Software Factories for Real Engineers - Matt Pocock

Build reliable AI software factories with Claude Code, domain modeling, specifications, automated review, testing, and GitHub Actions.

Build Reliable Software With AI Coding Agents

AI coding agents can generate code quickly, but without proper context, testing, and engineering controls, they can create more bugs and review work.

Software Factories for Real Engineers by Matt Pocock teaches developers how to build an end-to-end AI coding system for planning, delegating, implementing, reviewing, and improving software.

The two-week cohort combines established engineering principles with Claude Code and modern coding agents. You’ll learn to delegate well-specified work, enforce quality automatically, and create a software factory suited to your team and codebase.

What You’ll Learn

• Create shared domain language for developers and agents
• Write architecture decisions, specifications, and agent-ready tickets
• Plan work too large for a single coding session
• Build reliable review and feedback systems
• Enforce standards through tests, linters, and type checks
• Improve legacy and brownfield codebases
• Automate bug investigation and task delegation
• Create readable AI-generated pull requests
• Use GitHub Actions to track and review work
• Design a custom software factory for your project

What’s Included?

AI Coding Crash Course

A three-day foundation covering AI models, coding-agent harnesses, and the environments agents need to produce dependable results.

Domain Language and Wayfinder

Build shared glossaries, project maps, architecture decision records, specifications, and structured tickets. The Wayfinder skill helps break large projects into work an agent can execute systematically.

Review and Codebase Design

Learn three layers of review—automated checks, automated adversarial review, and human review—alongside deep modules, architectural boundaries, testability, and agent-friendly documentation.

Automated Feedback Loops

Use custom lint rules, type checks, mutation testing, frontend verification, and automated review to catch problems before code reaches a developer.

Software Factory Build

Follow Matt’s GitHub Actions workflow for investigating, triaging, assigning, delegating, tracking, and reviewing engineering tasks. You’ll build a working factory with an example codebase before designing one for your own team.

Practical Course Resources

The program includes:

• More than 60 video lessons
• Hands-on sandbox exercises
• A complete course code repository
• Six live Office Hours sessions
• Recordings and transcripts
• Private AI Hero Discord
• Completion certificate

The course is taught using Claude Code, but its engineering principles can also be applied to Cursor, GitHub Copilot, Windsurf, and other coding assistants.

Who Is This Course For?

Software Factories for Real Engineers is designed for:

• Professional software engineers
• Senior developers and technical leads
• Engineering managers
• Claude Code users
• Solo developers managing complex projects
• Teams adopting AI coding agents
• Engineers working with existing codebases
• Developers responsible for architecture and code quality