Capability Map

bijux-proteomics-intelligence evaluates candidates and scenarios under declared decision policy. Its outputs are inspectable recommendations, refusals, escalation signals, and review packets—not autonomous decisions or newly manufactured evidence.

flowchart LR
    Q[Decision question] --> C[Candidate cohort]
    C --> E[Evidence posture]
    E --> X[Contradictions and falsifiers]
    X --> S[Scenario evaluation]
    S --> P[Policy and ranking]
    P --> O[Recommendation, hold, or refusal]
    O --> R[Review and learning records]

Decision capabilities

Capability	What remains visible
Candidate preparation	schema validation, fingerprints, quality, filters, lifecycle state, cohort membership
Ranking and selection	policy, criteria, scores, ties, exclusions, and selected candidates
Evidence posture	support strength, gaps, provenance, readiness, and the distinction between evidence and interpretation
Skeptical challenge	contradictions, falsifiers, alternative explanations, and belief audits
Scenario reasoning	progression, synthesis, scale-up, and redesign outcomes with confidence and unresolved questions
Recommendation	action, reasons, downgrade chain, gate result, and human-review requirement
Refusal	stable reason and minimum missing evidence when a strong claim or action is unsupported
Review	decision briefs, boards, outsider packets, rerun kits, public-scrutiny records, and release candidates
Evaluation	blinded challenges, counterfactuals, confidence calibration, sensitivity, quality, and regret
Learning	reviewed adaptation proposals tied to outcomes rather than silent policy mutation

Intelligence can compare what would happen under different assumptions and policies. It cannot certify source truth, perform the underlying proteomics analysis, allocate laboratory resources, or execute a workflow. Those boundaries keep a recommendation reversible and reviewable.