§ ── THE SUBSTRATE ── ARCHITECTURAL DOCTRINE ── v1.0
Probabilistic computation needs
deterministic continuity.
The substrate is the layer that constrains entropy in institutional reasoning. It exists because modern AI introduces non-determinism, model volatility, schema mutation, and confidence drift — and companies require coherent, replayable, governable decision-making despite all of that. Diana Hu's YC framing names the destination: AI as the operating system the company runs on, not a tool layered on top. The substrate is what makes that operational. Read end-to-end. The audience for this page is the audience for the product.
§ 01 ── THE CRISIS BEING SOLVED
Companies are open loops + institutional reasoning is fragmented.
The hidden crisis is not data fragmentation. Fragmentation is the symptom. The crisis is that organizational reasoning is irreproducible — decisions made in 2024 cannot be reconstructed in 2026. The trade-off rationale lived in one Slack thread. The override logic lived in a calendar invite. The customer pain signal lived in a Zoom call no one transcribed. None of it survives team turnover or model upgrades.
01 · COMPANIES LOSE INSTITUTIONAL MEMORY
Rationale, intent lineage, scoring logic, failed-experiment memory, leadership judgment continuity. Most companies rebuild institutional memory every 3–5 years. The reasoning lives in routing-humans up and down the org chart — and walks out the door when those humans do.
02 · REGULATORS AND CUSTOMERS SCRUTINIZE
EU AI Act audit defensibility is increasingly formal. NIST AI RMF · SOC 2 · ISO 27001 · customer data-residency · and internal AI-use policies all require lineage that doesn't exist in current stacks. Architectural controls demanded, not policy controls.
03 · MODELS CHURN
Every ~6 months a new frontier model arrives. The prompts that worked stop working. The fine-tunes become stale. AI investment depreciates.
Companies need a layer that survives all three. That layer is the substrate.
§ 02 ── POST-SAAS · WHY THIS NOW
SaaS fragmented institutional reasoning. Substrate recombines it.
Modern companies don't lack software. They have Slack, Linear, GitHub, Notion, Salesforce, Zoom, Datadog, HubSpot, and twenty more. The constraint is not capability access. The constraint is continuity, coherence, and shared reasoning state. SaaS sliced institutional cognition into 30 application silos. Each silo owns its truth. None of them compose. Routing-humans glue them together at the cost of velocity — every layer of human routing is, as Jack Dorsey puts it at Block, a direct speed loss.
Substrate architectures don't replace SaaS — they recombine the reasoning that SaaS fragmented. One queryable operational state. One audit chain. One overlay that defines what your company knows, encoded as version-controlled YAML. Almost no human middleware in routing. Humans at the edge guiding. Agents in the middle synthesizing. This is the architectural transition modern infrastructure goes through every 20 years. Git did it for code. AWS did it for infrastructure. Stripe did it for payments. Snowflake did it for data. We do it for organizational reasoning.
§ 03 ── FIVE DISCIPLINES · NON-NEGOTIABLES
Five disciplines. Architectural, not declarative.
Five non-negotiable disciplines define what makes the substrate substrate-grade. Generic AI tools have none of these. Substrate-grade architecture has all five, architecturally enforced — not asserted as policy.
#
DISCIPLINE
OPERATIONAL DEFINITION
D1
AUDIT CHAIN BY DEFAULT
Every substantive action audit-emitted. Sole-emission authority per tier (one Audit Curator agent per tier · chain integrity by construction, not policy). Hash-chained · severity-classified · NEVER-deletable for severity:critical events at filesystem level. F4 is L0 doctrine · always applies.
D2
L8 ARCHITECTURAL FLOOR
Commitments that cannot be lifted by policy under pressure. Encoded as code-path absence — the escalation path simply does not exist in the substrate's code. L8-Enforcer surveils continuously · refusal events emit severity:critical · self-tests quarterly. 6 canonical floors F1-F6 + 0-3 optional F7-F9 per customer pattern. Floor amendment requires MAJOR cycle + domain-architectural redesign evidence.
D3
TETHER-PAIR DISCIPLINE
Every abstract claim ships paired with a concrete operational mechanism. The phrase doesn't ship without the tether. Policy without architecture is rejected at authoring time. Visible on this page in § 12.
D4
3-LAYER ONTOLOGY
Methodology · Substrate · Model never blurred. Methodology is the institutional invariant. Substrate is the continuity layer. Model is the replaceable execution engine. L0 / L1 / L2 IP boundaries enforced contractually and architecturally.
D5
SUBSTRATE-MIRRORING (I15)
Every commitment the substrate enforces on customers applies recursively to the substrate's own operations. The operator (us) is subject to the same audit chain · L8 floors · κ verification · D7 §12 inspection that the customer is. The substrate has no outside.
Without these five, “substrate” is marketing. With all five, “substrate” is what the word claims.
§ 04 ── INFRASTRUCTURE PATTERN
Every enduring infrastructure company solves an entropy problem.
FIG.01 — Six entropy-solving infrastructure companies · 2005 → 2026 · The Queryable Company solves institutional reasoning entropy.
FIG.01 — Six entropy-solving infrastructure companies · 2005 → 2026 · The Queryable Company solves institutional reasoning entropy.
COMPANY
ENTROPY SOLVED
YEAR
GIT
Codebase entropy
2005
AWS
Infrastructure entropy
2006
STRIPE
Payments complexity entropy
2010
SNOWFLAKE
Data fragmentation entropy
2014
LINEAR
Workflow entropy
2019
QUERYABLE.COMPANY
Institutional reasoning entropy
2026
Feature companies sell capability. Infrastructure companies solve entropy. The strongest companies in the history of software are entropy-solvers — they exist because some critical organizational resource accumulates disorder under usage, and they impose deterministic order. Reasoning under probabilistic computation is the next entropy frontier — and the substrate is what carries reasoning across model generations.
§ 05 ── ONTOLOGY HIERARCHY · L0/L1/L2
Three layers. Three IP boundaries. Stable across all surfaces.
The architecture is canonically defined by three layers (Methodology · Substrate · Model). The IP boundary is canonically defined by three levels (L0 · L1 · L2). The relationship between them is the source of the manifesto.
FIG.02 — L0 universal doctrine · L1 domain framework (TQC · Fund AI OS) · L2 per-customer instance.
FIG.02 — L0 universal doctrine · L1 domain framework (TQC · Fund AI OS) · L2 per-customer instance.
ONTOLOGY · WHAT EACH LAYER IS
METHODOLOGY
Institutional invariant. Your company's reasoning — strategy, planning frameworks, decision rights matrix, performance criteria, customer engagement playbooks, engineering principles, OKR composition, postmortem norms. What persists across everything else.
SUBSTRATE
Continuity + governance layer. The layer that preserves methodology across model evolution. Audit chain, consent gateways, evidence-mode framework, schema migration replay, customer overlay system. The queryable operational state.
MODEL
Probabilistic execution engine. The current frontier LLM running the reasoning. Replaced every ~6 months by the next one. The substrate makes this replacement non-destructive to methodology. Replay-tested on every swap.
IP BOUNDARY · WHO HOLDS WHAT
LEVEL
SCOPE
IP STAYS WITH
L0
Universal substrate doctrine. Five disciplines · audit chain · L8 mechanism · tier framework · κ verification · D7 §12. Genuinely domain-agnostic.
The framework itself. Open methodology.
L1
Domain-specific framework. The Queryable Company is one L1 (companies). Fund AI OS is another (specialist VC funds). Each L1 encodes how the domain's specialists actually reason.
Queryable.Company (operator IP). L1 frameworks travel across customers in the same domain.
L2
Per-customer instance. customers/[org].yaml encodes your specific methodology, decision rights, scoring weights, L8 floor calibrations, voice tiers.
The customer. Your overlay is yours. Versioned. Portable. Owned. The MSA § 8.3 spells it out.
Methodology is the invariant. Models are replaceable execution layers. The substrate carries continuity across model volatility — and the L0/L1/L2 boundary carries IP clarity across the engagement.
§ 06 ── OBSERVE → EXPLAIN → CONSTRAIN → GOVERN
Observe → Explain → Constrain → Govern.
Infrastructure systems evolve through four stages. Most current AI deployments are at stage 1 or 2. The substrate operates at stage 4.
FIG.03 — Four-stage architectural progression · most AI deployments stop at 02 · the substrate operates at 04.
FIG.03 — Four-stage architectural progression · most AI deployments stop at 02 · the substrate operates at 04.
STAGE
NAME
DESCRIPTION
01
OBSERVE
System captures what happens. Logs, metrics, events. Current state for most AI bolt-ons.
02
EXPLAIN
System reconstructs why something happened. Replay, evidence lineage, audit chain. Current state for mature audit-aware AI.
03
CONSTRAIN
System refuses outputs below evidence thresholds. Tier caps prevent autonomous action where consequences are highest. Confidence escalation routes uncertain outputs to humans. Humans at the edge; agents constrained by code-path absence in the middle.
04
GOVERN
System shapes what decisions are permissible. L8 architectural floors · authorization topology · schema lock discipline · reversibility by design. The substrate doesn't just explain past decisions — it constrains future ones. The org chart of routing-humans is replaced by an intelligence layer that humans guide from the edge.
Defensibility is retrospective. Governance is prospective. Most infrastructure stops at stage 2. The substrate operates at stage 4 by design.
§ 07 ── FOUR-TIER FRAMEWORK · HUMANS AT THE EDGE
Every agent in the substrate has a declared tier ceiling. Humans at the edge guiding.
The substrate's 12 agents (7 Class A customer-operational + 5 Class B architectural) each operate at a declared tier ceiling. For agents capped by L8 floors, the higher-tier code path does not exist in the substrate runtime. Tier ceilings are architectural, not policy. Per Jack Dorsey at Block: the intelligence layer routes information; humans at the edge guide the company. The tier framework is how that division is enforced architecturally.
FIG.04 — Authority ladder · 12 agents · tier ceiling architecturally enforced.
FIG.04 — Authority ladder · 12 agents · tier ceiling architecturally enforced.
TIER
NAME
AUTHORITY
EXAMPLES
T1
OBSERVER · 0 ACTIVE
Read-only · reports findings · cannot draft outputs.
(none currently in canonical fabric · reserved for future compliance roles)
T2
DRAFTER · 7 AGENTS
Drafts outputs for human review · cannot commit · advisory only.
All 7 Class A · Customer-Operational: Decision Brief Drafter (F3 cap) · Risk & Compliance Lead (F3 cap) · Customer Operations Lead (F2 cap) · Finance Operations Lead (F1 cap) · Knowledge Curator · Strategy Lead · Engineering Steward (F6 cap · F7 if unregulated).
T3
OPERATOR · 0 ACTIVE
Drafts AND executes defined operations within bounded scope. Deliberate architectural choice: substantive decision-making remains with humans at the edge.
(none in canonical fabric · architectural choice)
T4
STEWARD · 5 AGENTS
Holds authority over substrate discipline itself · L8 enforcement · audit chain integrity · versioning.
All 5 Class B · Architectural: Audit Curator (F4 sole-emit) · L8-Enforcer · Registry Maintainer · Eval Suite Runner · Version Controller.
No agent self-promotes across the tier ceiling. The lifting code does not exist. Where Diana Hu's YC framing names "almost no human middleware," the tier framework is what makes that architecturally honest — agents route within their tier · humans judge at the edge where consequences cross thresholds. The absence of T3 OPERATOR in the canonical fabric is deliberate.
YEAR-2 TRAJECTORY · 6-LOOP ACCRETION
Phase 1 starts with 12 agents in 1 loop. Phase 3 quarterly cadence accretes loops 2-6. Each loop adds 2-3 Class A agents. Class B stays at 5 (architectural · stage-invariant). By Year 2: ~25 agents across 6 loops · all operating in the same substrate · one audit chain · κ ≥ 0.85 maintained.
§ 08 ── LLMs AS COMPILERS · SOFTWARE FACTORIES
LLMs operate as compilers over typed specifications. Not as autonomous deciders.
The architectural choice that makes everything else possible: the frontier model is not given free-form authority over institutional decisions. It compiles structured intent — defined in the customer overlay — into structured outputs that flow through the audit chain. The model is execution, not judgment.
ABSTRACT CLAIM
OPERATIONAL TETHER
Constrained reasoning
LLM call wrapped in schema-validated input + output. Free-form outputs rejected by the validator. Every call audit-chain-logged with overlay version.
Typed specifications
customer.yaml defines decision rights, scoring axes, thresholds, override conditions, evidence weighting. The model reads this at bootstrap. Cannot exceed it.
Audit-chain-logged
Every model invocation: timestamp + agent ID + overlay version + model version + input hash + output hash + evidence set referenced. Append-only. Sole-emission per tier via Audit Curator.
§ SOFTWARE-FACTORY PATTERN
The compiler framing is also what makes AI software factories possible (per Diana Hu's YC framing: TDD's next evolution). Humans write specs and tests; agents generate code that iterates until tests pass. The Engineering Steward operates the factory at T2 with F6 ceiling (or F7 if customer is unregulated · per Customization Decision Tree) — no autonomous merge to main. The 1000x-engineer pattern is not a single agent that codes faster; it is a substrate that surrounds an engineer with spec / test / iterate capability under audit, with humans approving at the edge.
§ 09 ── THE SEVEN-STEP CLOSED LOOP
Observe → Define → Plan → Execute → Measure → Reconcile → Recalibrate.
Seven steps. Every step traceable. Every action graded against reality. Loop latency under 5 minutes for standard workflows. The substrate doesn't run open-loop — outputs are continuously validated against observed outcomes, and the methodology overlay updates when calibration drifts. Diana Hu's YC framing: every important process in the company captured by an intelligent closed loop. This is what that looks like architecturally.
FIG.05 — Seven-step closed loop · agent-attributed · RECALIBRATE → OBSERVE returning arc closes the loop.
FIG.05 — Seven-step closed loop · agent-attributed · RECALIBRATE → OBSERVE returning arc closes the loop.
RECONCILE detects drift between projected and actual outcomes. RECALIBRATE feeds back into the methodology overlay when measured drift exceeds threshold. Structural replay-test verifies substrate integrity on every model swap. Your company's reasoning improves under load, not in spite of it.
§ 10 ── DIVERGE-AND-RECONCILE
Load-bearing decisions are solved twice.
For any decision where being wrong is expensive — substantive strategic commitments, customer escalation responses, financial commits above threshold, hiring decisions with material consequence — the substrate solves the problem twice via independent routes. Different reasoning paths, different evidence sets, different model invocations. Then compares.
If both routes agree (within threshold), the decision proceeds with both traces logged. If they disagree, the substrate stops and surfaces — never silently picks one. Hidden disagreement is the most expensive failure mode in AI deployment. We refuse to allow it. Methodology Council adjudicates persistent divergence.
FIG.06 — Diverge-and-reconcile · two routes · agreement proceeds with both traces · disagreement surfaces to humans.
FIG.06 — Diverge-and-reconcile · two routes · agreement proceeds with both traces · disagreement surfaces to humans.
SCHEMATIC · DIVERGE-AND-RECONCILE EVENT
# Schematic: Class A drafter diverge-and-reconcile event decision_type: [SUBSTANTIVE_CLASS_PER_DECISION_SCORER] class_a_drafter: [AGENT_NAME] route_A: method: [PRIMARY_SYNTHESIS_PATH] evidence_set: [bounded source set 1] output: claim_A · confidence_band: [bounded] route_B: method: [COUNTER_EVIDENCE_SEARCH + PRIOR_PATTERN_MATCH] evidence_set: [bounded source set 2 · disjoint or partial overlap] output: claim_B · confidence_band: [bounded] reconcile: agreement: [true | false] delta: [computed] action: [PROCEED_BOTH_LOGGED | SURFACE_TO_HUMAN | METHODOLOGY_COUNCIL_REVIEW] routing: [per overlay decision rights matrix]
Full event schema + per-agent diverge-and-reconcile triggers available under NDA — see § 18.
§ 11 ── CONFIDENCE TOPOLOGY
Trust is distributed, not scalar.
Every output carries explicit evidence mode labels. When you read a substrate answer, you see what kind of trust produced it. Calibrated trust is bounded uncertainty.
FIG.07 — Six evidence modes · trust character per mode · contradictory triggers reconcile-flow.
FIG.07 — Six evidence modes · trust character per mode · contradictory triggers reconcile-flow.
EVIDENCE MODE
TRUST CHARACTER
EXAMPLE
[verified]
Deterministic
Specific Slack / Linear / GitHub citation inline.
[inferred]
Partially-bounded
Cross-source pattern match across N events.
[model-derived]
Probabilistic
LLM synthesis without specific grounding.
[stale]
Time-degraded
Source older than freshness threshold.
[contradictory]
Requires-resolution
Counter-evidence detected · reconcile-flow per § 10.
[human-override]
Authoritative-with-audit
Explicit human override logged with rationale.
Every answer decomposes into the trust-state distribution that produced it. “72% verified · 18% inferred · 8% model-derived · 2% contradictory-flagged.” Trust is bounded, not assumed.
§ 12 ── TETHER-PAIR DOCTRINE TABLE
Every abstract claim ships with operational tether.
We have a discipline that prevents drift into systems philosophy. Every category-naming abstraction is paired with a concrete operational mechanism. The phrase doesn't ship without the tether. Below: the doctrine table that governs every page of this site.
ABSTRACT CLAIM
OPERATIONAL TETHER
Epistemic trust
Evidence mode labels [verified] / [inferred] / [model-derived] with confidence bands on every output.
Replayability
restore_to(audit_id) reproduces past state bit-for-bit against past overlay version. Structural replay-test verified on every model swap.
Institutional cognition
YAML overlay loaded at agent bootstrap · audit-chain-tagged on every read.
Bounded uncertainty
Escalation thresholds · low-confidence outputs route to humans, never auto-execute.
Governance
L8 architectural floors · permission topology · refusal-under-low-confidence · L8-Enforcer (T4 Steward · recursively self-protected) · quarterly self-tests (6/6+ PASS for F1-F6 canonical · plus 0-3 optional F7-F9).
Continuity
Schema migration replay tests · past decisions survive overlay updates. MAJOR/MINOR/PATCH versioning · stakeholder concurrence for MAJOR.
Anti-fragility
Every replay improves calibration · κ ≥ 0.85 tracked weekly · pattern migration L2 → L1 across the cohort.
Confidence topology
Evidence-state distribution across verified / inferred / model-derived / stale / contradictory / human-override.
Humans at the edge
Tier framework (T1–T4) · L8 floor caps encoded as code-path absence · no agent self-promotes across its ceiling.
Substrate-mirroring (I15)
Every commitment applies recursively to our own operations. We operate under the same audit chain · same L8 floors · same κ verification · same D7 §12 inspection. The substrate has no outside.
If we ever ship a phrase without a tether, we've drifted into theater. This doctrine prevents that.
§ 13 ── D7 §12 VERIFICATION RIGHT
Customers, auditors, and regulators have standing inspection authorization. Five steps. Architecturally cooperative.
Every stakeholder of the substrate retains standing authorization to invoke a five-step verification protocol at any time. Reasonable notice for coordination-required steps (typically 3–5 business days); immediate for the others. The substrate is built to cooperate with verification, not to resist it. D7 §12 is canonical per L0 doctrine. Standing authorization for board · investors · regulators · acquirers · customer Methodology Council seats. Phase 0 W3-W4 D7 §12 inspection drill validates protocol operational before Phase 1 KPI gate.
FIG.08 — D7 §12 verification protocol · five steps · standing authorization for stakeholders.
FIG.08 — D7 §12 verification protocol · five steps · standing authorization for stakeholders.
STEP
INSPECTS
NOTICE
PASS CRITERIA
01 · INSPECT
Read access to substrate + audit chain. Any agent contract · any audit chain entry · methodology overlay · L8 floor implementations · Council decisions log · KPI evidence bundles.
IMMEDIATE / 3–5 BD CROSS-TIER
Substrate state queryable end-to-end. No hidden surfaces.
02 · REPLAY
Re-execute any past decision via audit chain query. Decision context · evidence set · 5-lens scorer output · diverge-and-reconcile trace · approver routing reproduced bit-for-bit.
IMMEDIATE (≤15 MIN SLA)
Audit Curator + Knowledge Curator deliver queryable interface. Replay matches historical output.
03 · VERIFY
Confirm code-path absence for each L8 floor. L8-Enforcer's quarterly self-test logs reviewable. Pressure-test refusal traces available on demand.
IMMEDIATE
Engineering Steward + Eval Suite Runner walkthrough. Lifting code provably absent.
04 · CROSS-CHECK
Verify κ against Methodology Council-ratified baseline. Eval Suite Runner re-computes κ on sampled decisions.
IMMEDIATE
Tolerance ±0.02 PASS · ±0.05 WARNING · beyond CRITICAL.
05 · ESCALATE
Route critical findings to operator-tier recursive anchor. Authority on bypass-discipline refusal · floor breach · F4 sole-emit violation · κ collapse.
72H CRITICAL / 1W WARNING
Escalation traceable end-to-end. Operator response logged.
Trust-by-architecture is verifiable. Trust-by-policy is asserted. We build the verifiable kind.
§ 14 ── SCHEMA MIGRATION
Overlay versioning replays bit-for-bit against historical decisions.
FIG.09 — Overlay semver · notice windows · MAJOR migration carries κ ≥ 0.85 replay-test gate.
FIG.09 — Overlay semver · notice windows · MAJOR migration carries κ ≥ 0.85 replay-test gate.
01 · OVERLAY VERSIONED
Every customer overlay versioned. MAJOR/MINOR/PATCH cycle per semver. PATCH (24h notice · no semantic change). MINOR (7-day notice · backward-compatible additions). MAJOR (30-day notice · stakeholder concurrence required).
02 · TAGGED AT WRITE
Every audit chain entry tagged with the overlay version active at write time. Replay against any historical entry uses the historical overlay version.
03 · REPLAY-TESTED
MAJOR migrations require structural replay-test against last 100 substantive audit entries. If replay outputs diverge by more than threshold (κ < 0.85 against historical outputs), migration is blocked. Replay-test infrastructure available under NDA — see § 18.
§ 15 ── PATTERN MIGRATION · NETWORK EFFECT
L2 patterns migrate up to L1 with concurrence and attribution.
01 · CONCURRENCE GATED
A pattern proven in customer L2 that may generalize to L1 requires explicit customer concurrence. The customer can decline. If granted, attribution is preserved. Concurrence gate is architectural, not policy.
02 · ATTRIBUTION + COMPENSATION
Customers whose L2 patterns migrate to L1 receive attribution in the L1 framework documentation + compensation per their MSA pattern-migration clause (typically: pricing credit · framework attribution · or both).
03 · NETWORK EFFECT
L1 frameworks improve over time as L2 patterns migrate up. Every customer in the cohort benefits from every other customer's L2 → L1 migrations. The substrate compounds across the cohort. Loop composition is multiplicative across the cohort, not additive within a single customer.
Cohort-level intelligence (k-anonymity ≥ 5 floor) activates once the cohort reaches scale. Cross-engagement κ benchmarks · decision archaeology · regulatory pattern propagation. Details under NDA — see § 18.
SEE FIG.10 BELOW · COHORT NETWORK EFFECT VISUALIZED INLINE WITH ANTI-FRAGILITY
§ 16 ── ANTI-FRAGILITY (COMPOUNDING)
The substrate compounds. AI tools decay.
Most AI deployments decay over time. Prompts age. Models churn. Fine-tunes go stale. Datasets drift. The marginal value of an AI tool today is lower than it was 18 months ago when you deployed it.
The substrate compounds. Every replay improves calibration. Every decision deepens the institutional record. Every L1 pattern migration improves the framework for everyone in the cohort. Your company's reasoning gets more queryable, more defensible, and more inspectable with time and use.
FIG.10 — AI tools decay · substrate compounds · cohort network effect visualized below the chart.
FIG.10 — AI tools decay · substrate compounds · cohort network effect visualized below the chart.
Every replay improves calibration. Every decision deepens the substrate. Every cohort engagement strengthens the L1. What you build never decays.
§ 17 ── THESIS
AI systems become institutionally valuable only when probabilistic computation is constrained by deterministic continuity layers.
── QUERYABLE.COMPANY · CORE THESIS
Most AI companies optimize output quality. We optimize institutional stability under model volatility. The historical pattern: the largest infrastructure companies are built around stability problems, not feature problems. Diana Hu's YC framing identifies the destination — AI as operating system. The substrate is the path that gets there with audit, governance, and provenance intact.
§ 18 ── UNDER THE HOOD · NDA-TIER
Substrate-grade depth available to qualified prospects.
What's public on this doctrine page: the five disciplines · the L0/L1/L2 ontology · the four-tier framework · the seven-step closed loop · the diverge-and-reconcile protocol · the six evidence modes · the tether-pair doctrine · the D7 §12 inspection right · schema migration semver · pattern migration concept · anti-fragility framing. What's available in a 30-min technical deep-dive under NDA:
CUSTOMIZATION DECISION TREE
How customer interview answers route to per-customer agent fabric (5-9 Class A) + L8 floor selection (F1-F6 canonical + 0-3 optional F7-F9) + stage-adapted phase durations + KPI gate timing + Methodology Council seat composition.
13-STEP GENERATION PROTOCOL
The exact sequence that produces a complete L2 substrate in 1-3 weeks of Phase 0.
CANONICAL EVAL CASES PER AGENT
The κ verification tests + L8 ceiling pressure tests + diverge-and-reconcile triggers + audit emission patterns per agent.
METHODOLOGY COUNCIL WORKSHOP
7-seat composition + κ baseline + monthly + quarterly + annual cadence playbook.
D7 §12 PROTOCOL MECHANICS
Drill format · pressure-test procedures · cross-tier composition verification · escalation routing.
PRE-FLIGHT VALIDATION TOOLKIT
Operator-side substrate discipline tooling.
COMPLIANCE MAPPING LIBRARY
Regulatory framework → L8 floor + audit chain tethering · per-jurisdiction (GDPR · HIPAA · SOC 2 · ISO 27001 · PSA · APPI · JFSA · FCA · sector-specific).
REPLAY-TEST INFRASTRUCTURE
Structural fingerprint methodology · model swap survival mechanics.
PATTERN MIGRATION L2 → L1 WORKFLOW
Concurrence solicitation gate · attribution + compensation protocol · anonymization pipeline · cohort intelligence activation gating.
Sophisticated buyers · YC partners · M&A diligence teams · procurement teams · regulator-facing audit teams · all welcome. The substrate is auditor-ready any time per D7 §12 protocol.
§ END ── THE SUBSTRATE TRAVELS
Yours. Versioned. Portable. Compounding.
At a different stage?