What it is
PriorBench is an independent, pre-registered evaluation of TypeSafe AI's Jev, a System One model. It runs 21 experiments and 5,721 calls, publishing raw data, predictions, and a report for researchers and practitioners.
How it uses Jev
The README does not describe how Jev is used in code. It reports measurements of Jev's Choice, Score, and Noul primitives across experiments, but does not show integration or decision-making logic.
Primitives:choicescorenoul
Technique worth stealing
Pre-register predictions before data collection and publish falsified results.
Try it
git clone, install requirements, set OpenRouter key, run latency/bench.py, probes/t6_banc.py, analysis/recount.py.
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-centricyes0.86
- Shows a System One patternunclear0.58
- Handles uncertaintyno0.06
- Measuredyes0.96
- Runnableno0.04
- Worth recommendingno0.35
- Model replicano0.03
- Problem scopescore on a 0–2 scale1.15
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.