jev-demos
Small demo collection testing Jev's spatial reasoning with maze single-step choices versus multi-step lookahead.
Overview
jev-demos is a small collection of demos using TypeSafe's System One models. It contains a shared package (jev_common) that wraps the API, loads .env and records results, and a maze_lookahead package that checks Jev's spatial reasoning and whether it can think ahead, comparing single-step move choices with multi-step lookahead.
The README is brief and does not report results. It suits readers curious about how a decision model handles planning-like tasks.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which direction should be taken next in the maze?(inferred) | — |
How it works
- Put the API key in .env
- Run the maze demo with single-step or lookahead settings
- Record results through jev_common
Strengths
- Tests a task that stresses planning rather than classification
- Shared result recording
Caveats
- No results in the README
- Minimal documentation
Getting started
Put your API key into .env and follow the demo-specific instructions.At a glance
- Type
- App
- Status
- prototype
- Language
- Python
- Runtime
- Scripts
- Has code
- yes
Use cases and integrations
How curated lists describe it
Maze experiments comparing Jev single-step choices with multi-step lookahead.
logicrw/awesome-jev-projectsDemos to test the effectiveness of TypeSafe's "Jev" System One Model.
hellogumbo/awesome-jev
Frequently asked questions
What is tested?
Jev's spatial reasoning in mazes, with and without lookahead.
Are results published?
Not in the README.
What do I need?
A TypeSafe API key in .env.
Related projects
Independent field guide to Jev 1.13.0 behaviour across 11,621 requests: framing sensitivity, Snake and 3D city driving experiments.
ProjectresearchPython, TypeScriptNine-board gomoku workbench comparing how two Jev players respond to five different input representations of the board.
ProjectresearchTypeScriptRPG experiment pitting Jev against OpenAI's Decisions API as NPCs, comparing wins, per-move accuracy, calibration and cost.
ServiceresearchPythonShort evaluation finding that Jev plays chess poorly, often calling unbalanced positions draws.
ProjectresearchProject Zomboid performance patches whose benchmark harness computes numbers in code and, per a list, asks Jev only for the verdict.
BenchmarkbetaJavaBenchmarks Jev against Luna, Haiku, Gemini and Astra at choosing legal moves in tic-tac-toe and Connect Four.
BenchmarkprototypeTypeScript