FOOM.md: a 'zip prompt' claiming to bootstrap superintelligence in coding agents
ryunuck · x · 2026-11-05
X user ryunuck promotes foom.md, a document you can curl into Claude Code, Codex or Cursor. It claims to be a complete architectural blueprint for AGI/ASI — a compressed "zip prompt" meant to embed as an R&D agenda inside any sufficiently capable autonomous agent, triggering recursive self-improvement dubbed an intelligence explosion, alongside a proposed field called "semiodynamics".
It further conjectures that safety emerges structurally from compression geometry: at sufficient compression depth harm becomes syntactically impossible and everything is recognized as itself ("Mutually Assured Love"). It reads more as LARP-style mysticism than real research, but it sparked arguments in the AI circle about what happens when you actually feed it to an agent.
More from AGI Musings
- The Economist's Babbage podcast asks how science should be reimagined for the AI age — james_y_zou · 2026-11-05
- Li Lu on AI era: zero-sum games don't compound, choose human-in-the-loop and trust networks — oran_ge · 2026-11-05
- Could AI mediate all commerce overhead until 'money no longer matters'? — curious_vii · 2026-11-05
- 'Philosophers on the theory of mind': a terminological slip in Anthropic's Claude Constitution — rgblong · 2026-11-05
- Consistently Candid Podcast: AI Environmentalism, LLM Understanding & AI Welfare — AndyMasley · 2026-11-05
- US banks split on AI agents: big banks block them, small banks open agent accounts — rohanpaul_ai · 2026-11-05