Deck B · Truth Units

A reasoning cache for enterprise AI.

Twelve slides. For funders, buyers, and design partners.

PoSoT · Deck B 2026-05-20 branch · feat/posot-mode-a 1920 × 1080 · letterboxed

1. Title — Truth Units

   A reasoning cache for enterprise AI.

2. The problem

   Every LLM query is a fresh act of reasoning — even when the corpus hasn't changed.

3. The insight

   The reasoning should be cached. The LLM call shouldn't repeat.

4. The Truth Unit

   Three locked components. Built once per domain. Re-used forever.

5. The cost curve

   Same workload, two architectures. 3–5× cost reduction at scale (N ≥ 1k, stable-precedent domain).

6. Domain-agnostic substrate

   Three requirements: a real graph, expert consensus, recurring queries.

7. Proof in law

   Dormant Commerce Clause as the canonical Truth Unit — locked, frozen, falsifiable.

8. How buyers deploy

   Five plays. Same primitive. Different domain corpora.

9. Two decks, one substrate

   Deck A earns the credibility. The substrate is what scales.

10. Where we are

    Substrate proven. Domain proof in flight. Buyers next.

11. The ask

    Three ways to engage — design partner, capital, reviewer.

12. Appendix — A vs B

    Same substrate. Two audiences. Different burden of proof.

Companion deck: Deck A — PoSoT in law (peer reviewers). Same substrate. Skeptic-first framing.