Jev validates Near Here event listings at 96 percent on 50 cases

Author: Jon Reed

Classification

Jon Reed

Original source

Classify this use case

  • Classification0.95
  • Shipped0.81
  • Measured0.98
  • Usefulness0.54
  • Community post0.97
  • Ready0.93
  • Specific case0.61
  • Copy quality0.47
  • Creativity0.82
  • Distinct intent0.91
Cost
$0.043 per 1,000 decisions
Latency
0.59s average; 0.58s median on the 21-record run
Result
48 of 50 expected decisions on the main test set

Ask Jev a Choice on whether a listing title and optional description should be excluded from Near Here after deterministic checks.

Ask Jev a Choice on whether a listing title and optional description should be excluded from Near Here after deterministic checks.

The cited post compared that Choice with Mistral Small 4 and Gemini 3.5 Flash-Lite on the same 50 listings. Jev matched 48 of 50 expected decisions and rejected none of the 13 expected-valid events. Mistral matched 42. Gemini matched 43.

Speed and cost come from a separate 21-record run: Jev median 0.58s and $0.000904 for those 21 listings, or $0.043 per 1,000 decisions. The post says the prompts were selected for prospective production use, not that they were already deployed.

How it works

  1. Run deterministic checks on each listing
  2. Send the title and optional description to Jev as a Choice
  3. Exclude listings Jev assigns to a rejection category
  4. Keep listings with no explicit exclusion