Automated README summary
Automatically summarized from the README on 28 Sept 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A companion repo for the 'DIY Jev' blog post. It builds a synthetic data-generating process with an exact closed-form posterior and uses it to probe TypeSafe AI's Jev API, train a small Jev-like toy model, and run head-to-head comparisons.
How it uses Jev
Jev is probed via its API on six pre-registered test suites (S1–S6) using synthetic emails. The generator's known true posterior grades Jev's returned probabilities, rather than noisy observed labels. The toy model mimics Jev's architecture for a soft-vs-hard-label training comparison.
Primitives:choicescorenoul
Technique worth stealing
Use a synthetic data-generating process with an exact closed-form posterior as ground truth to evaluate calibration.
Try it
uv sync --dev; uv run pytest tests/ -q; uv run python scripts/run_toy_experiment.py; uv run python -m jev.probe.run_probe --backend live --suites S1,S2,S3,S4,S5,S6 --n 200
Model judgments from README excerpts
Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.
Models & evaluation dates
- Inclusion gate
jev-1.13.0- Project signals
jev-1.13.0- Category
jev-1.13.0- Pattern tags
jev-1.13.0
Evidence
These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.
- Jev is core to the projectFound in material0.91
- Shows a System One patternnot found in the excerpt0.22
- Describes handling uncertaintynot found in the excerpt0.06
- Reports measured numbersnot found in the excerpt0.16
- Is a runnable toolnot found in the excerpt0.07
- Top-10 learning shortlist signalnot found in the excerpt0.39
- Is a replica of the modelFound in material0.88
- Solves a common problem (0–2)score on a 0–2 scale0.76 / 2
- About JevFound in material0.99
no-judgment band (0.40–0.60)
no-judgment band (0.40–0.60)
The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 28 Sept 2026.