Directory

Every entry here was opened and read before it was listed. Each one carries the URLs that were checked.

Showing 23 of 23

• 
  Guides & coverage

  What Is Jev? (Firecrawl)

  The best single starting point among the third-party explainers. It gets past the headline numbers into six concrete jobs a decision model is actually good at, including reranking retrieval results, checking an agent's citations against the page they came from, and screening fetched pages before they reach a model that will act on them.

  Firecrawl2026-09-23
• 
  Model & officialOfficial

  Jev

  The model this site is about. Jev takes your program state plus a set of typed questions and answers all of them in one pass, returning a choice, a score or a probability rather than a sentence. Because the answer is picked from a schema you wrote, it cannot come back off-schema. TypeSafe describes it as the first of a class they call System One models, and publishes speed and price comparisons against chat models; those figures are the vendor's own.

  TypeSafe AI2026-09
• 
  Evaluations

  Jev out-of-distribution calibration test

  Asks the question that matters for a model sold on calibration: does the calibration survive on a task the model cannot have seen? It pairs public benchmarks with rule-generated tickets whose labels never appear in the text, and reports error against a noise floor rather than in isolation. The useful conclusion is that miscalibration differs by question type, so one temperature for the whole model is the wrong unit.

  scienthoon7 stars2026-09-19
• 
  Open implementations

  LocalJev

  A bridge rather than a reimplementation: it keeps the request and response shapes, but gets its probabilities by asking an ordinary local chat model for a JSON number and then validating and normalising the result. That lets it run against a plain OpenAI-compatible server with no patched backend, and the README is explicit that a self-reported probability is not the same thing as a logit read. Check the calibration before you threshold on it.

  GitHub Next776 stars2026-09-18
• 
  Open implementations

  OpenJev

  An open decision server that speaks the same wire protocol on an open model, so the official SDKs can be pointed at it. Its distinguishing choice is to read answers out of the model's own token probabilities instead of parsing generated text, which is closer to what the hosted model does. The price is a patched inference backend and a GPU or Apple silicon machine to run it on.

  razorback16427 stars2026-09
• 
  Evaluations

  PriorBench: pre-registered evaluation of Jev

  The methodologically strictest evaluation listed here: fifty predictions were written down and timestamped before any data was collected, and the repository reports how many of them were falsified as prominently as how many held. The finding that travels furthest concerns the confidence value, which the authors found flat across most of its range and informative only at the very top.

  PriorBench0 stars2026-09-20
• 
  Model & officialOfficial

  Introducing System One Models & Jev

  The launch post, and the clearest statement of the design bet: give up string generation entirely, and train instead for probabilities that mean what they say. It names the training approach Reinforcement Learning for Calibrated Decisions, and frames the model as a function call rather than a conversation. Read it for the reasoning; read the evaluations elsewhere on this site for how it holds up.

  TypeSafe AI2026-09
• 
  DocumentationOfficial

  TypeSafe AI documentation

  Start here if you are writing code. The docs are organised around the three question primitives, then around what to do with the confidence value: threshold it, route on it, or decompose a hard question into several easy ones. There is an llms.txt index if you would rather hand the whole thing to an agent.

  TypeSafe AI2026-09
• 
  Gateways & access

  Jev on Vercel AI Gateway

  A route to the model that does not need a TypeSafe account of your own: the gateway exposes Jev under its own model id, and the AI SDK calls it through a dedicated evaluate function rather than a chat completion. The gateway lists the model id typesafe-ai/jev and the same published input price as TypeSafe. Worth knowing about, because one of the independent evaluations listed here — the out-of-distribution calibration test — ran its calls through it.

  Vercel2026-09-16
• 
  DocumentationOfficial

  Jev 1.13 known failure modes

  A vendor page documenting where its own model is uneven, which is rarer than it should be. The examples matter in practice: the same question asked as a boolean and as a choice can disagree sharply, and a question and its negation, asked separately, do not have to sum to one. If you are about to put a threshold on a probability, read this first.

  TypeSafe AI2026-09-17
• 
  Model & officialOfficial

  Jev model card (ids, context, limits)

  The page to check before you pin anything: the current model id and the moving alias, how the context budget is split between the state and the longest question, the published rate limits, and the note that English is where accuracy is currently best. It also links out to the known-failure-mode page for the specific version you are calling.

  TypeSafe AI2026-09
• 
  Evaluations

  Jev on agent tool-call risk

  A small, honest benchmark on one realistic job: sorting an agent's tool calls into risk classes before they run. The set is hand-labelled and deliberately mixed in difficulty, several backends are run against an identical task definition, and the write-up is clear that this measures one-shot classification, not the multi-step workflows the vendor's comparisons are drawn from.

  themsquared1 stars2026-09
• 
  Integrations & tools

  TypeSafeClassifier for LangChain agents

  LangChain's own write-up of putting a decision model inside the agent loop, with two concrete uses: routing a request to a model that matches its difficulty, and a middleware that inspects a tool call before it runs and flags the risky ones. Useful as a pattern even if you do not use LangChain, because it shows where in a loop a fast typed judgement actually pays for itself.

  LangChain2026-09-17
• 
  Gateways & access

  LiteLLM TypeSafe pass-through

  If you already run LiteLLM as a proxy, this forwards the decision endpoint through it: clients hold a virtual key, the real key stays in the proxy environment, and usage is counted like any other model. The same page documents using Jev as the chooser inside LiteLLM's auto-router, which is a neat fit for a model whose output is already a choice.

  BerriAI (LiteLLM)2026-09
• 
  Evaluations

  LocalJev model bake-off (2026-09-18)

  Not a test of Jev, but of what you give up by running a compatible server on local open models. The same engine is run against five installed checkpoints on three public gold-labelled tasks, twice: once on short inputs, and once with the same target buried in thousands of words of unrelated text. The second condition is where the smaller models come apart.

  GitHub Next2026-09-18
• 
  Guides & coverage

  A coding guide to Jev (MarkTechPost)

  A hands-on walkthrough that builds up from a single question to the patterns you end up needing anyway: thresholds kept in code so risk tolerance stays reviewable, many questions batched into one call so the state is only sent once, and narrow atomic judgements combined with weights you control instead of one broad question.

  MarkTechPost2026-09-23
• 
  Guides & coverage

  TypeSafe AI releases Jev (MarkTechPost)

  A short news write-up of the launch that is careful about whose claim is whose. It repeats the published price and latency figures, notes that the comparison demo is the vendor's own, and points out that a price cannot be shown to be unsubsidised from the outside. Good for a quick factual picture of what actually shipped.

  MarkTechPost2026-09-19
• 
  Integrations & tools

  pi-typesafe

  A sidecar for the Pi coding agent that runs a handful of hazard checks before a shell command, a write or an edit goes through, and can be set to advise, to watch silently, or to block. Small and early. It is listed because it is the library that the pi-warden guardrail tools build on; pi-warden itself exists as many near-identical forks whose original could not be established, so none of those are listed here.

  twilwa1 stars2026-09
• 
  DocumentationOfficial

  System One HTTP API reference

  One endpoint, one POST, three top-level fields: the state to judge, the model, and a map of named questions. The response comes back keyed by the same names you sent, with token usage attached. Worth reading even if you use an SDK, because it is the shortest complete description of the protocol, including which status codes you are expected to retry with backoff.

  TypeSafe AI2026-09
• 
  DocumentationOfficial

  TypeSafe cookbooks

  End-to-end recipes grouped by the shape of the problem: self-consistency, batching many questions into one call, re-ranking a retrieval shortlist, recovering structure from flattened text, checking citations against their source, and screening input before it reaches a model that will act on it. The batching recipe is the one that most changes how you write the first call.

  TypeSafe AI2026-09
• 
  SDKs & clientsOfficial

  @typesafe-ai/sdk (JavaScript / TypeScript)

  The official JavaScript and TypeScript client. Helper functions build each question type, and the client reads the API key from the environment, so a first typed call is a few lines long. Node 20 or newer is required.

  TypeSafe AI2026-09
• 
  SDKs & clientsOfficial

  typesafe-sdk (Python)

  The official Python client, with both a synchronous and an asynchronous client class. Questions are built as typed objects rather than dictionaries, so the shape of the answer is known before the call returns, and the response exposes each answer under the name you gave its question.

  TypeSafe AI2026-09
• 
  Guides & coverage

  How to use Jev: a practical guide

  A guide that is unusually direct about the limits. Its framing is that a decision model is a router, not a replacement: cheap deterministic code for the easy path, a large model for the genuinely hard one, a person for the ambiguous middle. It also states the failure mode plainly, which is that a schema stops an invalid answer but never a wrong one.

  Valyu AI2026-09

Nothing matches that. Try a shorter word, or clear the category filter.