WROITER
live
A deterministic writing diagnostic with a public method, API and MCP surface.
- My role
- Product direction, method decisions, review gates and publication decisions.
- Last reviewed
Problem
Black-box AI detectors and humanisers provide scores or rewrites without showing a stable, inspectable method.
My contribution
- Product framing and deterministic diagnostic method
- Public specification and evidence boundaries
- Review of revision behaviour, API and MCP publication
Constraints
- Diagnostic signals must be deterministic and documented rather than presented as proof of authorship.
- Revision output must preserve meaning and expose limitations.
- Implementation was heavily AI-assisted and cannot be framed as sole hand-written engineering.
What I worked on
- Defined the product direction and public Pattern Profile Specification.
- Directed a diagnostic covering 33 detectors across six signal families.
- Set review and publication gates for the revision engine, HTTP API and MCP tool.
Verification
Present state
WROITER is live with an inspectable method, deterministic diagnostic, revision endpoint and MCP surface.
Limits
- This case does not claim current users, pilots, revenue or adoption.
- Alex owns product and method decisions; implementation was heavily AI-assisted.
- Diagnostic patterns do not prove whether a person or model wrote a text.