The Wizard NexusProvisional by design
Actor, confidence, disposition, notes, and contrary facts remain open to accountable review.
In developmentThe standalone Sentinel app is not yet available to download or purchase.
Safeguards · Human accountable
The system can organize a theory. It cannot make the theory true.
The essentials
Sources stay traceable. Decisions stay human.
Machine output needs review. Uncertainty and contrary evidence stay visible. Private records belong in authorized workspaces.
Human review and uncertainty
Analytical boundary
Consequential language stays provisional until a person earns it.
Allegations are leads, not facts.
Contradictions are not automatically crimes.
Motive is not proof.
Reports are analytical work product, not proof.
Uncertainty stays visible
Ambiguous or unresolved provenance is labeled; the system does not silently select a convenient source.
Decisions stay human
Charging, admissibility, credibility, and guilt remain with the people and agencies authorized to decide them.
Limits of automation and reports
Built-in limits
A trustworthy tool says what it cannot know.
Automation can reduce review burden. It can also reproduce omissions, bad framing, and bias at scale if its outputs are treated as conclusions.
Automation is screening
Machine output may prioritize, classify, compare, or summarize. It does not become a finding until a person checks the source and records a disposition.
Contrary evidence belongs beside the theory
Defenses, innocent explanations, unresolved gaps, and disconfirming records remain visible rather than being buried in notes.
Confidence is not certainty
A confidence label communicates review state and support. It is never a probability of guilt or a substitute for proof.
Exports remain work product
Reports include sources, limits, open questions, and review history. Exporting a report does not transform it into evidence.
Private records and public information
Publication boundary
The public site contains no case material.
Evidence, filings, contacts, review state, exports, and reports belong in an authorized case workspace, not in Git, not in a public issue, and never in the Pages artifact.
Product explanations, synthetic examples, roadmap status, and safety rules.
Versioned Sentinel code and approved synthetic fixtures after extraction.
Real evidence, personal data, contacts, reports, and workspaces under the user’s authority.
Git and package exclusion reduce accidental publication risk. They are not encryption, authorization, or access control.
Responsible use and prohibited conduct
Responsible use
Investigation cannot become a license to harm.
High-risk actions, connectors, and exports require greater control than ordinary record review.
Sentinel must not be used for
- Stalking, doxxing, harassment, or coercive monitoring
- Discriminatory profiling or automated bulk targeting
- Unauthorized access, pretexting, impersonation, or unlawful surveillance
- Protective-order violations or the fabrication or alteration of evidence
- Publishing an accusation as though it were an adjudicated fact
Accountable use requires
- Authority to possess and review the records
- Source inspection before a machine-surfaced lead is adopted
- Recorded uncertainty, contrary evidence, and next verification steps
- Appropriate access controls, retention, export, and deletion decisions
- A human decision-maker responsible for every consequential output
Access follows responsibility
The same integrity safeguards belong in every plan.
Provenance, citations, uncertainty, contrary evidence, human review, and complete export are not premium features.
See planned pricingReview the method