Three loops, one model.
Syntauros is built around three loops that share one underlying model. The first builds the picture. The second curates it. The third keeps it honest as the world changes underneath.
- 01
Ingest
Connect a source — a repo, a website, a folder of docs. Syntauros scrapes or reads it, runs an AI analyzer over the content, and proposes candidate elements (services, data stores, integrations, decisions) with the evidence it used to find each one.
- 02
Model
A human curates: accept, edit, merge, reject. Accepted elements land in a workspace, organized into packages (TOGAF-style hierarchy) and connected through typed relationships. Views are diagrams; they reference elements but never own them. Edit once, update everywhere.
- 03
Detect drift
When the source changes, ingestion runs again. Syntauros compares the new picture to the current workspace and surfaces what moved — new services, removed dependencies, integrations that no longer match what was approved. The human decides whether the change is acceptable.
The split that runs through everything
Each loop assumes the same thing: humans do judgment, AI does execution. The AI catalogs five hundred elements from a repo; the human decides which ones matter. The AI surfaces drift; the human decides whether the drift is a problem or progress.
That split is the rubric every Syntauros feature has to pass. Features that ask the human to do AI work (manually tagging hundreds of elements) get cut. Features that ask the AI to make judgment calls (autonomously rewriting a target architecture) get cut.