What it is
A web lab for 2048 where every move is decided by TypeSafe AI's Jev System One model, exposing probability, confidence, latency and token cost. Aimed at researchers studying how context design affects decision models.
How it uses Jev
Each turn sends the board state plus a Choice question over the four directions; Jev returns calibrated probabilities. A decision strategy converts the distribution into an actual move, gated by a user-set confidence threshold that pauses play for human review. No heuristic fallback.
Primitives:choice
Technique worth stealing
Expose uncertainty instead of silently swapping in a rule algorithm, so every move stays attributable to the model.
Try it
git clone, npm install, ./start.sh
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-centricyes0.97
- Shows a System One patternyes0.91
- Handles uncertaintyyes0.89
- Measuredno0.07
- Runnableyes0.66
- Worth recommendingyes0.67
- Model replicano0.03
- Problem scopescore on a 0–2 scale0.09
- About Jevyes0.99
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.