Educational demonstrationSep 2026
Cold cases persist. So does the record.
Explore how Second Trace organizes cold-case information into source-linked timelines, evidence, relationships, and hypotheses. The current release is an educational demonstration and is not for real casework.
The instruments, running on the demo case
Workspaces
One case, many instruments
Eight featured workspace views from the demonstration case, shown as screenshots. Select a view to preview it. The interactive workspace requires sign-in; the capabilities page lists the fuller inventory.
View full-size screenshot (opens in a new tab)
The demonstration case combines public historical records with clearly labeled synthetic examples. No private case material is shown.
Principles
01
Every fact traceable
Claims, accounts, and relationships carry their source and exact location.
02
Every uncertainty visible
Disputed, approximate, and unknown states are first-class — never smoothed over.
03
AI output requires review
AI-assisted findings must remain linked to their sources and be checked by a person before they are accepted.
Worked example
One example, end to end.
Follow a fictional record through the workspace. Every detail below is synthetic and illustrates the interface rather than a historical event.
Source
A fictional record
An example archive entry reports a vehicle near a library on 14 June 1998.
Evidence
An item with a source
The entry is saved with its origin and page number so later notes can link back to it.
Claim
A report with context
The entry reports a vehicle near the library. Keep the claim labeled as a reported account, not a verified fact.
Contradiction
Two accounts, preserved
A second fictional record gives a different date. Keep both accounts and mark the date as disputed.
Hypothesis
An explanation to test
Record a possible explanation with its supporting and contradicting evidence, assumptions, and the checks it would require.
Synthetic illustration for educational use.How neutrality is enforced →
Trust
Built to be trusted with the record.
Case material is treated as highly sensitive by default. The architecture is defensive in depth — nothing here depends on trust in the interface alone.
Isolation at the database
Row-level security on every table. Cross-tenant reads and writes are denied by the database itself — and an automated suite proves it for every resource class.
Evidence that can't be quietly changed
Uploaded originals are immutable and content-addressed (SHA-256). Audit events are append-only: the database rejects updates and deletes.
Scoped AI, local by default
AI processing sends only explicitly scoped, case-authorized content to the configured provider — local by default, so case content stays on your own hardware.
Encrypted, recoverable
Backups are encrypted and restoration is tested. Version history is preserved for claims, notes, entities, and timeline events.
Controls are documented against common frameworks such as the NIST 800-53 control families. Second Trace does not claim formal compliance or certification — those require independent audit, and we say so plainly. This is the current public security posture; it is not a claim that this audit proved model correctness or production security.
Responsible by design
Presumption of innocence is not a banner. It is the system.
Second Trace describes people neutrally until a source establishes a role. It never scores guilt, never ranks suspects, and refuses to convert allegations into facts. The current release is an educational demonstration of these methods — not for real investigations or legal decisions.
Read the responsible-use policy →