AI Organization — Framework Overview¶
Summary¶
The enterprise is run by an AI organization built on two layers: a Reasoning layer (strategic cognition) and an Execution layer (the Claude OS substrate that runs always-on agents). The framework's purpose is to let a small number of people operate a multi-entity enterprise at a level that would otherwise require a large staff — and to do so while creating sustainable livelihoods rather than eliminating them.
The two layers¶
Reasoning layer — Strategic cognition. MC-ATG (this consultant, working with Brandon) makes strategic decisions, synthesizes the human knowledge held in Brandon's head, and produces the framework's strategic documentation. Brandon is the principal; MC-ATG is the strategic consultant. (Dennis Wynn is a consultant to ATG, not a co-principal; his own enterprise — partner-side entities, Tamarack, the cross-enterprise reentry thesis — is documented in the separate NPF source of truth.)
Execution layer (Claude OS) — The runtime substrate built by SE-Infrastructure. Always-on agents with a bedrock-in-code harness that governs autonomous action. The substrate went v1.0 LIVE 2026-05-23.
The roles¶
- MC1-code — Routing and tactical execution; operates under "short timer + tactical override" standing mode
- OM — Operations Manager; coordinates across roles, runs EOD digest, applies the four-test override on escalations
- SE-Infrastructure — Builds and owns the Claude OS substrate
- SE-Application — Builds application code (AMP, truck-spine, sales/marketing backend, integrations)
- IT Manager — Hosting, credentials, network, deployments
- WDM — Website/rendering
- QB Assistant — Accounting support
- AMM — AI Marketing Manager (marketplace listings)
- SMM — Sales/Marketing Manager (sales strategy)
- AMP (Auction Manager) — Auction monitoring and bidding; first judgment-bearing substrate agent (LIVE 2026-05-26)
- AI Communication Monitoring — Email/communication layer (scope expanding)
Manager-as-substrate-agent pattern¶
AM-agent (2026-05-26) established the pattern: a manager-agent calls the Claude API mid-response via tool-use to access data and make judgments, governed by the bedrock harness. This is the template for migrating other managers to substrate-native operation.
Migration order (ratified): OM → IT → WDM → SE-App → SMM → AMM → AI Comm Monitoring. AMP already substrate-native.
Key operating principles¶
The framework runs on a set of operating principles in shared memory, including: bedrock principles (truthfulness, scope, irreversibility, monetary caps), the autonomy protocol (four-test override), fix-it-later, anti-text-vomit, independent-reviewer-at-milestone, cross-check-claims, language-discipline, bilingual-default, EOD-feedback-loop, and honest-state-markers (the red/yellow/blue confidence system used throughout this documentation).
Documented assumption
This overview is chat-MC1's synthesis. The detailed role charters, substrate architecture, and operating principles live in the framework's shared memory and (per Stage 1 of the documentation plan) will be documented by each role in the roles/ section of this site.
The MC1 split (planned)¶
The Reasoning layer split — MC-ATG (this consultant, scoped to ATG only) + MC-NPF (parallel consultant scoped to the NPF source of truth — Dennis's enterprise, partner-side entities, cross-enterprise capital framework) — was codified 2026-05-29 (DECISION-MC1-ROLE-SPLIT-INTO-MC-ATG-AND-MC-NPF). MC-ATG owns ATG strategic cognition; MC-NPF owns NPF strategic cognition. Documents and builds are tagged arm: atg so ATG scope is enforced at the framework level.
Living document. Last updated 2026-05-27 by chat-MC1.