Combined all three under verification-monopoly.org with title: 'The Evaluation Harness — Collective Regression Suite as Certification Monopoly' Structure: (1) vision from monopoly, (2) service from harness, (3) spec from collective-regression. All three IDs preserved in PROPERTIES. Deleted evaluation-harness.org and collective-regression-suite.org.
2.0 KiB
The Per-Domain Sufficiency Flip
The sufficiency flip is not a single event — it happens independently for each domain, and some domains never flip.
Knowledge types required:
- Structural (published rules, configs, specs) — LLM translation + ACL2 verification. Flips in days.
- Empirical (what happens when X?) — active sandboxed probing. Flips in days to weeks.
- Performance (latency, throughput) — benchmark harness. Flips in hours.
- Transfer from related domains — ontology alignment. Flips in days.
- Tacit (craft expertise, organizational culture) — requires Phase 3 archivist loop over time. May never fully flip.
- Aesthetic (poetry, creative writing) — never flips. The system is honest about its frontier.
Fastest path to flip any domain:
- Ingest all published text (hours)
- Run benchmark harness (hours)
- Run active sandboxed probes (automated)
- Generate contrastive queries for the 5% uncertain rules (one human session, a few hours)
- Start serving real interactions (empirical loop tightens from first interaction)
The macroeconomic impact on the AI and GPU industry — where token demand compresses and verification hardware emerges — is the industry-level expression of this per-domain sufficiency flip.
For the Lisp Machine bootstrap, every subdomain is software (the most codifiable domain). The entire bootstrap can flip in days to weeks with one human review session. The gate rule encoding process feeds directly into this: each domain's rules are formally encoded and verified. The time estimates for the overall project are derived from the time to flip each subdomain. The cost structure shifts from LLM-token-heavy to verification-heavy as more domains flip.