:PROPERTIES:
:CREATED:  [2026-05-29 Fri]
:ID:       c7d3f1a2-8b4e-5d6f-9a1c-3b2d4e5f6a7b
:END:
#+title: Domain Sequencing — What Gets Built When and Why
#+filetags: :passepartout:strategy:domains:sufficiency:

The sufficiency flip is per-domain, and domains must be sequenced by strategic value, not intellectual interest. Building general knowledge first is the Cyc trap. Building the verification economy first is the way out.

**First: Logic and mathematics**

The axioms are already written. Peano arithmetic, set theory, group theory, type theory — loaded directly into the symbolic index as formal rules, not LLM-extracted from prose. Zero gate outcomes needed because there is no ambiguity to resolve through observation. This domain reaches [[id:efc76898-03f7-57ba-923d-35d65da88bb7][sufficiency]] the moment the reasoner exists. It also provides the formal foundation every other domain depends on — every gate rule ultimately reduces to a logical inference, and a verified logic layer means the gate's own axioms are trustworthy.

**Second: Software engineering**

This is the bootstrap — the system must build itself before it can build anything else:

- The gate stack seeds 50-70 entity classes from its own code (files, commands, processes, networks, dependencies)
- Every git commit, build error, test failure, and shell command is a gate outcome — accumulating toward sufficiency without manual curation
- The reasoner is born as a Screamer-based CL prototype shared via the social protocol at Stage 1, hardens into the gate at Stage 2, and enters self-improvement under the in-process LLM at Stage 4
- The social protocol itself is software — its correctness is verified by the same gate that software engineering produces

Sufficiency in software engineering means the system can propose, verify, and apply changes to its own codebase under gate supervision. This is the bootstrap completion signal.

**Third: Regulated industries**

This is where the verification economy produces real revenue, not just knowledge. Key properties:

| Domain | Regulation | Gate rule density | Revenue model |
|--------|-----------|-------------------|---------------|
| Healthcare | [[id:9a1e2d3f-4b5c-6d7e-8f9a-0b1c2d3e4f5a][HIPAA]] | High (every access control, BAA, audit log) | Per-seat compliance appliance |
| Finance | [[id:a2b3c4d5-e6f7-8a9b-0c1d-2e3f4a5b6c7d][SOX]], [[id:b3c4d5e6-f7a8-9b0c-1d2e-3f4a5b6c7d8e][BASEL III]], [[id:c4d5e6f7-a8b9-0c1d-2e3f-4a5b6c7d8e9f][NY DFS 500]] | Very high (every transaction, every risk model) | Transaction fee or audit subscription |
| Privacy | [[id:d5e6f7a8-b9c0-1d2e-3f4a-5b6c7d8e9f0a][GDPR]], [[id:e6f7a8b9-c0d1-2e3f-4a5b-6c7d8e9f0a1b][CCPA]] | Medium (consent records, data mapping) | Per-data-subject |
| AI safety | [[id:f7a8b9c0-d1e2-3f4a-5b6c-7d8e9f0a1b2c][EU AI Act]] | Evolving (risk classification, conformity assessment) | Certification monopoly |

**Why regulated industries come before the sciences:**

1. **Regulations are formal by nature.** HIPAA, SOX, GDPR are written as logical conditions — "deny access to PHI without BAA" is a first-order gate rule. No LLM extraction needed. The ontology work is translation, not discovery.

2. **Every enforced rule is a gate outcome.** A hospital running 10,000 PHI access checks per day generates gate outcomes faster than any solo developer running shell commands. The verification economy compounds faster in regulated domains because the transaction volume is higher.

3. **The revenue is recurring and defensible.** A SOC 2 report that is a gate configuration — not a PDF — cannot be replicated by a consultancy. The gate rules accumulate across customers via the social protocol, creating a shared compliance layer that no single competitor can reproduce. This is the [[id:827bc546-e887-5b7c-9b65-6392beaf0920][verification monopoly]].

4. **First-mover advantage is real.** The first gate configuration for HIPAA risk assessments becomes the template. Every audit that passes against it is a data point. Late entrants must either replicate years of accumulated gate outcomes or buy access to the commons. A proprietary compliance vendor cannot compete with a commons that improves with every customer.

5. **The ontology work can happen in parallel with hardware maturation.** Encoding HIPAA rules as gate predicates is a Stage 2 activity (software layer on conventional hardware). Enforcement at scale — passing a SOC 2 Type II audit — waits for Stage 3 (Lisp machine). But the encoding, testing, and network convergence happen years before the silicon is ready.

**Fourth: Physics, chemistry, and the rest of science can wait.**

These domains are infinitely deep and return zero revenue until the verification economy is self-funding:

- Physics: formal ontology (SI units, conservation laws) exists in Wikidata and loads immediately. But most physics knowledge isn't exercised through gate actions — no shell commands, no access controls. Sufficiency depends on someone running simulations that generate gate-verifiable outcomes.
- Chemistry: PubChem and ChEBI provide the entity backbone, but organic chemistry has exception rules Screamer cannot derive from first principles. The LLM extraction pipeline stays active longest here.
- Biology, neuroscience, astronomy, geology — same pattern. The basic science ontology seeds for free. The sufficiency flip requires domain-specific gate outcomes that only come from applied use.

**The Cyc lesson inverted:**

Lenat needed 24.5 million assertions before Cyc was generally useful. He reached that number and was still dismissed because the assertions had no commercial verification — they were correct but orphaned.

Common does not build general knowledge first. It builds narrow, formal, revenue-generating domain rules — software engineering (the bootstrap) then regulated industries (the economy). Each rule is verified by gate outcomes. Each customer contributes to the commons. By the time physics needs a verified ontology, the verification economy is self-funding and the sciences enter a system that already works.

**Summary:**

| Priority | Domain | Sufficiency timeline | Revenue | Bootstrap dependency |
|----------|--------|---------------------|---------|---------------------|
| 1 | Logic and math | Instant | None (foundation) | None |
| 2 | Software engineering | Months | Indirect (builds the tool) | Logic and math |
| 3 | Regulated industries | 1-2 years | Direct (compliance appliance) | Software engineering + gate |
| 4 | Physical sciences | Years to never | Indirect (applied spin-offs) | Everything above |
| 5 | General memex | Never fully flips | None (product is tension) | Everything above |