Build · Module
Factory
Governed idea-to-spec-to-code-to-release for Packs and external repositories
Factory is the planned W5 path from idea to approved spec, task packets, reviewed code, tests and approved releases, producing One3 Packs or external repositories. Humans approve specs, merges and production deploys.
What it does
Factory is planned for W5 (Years 4–5). It turns ideas into released software with human gates at each critical step. A SoftwareProject workspace has an owner, a target — a One3 Pack or an external repository — a budget, repositories and environments.
From idea to spec
Idea intake links ideas and clustered feedback from Desk, Survey and Work into a project brief. Spec generation drafts a PRD, user stories with acceptance tests, a data model and an Action list — for Packs, a draft manifest. A human must approve the spec.
From spec to code
The approved spec is split into task packets with allowed changes, interfaces and pre-written acceptance tests. Executor orchestration assigns packets to AI executors or human developers with budgets, timeouts, retries and traces. Code lands on branches as pull requests, and merging always needs a human code review.
Test, deploy, release
The build and test pipeline runs unit, contract and end-to-end tests, plus static analysis, dependency, secret and licence scans; acceptance tests stay unchanged from the spec. Environments include a preview per PR, staging and production; production deploys require approval. Cost and budget tracking posts AI and compute cost per project and packet to the ROI Ledger and auto-pauses at budget.
Traceability and maintenance
At GA, a traceability matrix links stories to acceptance tests, PRs and releases. Release management covers changelog, versioning, rollback and feature flags. A provenance and licence record per change captures executor, model and version, prompt reference, reviewer and dependency licences; this item remains partially specified in the plan. Maintenance mode proposes dependency updates and vulnerability fixes as PRs with tests. Agency delivery lets clients approve specs and releases through a portal, and Factory quality evals track first-attempt pass rate, rework and escaped defects. Legacy modernisation is a W6 item.
Features
What Factory includes
SoftwareProject workspace
Owner, Pack or external repo target, budget, repositories and environments.
Idea intake
Ideas and clustered feedback from Desk, Survey and Work linked into a brief.
Spec generation
Draft PRD, user stories with acceptance tests, data model and Action list; human approval required.
Task packets
Spec split into packets with allowed changes, interfaces and pre-written acceptance tests.
Executor orchestration
AI executors or human developers with budgets, timeouts, retries and traces.
Repository and pull requests
Code on branches with mandatory human code review for merge.
Build and test pipeline
Unit, contract and E2E tests with SAST, dependency, secret and licence scans.
Environments
Preview per PR, staging and production with approval for production deploy.
Cost and budget tracking
AI and compute cost per project and packet to the ROI Ledger with auto-pause.
Spec-to-test traceability
Stories linked to acceptance tests, PRs and releases.
Release management
Changelog, versioning, rollback and feature flags.
Provenance and licence record
Executor, model, prompt reference, reviewer and dependency licences per change.
Maintenance mode
Dependency updates and vulnerability fixes proposed as PRs with tests.
Agency delivery and quality evals
Client approvals via portal, plus first-pass rate, rework and escaped-defect tracking.
Workflow
Spec to release
- 1
Brief created from ideas and feedback
- 2
Spec drafted and approved by a human
- 3
Task packets assigned to executors
- 4
PRs reviewed and merged by humans
- 5
Tests and scans pass
- 6
Production deploy approved and released
AI
AI skills and Digital Employees
Spec generation
Drafts specs with acceptance tests for human approval.
Feedback clustering
Clusters Desk and Survey feedback into project briefs.
Code execution in packets
AI executors implement task packets within budgets and allowed changes.
Digital Employees that use this module
They act through the same governed Actions as people; risky steps wait for approval.
Frequently asked questions
When will Factory be available?
It is planned for W5, Years 4–5.
Does AI deploy code to production on its own?
No. Specs, merges and production deploys all require human approval.
How is cost controlled?
Each project has a budget. AI and compute cost per packet posts to the ROI Ledger, and work auto-pauses at budget.
See Factory in your workflow
Tell us about your team and the outcomes you want. We will walk you through the apps, the Digital Employees and the plan for your industry.