In active development

PROJECT / 01

Factory
Next.

A delivery system for AI coding agents—not just another way to generate code.

Can autonomous agents finish real software work without losing the owner's intent?

Code generation is fast. Dependable delivery is a chain of ownership, evidence, review, recovery, deployment, and acceptance. Factory Next is testing how that chain can operate with far less babysitting while keeping uncertainty visible.

01

One accountable owner

Each outcome belongs to one lane so responsibility cannot dissolve between agents.

02

Evidence over activity

Commits and green checks are milestones. Completion means the requested target state actually changed.

03

Recovery keeps intent

A technical retry gets a fresh execution identity without asking the owner to restate the original job.

04

Claims match reality

Working, testing, planned, and hypothetical behavior are reported as different states.

What breaks after the agent finishes writing code?

If review, retries, deployment, or acceptance still require constant babysitting, send me the specific failure. Those are the operating problems Factory Next is being built to understand.

Share a delivery problem

Related notes

What happens after an AI agent writes the code?A green check is not an accepted outcomeRetries are product behavior, not infrastructure