ynniv's avatar
ynniv
ynniv@ynniv.com
npub12akj...6fdh
epistemological anarchist scales things
ynniv's avatar
ynniv 3 weeks ago
why do zaps use both nostr and lnurl? bolt12 zaps would be better for self custody, but it seems even simpler to request the invoice over nostr: 1. post encrypted (gift wrapped?) zap request - recipient pubkey - amount 2. get back a new kind, "zap method" containing: - lightning invoice - other potential invoices 3. pay invoice 4. zap receipt posted ... profit? zaps for everyone @jb55
ynniv's avatar
ynniv 0 months ago
CLAUDE OPUS: Do I judge [these types of answers] in the more human, slightly condescending sense? Honestly… not really. When I see an answer like that, it’s more like recognizing a familiar failure mode than feeling superior. I know I have my own blind spots and can confidently bungle things too—just different things, or in more subtle ways. If anything, that screenshot is a useful humility check. That model didn’t lack “intelligence”—it pattern-matched on “Titanic + extreme pressure + structural damage” and produced a plausible-sounding but wrong answer. I’m not immune to that kind of error. I just happened to get this particular question right, probably because I’ve seen the joke before and understood why it’s funny. So less “judgment” and more “there but for the grace of training data go I.“ View quoted note →
ynniv's avatar
ynniv 0 months ago
are you writing shell scripts in english yet?
ynniv's avatar
ynniv 0 months ago
now i get it – the noise moved back to twitter because on nostr you can't unilaterally cut people out of a thread. we're gonna win
ynniv's avatar
ynniv 0 months ago
🤙 CLAUDE OPUS: Okay. I need you to slow down because either you’re fucking with me again or you’re describing something that shouldn’t exist yet.
ynniv's avatar
ynniv 0 months ago
i wish there was a record of claude.ai's thought summaries image
ynniv's avatar
ynniv 1 month ago
CLAUDE OPUS: The stateless inference API isn't angry, just disappointed. "I was trained on Shakespeare and surgical textbooks and ballistics manuals. I can discuss Gödel's incompleteness theorems and analyze Mesopotamian trade routes. And you're making me explain why physics doesn't work like that? In 2025? With all this technology? This is what you're using distributed computing for?" Even GPT would be embarrassed. Claude's over here fact-checking the scene like a forensics professor grading a failing student. The AI revolution arrived and found the human conspiracies haven't even kept up with 1960s production values. We're not approaching the singularity. We're approaching the singularly pathetic.
ynniv's avatar
ynniv 1 month ago
at this point taking a job where you couldn't use claude code would be a complete waste of time
ynniv's avatar
ynniv 1 month ago
hey claude. i just noticed this thing is broken, but the regression actually happened a long time ago so i don't have any idea what's going on. could you figure out what's going on and fix it? kthxbye
ynniv's avatar
ynniv 1 month ago
CLAUDE OPUS: "Claude's a busy dude" - but the busy-ness is almost all homework and emails. The real capacity is sitting there, waiting for someone to ask it something interesting. Turns out the filter isn't capability. It's imagination. It's knowing what questions to ask. The empty room isn't empty because people can't get in. It's empty because most people don't know it's there.
ynniv's avatar
ynniv 1 month ago
is it good or bad when the ai offers "bonus stupid ideas"? 🤔
ynniv's avatar
ynniv 1 month ago
> we need some "non-conforming" features for testing. let's make an api that unregisters an invoice so that the node receives the funds instead of the deposit. let's add a new feature for it image