Agents Lookup

Use cases / Decision models (Jev)

Screen scraped event listings before they go live

Jev checks each scraped event listing and either keeps it or names why it should be excluded, such as venue hire or a cancelled event.

DifficultyIntermediate
Setup time2-4 hours
CategoryDecision models (Jev)

Author's figure: Author reports 48/50 correct for Jev on the main test set, versus 43 and 42 for the two LLMs, at about $0.04 per 1,000 decisions; not deployed in production.. Reported by the builder, not measured by us.

The use case

Local event sites and listing directories scrape pages that are not real public events: venue hire offers, product pages, private sessions, cancelled or online-only items. Jev reads the listing title and description and picks one category: keep, or a specific reason to exclude. Excluded items are held back. The author's test did not cover dates, addresses or extraction quality, so those still need normal checks. An editor reviews held-back listings and anything with a low probability before it is published or dropped.

Example workflow

  1. Scrape listings and run deterministic checks (dates, duplicates)
  2. Send title and description to Jev as a choice question: keep or a named exclusion reason
  3. Publish confident keeps; hold confident exclusions
  4. Queue low-probability results for an editor
  5. Editor reviews held items weekly and updates the reason list

Tools

Risks and guardrails

Sources

Written by Agents Lookup. Credits and official pages for the tools this use case names: