Mastery console96-source curriculumGoverned architecture

Engineering autonomous delivery beyond the coding agent

Build the system around the agent.

Learn how human intent becomes validated software through bounded authority, durable execution, independent evidence, controlled delivery, and governed improvement.

Start learning Explore architecture

01 / SystemThe agent executes.

It reasons, proposes, and acts inside a bounded runtime.

02 / GovernanceThe factory authorizes.

It owns policy, state, admission, recovery, and decisions.

03 / ProofIndependent evidence decides readiness.

The producer cannot certify its own material work.

Canonical architecture explorer

Follow capability down. Trace evidence back up.

Use the flow lens, then select any component to inspect its responsibility, contracts, authority, failures, and proof.

Governed orchestration

Control plane

Owns durable state, policy, admission, authority, dispatch eligibility, approvals, and recovery decisions.

Owns

Durable workflow state
Admission, policy, and authority

Does not own

Model reasoning
Independent quality judgment

Receives: Intent decision · Qualified capability inventory
Produces: Authorized dispatch · Decision and state history
Authority: The control plane grants only policy-eligible capability and records every material transition.
Failure modes: Unauthorized dispatch · Split-brain state
Evidence: State-transition log · Policy and approval receipts

Upstream

Downstream

An executor cannot grant itself authority or independently certify its own material work.

The governed value stream

How work moves through the factory.

The six-stage rail is a navigation abstraction. Canonical chapter terminology remains intact.

Canonical mappingIntent

Intent

Human purpose, outcomes, constraints, policy, risk, and acceptance expectations.

Human intentStrategyPolicyRiskAcceptance

Explore all intent material

Start with these chapters01AI Software Factory and Mission ControlStart Here · 12 min 02Canonical GlossaryStart Here · 38 min 03Platform Blueprint and Operating PlaybookStart Here · 16 min

Start at the right altitude

Four paths. One system.

Compare the paths

01Light

Executive

Explain value, risk, governance, and accountable autonomy.

For: Engineering and business leaders
Effort: ~45 min
Curriculum: 6 chapters

Open path

02Medium

Architect

Whiteboard the complete system and defend every boundary and contract.

For: Architects and senior technical leaders
Effort: ~6 hours
Curriculum: 17 chapters

Open path

03Deep

Builder

Build and exercise a governed delivery path through failure and recovery.

For: Platform, AI, quality, and delivery engineers
Effort: ~12 hours
Curriculum: 13 chapters · 4 labs

Open path

04Complete

Deep Study

Defend architecture and implementation decisions from first principles.

For: Practitioners preparing to design, build, and defend the system
Effort: Complete corpus
Curriculum: 96 chapters · 13 labs

Open path

Explore the curriculum

Find the layer behind the failure.

View the topic index

Capability supply chain

Registries, packaging, versioning, certification, discovery, and revocation.

Autonomous workflows

Repository onboarding plus feature, defect, test, security, incident, and knowledge work.

Runtime and platform

Harnesses, environments, scheduling, cost, resilience, and human control surfaces.

Verification and delivery

Testing, artifacts, migrations, progressive delivery, rollback, and production outcomes.

Security and identity

Agentic threats, workload identity, secrets, privacy, provenance, and policy.

Evaluation and learning

Datasets, calibrated graders, controlled experiments, optimization, and regression control.

The core thesis

Reliable autonomy comes from a trustworthy system around fallible agents.

HumansIntent, judgment, and material risk

AgentsBounded reasoning and execution

SoftwarePolicy, state, recovery, and control

EvidenceIndependent proof of readiness

Truth before breadth

See what is covered—and what is not yet proven.

Every document carries a maturity label. The coverage map separates review-ready architecture from validated and operational evidence.

Inspect coverage and maturity

Reference depth

From one map to every contract.

Trace lifecycle, planes, runtime components, governance, inventory, autonomy patterns, monitoring, data flow, and evidence—then open the canonical specification behind each item.

Explore the architecture hub

Living case study

Vision stays separate from proof.

Mission Control grounds the curriculum in a versioned implementation while keeping current capability, enduring architecture, and future vision distinct.

Inspect the evidence map