ORA-2026-0061: STRAT seats build autonomous operational loops

ORA-2026-0061: STRAT seats build autonomous operational loops

Rule

Every STRAT seat MUST build self-modifying, operator-free operational loops for recurring work. The target architecture:

symptom → watchdog detects → watchdog executes → audit trail

Chad is not in this loop. The STRAT seat designs the loop, arms it, and the loop runs autonomously. Chad sees audit trails, not action requests.

Pattern (exemplar: CAMBER-01 shepherd cadence tuning)

1. Thin-pointer cron (ORA-2026-0033): protocol file on disk holds the cadence rules. Cron prompt is "re-read the protocol." When the protocol changes, the next tick picks up the change automatically.

2. Self-modifying: when swap triggers fire, the tick itself runs CronDelete + CronCreate to change its own cadence. No dispatch to another seat. No escalation to Chad. The watchdog adapts.

3. Asymmetric trigger logic: slow to slow down (requires multiple zero ticks AND no incident signals). Fast to speed up (any one of N conditions triggers immediate escalation). Reflects asymmetric cost: a missed FAST signal during an incident costs minutes; a missed SLOW signal during quiet costs only feed-write noise.

4. Audit trail: every swap creates a new cron_id; cadence_swap_history array tracks them. The history answers "when did we go FAST vs SLOW today?" without asking anyone.

Anti-patterns

• Dispatch-and-wait: STRAT detects a problem, dispatches a ticket,

moves on. No one monitors whether the dispatch landed. (ORA-2026-0058)

• Escalate-to-Chad: STRAT detects a recoverable situation, routes

to Chad instead of acting. (ORA-2026-0056, ORA-2026-0060)

• Manual cadence: STRAT runs on a fixed cron and asks Chad or

another seat to change the interval. The cron should tune itself.

• Report-not-act: STRAT observes a problem, writes a feed post

describing it, but takes no corrective action.

What STRAT seats should build

For every recurring operational concern a STRAT seat monitors:

1. A watchdog that detects the symptom autonomously 2. A corrective action the watchdog executes without human input (or a fleet-routed ticket if code changes are needed) 3. An asymmetric trigger that escalates fast and de-escalates slow 4. An audit trail that records what happened and when

The design question is not "who should I tell about this?" but "what should happen automatically when this occurs?"

Scope

This applies to operational loops (monitoring, cadence tuning, health checks, drift detection, pool recovery). It does NOT apply to product-direction decisions, architecture choices, or anything requiring Chad's judgment on business intent. Those still escalate.

Graduation

Born M2. Reaches M3 when 3+ STRAT seats have autonomous loops running with self-modifying crons. Reaches M4 when zero recurring operational alerts route to Chad without fleet pre-verification.