Automated README summary
Automatically summarized from the README on 20 Oct 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.
What it is
A reproducible benchmark comparing Jev (TypeSafe AI's System One model) against Claude Haiku 4.5 on one security decision: should an email agent click the link in this email? It measures accuracy, calibration, latency and cost on 2,000 PhishNChips v5.2 emails.
How it uses Jev
Each email is sent as a JSON state with nine typed questions in one request: a verdict Choice, a mirror Noul, five signal Nouls (domain mismatch, free hosting, lure, urgency, generic sender) and two alternative verdict wordings. The returned probabilities are compared against the LLM and used in rule and regression baselines.
Primitives:choicenoul
Technique worth stealing
Ask all verdict wordings and signal questions in one call so wording sensitivity is measured without post-hoc selection.
Try it
cp .env.example .env, fill keys, then uv run prepare_data.py and uv run run_jev.py --limit 10.
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 material0.63
- Shows a System One patternnot found in the excerpt0.19
- Describes handling uncertaintynot found in the excerpt0.10
- Reports measured numbersnot found in the excerpt0.27
- Is a runnable toolnot found in the excerpt0.06
- Top-10 learning shortlist signalnot found in the excerpt0.30
- Is a replica of the modelnot found in the excerpt0.09
- Solves a common problem (0–2)score on a 0–2 scale1.72 / 2
- About JevFound in material0.98
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 20 Oct 2026.