Make preclinical portfolio decisions reconstructable.
PredX is planned to support preclinical evaluation and portfolio triage by preserving why an asset advanced, paused, or stopped when evidence is incomplete, conflicting, or dependent.
In portfolio triage, the people who benefit from a good decision also bear the cost of a false one. Reconstructable reasoning is operational protection, not compliance overhead. PredX will not treat a model output, computational hit, or process receipt as biological validation: supporting, conflicting, negative, and null evidence remain visible to the decision.
Candidate evaluation
Keep supporting, conflicting, and negative evidence attached to the decision.
Portfolio triage
Make advancement and termination reasoning inspectable before capital moves.
Claim discipline
Do not write computational predictions as measured biological results.
Planned capabilities, and the limits that bound them.
Both exhibits travel together: what PredX is planned to do, and the claims discipline that no capability may override.
Planned capabilities
Evidence-aware evaluation
Governed external evidence retrieval
Multi-channel disease and target representation
Structured negative evidence and failed-asset memory
Specialist evaluation instead of one overloaded score
Explicit uncertainty, calibration, and validation receipts
Non-negotiable limits
Computational output is not biological truth
Code is implementation evidence, not proof of efficacy
A structural test proves only the behavior it asserts
Duplicate sources cannot become independent confirmation
Negative and null results remain part of the record
No current public demo performs live scientific evaluation
CageOS first. PredX second.
PredX is not presented as a deployed pharmaceutical platform. Its credible path is to inherit a validated CageOS authority and evidence substrate, then prove value in bounded preclinical evaluation and portfolio-triage workflows with design partners.