Articles
Notes on belief states, dissonance and reasoning under uncertainty.
The Physics of IntelligenceIntelligence is not a mapping learned from data. It is a state that evolves under forces. An overview, and the start of a series.
FAQCognitive infrastructure, belief states, dissonance and how to build with them.
Cognitive infrastructure, in one pageReasoning needs state. Cognitive infrastructure keeps it.
What you can do with a belief stateInterrogate it, listen to it, diagnose it, analyze its coherence.
APIReasoning primitives you compose around your own data.
Humans: conversations on a belief stateA conversation that keeps what it concluded, and shows why. Where agents get agencyAgents that act from what they believe, not from prompt files.
Tax: phase-outs, thresholds and a yearly checkOne IRS publication, nested conditions, and rules that expire every January.
Tax: answerable means extractedTwo runs, 44 to 46 beliefs, two missing tax bands, and a headline question taken off the shelf.
Security: one zero-day across 9 public sourcesA critical firewall CVE, four vendor advisories, and guidance that reversed itself. Security: scope boundaries and one fact said four waysWho is exposed, who is not, and what 72 beliefs from nine sources get right and repeat.
Trading: a strategy desk's decision journalSix dated entries, two crossover signals, and a position that changed for stated reasons. Trading: base rates against the desk's own backtestTwo kinds of evidence for one decision, and which numbers an answer is allowed to cite.
43 AI agents and zero shared realityThe next enterprise AI primitive is not another chatbot, copilot or agent. It is a belief layer.
The problems we're trying to solveLLMs excel when the world is explicit. The interesting problems are implicit.
Belief-conditioned agencyAgency from a persistent belief state, without the harness.