AI & data handling
No live AI in your review — yet.
ShipForge's Founding Beta Launch Doctor Review does not run uncontrolled AI over your code. Every review is deterministic checks plus a human — Nei Cescon, founder — who verifies each finding before it reaches you. This page exists to tell you exactly what that means.
Current status
Founding Beta reviews are deterministic + human. No model call is made against your source code. AI may be used later as an explanation layer — to help prioritize and word fixes from already-verified evidence — but only after the safeguards below are in place and you have opted in.
How your code is handled
- You submit a sanitized project to your reviewer (ZIP or repo access). Automated self-serve upload is planned, not live.
- Source is reviewed by a person, then deleted on request. Automated deletion is planned.
- ShipForge never auto-fixes, auto-configures, or auto-deploys your app. Every recommendation is reviewed before it reaches you.
- Reports and share links never expose your repository, file names, source code, or secrets.
Before any live AI is used
Live AI in the delivery path is gated behind all of the following. None are met today:
- The evidence schema rejects unsupported file paths and line numbers — AI can never invent them.
- The model receives only normalized, minimized evidence — never your raw source.
- Prompt logging is redacted; provider retention and training terms are reviewed and approved.
- A fixed token and cost budget, with a stop condition, is set before any run.
- A live-provider evaluation is owner-approved and measures false positives, unsupported claims, unsafe advice, duplicate findings, and missed evidence.