JevMade hello@JevMade.com
← Back to experiments

Benchmarks & research

jev-labs

An experimental pharmacy checker that asks Jev the same question five ways and looks for agreement.

Source screenshot of jev-labs
SOURCE SCREENSHOT · source ↗ · captured 2026-09-22Full screenshot ↗

What it does

Unresolved cases go to a person. This is a safety study, not a clinically validated system.

Maker-reported (not independently measured by JevMade): 1,680 simulated pharmacy rounds: 0 wrong verdicts in 1,080 golden rounds, with escalation rising from 5.0% to 18.0% under severe chaos (author-reported) · 1,490 hash-verified live calls to jev-1.13.0: identity noise floor 0.042, question-reorder 0.059, paraphrase cohort 0.073 (author-reported) · Latency flat in question count: 96.7 ms mean for one question (n=1,161) and 98.0 ms for 38 (author-reported) · TLC model check at 5 agents / quorum 3 / 2 crashes: 1,049,750 distinct states, 0 errors (author-reported) · Zero wrong verdicts in 1,080 golden rounds bounds the violation rate below 0.28% at 95% confidence by the rule of three (author-reported)

Primitives
noul
Platform
Rust
Added
Project created
GitHub stars
1 · snapshot 2026-09-22

Source checked 2026-09-22 — opened the primary source directly.