RESEARCH · METHODOLOGY
State of AI-built software.
This page defines a possible research programme, not a report of findings. Bynilson has not aggregated ShipCheck responses into a public dataset.
Questions a future dataset could answer
- Which production checks builders say they have verified.
- How readiness gaps differ by product stage and architecture category.
- Which failures recur across independent, consented submissions.
Method required before reporting
- Explicit opt-in before any structural diagnostic data is retained for research.
- Collect only product category, stack category, stage and check outcome; no source code, credentials, private URLs or customer data.
- Publish sample size, period, definitions, collection method, missing-data handling and limitations with every result.
- Keep individual results private and allow consent withdrawal for future aggregation.
Current dataset status: no public aggregate and no sample size to report. ShipCheck does not send response values; interaction analytics record assessment starts and the number of steps answered.