AI Software Factory with Codex: Plan, Code, Test - Indie Hacking Blog #4
Video description
I'm building an AI software factory with Codex to automate planning, coding, and testing. This architecture walkthrough and prototype demo covers a custom workflow engine and a previous run that built a small Rust job queue.
The AI coding agents run a brainstorm, implement, test workflow. I explain the control plane, engine, and worker; YAML process definitions; PostgreSQL state; retries and timeouts; and how structured outputs pass context between steps.
This is an early prototype. The current coding-agent integration uses Codex and a local working directory. The example run needed fixes and restarts while I was developing the engine, and the generated code still needs human review.
I also explain where I want to take it next: changes to a system design document triggering implementation updates and a pull request for an engineer to review. That document-to-PR workflow is the roadmap, not something the prototype does today.
Timestamps: 00:00 What is an AI software factory? 05:56 Modeling development as a workflow 09:00 Architecture and process data models 19:41 Prototype demo: YAML and Codex workflow 28:57 Previous execution and Rust job queue 33:19 Roadmap and why human review matters
Which part of your development workflow would you automate first, and where would you keep a human review step?
#AIEngineering #Codex #SoftwareArchitecture
Source code
View the AI Development Orchestrator repository on GitHub
Watch on YouTube
Watch AI Software Factory with Codex: Plan, Code, Test - Indie Hacking Blog #4 on YouTube
š§ Stay Updated
Get weekly insights on backend development, architecture patterns, and startup building directly in your inbox.
Free ā¢ No spam ā¢ Unsubscribe anytime
Share this article
Related articles
How To Keep Building Side Projects in AI era?
How do you stay motivated to build projects when AI tools are taking over more of the coding? Building projects is an essential part of the indie hacking journey. But as a software engineer, it can be difficult to keep going when AI is changing the work that made you excited about building software in the first place.
A Better Way to Brainstorm Startup Ideas | Indie Hacking Vlog #3
In this video, Iām improving my indie hacking brainstorming process by using the How Might We technique to move from problems and product vision toward concrete product ideas. Instead of immediately brainstorming features or startup ideas, I start by defining the product vision: what problem I want to solve, who I want to help, and what outcome the product should create. I also explain the criteria I use to evaluate and vote on different directions.
