Explore the interaction model: source-aware answers, inspectable provenance, explicit uncertainty and a visible evidence boundary. The scenarios are illustrative and do not resolve live regulatory or firm data.
INTERFACE DEMONSTRATION · Static scenarios are specimens, not legal advice, live FCA content, or cryptographic proof. A connected response is certified only when its AVE token and ledger chain resolve.
1 · The question
A probate matter missed the CGT reporting deadline. What is my client-duty exposure and the penalty position?
2 · Specimen response — verify every source
Illustrative response, not legal advice. The interface can place a recorded filing-penalty excerpt from
● SPECIMEN · REVIEW PENDING
Specimen · verify current force and applicability
Effective from: not supplied · Verified: not AVE-certified
Uploading confidential documents into an open-source AI tool is to place this information in the public domain, and thus to breach client confidentiality and waive legal privilege.
Source status is not an AVE coverage proof. Official links and resolvable
proof tokens are exposed only when supplied by the provisioned console.
and
● SPECIMEN · REVIEW PENDING
Specimen · verify current force and applicability
Effective from: not supplied · Verified: not AVE-certified
The benefit of a taxable cheap loan is treated as earnings where it is made by reason of employment, including through connected persons.
Source status is not an AVE coverage proof. Official links and resolvable
proof tokens are exposed only when supplied by the provisioned console.
.
⤷ Typed assertion with a confidence score and a source hash — not a passage left for you to interpret.
3 · Governance format — specimen until resolved
The static events below illustrate the intended format. They are not live ledger events or signatures.
specimen-01Example query eventnot live
specimen-02Example retrieval eventnot live
specimen-03Example proof boundarynot signed
—No current corpus claim
2Specimen citations
3Specimen events
—No live response
Specimen mode · no backend or proof has been resolved
— No public performance metricInternal research results are not product evidence
— No deployed evaluationNo live study or same-job result is represented
— No current serving snapshotState Authority Lock controls current corpus state
— No product configuration claimCandidate controls require independent proof
Where this sits
Generic chat. Aprentiz. Manual research.
Generic AI chatAprentizManual research
AccuracyVaries by model, retrieval and controlsRequired to ground or preserve an incomplete stateDepends on source selection and reviewer judgement
AuditabilityImplementation-dependentAVE token and verified chain required for machine certificationDepends on the firm’s documented process
SpeedWorkload-dependentMeasured per deployment; no public SLA claimedMatter and reviewer-dependent
Regulatory defensibilityDepends on data, source and governance controlsEvidence support only; not a compliance determinationDepends on competent review and retained evidence
The compounding system
Intelligence that can improve without losing accountability.
Institutional knowledge depth
Provenance-bearing memory is designed to preserve useful organisational context while keeping validation, retention and operating boundaries explicit.
From retrieval to cognition
LoRA-based domain adaptation is an AIRR research track. It is not generally available and no customer-owned weight or performance outcome is promised.
Time is the non-fungible input
Any compounding advantage must be demonstrated with longitudinal customer evidence. No 24-month result or competitor-impossibility claim is made.
“A roadmap is not evidence until its gate passes.”
Static scenarios are illustrative and may not reflect current law. Verify official sources and firm applicability. No FCA endorsement, legal advice or compliance guarantee.