Jev 1.13 jaggedness
TypeSafe's documentation of where Jev 1.13 is known to fail: counting, dates, indirection, distractors, adversarial state and more.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
This is TypeSafe's own documentation page on the jaggedness of Jev 1.13, meaning the areas where the current public model is known to be weak. According to the list summaries, it covers literal understanding, unreliable counting and arithmetic, date comparison, multiple levels of indirection, large state with irrelevant details, adversarial content, contradictions between instructions and criteria, structural invariants, and generation.
The official recommendation, quoted in one list, is to split such problems into atomic questions and combine the answers in code. It is essential reading before relying on Jev for anything involving numbers, dates or long noisy inputs, and it is a first-party admission of limits rather than a third-party critique.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Which task types does Jev 1.13 handle unreliably?(inferred) | — |
How it works
- Read the known failure categories
- Identify which apply to your questions
- Split risky problems into atomic questions
- Combine answers deterministically in code
Strengths
- First-party list of failure modes
- Gives a concrete mitigation strategy
Caveats
- Specific to version 1.13; later versions may differ
At a glance
- Type
- Article
- Status
- production
- Has code
- no
Use cases and integrations
How curated lists describe it
Documentation: TypeSafe's own page on model jaggedness for the 1.13 release.
yibie/awesome-jevLiteral understanding, **unreliable counting**, date comparison, multiple levels of indirection, large state with irrelevant details, adversarial content, contradictions between instructions and criteria, structural invariants, and generation. The official recommendation is to "split them into atomic problems and combine them in code".
wh000wh000/awesome-jev-liveFirst-party limitations covering arithmetic, dates, distractors, adversarial state, and structural inconsistencies.
cobanov/awesome-jevKnown failure modes of the current public model, documented by TypeSafe.
hellogumbo/awesome-jevWhere the current model is known to be weak.
MrJev/awesome-jev
Frequently asked questions
What is Jev 1.13 bad at?
Per the lists: counting, dates, multi-level indirection, large noisy state, adversarial content, conflicting instructions and generation.
What does TypeSafe recommend?
Split hard problems into atomic questions and combine the results in code.
Is this an independent review?
No, it is TypeSafe's own documentation.
Related projects
Library of 43 field-tested Jev question patterns with templates, thresholds and failure modes, plus 10 anti-patterns; CC0, EN/ZH.
ArticleproductionTypeSafe cookbook on why repeated Jev calls can differ, showing how an abstention rule raises repeat agreement.
ArticleproductionFlavio Copes' hands-on walkthrough of Jev and the System One idea, aimed at developers who have only seen LLM marketing.
ArticlediscussionIndependent multilingual guide site covering Choice, Score, Noul, Python SDK requests, thresholds, fallbacks and human escalation.
ArticlePythonAgent skill and CLI that chains Jev question design: find decision calls, check fit, draft, lint, calibrate thresholds and plan a cascade.
Agent skillprototypePythonPython sandbox turning Jev issue-category and reproduction judgments into testable routing proposals, with offline replay.
ProjectprototypePython