iambraun.com · Jev reports

Jev reports

Two measured studies and two running indexes of what is built on Jev, TypeSafe AI's typed-decision model.

Studies

  1. BenchmarkPublished 17 September 2026

    Worth asking?

    A typed-decision model wired into five places in a Dungeons & Dragons co-DM, measured on four benchmarks (table rules, routing, contradictions, asset kinds) against the code it replaced, a frontier model called as a classifier, and Laya, the open-weight model marketed as its replacement.

  2. BenchmarkPublished 17 September 2026

    Skill Selection, Measured

    Six arms on 8,916 skill-selection decisions each (743 skills across 12 job postings), plus 240 of the hardest pairs adjudicated twice: Jev, GLM-5.3-flash, Haiku 4.5, two embedding arms, a substring matcher, and Laya.

Indexes

  1. IndexRefreshed twice daily since 19 September 2026

    Jev radar

    Tools, demos and findings from the TypeSafe Discord, GitHub and X, each scored 1 to 5 and tagged; searchable by text, tag, source, score and repository stars.

  2. IndexRefreshed twice daily since 19 September 2026

    Jev in production

    Named products and companies running Jev in production, each with a professional-usefulness score and its source, and a weekly highlight.