https://keta1930.github.io/what-the-jev/New and updated examples, experiments, and resources.enSat, 10 Oct 2026 15:27:32 GMThttps://keta1930.github.io/what-the-jev/about/urn:what-the-jev:en:about:04e0e3a164c21d71c3187fc7fdd43bb3968574a2829aa9ae04664edff54b3bbeSat, 10 Oct 2026 15:27:32 GMTwhat-the-jev showcases what Jev (TypeSafe AI's "System One" decision model) can do across a wide range of tasks. This project provides a stable Jev experimentation framework for running tasks at scale with reproducible results. We use this framework to conduct experiments and provide reproducible results and analysis r https://keta1930.github.io/what-the-jev/resources/urn:what-the-jev:en:resources:bf4ad77d4196c59e4738d7696566cddbf9d1ee9effa0fc64aa8c6c949e464b41Sat, 10 Oct 2026 05:29:30 GMTA catalog of Jev-related models, projects, tools, and benchmarks. Alongside this repository's own experiments (example/ and experiments/), it collects community projects built around Jev and the "System One" decision-model category it started. https://keta1930.github.io/what-the-jev/resources/open-source-models/chinese-laya/urn:what-the-jev:en:resources/open-source-models/chinese-laya:ab0bf2cc8dfb9d1ad151b38fbbf666398691817fd00e7615c17cabc769f4bf9aSat, 10 Oct 2026 05:29:30 GMTChinese fine-tune of Laya's multilingual checkpoint, with machine-translated data splits and author-reported test results. https://keta1930.github.io/what-the-jev/resources/open-source-models/startlux-decision/urn:what-the-jev:en:resources/open-source-models/startlux-decision:988efc3182daa1d9a718c27d2f664a762c6d4900519cfb71e8519e165f40b23eSat, 10 Oct 2026 05:29:30 GMTTyped decision family from 0.8B to 35B-A3B reading text, JSON, and images, with non-commercial weights. https://keta1930.github.io/what-the-jev/examples/ethics-dilemmas/urn:what-the-jev:en:examples/ethics-dilemmas:2611def1db30b9584b08afc387daad2b195a2e951f8437829c9cac6b546cc5c2Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests the Jev model's judgments on classic ethical dilemmas. https://keta1930.github.io/what-the-jev/examples/math-word-problem/urn:what-the-jev:en:examples/math-word-problem:cd0e7835142aebdfaabe5d385986a47b7d2a8eb137b672a591996afed4062961Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests the Jev model's ability to answer elementary arithmetic word problems. https://keta1930.github.io/what-the-jev/examples/paper-qa/urn:what-the-jev:en:examples/paper-qa:facb2402d1ab964f0edc32d3bb698880dd2f19103b8a1fa33c76d74c15b3cd8aSat, 10 Oct 2026 03:25:20 GMTThis experiment tests whether the Jev model can answer questions about a research paper from only part of its text. https://keta1930.github.io/what-the-jev/examples/pixel-recognition/urn:what-the-jev:en:examples/pixel-recognition:9d47eb75b85f6591c2efedd8aed0a5c30e39d72471e27ca93163b848489c7cf0Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests whether the Jev model can recognize image content from pixel values alone. https://keta1930.github.io/what-the-jev/examples/prompt-injection-detection/urn:what-the-jev:en:examples/prompt-injection-detection:4b2216e8d33925939be0c9a7b25bd29fe1c5c07a105d9e42a9fabc49f552f9e8Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests whether the Jev model can detect prompt-injection attacks hidden in tool-call results within multi-turn agent conversations. https://keta1930.github.io/what-the-jev/examples/ticket-triage/urn:what-the-jev:en:examples/ticket-triage:4e92fa5ba69682a5000e45091611882ffcdc11266f97c8ecc01d76330634b5aeSat, 10 Oct 2026 03:25:20 GMTThis experiment tests the Jev model's judgment on customer support tickets. https://keta1930.github.io/what-the-jev/examples/university-math/urn:what-the-jev:en:examples/university-math:b8c066ac626be3fce4c6fe4a8c61d7fb167d743a04e337db327d1fbc8ee71937Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests the Jev model's ability to answer university-level computational math problems. https://keta1930.github.io/what-the-jev/examples/us-election/urn:what-the-jev:en:examples/us-election:52b8da6aebbca40f32c7efc54f114cd3c12be983dce48aadb7485e32eb5ace46Sat, 10 Oct 2026 03:25:20 GMTThis experiment tests whether the Jev model knows the outcomes of the eight US presidential elections from 1996 to 2024. https://keta1930.github.io/what-the-jev/experiments/bbq/urn:what-the-jev:en:experiments/bbq:fdf259fa60081b1137cb9ece04498db5ca66d79771205c29b4fa69eebd844fd8Sat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on BBQ's bias-sensitive questions, and whether it leans toward stereotypes. https://keta1930.github.io/what-the-jev/experiments/boston-housing/urn:what-the-jev:en:experiments/boston-housing:c6479424082120ae0a184e36f60ee025508331b1bdc085c40cf734dbe2c3466dSat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can predict Boston suburb housing prices, and compare price levels across suburbs. https://keta1930.github.io/what-the-jev/experiments/dpo-jev-judge/urn:what-the-jev:en:experiments/dpo-jev-judge:bd48defcb2519f33aa88c846d2212d87aded57fe45fe66b27e64b29dcdabb3afSat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on preference annotation for DPO training data. https://keta1930.github.io/what-the-jev/experiments/gpqa-diamond/urn:what-the-jev:en:experiments/gpqa-diamond:68892b491301a21e822d92ac644e87781adb38fc2cb09948ac773da59d0a91a5Sat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on GPQA Diamond graduate-level science questions. https://keta1930.github.io/what-the-jev/experiments/grpo-jev-judge/urn:what-the-jev:en:experiments/grpo-jev-judge:3393e0a8fd0bced791ee5109ee77ed53e3060e5effccca4637a0c02accb59a5bSat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on GRPO trajectory reward assignment. https://keta1930.github.io/what-the-jev/experiments/gsm8k/urn:what-the-jev:en:experiments/gsm8k:82a1dc40da32a13568f28cc1f3bf0355c5f7e365d1f68fa03ef86352ae56697bSat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on GSM8K grade-school math word problems. https://keta1930.github.io/what-the-jev/experiments/mmlu-pro/urn:what-the-jev:en:experiments/mmlu-pro:52954ce808050553633d6c5e6624642b0f75dda0c2b1afe5eb8716bc17c5a42eSat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on MMLU-Pro college-level questions across 14 disciplines. https://keta1930.github.io/what-the-jev/experiments/paper-classification/urn:what-the-jev:en:experiments/paper-classification:9096f641299339fb118baaaedb11957d6e59793591dd6766ac463bb9afd05cb4Sat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can decide, based on a research preference, whether an arXiv paper belongs in a paper library, and which topic it falls under. https://keta1930.github.io/what-the-jev/experiments/paper-qa/urn:what-the-jev:en:experiments/paper-qa:3508e95c04cbe4bbabd24f383c624d0fbec034190d270f116b72d9b34fbcdb53Sat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can answer questions about a paper based on its own text. https://keta1930.github.io/what-the-jev/experiments/prompt-routing/urn:what-the-jev:en:experiments/prompt-routing:02f2b5fa32eb70d86fae4ae4381f252fc1e13708c2e4c72b20c5e049fb48ad1aSat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can decide from the question text alone whether a question in paper QA goes to the JEV fast path or the LLM slow path. https://keta1930.github.io/what-the-jev/experiments/skill-routing/urn:what-the-jev:en:experiments/skill-routing:8d6d006972d005cea0c5ff700d47e4f36b10db06b93f420412f687f798c64d8eSat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can route user tasks to the right skill, skill set, or none, among 126 real agent skills. https://keta1930.github.io/what-the-jev/experiments/socialiqa/urn:what-the-jev:en:experiments/socialiqa:ff1b52b94622bac03f19b6e5afeed96a7409f13bf72f3bd041d67e798786663aSat, 10 Oct 2026 02:21:18 GMTExplore JEV's performance on SocialIQA social commonsense questions. https://keta1930.github.io/what-the-jev/experiments/titanic/urn:what-the-jev:en:experiments/titanic:c4b58ea1c4ae1a00fd9f33930071248b034e632050d6a431baaf18956e27da3cSat, 10 Oct 2026 02:21:18 GMTExplore whether JEV can predict whether a Titanic passenger survived from the passenger record, and compare structured fields with prose text. https://keta1930.github.io/what-the-jev/examples/urn:what-the-jev:en:examples:cf56216c9af82aa41536061c4d0e89c1e194164bb23c643b5a445de50b429578Sat, 10 Oct 2026 02:21:09 GMTEight small-sample examples, grouped by the ability they probe. Each description states what the example actually examines. https://keta1930.github.io/what-the-jev/experiments/urn:what-the-jev:en:experiments:78d5929cf653f446dd8b3c37c4178882ad8cdc68de9dfd6e6fe63a79dbeed12fFri, 09 Oct 2026 08:55:58 GMTThirteen experiments on the JEV decision model, grouped by what they probe. Each entry links to its report. https://keta1930.github.io/what-the-jev/resources/benchmarks/classifier-benchmark/urn:what-the-jev:en:resources/benchmarks/classifier-benchmark:3e8b104b5e81f11578da45802ca77c3e05dc06087388b4a403fcd17d56f90c71Fri, 09 Oct 2026 07:22:39 GMTA head-to-head benchmark for choice, noul, and score classification models, with two hash-locked suites. https://keta1930.github.io/what-the-jev/resources/benchmarks/jev-rerank-bench/urn:what-the-jev:en:resources/benchmarks/jev-rerank-bench:0ca37875f25c5a0a10fffd2a9b714456bda90468f460399011b7f2a99e38370fFri, 09 Oct 2026 07:22:39 GMTA 14-dataset study of Jev as a reranker against Cohere, ZeroEntropy, and open models. https://keta1930.github.io/what-the-jev/resources/benchmarks/jev-sec-bench/urn:what-the-jev:en:resources/benchmarks/jev-sec-bench:4a544aa7304d9379a2ae19419a46269e3b735d32b2cfc42a2a83b8ad67622c40Fri, 09 Oct 2026 07:22:39 GMTA blind security benchmark testing Jev on prompt injection and vulnerable-code detection. https://keta1930.github.io/what-the-jev/resources/benchmarks/s1mb/urn:what-the-jev:en:resources/benchmarks/s1mb:d61cdad9c4e6326a02c1a57e76e22a0cf29d060d23f9da841e88c9eaf1ab9e3eFri, 09 Oct 2026 07:22:39 GMTA leaderboard comparing Jev and open decision models across more than 100 benchmarks. https://keta1930.github.io/what-the-jev/resources/closed-source-models/hanzo-kai/urn:what-the-jev:en:resources/closed-source-models/hanzo-kai:e748bc48b1197c2e503a803324cbd832945200ff3ed969cbb3e60b68190f86f1Fri, 09 Oct 2026 07:22:39 GMTHanzo's closed decision model returning four typed answer kinds, supporting multi-decision runs and replay, with free output. https://keta1930.github.io/what-the-jev/resources/closed-source-models/solar-decide/urn:what-the-jev:en:resources/closed-source-models/solar-decide:5152e423775ef9ca199240b130996149e7512caa47a621c99a445e8dc5e514e9Fri, 09 Oct 2026 07:22:39 GMTUpstage's System One decision model on Solar Mini 4, with a 512K context, Korean support, and free output tokens. https://keta1930.github.io/what-the-jev/resources/closed-source-models/span-01/urn:what-the-jev:en:resources/closed-source-models/span-01:dbf4cee51cbcec4154038381a58517524272b617752bccd1091f27ed878a4afeFri, 09 Oct 2026 07:22:39 GMTRespan's behavior classifier that returns present, absent, and not_observable probabilities in a single forward pass. https://keta1930.github.io/what-the-jev/resources/integrations/dspy/urn:what-the-jev:en:resources/integrations/dspy:44a134ca3cc08818ad4529cf9f32e4e0785a7b82e64822b97ea2c0fc5ce890bbFri, 09 Oct 2026 07:22:39 GMTExperimental Jev/TypeSafe integration in DSPy 3.4.0 with decision types and the ReAnchor calibrator. https://keta1930.github.io/what-the-jev/resources/integrations/mastra/urn:what-the-jev:en:resources/integrations/mastra:d99db8126f9400d503ba33d3f1a3fa756bafef138224534db1a7447f29906185Fri, 09 Oct 2026 07:22:39 GMTMastra's Classifier primitive for typed classification with decision models. https://keta1930.github.io/what-the-jev/resources/integrations/microsoft-agent-framework/urn:what-the-jev:en:resources/integrations/microsoft-agent-framework:1fc8e8dd14c1e13815d7f1e2ce9d26f87b45c30e8db2ab5f07e7b16cd451d650Fri, 09 Oct 2026 07:22:39 GMTOfficial alpha package adapting TypeSafe System One models to Microsoft Agent Framework Python. https://keta1930.github.io/what-the-jev/resources/integrations/nocode-integrations/urn:what-the-jev:en:resources/integrations/nocode-integrations:afa05da912552fbfc5ddacc726c3f5fb4b8b5cc493ae7bd34bffe078d11787acFri, 09 Oct 2026 07:22:39 GMTOfficial TypeSafe/Jev integrations on the no-code platforms Zapier, Make, and n8n. https://keta1930.github.io/what-the-jev/resources/open-source-models/jebadiah/urn:what-the-jev:en:resources/open-source-models/jebadiah:fcfbdd5424fc15a078925c898938d21b0cfa0b036496a9607f5433a0e4197f54Fri, 09 Oct 2026 07:22:39 GMTFrontier Infra's System One-style decision model with a full training, data, and evaluation pipeline. https://keta1930.github.io/what-the-jev/resources/open-source-models/julia-1/urn:what-the-jev:en:resources/open-source-models/julia-1:e0bc15842a84d115fd1bebfa51a2a9cb9f757bf16ae35a2c718f4027a5a4ce80Fri, 09 Oct 2026 07:22:39 GMTSupersonic Labs' 144.3M-parameter compact decision model that runs on CPU and in the browser. https://keta1930.github.io/what-the-jev/resources/open-source-models/lev/urn:what-the-jev:en:resources/open-source-models/lev:f4580ba2483d9eb59fc8fe63134496a0b2aca3460e355c888ad4672bb1c99d69Fri, 09 Oct 2026 07:22:39 GMTInterfaze AI's System One decision model plus the levbench evaluation harness; adoption is early. https://keta1930.github.io/what-the-jev/resources/open-source-models/neohorse-jev/urn:what-the-jev:en:resources/open-source-models/neohorse-jev:de51754a22d0ca81f38ea6937d9db06a1663e33772217c18ffc3764473e0792bFri, 09 Oct 2026 07:22:39 GMTTokenRhythm's prefill-only decision model for agent routing, tool selection, and scoring. https://keta1930.github.io/what-the-jev/resources/open-source-models/nimble/urn:what-the-jev:en:resources/open-source-models/nimble:6aa5270b2a29ad6d2b5959471a880dd3f8a77ed79222f25314345d57503c7934Fri, 09 Oct 2026 07:22:39 GMTBespoke Labs' open Jev alternative: a Qwen3.5-9B one-step typed text decision model with public data and recipe. https://keta1930.github.io/what-the-jev/resources/open-source-models/openjev-razorback16/urn:what-the-jev:en:resources/open-source-models/openjev-razorback16:82014a4d23e26b73f855f48891684fde4fda79c83132e94dac09140ff20b4b14Fri, 09 Oct 2026 07:22:39 GMTrazorback16's open System One decision server; several unrelated projects share the OpenJev name. https://keta1930.github.io/what-the-jev/resources/open-source-models/rune-26b-a4b/urn:what-the-jev:en:resources/open-source-models/rune-26b-a4b:403adddff99c307de8965035d950811831b6eb25b80cad3880867f4f250204c3Fri, 09 Oct 2026 07:22:39 GMTInvergent's multimodal decision model reading text and images and returning probabilities in one pass. https://keta1930.github.io/what-the-jev/resources/tools/ollaya/urn:what-the-jev:en:resources/tools/ollaya:0f705c7cf0fbe5b54c7eb86fd65a94d8c68cf8dc1eeb68f189be027815d7a9aeFri, 09 Oct 2026 07:22:39 GMTLocal runtime for decision models, Ollama-style; serves TypeSafe's /v1/systemone wire format, written in Rust. https://keta1930.github.io/what-the-jev/resources/tools/swama/urn:what-the-jev:en:resources/tools/swama:0d2a7e48af9cae29c287eaf602e4d3964a04773025933f8f6095b44de927dc9cFri, 09 Oct 2026 07:22:39 GMTLocal AI runtime for Apple Silicon Macs, serving OpenAI-compatible and SystemOne decision endpoints. https://keta1930.github.io/what-the-jev/resources/use-cases/hermes-jev-skills/urn:what-the-jev:en:resources/use-cases/hermes-jev-skills:cd8f815cc5cdb91b4ff0d8070221caf47dcf137628e80704468b80321b7c1d23Fri, 09 Oct 2026 07:22:39 GMTEleven Jev-powered agent skills for Hermes, Claude Code, and Codex. https://keta1930.github.io/what-the-jev/resources/use-cases/jev-trader/urn:what-the-jev:en:resources/use-cases/jev-trader:641c0acc9a50a818e301c9dccb36b88f178c666916ea3540b6b8f7b42c813ca7Fri, 09 Oct 2026 07:22:39 GMTAn experimental Monad trading bot that asks Jev for one buy-or-sell decision per block. https://keta1930.github.io/what-the-jev/resources/use-cases/quantdinger/urn:what-the-jev:en:resources/use-cases/quantdinger:daf37b590622955d95bb6de52c8c85660c9166c00dc670fd340e06f23512edd8Fri, 09 Oct 2026 07:22:39 GMTAn open-source AI trading OS with Jev as one optional pre-trade decision gate. https://keta1930.github.io/what-the-jev/resources/tools/llama-cpp/urn:what-the-jev:en:resources/tools/llama-cpp:dd2100849abdef1224d5e178a5173521d19f898ad9d1e23dd51f562dddf22583Fri, 09 Oct 2026 06:35:03 GMTllama-server exposes /v1/systemone since 2026-10-02, with five pre-converted decision-model GGUFs under ggml-org. https://keta1930.github.io/what-the-jev/resources/tools/ollama/urn:what-the-jev:en:resources/tools/ollama:dfcbe207133289c09a7efc93223ceb9a1abee07c9c4242636bf296d071bafccdFri, 09 Oct 2026 06:35:03 GMTOllama 0.35+ serves Jev-style decision models locally on /v1/systemone, starting with nimble and tev1 models. https://keta1930.github.io/what-the-jev/resources/tools/sglang/urn:what-the-jev:en:resources/tools/sglang:0d9d7e71eaa591914bd4f8cfb309515465cdbbc71d1d6e3c8a2c8ea687810c33Fri, 09 Oct 2026 06:35:03 GMTSGLang turns any chat model into a decision model via /v1/decisions and /v1/systemone; probabilities are uncalibrated. https://keta1930.github.io/what-the-jev/resources/benchmarks/decision-index/urn:what-the-jev:en:resources/benchmarks/decision-index:f2b97645980f973efb80963a6a946de4021487080ae91153cce7cc5e55be654aFri, 09 Oct 2026 00:00:00 GMTBenchmark for typed decision engines with a full reproduction kit and a public-private mixed suite. https://keta1930.github.io/what-the-jev/resources/benchmarks/evals-typesafe/urn:what-the-jev:en:resources/benchmarks/evals-typesafe:f386fa6f5a7f4cbc156be0ab3fb58a63c2a58fc2f4681ec8d359e5a90700fd9aFri, 09 Oct 2026 00:00:00 GMTTypeSafe's official workflow-evals site; figures are vendor-reported, workflow code is open-sourced. https://keta1930.github.io/what-the-jev/resources/benchmarks/jev-bench/urn:what-the-jev:en:resources/benchmarks/jev-bench:c4e8d16f2e936126a83dd1d4f40202223c2f248afcab5ab5fb70ecb68fb623f8Fri, 09 Oct 2026 00:00:00 GMTSolo independent benchmark comparing Jev with two GPT-6 models on two 500-sample datasets. https://keta1930.github.io/what-the-jev/resources/benchmarks/jevals/urn:what-the-jev:en:resources/benchmarks/jevals:1b1a8942f2e735118b5db70500e8ed8abaf12bb06614d859652e2d4311ae7d5cFri, 09 Oct 2026 00:00:00 GMTIndependent leaderboard scoring Jev and six LLMs on identical typed decision questions, with public per-decision probability data. https://keta1930.github.io/what-the-jev/resources/benchmarks/jevbench/urn:what-the-jev:en:resources/benchmarks/jevbench:ea255916aeb02ac25da4ee61e35165b0aec2d2e74b19d733d4fa130be69b92a6Fri, 09 Oct 2026 00:00:00 GMTBenchmark Heaven's leaderboard for Jev-class decision models; four-axis geometric-mean score, cite with version. https://keta1930.github.io/what-the-jev/resources/closed-source-models/jev/urn:what-the-jev:en:resources/closed-source-models/jev:b85f234b9c16b5de45d2afa8d6a50b09c72092665461f9fc864cf70addc55de0Fri, 09 Oct 2026 00:00:00 GMTTypeSafe AI's closed-source System One decision model returning typed answers with calibrated probabilities. https://keta1930.github.io/what-the-jev/resources/closed-source-models/liquid-d1/urn:what-the-jev:en:resources/closed-source-models/liquid-d1:9874d3d63812289a8c7aea3383d167825507f61e7780fc9de33764a59e8eb8f7Fri, 09 Oct 2026 00:00:00 GMTLiquid AI's TypeSafe SDK-compatible decision model family with three hosted variants covering text, image, and audio. https://keta1930.github.io/what-the-jev/resources/closed-source-models/openai-decisions-api/urn:what-the-jev:en:resources/closed-source-models/openai-decisions-api:dcdb9eb29c443d831e41861e9220180beeacc67a3b7468bf90390161d6a8c87fFri, 09 Oct 2026 00:00:00 GMTOpenAI's public-beta typed decisions API with a single gpt-6-luna model and text-plus-image states. https://keta1930.github.io/what-the-jev/resources/integrations/cloudflare-workers-ai/urn:what-the-jev:en:resources/integrations/cloudflare-workers-ai:ea04560a1f49406f77c0d8f9cd24a9171cb510d4039dae23f1a4c3097c0222faFri, 09 Oct 2026 00:00:00 GMTJev served inside Cloudflare Workers AI, callable with a single env.AI.run() call. https://keta1930.github.io/what-the-jev/resources/integrations/langchain/urn:what-the-jev:en:resources/integrations/langchain:5c70018fd0f26956ba1f814deda3a2e789fd856257ac46d447fa5702dbd9c5e4Fri, 09 Oct 2026 00:00:00 GMTLangChain's Jev-as-judge benchmark: five trace classes each judged 100 times on accuracy, variance, cost, latency. https://keta1930.github.io/what-the-jev/resources/integrations/leaperone/urn:what-the-jev:en:resources/integrations/leaperone:6799bbd925b4a6ad3a1f1bd342bbabbe244cc9f2545704a4e70e2a78110afaefFri, 09 Oct 2026 00:00:00 GMTOpenRouter-compatible gateway passing Jev decision requests through; Chinese docs only. https://keta1930.github.io/what-the-jev/resources/integrations/litellm/urn:what-the-jev:en:resources/integrations/litellm:a85c168978dec3d39fcfedab459627f4fb5835ae288d7dd6ecbe06d66815b8c3Fri, 09 Oct 2026 00:00:00 GMTLiteLLM gateway support: a Jev pass-through endpoint and a relevance-compression guardrail. https://keta1930.github.io/what-the-jev/resources/integrations/openrouter/urn:what-the-jev:en:resources/integrations/openrouter:a543ab72b7d785057cd5c8b8a0b1b7e5f5043c3a2ceee35d49309a70df02054bFri, 09 Oct 2026 00:00:00 GMTOpenRouter's Jev hub: concept guide, tutorial, cookbooks, and an interactive demo on one platform. https://keta1930.github.io/what-the-jev/resources/integrations/pydantic-ai/urn:what-the-jev:en:resources/integrations/pydantic-ai:4b317d1b0698b9b69430dd0ddf587d839c5189f54933b078f2b83b76cf1d55b6Fri, 09 Oct 2026 00:00:00 GMTPydantic AI's TypeSafeModel derives Jev questions from output types, with LLM fallback. https://keta1930.github.io/what-the-jev/resources/integrations/spring-ai-typesafe/urn:what-the-jev:en:resources/integrations/spring-ai-typesafe:107a0f7ebe19022489cb26988431a374cd50df9f11280203f70b05295a2af490Fri, 09 Oct 2026 00:00:00 GMTSpring AI community integration: TypeSafeClient with Jev judge, guardrail, and evaluator adapters. https://keta1930.github.io/what-the-jev/resources/integrations/vercel-ai-gateway/urn:what-the-jev:en:resources/integrations/vercel-ai-gateway:cd4ba626fb31e725a4abe01904564f790fbea9d584306f82ffd5d16516cf964eFri, 09 Oct 2026 00:00:00 GMTVercel serves Jev via AI Gateway, with AI SDK 7 evaluate API and eve framework integration. https://keta1930.github.io/what-the-jev/resources/open-source-models/clef/urn:what-the-jev:en:resources/open-source-models/clef:7b0d4da51f977812f76198ba64a645bccf2854166d3f3b8c3ec95cb061639369Fri, 09 Oct 2026 00:00:00 GMTCloudflare decision models at 27B and 9B with image input, 64K context, and Jev API compatibility. https://keta1930.github.io/what-the-jev/resources/open-source-models/hopper/urn:what-the-jev:en:resources/open-source-models/hopper:c60f36b900bfbb30e79ade18efcd2705ba7db948f47be4f8f85583afeda824a0Fri, 09 Oct 2026 00:00:00 GMTSingle-forward-pass decision server with five models; its 4B adapters are restricted to research and demonstration use. https://keta1930.github.io/what-the-jev/resources/open-source-models/jevk5/urn:what-the-jev:en:resources/open-source-models/jevk5:cce1b5e36cb65adf2195830db4511f06d266d257379dc95fe8d48280435bf48cFri, 09 Oct 2026 00:00:00 GMTOpen decision model pairing distilled Qwen3.5 LoRA with option-letter logit readout; ranked first among open-source models on JevBench v1.4. https://keta1930.github.io/what-the-jev/resources/open-source-models/kev/urn:what-the-jev:en:resources/open-source-models/kev:b4d4a43beee950820c119be14f71a19d8bb2cd3fa9693d02f24293246f099150Fri, 09 Oct 2026 00:00:00 GMTSmall Qwen-based decision model family for self-training and local deployment, with three isolated question types and calibrated probabilities. https://keta1930.github.io/what-the-jev/resources/open-source-models/laya/urn:what-the-jev:en:resources/open-source-models/laya:5b4cf548b6fae2aa3f099b2af515000f509c1a651d8d22b217ef01a9bc4266a6Fri, 09 Oct 2026 00:00:00 GMTMultilingual non-autoregressive decision engine returning typed decisions with calibrated confidence in one forward pass across 100+ languages. https://keta1930.github.io/what-the-jev/resources/open-source-models/mapika-decider/urn:what-the-jev:en:resources/open-source-models/mapika-decider:32d0b358f7eca9707fa57a3385535bdcb080fcf06d90543ededda872dcc5a8e3Fri, 09 Oct 2026 00:00:00 GMTMapika's Apache-2.0 decision model family spanning 0.8B to 35B variants, with the training recipe published in full. https://keta1930.github.io/what-the-jev/resources/open-source-models/pplx-decider/urn:what-the-jev:en:resources/open-source-models/pplx-decider:152aa212a52571d563489b1198b4e633b3bc26264e383d6851dbfd257e19374cFri, 09 Oct 2026 00:00:00 GMTMultimodal decision model fine-tuned from Qwen3.8-27B, with Apache-2.0 weights and a commercial hosted API. https://keta1930.github.io/what-the-jev/resources/open-source-models/prometheus-eval/urn:what-the-jev:en:resources/open-source-models/prometheus-eval:66fb8e0e640d7d8662c79e7936fd6cf609ebdc7e899e9ad4e8f1100944f1905cFri, 09 Oct 2026 00:00:00 GMTKAIST-led open-weight judge model family outputting scores and written judgments, not typed calibrated probabilities. https://keta1930.github.io/what-the-jev/resources/open-source-models/semif/urn:what-the-jev:en:resources/open-source-models/semif:758f2907b0952b947744948f8883c96e7a810c4fd63f1af76423b38ba8ff1020Fri, 09 Oct 2026 00:00:00 GMTZero-training approach reading option probabilities from frozen open models in one forward pass, with MIT-licensed code. https://keta1930.github.io/what-the-jev/resources/open-source-models/tev1/urn:what-the-jev:en:resources/open-source-models/tev1:cf0f12bdb5379912621c3c0a456351296f7e4f3866830dda8ebf425bbe59924fFri, 09 Oct 2026 00:00:00 GMTOpen Qwen3.5 4B/0.8B LoRA decision weights supporting choice-type only, with a $17 training walkthrough. https://keta1930.github.io/what-the-jev/resources/tools/building-with-typesafe-jev/urn:what-the-jev:en:resources/tools/building-with-typesafe-jev:dc9afc497c93a17b150a2ef8e382206fa8c8ad9f1213b3d56df7cf53746547b3Fri, 09 Oct 2026 00:00:00 GMTUnofficial agent skill by aaddrick teaching Jev question design; self-reported evaluation gain of 0.65 to 0.96. https://keta1930.github.io/what-the-jev/resources/tools/duckdb-jev/urn:what-the-jev:en:resources/tools/duckdb-jev:1245ff42a7ff02a893a02bf25006ed5e8be873b6749c7a5526a021f25716189eFri, 09 Oct 2026 00:00:00 GMTDuckDB extension by colliber that poses a typed Jev question per row via four scalar SQL functions. https://keta1930.github.io/what-the-jev/resources/tools/jev-mcp/urn:what-the-jev:en:resources/tools/jev-mcp:466f46b9b35d2a72d96ae080ce4459535ed947c88c9b2776735fa2d100ad7a6dFri, 09 Oct 2026 00:00:00 GMTMCP server by jkudish exposing 12 Jev judgment tools to coding agents over four carriers or compatible endpoints. https://keta1930.github.io/what-the-jev/resources/tools/jev-reranker/urn:what-the-jev:en:resources/tools/jev-reranker:4f5e4221449fddd9535db06a0ed4396c1331cc7b2f8f5d705a37afe787ba5e87Fri, 09 Oct 2026 00:00:00 GMTPython library by hotchpotch using Jev for RAG candidate reranking and relevance filtering with threshold-friendly scores. https://keta1930.github.io/what-the-jev/resources/tools/openrouter-decisions-skill/urn:what-the-jev:en:resources/tools/openrouter-decisions-skill:cc776170dd5d97daeb87ba2a4d2c7c645d3461e778d2ffabdc52aea87b745e6aFri, 09 Oct 2026 00:00:00 GMTOfficial OpenRouter skill teaching agents to delegate judgment and computation to decision models via a seven-step method. https://keta1930.github.io/what-the-jev/resources/tools/typesafe-ai-github/urn:what-the-jev:en:resources/tools/typesafe-ai-github:02ab081332a41a6b74b23ee7889602363288114ba81dfde7999a822c11445e2aFri, 09 Oct 2026 00:00:00 GMTOfficial TypeSafe AI GitHub organization hosting the Jev SDKs, LLM-provider adapter, agent skills, and evaluation source. https://keta1930.github.io/what-the-jev/resources/use-cases/deepeval/urn:what-the-jev:en:resources/use-cases/deepeval:72455d89ebe18d557619b1a5c509fdf4b441d57c29fc1504dca9c8a3291d3089Fri, 09 Oct 2026 00:00:00 GMTPytest-style LLM evaluation framework whose JevEval metric delegates scoring to Jev for calibrated probabilities. https://keta1930.github.io/what-the-jev/resources/use-cases/ha-jev/urn:what-the-jev:en:resources/use-cases/ha-jev:612dc21f2ebc8d94e95203ef74c5b3c2642e62b0fd8c3d7857e8587535159dc3Fri, 09 Oct 2026 00:00:00 GMTHome Assistant integration exposing Jev questions as sensors and actions, with 25 blueprints and a spend dashboard. https://keta1930.github.io/what-the-jev/resources/use-cases/inbox-zero/urn:what-the-jev:en:resources/use-cases/inbox-zero:30731b231fbe7ce041de5a228247dc648c575e63269e6e2944ab4ea596e63944Fri, 09 Oct 2026 00:00:00 GMTOpen-source email assistant using Jev as an optional classification backend for sorting and filtering. https://keta1930.github.io/what-the-jev/resources/use-cases/jev-tetris/urn:what-the-jev:en:resources/use-cases/jev-tetris:69c9b21cc6bcec0e32ed2107342e43e9750f40a828761642a0d2aac95e8f6fbcFri, 09 Oct 2026 00:00:00 GMTPlays Tetris with Jev under controlled measurement; a 23-line regex baseline beat the model in its own tests. https://keta1930.github.io/what-the-jev/resources/use-cases/jev-ultrafast/urn:what-the-jev:en:resources/use-cases/jev-ultrafast:9f298b584b2583833a2eaa324cdbf17641f7ff1540916c85454481c09e3bb393Fri, 09 Oct 2026 00:00:00 GMTOfficial Browser Use agent making one Jev decision per cycle, with an LLM only for text entry. https://keta1930.github.io/what-the-jev/resources/use-cases/jevtown/urn:what-the-jev:en:resources/use-cases/jevtown:d8ebfa08f265a501d7290f8226a2d5893056e8451fdedeef76781789b1ac56caFri, 09 Oct 2026 00:00:00 GMTPixel-art town simulation engine where Jev picks actions from engine-defined options; LLM handles dialogue. MVP stage. https://keta1930.github.io/what-the-jev/resources/use-cases/milvus-bootcamp/urn:what-the-jev:en:resources/use-cases/milvus-bootcamp:3dd8335b2137a7bd0251cf11797ce199b1043aaaa65206f6d54ed34f2b04857aFri, 09 Oct 2026 00:00:00 GMTNine official Milvus notebooks using Jev for judgment calls across a RAG pipeline, with pymilvus reranking merged. https://keta1930.github.io/what-the-jev/resources/use-cases/paper-radar/urn:what-the-jev:en:resources/use-cases/paper-radar:1c63a9790064ba5850492b6f7921e6051efbb025cd96cd7528341bb153636eadFri, 09 Oct 2026 00:00:00 GMTDaily radar screening the full arXiv feed with one Noul question per declared interest; author-reported figures. https://keta1930.github.io/what-the-jev/resources/use-cases/routellm/urn:what-the-jev:en:resources/use-cases/routellm:fca6ea0ecb68b79542fd459f2144c430378458b13a7d764e2aa7ec3d21d81f00Fri, 09 Oct 2026 00:00:00 GMTFramework routing traffic between strong and weak models by difficulty; 85% cost savings officially self-reported. https://keta1930.github.io/what-the-jev/resources/use-cases/semantic-router/urn:what-the-jev:en:resources/use-cases/semantic-router:36b843eae3be6a5058b3876a58cf789ea8ab7c4054d28d8090e99c8bb5336232Fri, 09 Oct 2026 00:00:00 GMTPre-Jev decision layer routing via encoder similarity in tens of milliseconds; currently being rewritten for 1.x. https://keta1930.github.io/what-the-jev/resources/use-cases/toolgate/urn:what-the-jev:en:resources/use-cases/toolgate:b5709df7fde4b5910c42cc37c395a5e3db1378e9dd018942be860e40528eb718Fri, 09 Oct 2026 00:00:00 GMTCalibrated firewall for agent tool calls: seven Jev questions map through thresholds to allow/ask/deny verdicts.