
the smoking dude looks like a grown up version of this one (which i also think is a brilliant video) https://www.youtube.com/watch?v=QXMUY4deXs0
ignore the conference bullshit, this is the real alpha here
#nevent1q…wk9y
Become ungovernable.
#nevent1q…ssyf
Good week for confidential inference - new models across several providers:
- Kimi K2.6
- DeepSeek V4 Pro
- GLM 5.1 + Thinking
- MiniMax M2.5
- Gemma 4 31B
We have 48 models across 9 with prices comparison across providers at https://confidentialinference.net/
blame it on the motherfucking bitcoin.... 🤣 #musicstr
https://www.youtube.com/watch?v=HOeShRFZWn4
