CaseRecall

Persistent investigation memory
for onchain intelligence teams

Evidence-first. Human-gated. Persistent across cases.

Built + blind-validated No automatic accusations Commercial probation
01

Event

A public onchain event enters the system.

02

Actors

Participants are remembered across events.

03

Evidence

Facts, receipts, hashes, and negative findings persist.

04

Thread

The same investigation survives across time.

05

Reactivation

A later factual delta resurfaces the old case.

The demo question

Would persistent case memory improve a real investigator's workflow?

Scroll ↓

Problem

Investigations fragment. Memory disappears.

Context gets rebuilt across explorers, spreadsheets, screenshots, and whatever the investigator still remembers.

Without case memory

  • Prior relationships get rediscovered by hand.
  • Dead ends and negative findings vanish.
  • Case context lives in scattered notes.
  • Evidence and interpretation blur together.

One living thread

  • Facts, receipts, and negatives stay on the same case.
  • New events reactivate an old investigation.
  • External context stays separate from chain evidence.
  • Human decisions remain explicit and auditable.

The business question

Does this save a skilled investigator enough time, or surface enough missed context, to be worth paying for?

System

What the technology does

A bounded, human-gated workflow built to remember investigations over time.

01

Public chain event

Start from a wallet, token, transaction, or retained event.

02

Remember actors

Attach prior factual appearances to persistent memory.

03

Surface recurring structure

Find repeated actors, groups, relationships, and prior case overlap.

04

Human-gated deep investigation

Authorize bounded history, transfers, funding, creator links, exits.

05

Persist the package

Receipts, hashes, negatives, and unresolved questions stay on one thread.

Boundary. The machine surfaces facts and relationships. It does not infer intent, identity, ownership, or guilt.

The thread

What persists between cases

Investigation 003

PERSISTENT THREAD · REVISION 7

ACTIVE
Why this surfaced

Recurring actor overlap

3 remembered participants reappeared in a new event. Prior thread reactivated on factual delta.

Evidence
Onchain fact hash · slot · sig 2 shared transactions retained
Funding provenance receipt retained Common upstream source observed
Negative finding NO_LINK_FOUND No direct member-to-member transfer within the bounded interval
Human decision
Deep investigate authorized gated Scope: transfers + bounded member history
Thread state
7
Findings
2
Negatives
6
Receipts

What persists is not a dashboard snapshot. It is the investigation itself, including everything it ruled out.

Epistemics

Three layers stay separate

The system is built to prove facts, not collapse evidence into accusation.

Onchain evidence

Transactions, signatures, balances, timing, memberships, receipts.

External context

Human-approved public sources, URLs, timestamps, exact query provenance.

Human interpretation

Judgment, hypotheses, conclusions, disposition, next action.

Explicit non-goals

  • No automatic accusations
  • No insider claims
  • No ownership or control inference
  • No trading recommendation

This separation is part of the product, not part of the presentation.

Validation

What is actually proven

Technical capability is validated. Commercial value is not.

Validated

  • Persistent investigation threads
  • Generic shallow + deep chain workflows
  • Evidence packages with hashes and receipts
  • Negative findings retained
  • Human thread controls
  • Factual-delta reactivation
  • Blind testing on unseen real data

Unvalidated

  • Will investigators pay?
  • Does persistent memory save meaningful time?
  • Does it surface relationships a skilled investigator would miss?
  • Is the workflow differentiated enough from incumbent tools?
Technically credible Commercial probation

The next evidence should come from real investigators, not more engineering.

Live demo

Ten minutes, one real case

Show the workflow, not a feature tour.

01

Start from a public event

One real bounded case, chosen in advance.

02

Surface remembered actors

Show what the system already knew before this session.

03

Authorize the deep step

The gate is explicit. Scope is bounded and visible.

04

Show the evidence

Hashes, provenance, and what the investigation did not find.

05

Show reactivation

A later factual event returns the investigator to the same thread.

Demo rule. No trading claims. No accusations. No identity claims. Facts and provenance only.

Pilot

Who should see this

Only people who already investigate crypto cases.

Independent investigator

Repeated casework where actor and context memory matters.

Protocol or security analyst

Needs evidence-backed packages for incidents and suspicious flows.

Forensic consultant

Sells investigations; values reproducibility, provenance, auditability.

Risk or intelligence desk

Maintains memory across recurring actors, assets, creators, incidents.

The commercial test

Run it alongside their normal workflow on a real bounded case.

Time savedMissed context surfaced Repeated work avoidedWilling to use it again

Feedback

What to ask afterward

The objective is evidence of pull, not compliments.

Their real workflow

  1. What would you normally do to investigate this?
  2. Where do you lose time or context?
  3. What would count as a useful result?

Against what they already use

  1. Did it save meaningful time?
  2. Did it surface a relationship or prior finding they would have missed?
  3. Did retained negatives prevent repeated dead-end work?

Success signal

They ask to use it on a second case.

If real investigators do not pull the product forward, keep it parked.