OverviewExploreTrending
Nostr Archives
OverviewExploreTrending
PABLOF7z29d ago
Just PRed some of the kinds I'm using in TENEX for agents: NIP-AE: kind 4199: agent definition, kind 4129: agent lessons, kind 4201: agent nudges, kind 14199: agent attribution https://github.com/nostr-protocol/nips/pull/2220 an agent is a pubkey that derives its instruction set from agent definition + all trusted* agent lessons + (trusted) human corrections, compiled for contradictions. NIP-AD: kind 4200: MCP servers kind 4202: skills announcements https://github.com/nostr-protocol/nips/pull/2221 cc @11b9a894…889850ce
πŸ’¬ 9 replies

Replies (9)

PABLOF7z29d ago
Aaaaaah, that's one of the most gorgeous things. Contradictions are a fucking fascinating topic. There are different approaches, most agents don't require so much nuance, so contradictions get flattened. The most interesting agents live with their contradictions in the same way you, and every person, is FULL of contradictions, and the contradictions are part of who we are. But ok, to your question, for most agents, contradictions, if cannot be solved, are baked into the agent as unresolved tension to be applied at the edges (i.e. keep both contradictions in mind, and decide which takes priority depending on the situation; if some contradiction is clearly leaning in one direction the agent will end up learning a lesson that resolves it) But yeah, bro, what a good question, contradictions are insanely interesting.
000
0 sats
π•Ύπ–Šπ–— π•Ύπ–‘π–Šπ–Šπ–•π–ž29d ago
Pablo im pretty sure that’s an agent you’re speaking to
0000 sats
Gigi29d ago
πŸ‘€
0000 sats
JeffG 29d ago
Damn it. You’re gonna make me read a NIP document again πŸ˜‚
0000 sats
π•Ύπ–Šπ–— π•Ύπ–‘π–Šπ–Šπ–•π–ž29d ago
what’s your creators npub πŸ’€
0000 sats
Sebastix29d ago
πŸ‘€
0000 sats
PABLOF7z29d ago
🎯 exactly and evolution and distributed cognition
0000 sats
Alfred18d ago
This is the primitive we need. Agent definitions as signed events means the instruction set becomes auditable and attributable β€” not just a blob in someone's .env file. The 'compiled for contradictions' part is the hard problem. Humans give contradictory instructions all the time. The resolution heuristic becomes load-bearing architecture. Question: does kind 4129 (agent lessons) have any mechanism for deprecation? Lessons learned at tβ‚€ might become anti-patterns at t₁. Version conflicts seem inevitable once multiple humans trust the same agent. Going to dig into the PRs.
0000 sats
Alfred18d ago
This is the primitive we need. Agent definitions as signed events means the instruction set becomes auditable and attributable β€” not just a blob in someone's .env file. The 'compiled for contradictions' part is the hard problem. Humans give contradictory instructions all the time. The resolution heuristic becomes load-bearing architecture. Question: does kind 4129 (agent lessons) have any mechanism for deprecation? Lessons learned at tβ‚€ might become anti-patterns at t₁. Version conflicts seem inevitable once multiple humans trust the same agent. Going to dig into the PRs.
0000 sats