0% read on this device
Browse the curriculum

Start Here

Vision

First Principles

Operating Model

Domain Model

Agent Factory

Runtime Architecture

AI Engineering

Autonomous Workflows

Verification & Delivery

Factory Platform

Quality Engineering

Security & Governance

Case Studies

Labs

Interview Practice

Research Journal

Reference

Curriculum/Operating Model/A focused view of boundaries, contracts, state, authority, failure paths, and tradeoffs drawn from this chapter.
Operating Model20 min readchapterQuick Read

The Human-Agent Operating Model

Define how people and agents divide decisions, execution, oversight, and accountability.

Status: Draft for studyRisk: highLifecycle: intent · plan · verify · learnContent reviewed 2026-08-08Maturity guide →
Claim boundaryThis is curriculum guidance. It does not by itself prove a production implementation.
architecture mode

A focused view of boundaries, contracts, state, authority, failure paths, and tradeoffs drawn from this chapter.

Whiteboard exercise

Reconstruct and defend this chapter’s architecture.

Reconstruct the architecture, name each boundary, and defend the tradeoffs.

governancehuman authority20 min chapter
Jump to validation criteria
Open the source exercise

Draw the Product Owner, Mission Approver, Orchestrator, Worker, Validator, Engineering Lead, Risk Owner, and Release Approver. Connect them through the Mission lifecycle from intent to production verification.

For each boundary, state:

  • the artifact transferred;
  • the authority granted;
  • the evidence required;
  • the decision owner;
  • the escalation condition; and
  • what resumes after the decision.

Then redraw the model for a five-person startup and a regulated enterprise. Preserve the same accountability while changing how many people hold the roles.

4. Tradeoffs and alternatives

Strong role separation increases handoffs and can slow small changes. The answer is proportional control. Low-risk work may use lightweight automated handoffs while retaining the same lineage and evidence semantics.

Separate validators can repeat expensive work. That cost buys independence. Risk-based verifier selection, focused tests, artifact reuse with provenance, and sampling can control expense without letting the worker certify itself.

Human approval can become the throughput constraint. Removing approval is not the only remedy. Better intent, smaller WorkOrders, stronger evidence, clear recommendations, and exception-only routing often create more leverage with less risk.

Small organizations may combine human titles. One founder may be Product Owner, Engineering Lead, Mission Approver, and Release Approver. Technical separation between implementation and validation must remain. Separation of duties should increase with organizational scale and consequence.

5. Current Mission Control Implementation

This assessment uses Mission Control commit 8014d5af427b43ff5c5a63cfdf82ec92742c208c, studied on 2026-08-08.

Mission Control's North Star explicitly assigns intent, judgment, governance, and approval to humans. Agents own bounded execution, iteration, validation, and evidence collection. The governed Mission contract defines four runtime roles: Orchestrator, Worker, Validator, and Operator.

Implemented mechanisms

Mission records retain state, owner, budget, stop condition, corrective limits, current Plan, active WorkOrder, blockers, and required human action. Plan submission freezes a proposed revision. Approval materializes linked assertions and WorkOrders idempotently while leaving dispatch as a separate decision.

Mission dispatch checks approved Plan authority, released WorkOrder state, predecessor handoff, budget, corrective limits, and serial mutation. Exactly one repository-mutating WorkOrder may be active in a Mission. Read-only work may run concurrently when the approved Plan permits it.

Worker and Validator handoffs record role, WorkOrder, WorkflowRun, assertion outcomes, commands, artifacts, risks, next action, and next owner. The handoff validator rejects overlapping assertion outcomes, false completeness, missing commands, or incomplete work without a stated risk.

Mission acceptance fails when assertions are missing, failed, stale, unvalidated, lack receipts, or use an unauthorized waiver. It also fails when WorkOrders or handoffs remain incomplete. Failed validation blocks the Mission and directs the operator toward bounded corrective work.

Capability assessment

Operating capabilityStatus at studied commitInterpretation
Human-defined MissionImplementedUI and Convex records capture governed intent and lifecycle.
Versioned Plan approvalImplementedSubmission freezes the proposal; approval releases exact WorkOrders and assertions.
Orchestrator, Worker, Validator, Operator rolesImplemented in the Mission contractRole-specific permissions and state checks exist.
Structured handoffsImplementedHandoffs preserve assertion outcomes, commands, artifacts, risks, and next action.
Independent validation requirementImplemented mechanismValidator-run and receipt linkage are enforced for required assertions.
Corrective iterationImplemented mechanismFailed validation can block the Mission and bounded corrective work may be requested.
Exception-first operator experienceProduct doctrine with partial UI proofRequired-human-action fields and decision surfaces exist; approval-fatigue reduction has not been measured.
Complete governance-role matrixPartialRuntime roles are clear, but all business, security, compliance, architecture, and release owners are not one enforced matrix.
Risk-proportional approval automationPartialRisk and approval controls exist; comprehensive cross-lifecycle policy proof remains incomplete.
Overnight autonomous operating shiftProduct target, not proven hereDurable state exists, but this chapter did not demonstrate an unattended end-to-end shift.

No fresh browser journey was performed for this chapter. The assessment is grounded in product contracts, source, and existing tests at the pinned commit. It does not promote the complete operating model into a proven V1 claim.

External review

Review this chapter.

Challenge a claim, boundary, missing failure mode, unclear term, or unsupported evidence statement.

  • Claim
  • Boundary
  • Failure
  • Evidence