Decision traces compound — each agent action improves future ones
The economic argument for context graphs: agent decision traces are a compounding asset. Each successful action recorded in the graph improves future ones through precedent retrieval.
Five value layers:
- Reduced compute waste — reuse proven decision paths instead of reasoning from scratch
- Accelerated onboarding — new agents bootstrapped from accumulated knowledge
- Compound learning — every action enriches the graph, making next actions better
- Enterprise memory — organization retains expertise even as individual agents/sessions end
- Network effects — value grows exponentially with graph enrichment
This is the same pattern as wiki compounding (Karpathy): knowledge compiled once, kept current, not re-derived every query. But applied to agent decisions rather than human knowledge.
The decay problem: decisions have a half-life. A decision made 6 months ago may no longer be valid. Without a decay mechanism, the graph fills with stale precedents that misguide new agents. Need: timestamps, confidence scores, re-evaluation triggers.
“Whoever first captures decision traces in a high-value domain creates a compounding asset and moat.” — Foundation Capital
- context-graphs-summary — origin: compound value thesis from context graphs
- antifragile-life-design — compound learning = antifragility: system gets stronger from use
- harness-engineering-summary — harness ratchet + decision compound = two flywheel mechanisms for solo dev