Every key fact has a source.
A source register with verification levels. Not every link could be retrieved in full; for preprints, the abstract was checked.
- S01Models ↗
TypeSafe AI · Primary source retrieved
Pricing, limits, versions and languages. As of 23 September 2026, the aliases jev-latest and jev-preview point to jev-1.13.0. This is vendor documentation, not the result of our own API test.
- S02Introduction ↗
TypeSafe AI · Primary source retrieved
- S03Choice ↗
TypeSafe AI · Primary source retrieved
- S04Score ↗
TypeSafe AI · Primary source retrieved
- S05Noul ↗
TypeSafe AI · Primary source retrieved
- S06Confidence ↗
TypeSafe AI · Primary source retrieved
- S07Jev 1.13 jaggedness ↗
TypeSafe AI · Primary source retrieved
Limitations acknowledged by the vendor. The page states a review date of 17 September 2026.
- S08API reference ↗
TypeSafe AI · Primary source retrieved
- S09AI primer ↗
TypeSafe AI · Primary source retrieved
- S10Introducing System One Models & Jev ↗
TypeSafe AI · Primary source retrieved
- S11Team ↗
TypeSafe AI · Primary source retrieved
- S12Legal ↗
TypeSafe AI · Primary source retrieved
- S13Master customer agreement ↗
TypeSafe AI · Primary source retrieved
- S14Privacy policy ↗
TypeSafe AI · Primary source retrieved
- S15Data processing addendum ↗
TypeSafe AI · Primary source retrieved
Do not add a legal conclusion on GDPR compliance to the publication without a legal review.
- S16Jev 1.13 – model catalog ↗
OpenRouter · Primary source via search index
When loaded directly, the catalog page returns 404; the 32,000-token context and the rate of $0.042 were confirmed by a search engine index on 23 September 2026. Operational availability percentages are not reproduced.
- S17What Is Jev? TypeSafe’s Decision Model Explained for Developers ↗
OpenRouter · Primary source via search index
The Decisions API interface and examples; these were not our API calls.
- S18Insights – classification benchmark summary ↗
OpenRouter · Primary source via search index
The index confirms the date, Banking77, n=3080, the accuracy, the cost and the median. It does not verify every detail of the original article’s methodology.
- S19TypeSafe emerges from stealth with a new way of doing AI ↗
DCVC · Primary source retrieved
Announcement by the lead investor: $40 million Series Seed, 15 September 2026. The investor’s performance claims come from an interested party.
- S20Laya – model card ↗
Convai Innovations · Primary source retrieved
A different model from Jev; without a controlled comparison, the author’s own benchmark figures are not used to declare a winner.
- S21Jev is the fastest-adopted model in AI Gateway history ↗
Vercel · Primary source via search index
Vercel’s own data. The promotion until 25 September 2026 is stated in the index; current account eligibility was not tested.
- S22Jev for Scientific Decisions: Evaluating Semantic Choices and Their Consequences ↗
Boyuan Deng et al. / arXiv · Preprint abstract
Abstract read, not the full paper or a replication of the experiment. Preprint. Abstract retrieved directly from arXiv on 23 September 2026.
- S23Visual Jev: Accurate and Efficient Decisions from Shared Visual Context ↗
Guanxu Yu, Yuhang Yao / arXiv · Preprint abstract
A separate research project. It does not imply image input in the TypeSafe Jev API. Abstract read. Abstract retrieved directly from arXiv on 23 September 2026.
- S24Fast Intent-Driven Service Orchestration with Jev for 6G Edge Networks ↗
Delong Li et al. / arXiv · Preprint abstract
Abstract read; live calls combined with simulation. Not proof of general operational accuracy. Abstract retrieved directly from arXiv on 23 September 2026.
- S25Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration ↗
Delong Li et al. / arXiv · Preprint abstract
Abstract read. Same authors as S24; not to be treated as two fully independent confirmations. Abstract retrieved directly from arXiv on 23 September 2026.
- S26Alternative dispute resolution rules – document page ↗
SK-NIC · Primary source via search index
A historical document page, effective from 1 September 2020. The current complete rules and any specific case should be checked by a lawyer.
- S27Jev: model, který nepíše, jen rozhoduje. Jak ho použít a kdy se vyplatí ↗
Vibecoding.cz / Patrick Zandl · Primary source retrieved
Full article retrieved on 23 September 2026. A test of internal linking across 209 articles on vibecoding.cz; the reference was the Opus and Sonnet models, not human evaluation.
- S28TypeSafe Console – login page ↗
TypeSafe AI · Access not confirmed
On a follow-up attempt, the login page could not be loaded. Neither access nor credit for a new account was verified.
- S29Is Jev as Accurate as Frontier Models at Classification? – full article ↗
OpenRouter · Primary source retrieved
Full article retrieved on 23 September 2026, including p95, confidence intervals and the cascade.
- S30Documentation index ↗
TypeSafe AI · Primary source retrieved
A map of the documentation; as of 23 September 2026, it lists no model newer than jev-1.13. Not proof that a paid account is available.
- S31Wilson Sonsini Advises TypeSafe AI on $40 Million Seed Round ↗
Wilson Sonsini · Primary source via search index
A direct load on 23 September 2026 returned 403; the investment and date are confirmed by a search engine index and by the investor announcement S19. Not an independent evaluator of the model.
- S32TypeSafe AI’s Jev now available on AI Gateway ↗
Vercel · Primary source via search index
The native integration uses the experimental evaluate and the boolean type. Do not carry over a raw TypeSafe noul without conversion. The promotion and the eligibility of any specific account were not tested.
- S33Current .sk domain rules – index page ↗
SK-NIC · Primary source via search index
Verified current index page. No trademark search and no legal conclusion regarding any specific domain.
- S34JEV-as-a-Judge: Accept When Confident, Escalate When Unsure ↗
arXiv (Yubo Li, Yidi Miao, Ramayya Krishnan, Rema Padman) · Primary source retrieved
- S35Confidence-Gated Routing ↗
TypeSafe AI · Primary source retrieved
- S36How to Use Jev: Moderation with the Jev API in TypeScript ↗
OpenRouter · Primary source retrieved
- S37Authentication ↗
OpenRouter · Primary source retrieved
- S38Jev Lab ↗
OpenRouter · Primary source retrieved
- S39Auto-Approve Coding Agent Permission Prompts with Jev ↗
OpenRouter · Primary source retrieved
- S40Hooks reference ↗
Anthropic · Primary source retrieved
- S41Prompt to questions ↗
OpenRouter · Primary source retrieved
- S42To Jev or Not? Evaluating the Accuracy and Efficiency of Structured Decision Models for Hate-Speech Moderation ↗
arXiv (Paschalides, Pallis, Dikaiakos) · Primary source retrieved
- S43AI Gateway adds confidence-based decision fallbacks ↗
Vercel · Primary source retrieved
- S44Decision Fallbacks ↗
Vercel · Primary source retrieved
- S45Labels Override Definitions in Jev-Style Typed Decision Models ↗
arXiv (preprint, University of Southern California) · Primary source retrieved
- S46Specialized Decision Models vs. General-Purpose LLMs: Benchmarking Jev Across Knowledge, Reasoning, and Multilingual Tasks ↗
arXiv · Primary source retrieved
- S47When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows ↗
arXiv (Ling, Chen, Fang, Yu) · Primary source retrieved
- S48Same-Number Citation Swaps: Stress-Testing Jev as a Financial Evidence Judge ↗
arXiv · Primary source retrieved
- S49Nudge the Judge: Uncovering Program Decision Hijacking in Real-World Jev Applications ↗
arXiv · Primary source retrieved
- S50How to build with TypeSafe ↗
TypeSafe AI · Primary source retrieved
- S51Images ↗
TypeSafe AI · Primary source retrieved
- S52Box ↗
TypeSafe AI · Primary source retrieved
- S53Changelog ↗
TypeSafe AI · Primary source retrieved
- S54Does a Hosted Decision Model Respect Irrelevant Alternatives? A Pre-Registered Test of Option-Set Dependence in Jev ↗
arXiv (Rajput, Chauhan, Kosaraju; AltSlate Labs LLP) · Primary source retrieved
- S55Jev-as-a-Judge for RAG Claim Verification: Confidence-Gated Cascading Between a Non-Generative Verifier and a Frontier LLM ↗
arXiv · Primary source retrieved
- S56Jev-LiteMAS: Separating Decisions from Generation in Multi-Agent Workflows ↗
arXiv (Wang, Zhou, Lou, Li) · Primary source retrieved