What it is
A Rust CLI called typesafe-ai built with usage-rs, offering experiments around Jev: PHI detection, code-comment review, parts-of-speech and load-bearing analysis, interactive tone analysis, and business and occupation classification.
How it uses Jev
Jev answers typed questions over state. Load-bearing defines its question and five rubric levels once in shared state, then gives each eligible line its own Score question in one batched request per file, sharing the full file as context. Scores are divided by four to normalize them.
Primitives:choicescore
Technique worth stealing
Define the question and rubric levels once in shared state, then issue per-line Score questions in a single batched request sharing full-file context.
Try it
cargo install --path . --locked; export OPENROUTER_API_KEY; typesafe-ai phi patient-notes.txt
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-centricunclear0.54
- Shows a System One patternno0.13
- Handles uncertaintyno0.04
- Measuredno0.04
- Runnableyes0.92
- Worth recommendingno0.25
- Model replicano0.04
- Problem scopescore on a 0–2 scale0.60
- About Jevyes0.97
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 22 Sept 2026.