Authority survives time.
Responsibility survives authority.
Consequences survive responsibility.
Risk survives organizations.
Evidence survives institutions.
Every signed proof in the H33 corpus earns one of these lines. Each line names something that survives the failure of the line above it. Four lines are proven. One is the capstone, waiting for a future proof to land.
Each rung names a thing organizations care about — and what it survives. Below each rung sit the signed, replayable proofs that earned it. Read top to bottom; the claim grows with each line.
The cryptographic floor that makes survival mathematical, not operational.
Every higher rung depends on signatures that remain valid against quantum adversaries. Without a post-quantum substrate, every claim above this one is operational, not mathematical — it survives only as long as the cryptography that produced it survives. H33 chose ML-DSA-65 + FALCON-512 + SLH-DSA-SHA2-128f from the start. The substrate does not "survive X" — it is what makes the four survival lines above possible at all.
scif-backend, not in any vendor serviceAuthority survives time.
Most systems record organizational activity. H33 reconstructs organizational reality. Years later, you can replay the canonical auth event log byte-identically and recover the exact authority graph that was in force at any prior moment — even if the vendor that produced it has disappeared. The chain does not depend on H33 still existing.
Responsibility survives authority.
Many systems can tell you who had permission. Very few can tell you who remained accountable after the permission disappeared. Even after authority is revoked, expired, suspended, or the holder leaves the organization, the responsibility chain that surrounded the decision at the moment of execution is permanently recoverable — actor, supervisor, policy owner, model owner, asset owner, delegated_from. Seven typed roles per decision.
Consequences survive responsibility.
Most systems record actions. Very few can reconstruct consequences. Even after the actor responsible for a decision is gone, the consequences that decision produced — losses, claims, lawsuits, regulatory actions, settlements, payouts, fraud events, breaches, operational failures — remain replayable. Each consequence is typed (9 kinds), severity-scored (6 levels), and tied back through triggering_decision_ids to the full responsibility chain.
Risk survives organizations.
When one enterprise absorbs another — through acquisition, merger, divestiture, roll-up, or carve-out — the open consequences cross the organizational boundary with the balance sheet. Lawsuits not yet settled. Regulatory actions not yet closed. Operational failures not yet remediated. Impairments not yet written down. show_inherited_risk(acquisition_001) returns the inheritance picture deterministically, five years after close. Every inherited consequence is tagged with origin_enterprise_id — the diligence reader can tell, line by line, whether the consequence originated before or after the acquisition.
Evidence survives institutions.
First Institutional Death Replay (#184) earns this line. The customer enterprise dissolved in 2031; the court trustee replays the chain in 2035; Claim #84711 still reconstructs; state_id at T=2035 is byte-identical. The institution died. The evidence did not. The continuity ladder is now complete in proven form.
The ladder above is one axis: continuity of evidence. After the capstone landed, a second axis emerged off the side: continuity of computation. The corpus question on this axis is different — not "What survives change?" but "Why did the computation happen?"
A future proof earns "Reasoning survives systems" only when it demonstrates a sharper claim than #174 + #167 already prove:
"Given the preserved reasoning substrate, the same system can re-run the reasoning path — not merely reconstruct the decision object."
Most enterprise software vendors prove one rung. A subset prove two. Almost none make a continuity ladder coherent, and the ones that try usually frame each rung as a separate product. Here, the same canonical event log, the same replay engine, the same signing stack, and the same independent verifier carry every line — from "Authority survives time" all the way down. The substrate is shared. The story is one story.
Authority can be revoked — responsibility survives. The responsible actor can leave — consequences survive. The responsible enterprise can be acquired — risk survives. The institution itself can dissolve — evidence survives. The ladder is structurally hierarchical: each rung is a deeper continuity claim than the one above it.
Cyber / AI buyers care about rungs 2–3 (authority and responsibility). Insurance / Legal buyers care about rungs 3–4 (responsibility and consequences). PE / M&A / Fund Admin / Big 4 buyers care about rungs 4–5 (consequences and risk). Regulators and auditors care about all five. The ladder lets one product serve all those conversations without forcing one frame on every buyer.
No rung is a marketing claim. Every rung is anchored to one or more proofs, each with a state_id that anyone can reconstruct independently with the public verifier. The ladder is not a positioning exercise — it is a list of mathematical claims, each backed by replay.
"Evidence survives institutions" is the endpoint. After it is earned, every future proof has to either deepen one of the five lines (more institutional-death scenarios, more cross-boundary inheritance patterns) or extend the ladder downward. The bar is no longer "another replay scenario." The bar is "does it move the endpoint?"
Issued by H33, Inc. · Eric Beans, CEO · 2026-06-03
This page is the map of the corpus. It does not introduce any new claim of its own — every line on it is anchored to one or more signed, independently-replayable proofs already on this site. The fifth line ("Evidence survives institutions") is the endpoint of the thesis, and is explicitly marked as the capstone that a future proof must earn.