Papers
Short, honest write-ups of the systems behind mowa — what's implemented today, the design decisions that got us there, and the directions the substrate is built to support next. Written by the engineering team, not the marketing team.
How mowa stores prompt identity, lineage, authorship, and per-version analysis — and the behavioral dimensions this substrate is designed to support next.
The analysis pipeline that runs on every draft — health re-score, structural diff, AI rewrites, test runs — and the deliberate decision to keep its output advisory rather than enforced at the merge button.
How the history timeline aligns mowa edits, scan-detected drift, and PR lifecycle on one axis — and the steps from human diff review toward automated counterfactual attribution.
On the framing
Each note is calibrated to what's actually shipped. Where a capability is design intent rather than running code, the paper says so plainly and names what would be required to close the gap. We'd rather show our work than oversell it.