How It Works

Project and chat intakes turn source material into reviewable dossiers and structured packets. Deterministic analyzers count the activity that can be counted reliably. Human review supplies the context those counts cannot provide, decides what belongs in the public record, and separates an interesting signal from an unsupported conclusion. Approved records are then written to the knowledge backbone that powers the project, note, evidence, report, and Map surfaces.

Those measurements have firm limits. A tool call is not a unit of productivity. Token volume does not measure quality or cost. A clock-hour bucket does not reveal elapsed working time. A delegated run is not automatically better or worse than a closely supervised one. Hyphenomenon currently makes those patterns visible; it does not convert them into an objective score.

That distinction is central to how the site works: automation structures the evidence, but interpretation remains human-accountable.

Recurring Publication Workflows

The weekly Codex publisher and twice-weekly Chronicle publisher turn eligible source material into public reports. Their operating paths include source-window selection, coverage checks, public-safety validation, additive-only publication, human-review stops, and verification of the live report and its discovery surfaces. Scheduler configuration alone is not proof that a report was published.

Relationship

This note is connected to the public Hyphenomenon project. The project tells the story of why the work began and what it became; this note explains the machinery underneath it.

Future Enhancements

Future work may add an explicit, reviewable framework for comparing collaborative and autonomous runs, assessing delegation, identifying over-iteration, and relating activity patterns to verified outcomes. Any such analysis should preserve the difference between observed telemetry, human evaluation, and inference.

Dataset Preview

• Raw CSV row/table content is available in the source artifact.