jevbooks

← All projects

jev-heart-risk-bench

rubinagentagi-tech/jev-heart-risk-bench

Benchmarking Jev on 5,000 CDC heart-risk survey respondents.

What it is

Scores Jev (TypeSafe System One) on 5,000 real CDC BRFSS 2015 respondents against logistic regression, a hand-written rule, a chat LLM, and the base rate. For researchers evaluating decision models.

How it uses Jev

Jev reads plain-English survey answers and judges heart-disease risk. It returns typed output with a checkable reason. The result is used as a ranking score, not a calibrated probability; the cut-off is set in code.

Primitives:choice

Technique worth stealing

Use Jev for selection among plausible options, not calibrated probability of a rare event.

Try it

Open the interactive demo: https://rubinagentagi-tech.github.io/jev-heart-risk-bench/

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.83
  • Shows a System One patternno0.25
  • Handles uncertaintyno0.05
  • Measuredno0.26
  • Runnableno0.11
  • Worth recommendingunclear0.42
  • Model replicano0.04
  • Problem scopescore on a 0–2 scale0.26
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.