← What I’m building

Active project

Active

Make AI coding work continuous, reviewable, and less dependent on one model.

An orchestration layer for planning, approving, building, reviewing, testing, and continuing software work across AI coding engines.

Parked until it has a next proof I’m actually running this week. The CLI work and DevHQ wireframes stay in the record.

01The problem

What makes this worth solving?

AI coding agents can produce useful software, but the plan, decisions, memory, and verification evidence often remain tied to one session or one engine. Switching tools can mean rebuilding context or trusting a summary that is difficult to audit.

For a solo builder, that creates a second job: repeatedly coordinating tasks, checking what changed, confirming tests, and deciding whether another agent can safely continue the work.

02What I’m building

How the product is meant to work

AI Dev OS is intended to own the durable project state while coding agents act as interchangeable engines. It keeps the proposed work, approval gates, task progress, review results, and verification evidence together.

A builder can define the work, approve the plan, let specialized agents implement and review it, inspect the evidence, and continue later with the project context intact. The coding engine can change without becoming the owner of the project’s memory or process.

03Why this project

Why I’m spending time on it

I want the leverage of multiple AI coding agents without giving up judgment, traceability, or continuity. A faster coding engine is useful, but a dependable way to direct, review, and continue the work is what makes that speed sustainable.

This is a problem I encounter directly as a solo builder with limited time. Solving it would let me spend less effort reconstructing context and coordinating tools, and more effort making product decisions that move the work forward.

Intended value

What should become better if this works?

The value is still a hypothesis until sustained use and payment confirm it.

01

Continuity

Project state survives a pause, a handoff, or a change of coding engine.

02

Control

Plans, approvals, reviews, and verification remain visible instead of disappearing into a chat session.

03

Engine flexibility

Different coding engines can contribute without any one of them owning the project’s memory or process.

Related field notes

The story behind the current evidence.

Read all notes