Start here. A “factory” is only one kind of system. The Atlas places it beside the data, identity, security, platform, evidence and Day 2 organs a complete department needs.
Use the rooms to orient yourself, not to crown a winner. A factory can make a change while another organ owns identity, evidence, deployment and the living service.
01Intentwhat needs doing→02Workerwho or what acts→03Evidencewhat proves it→04Day 2what survives launch
Issue → plan → code → reviewable PR or migration.
GitHub Copilot Cloud AgentOpenRewrite OSS
Signal → diagnosis → bounded repair → independent verification.
Stakpak AgentOpenTelemetry
Metadata, lineage, quality, memory and evidence with a source boundary.
Identity, policy, cost, contracts, access and recovery.
Microsoft 365Cloudflare Zero TrustPostgreSQL
THE FACTORY LANE
Where a factory actually sits.
The Atlas maps the whole department. A factory is the narrower loop between intent and an accepted change. Follow the lane to see what it produces, then compare only the machines doing the same work.
01Intentissue, event or migration→02Factoryworker, workflow or fleet→03ArtifactPR, patch, image or report→04Acceptanceindependent proof→05Day 2live service and recovery
Use this rail whenever a factory sounds impressive. The useful question is whether its evidence reaches the deployed service—or stops at a branch, PR, image, report or workflow result.
Use the inventory after the map. Search for a tool, open its passport, and keep its source and boundary attached to the claim.
588 named systems in the current lens
588 systems
USE IN REFERENCE STACKLAB TRIALDEFER UNTIL A GAPWATCH / RESEARCHNEGATIVE CONTROL
Ready for the full inventory?588 named systems with stance, features, provenance and a primary-source link.
YOU HAVE THE ROOM
Now hold the factory job still.
The comparison bench turns the map into a fair test: same mission, same questions, visible stopping point. The full cluster lesson remains available if you want the explanation before the shortlist.