Codex-readable bridge repository for the OC Parcel Anomaly GIS Comparator.
This repository separates ChatGPT proposals from validated model evidence.
ChatGPT proposals
-> validated request queue
-> Codex processing
-> refreshed evidence snapshots
-> ChatGPT-readable Gist raw JSON
Every Bridge task must explicitly separate:
Observations
Claims / hypotheses
Validation result
Evidence sources
Published state
Validation outcomes are limited to:
PASS
FAIL
UNKNOWN
Unknown is a valid result and must not be replaced with speculation.
- Do not write directly to model outputs from ChatGPT requests.
- Do not write directly to the public evidence Gist from ChatGPT.
- Treat user claims as notes or hypotheses unless independently sourced.
- Preserve provenance and validation errors.
- Use
unknowninstead of guessing. - Use
PASS,FAIL, orUNKNOWNfor validation outcomes.
Bridge separates observations, claims, validation, evidence graph, and published state.
Human observations
-> Claims / hypotheses
-> Request queue
-> Validation layer
-> Evidence graph
-> Published state
See:
docs/EPISTEMIC_FIREWALL.md
Public review page:
https://aevespers2.github.io/Bridge/evidence-graph/
manifests/codex_api_manifest.json
schemas/chatgpt_request.schema.json
exports/chatgpt_request_queue.jsonl
docs/GITHUB_CHATGPT_WRITEBACK.md
reports/reproducibility_report.md
reports/bridge_health_latest.md
docs/evidence-graph/index.html
docs/CHAT_BUNDLE_DISPATCH.md