Archive opportunity claims and revisit them with Jev

01 · Role in the system

What Jev does here

The flow targets a real failure mode: job and collaboration offers that turn out to be scams asking for money, credentials or code execution. Each opportunity's claims are extracted with exact quotes, archived on-chain via Walrus memory on Sui mainnet, and confirmed only on completion. When the next message from the same party arrives, the archived evidence is recalled and Jev classifies every prior/current claim pair with a typed scam Noul - the README shows the exact question: does this offer aim to take the candidate's money, credentials, personal data or run malicious code. DeepSeek explains the verdict with citations, corrections append and supersede without rewriting history, and when Jev and DeepSeek disagree the app says so instead of inventing an answer.

02 · Control boundary

Where Jev sits

Jev as the scam Noul between archived and fresh claims: on-chain memory recalls prior promises, typed judgments flag contradictions and fraud patterns, DeepSeek explains with citations, and an append-only history keeps the record honest.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

• Hackathon cut of the spec; later milestones (per the deferred table) are still plans.
• Scam detection depends on one Noul question per claim pair; no measured precision yet.
• 0 stars; deployed at revior.xyz with Google sign-in, single-maintainer project.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.

• Nuel-osas ↗Community · github · public · checked 2026-10-09