The AI Software Factory Guide / 04

Build the system around the agent.

The FDLC-native portal into the full field guide for designing, building, proving, operating, and improving an AI software factory.

44 chapters6 parts8 Guide stages

Open the full Guide ↗Understand FDLC in 5 minutes Framework →

Choose a reading path

Start with the outcome you need.

10 minutesGrasp the whole model quickly

BuilderAssemble capabilities, harnesses, and workflows

ArchitectDraw boundaries, records, and authority

OperatorRun the factory as a production platform

Executive / FundExplain or justify the system

Front to backRead the complete methodology

One system / different lenses

How FDLC and the Guide relate.

FDLC defines the operating system. The Guide teaches you how to build and run it. The models align without collapsing into one another.

FDLC / Factory lifecycle

How a factory line evolves.

The Guide shows how to discover, design, assemble, validate, deploy, operate, and improve a factory line.

View the canonical lifecycle →

FDLC / Runtime + artifacts

How work advances and persists.

The Guide supplies authoritative records, stage contracts, and the chain from intent to outcome.

Open the Guide stages ↗

FDLC / Architecture

What the factory must contain.

The Guide goes deep on intent, harnesses, capabilities, models, trust and verification, and learning.

Explore the six areas →

FDLC / Principles

What must remain true.

The Guide develops the first principles of trust, evidence, authority, and human–agent operating models.

Read the Manifesto →

These are the same system described at different levels of abstraction. The Guide’s eight stages are not a competing FDLC lifecycle.

Six parts / One arc

Six parts. Forty-four chapters. One path to verified delivery.

Use the six-part arc for a complete read, or enter through the role and topic indexes below. Every part opens directly in the authoritative Guide.

Eight stages / The runtime one line

Follow autonomous delivery from intent to software.

These technical deep-dives explain the contracts, records, and decision rights across each step. They map across the FDLC Runtime Lifecycle and Artifact Protocol; they are not a second lifecycle or a one-to-one artifact list.

Intent → Plan → Define Agent → Execute through Harness → Apply Skills → Evaluate → Improve → Deliver Software

FDLC topic map

Twenty-five entry points into the same source.

Use these routes when you already know the concept you need. First-time readers should start with the structured parts and stages above.

Full chapter links are maintained in the canonical Guide ↗

Mission Control and the Guide

A living implementation, with its boundaries visible.

Mission Control is the open-source reference implementation of FDLC. Guide chapters and appendices use it as a living case study for capability maturity, verification-first flows, and operator surfaces—while separating enduring architecture from point-in-time implementation evidence.

Review current limitations Read the case study ↗

Source provenance

Open, attributable, and visibly evolving.

The Guide is maintained in Git, while its dedicated reading site provides search, the complete table of contents, stage pages, role paths, appendices, and a public changelog.

View source on GitHub ↗Read the changelog ↗

Next steps on FDLC.ai

Continue with the depth you need.

01FDLC in 5 MinutesOne factory line end to end →02FrameworkCanonical definition and system views →03ArchitectureSix areas in depth →04ManifestoTen operating principles →05MaturityAssess your current state →06Mission ControlReference implementation →

Verified delivery, not merely generated code.

Build the system around the agent.

Start with the outcome you need.

How FDLC and the Guide relate.

How a factory line evolves.

How work advances and persists.

What the factory must contain.

What must remain true.

Six parts. Forty-four chapters. One path to verified delivery.

See the factory as a system

Define ownership, records, and authority

Assemble capability and execution

Verify exact outcomes

Run the factory as a platform

Learn without self-authorizing change