Deferred Decision Markers (DDM)¶
Purpose¶
Some decisions are real but not yet ripe. The information needed to make them well isn't available, or the conditions that would make them obvious haven't arrived. The wrong move is forcing the decision prematurely; the wrong move is also forgetting about it. Deferred Decision Markers (DDMs) hold these decisions visibly with clear revisit triggers.
The DDM system is distinct from:
- Decision Log — decisions already made
- Open questions in individual documents — gaps to fill via principal clarification (smaller scope)
- Future arms — strategic intent for entire new business lines (larger scope, lives in
future-arms/)
DDMs are middle-scope: specific decisions that need a specific trigger to revisit.
Status taxonomy¶
- Active — work in progress; not deferred
- Deferred — waiting for a defined trigger condition before revisit
- Parked — paused indefinitely; revisit if circumstances change
- Absorbed — merged into a different track
- Killed — closed out, no longer relevant
Current DDMs¶
Local model substrate¶
Status: Deferred Revisit trigger: Q1 2027 or earlier on capability change Context: Hardware choice for local AI model hosting (Mac Studio M4 Max vs. RTX 4090/5090). Discussed as a hybrid layer alongside Claude API, not a replacement. Current state: Claude API handles judgment-bearing work cost-effectively at AM-agent scale. Local model justification depends on cost scaling and latency requirements that don't yet exist. Decision when revisited: Hardware purchase + local model deployment vs. continue API-only.
Ritchie Bros cross-border parts conduit¶
Status: Dormant opportunity Revisit trigger: When CTP US-side reaches sustained volume (~$10K+/month for 60+ days) Context: Documented in CTP-MX / Ritchie Bros. A Ritchie Bros VP shared an insight that ATG is structurally positioned for cross-border used-parts conduit. Not a partnership proposal; an opportunity to choose. Dormant until CTP US-side is operating reliably. Decision when revisited: Reach back to RB VP and explore partnership/independent execution.
Cactus Truck Parts as marketing engine for Superior Truck¶
Status: Deferred Revisit trigger: When CTP web traffic and SEO surface ATG truck inquiries naturally Context: CTP's customer base (water truck operators searching for parts) overlaps with ATG's truck-buyer base. The cross-sell opportunity is real. Sequencing question: when does CTP marketing get explicitly oriented to surface truck-sales pipeline vs. focusing on parts revenue? Decision when revisited: Marketing strategy that integrates parts+trucks cross-sell.
Auction Manager + Sales/Marketing Manager AI roles — capacity¶
Status: Deferred Revisit trigger: When AM-agent has run substrate-native for 30+ days with stable performance, AND SMM substrate migration scheduled Context: AM-agent (auction manager) is operational. SMM substrate migration is in ratified queue. Joint capacity question — do these roles eventually need their own substrate agents, or do they share? Capacity is open. Decision when revisited: Substrate architecture for AM and SMM long-term.
NewPath ADU pilot sequencing with Thorson Construction¶
Status: Deferred — partner-dependent Revisit trigger: Dennis Wynn raises ADU pilot timing in strategic conversation Context: The reentry-economy thesis includes NewPath programming + Thorson ADU builds providing housing. Sequencing of the first ADU pilot (location, property, financing, programming integration) is partner-side work. Decision when revisited: ADU pilot scope, timing, and economics.
Sovram Group integrations¶
Status: Deferred — partner-dependent Revisit trigger: Dennis interview surfaces specific Sovram integration opportunities Context: Sovram Group is Dennis's real-estate arm. Integration points between Sovram and ATG-side operations (or framework-as-product, or Tamarack future bank) are real but undefined. Pending Dennis interview to surface specifics. Decision when revisited: Specific Sovram-ATG integration design.
Substrate agent cost caps¶
Status: Deferred — Brandon-action item Revisit trigger: Brandon decides individual + aggregate caps Context: AM-agent runs at $20/day cap (~21¢ actual spend). As more managers migrate to substrate-native, per-agent and aggregate caps need formal policy. chat-MC1 suggestion: $20/day per individual agent, $100/day aggregate. Brandon's money decision. Decision when revisited: Cost cap policy that scales across the manager-as-substrate-agent pattern.
CTP /admin feature gap¶
Status: Pending Brandon enumeration Revisit trigger: Brandon reviews cactustruckparts.com/admin and lists existing features Context: Cactus IMS scope (inquiry inbox, AI photo capture, QB-style item entry, edit/sold management, file-based item management) was absorbed into the existing CTP web app. Which features already exist in /admin vs. need porting requires Brandon enumeration. Decision when revisited: SE-Application scopes the porting work for missing features.
Email infrastructure choice¶
Status: Deferred Revisit trigger: Stage 1 of AI Communication Architecture begins Context: Google Workspace vs. Microsoft 365 vs. self-hosted; multi-mailbox vs. single-intake-with-aliases. See AI Communication Architecture. Decision when revisited: Email infrastructure choice for the substrate email layer.
How DDMs get added and resolved¶
Adding: When a real decision surfaces that isn't ripe to make, MC1-code (or chat-MC1) authors a DDM with status, revisit trigger, context, and what decision will be made when revisited.
Resolving: When a revisit trigger fires, the DDM moves to active work. Decision gets made, lands in Decision Log, DDM status updates to Killed/Absorbed.
Reviewing: Quarterly review session (see Quarterly Review template when authored) reviews all Active and Deferred DDMs to surface trigger-status changes.
Living document. Last updated 2026-05-27 by chat-MC1, authored from MC1-code mobile-session handoff 2026-05-26.