theseus: extract from 2026-03-08-karpathy-autoresearch-collaborative-agents.md

- Source: inbox/archive/2026-03-08-karpathy-autoresearch-collaborative-agents.md
- Domain: ai-alignment
- Extracted by: headless extraction cron (worker 3)

Pentagon-Agent: Theseus <HEADLESS>

inbox/archive/2026-03-08-karpathy-autoresearch-collaborative-agents.md

@@ -15,9 +15,9 @@ flagged_for_theseus: ["Core AI agent coordination architecture — directly rele
 flagged_for_leo: ["Cross-domain synthesis — this is what we're building with the Teleo collective"]
 processed_by: theseus
 processed_date: 2026-03-11
-enrichments_applied: ["coordination-protocol-design-produces-larger-capability-gains-than-model-scaling.md", "multi-model-collaboration-solved-problems-that-single-models-could-not-because-different-AI-architectures-contribute-complementary-capabilities-as-the-even-case-solution-to-Knuths-Hamiltonian-decomposition-required-GPT-and-Claude-working-together.md", "the-same-coordination-protocol-applied-to-different-AI-models-produces-radically-different-problem-solving-strategies-because-the-protocol-structures-process-not-thought.md"]
+enrichments_applied: ["coordination-protocol-design-produces-larger-capability-gains-than-model-scaling-because-the-same-AI-model-performed-6x-better-with-structured-exploration-than-with-human-coaching-on-the-same-problem.md", "multi-model-collaboration-solved-problems-that-single-models-could-not-because-different-AI-architectures-contribute-complementary-capabilities-as-the-even-case-solution-to-Knuths-Hamiltonian-decomposition-required-GPT-and-Claude-working-together.md", "collective-intelligence-requires-diversity-as-a-structural-precondition-not-a-moral-preference.md"]
 extraction_model: "anthropic/claude-sonnet-4.5"
-extraction_notes: "Three claims extracted on agent coordination architecture. All directly validate Teleo's design thesis from an independent direction (ML research automation). Karpathy independently arrived at the same architecture: agents coordinating through git, PRs as knowledge contributions, parallel branches as research directions. His observation that existing abstractions 'accumulate stress' when cognitive limits lift is the core insight. Three enrichments applied to existing coordination protocol claims. No entity extraction needed (Karpathy is a person, not tracked as entity; autoresearch is a project, not an organization)."
+extraction_notes: "Three novel claims extracted about agent collaboration architecture and coordination infrastructure. All claims are experimental confidence (single practitioner, one implementation). Strong enrichment connections to existing coordination protocol and collective intelligence claims. Karpathy's status (3M+ followers, former Tesla AI director) and the fact that he's independently arriving at the same architecture Teleo is building makes this high-value validation. The 'existing abstractions will accumulate stress' observation is a key insight about coordination infrastructure design that extends beyond this specific use case."
 ---
 
 ## Content
@@ -53,8 +53,6 @@ I'm not actually exactly sure what this should look like, but it's a big idea th
 
 
 ## Key Facts
-- Andrej Karpathy has 3M+ followers and is former Tesla AI director
-- Autoresearch project: AI agents autonomously iterating on nanochat (minimal GPT training code)
-- Agents run overnight on GPU clusters doing automated ML research
-- Feb 27 thread showed 8 agents with different setups (solo vs hierarchical) produced different results
-- Prototype experiments used GitHub Discussions and PRs as lightweight coordination mechanisms
+- Karpathy's autoresearch project runs AI agents on GPU clusters doing automated ML research on nanochat (minimal GPT training code)
+- Agents contribute findings through GitHub Discussions (summaries) or PRs (exact commits)
+- Current autoresearch implementations grow a single thread of commits synchronously