All tracks
Builder Track · engineers, technical evaluators, sales engineers, security reviewers

The PM33 platform, end-to-end

A 90-minute walkthrough of how PM33 actually works — governance and trust layer, the Brief lifecycle, harness setup with hands-on commands, outcome attribution, and the structural answer to "what does this buy us." Each module pairs a diagram with a workflow narrative.

This is the technical case. PM33 is named throughout — Briefs, the Pam orchestrator, MCP tools, specific RLS patterns, the AR(1) recalibration model. If you want the vendor-agnostic argument instead, read the PM Track.

The 6-module path

Total: ~95 minutes including the hands-on harness setup. Modules are self-contained.

Pick a path

Five common reasons people open this page. Each path is self-contained.

If you're a…WhyRead theseTime
First-time evaluatorCEO, CTO, head of product. Just want the thesis.25 min
Engineering leaderEvaluating for adoption. Needs governance + harness story.70 min
Practitioner / soon-to-be operatorAbout to use PM33 day-to-day. After this it's operational.55 min
Sales engineer prepping a demoAdd the outcome-attribution piece for objection-handling.40 min
Compliance / security reviewerLead with governance, add audit-trail context.30 min

The whole cycle in 90 seconds

If you read nothing else: PM33 is a closed-loop strategic execution platform. The loop has 8 stages — every transition is a structured event in an append-only audit log.

  1. 1
    Customer SignalA request, bug, idea, or VOC entry enters PM33.
  2. 2
    Strategic AlignmentPam scores the signal against active strategic objectives — the workspace's OKRs / roadmap themes.
  3. 3
    Brief AuthoredThe atomic unit of agent-executable work is created — machine-verifiable acceptance criteria, a TDD plan, and an outcomeHook that defines how success will be measured.
  4. 4
    Sprint PlannedCapacity-aware scheduler places the Brief in a sprint that respects dependencies + team availability.
  5. 5
    Agent ExecutesA specialist agent picks it up, working in a per-agent git worktree. TDD discipline (RED → GREEN → REFACTOR → DELIVERY) enforced by the harness skill.
  6. 6
    ValidateIndependent review agent inspects the diff. CI runs. Schema-drift gate fires. Security checks pass. Only then does the Brief move to in_review.
  7. 7
    ShippedPR merges. Lifecycle event pr_merged fires. Brief auto-flips to done if all verification signals pass.
  8. 8
    Outcome TrackedThe outcomeHook window opens. Metrics get attributed back to the strategic objective. If the metric moved → great. If not → the AR(1) forecast model recalibrates.

The loop is continuous, auditable (every transition is a structured event), and multi-tenant (every row in every table is tenant-isolated via RLS). This is the difference between "AI agents write code" (table stakes now) and "AI-driven strategic execution with closed-loop attribution" (PM33).

The Builder-track thesis

The loop is the product.

Everything else — the MCP tools, the Briefs, the harness, the agents — exists to keep the loop running and the data flowing. Strategy → code → outcome → recalibration, in one system, with one audit trail.

Want the vendor-agnostic argument that doesn't name-drop PM33? Read the PM Track →