Jev directory
Jev Practices & Patterns
296 curated Jev entries. How to design Jev states, questions, and confidence gates the right way, from the official patterns to the failure modes worth designing around.
By format
296- Docs18
- Repos & sites199
- Videos7
- Posts16
- Articles56
Updated
All 296 Jev Practices & Patterns
Search the full category, or filter by format and tag.
Showing 1–48 of 296 resources
An MIT-licensed case study with two working Python programs pairing Jev and Claude Code: a support-triage app with Jev inside it, and a Claude Code safety hook where Jev judges the agent's own shell commands. Tests and evals run offline, free.
Jev Practices & Patterns#community#github#python
A documentation-only review of where a typed, local SystemOne decision component could help the SeerScore prediction pipeline, with independent perspectives from Hermes Agent and Cursor Agent and no integration code included.
Jev Practices & Patterns#community#github#review
Community
martyschoff/systemone
martyschoff/systemoneA research workspace exploring visual System One models on 2D affordance tasks, predicting interaction regions for robot execution, with a five-document primer covering concepts, candidate models, prior work and first-round debugging of methods A and D.
Jev Practices & Patterns#community#github#robotics
Community
yutian929/SystemOneAffordancePython
yutian929/SystemOneAffordancePythonKyle Wild's October 13, 2026 post testing whether decision models inherit the primacy and recency effects of the LLMs underneath: Jev was not tricked by option ordering, while the other decision models he tested were. Reproduction code is linked on GitHub.
Jev Practices & PatternsPost#community#order-bias#evaluation
Community
Kanta Hayashi's calibration check using a hidden fair die: across 400 trials Jev picked '1' every time at an average 83% probability while being right 19% of the time, and handing forecast documents through Choice shrank a stated 30% risk to about 5%.
Jev Practices & PatternsPost#community#calibration#evaluation
Community
The experimental image-input reference: PNG or JPEG files in a top-level assets map, referenced by string anywhere in state, instructions, or criteria, behind the TypeSafe-Danger-Zone-Images header.
Jev Practices & PatternsDocs#official#docs#vision
Official
The box question type: one bounding box per question, returned as four edges that are fractions of the image, plus confidence, with no criteria field and no not-found value.
Jev Practices & PatternsDocs#official#docs#vision
Official
A code-repair workflow that delegates bounded decisions to Jev as a low-cost probabilistic judge: across 400 SWE-bench tasks it cuts mean repair-model token use by 21.6-50.6% with higher observed repair rates on three models.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A jailbreak framework that uses multi-dimensional probabilistic judge feedback to guide prompt optimization: 89.8% average HarmBench attack success across ten open-source settings under a 10-query budget.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A 34-page technical review and implementation framework by Av1dlive for using Jev's typed decisions inside agent workflows, separating observation, model judgment, deterministic policy, authorized action, and outcome verification.
Jev Practices & Patterns#community#catalog-import#logicrw
An open lab for using Jev correctly next to Claude Code: a patched jev-router with per-turn model routing, a health-gated fallback with alerts, and write-ups of two silent bugs and of where tokens actually go.
Jev Practices & Patterns#community#catalog-import#logicrw
Community
pasblinn/jev-labTypeScript
pasblinn/jev-labTypeScriptABTdomain's workflow design for brand-protection review: three layers of typed Jev questions turn each new domain that carries a brand word into Release, Watch or Flag, with 0.3 and 0.7 probability lines deciding what happens next.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
An 11-page working paper on placing a small commercial text judge beside a live trading lab: six tests for a text signal, what the judge may and may never touch, and the plumbing failures that corrupted its readings. No profitability claims.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
Kitsuno's engineering write-up on training an open-weight Laya model to take over parts of its Jev usage: eight days of work, run in shadow beside Jev on every job-side request, without training on Jev's outputs.
Jev Practices & PatternsPost#community#catalog-import#madewithjev
Community
Sayora's AI receptionist replaced production RAG with Jev for page picking and measured the change on 168 caller questions: wrong pages handed to the LLM fell from 33% to 3%, and the right page ranked first 88% of the time against 70%.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
A social-scheduler founder's accuracy test of Jev: four yes/no questions over 522 captions, then the same questions for Fable on a random 100. Agreement reached 94/100 on promotion, and all 520 scored captions cost $0.011.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
Tenuo's engineering write-up of safe-upgrade, a dependency-upgrade agent built on Jev, LangGraph and Tenuo: Jev turns repository evidence into typed probabilistic choices, the graph routes on confidence, and each action gets a short-lived task-scoped warrant.
Jev Practices & PatternsArticle#community#catalog-import#yibie
Community
A hands-on write-up of swapping an LLM routing call for System One decisions in a Quarkus LangChain4j app: a custom Planner asks one Choice question to pick the specialist agent, and an OutputGuardrail asks a Noul question to reject bad replies.
Jev Practices & PatternsPost#community#catalog-import#madewithjev
Community
A theoretical study of recalibrating many related yes/no questions that each have few labels, motivated by decision models such as Jev. It derives minimax compound-risk rates and shows pooled isotonic fitting with one per-question correction achieves them, while tuning questions separately loses a factor.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A study of option-order sensitivity in System One models and a training-free fix called Runoff. It scores each option alone to keep a few candidates, then compares them with each candidate in every position, removing order dependence in two forward passes.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A test of Jev on reflection tests, one-shot matrix games, ALFWorld, and robot control that finds a sharp boundary: it succeeds when the right option can be judged from what the input describes, but fails when the right option depends on predicting something not in the input.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A conceptual article that formulates a conditional Jevons hypothesis for machine evaluation: sufficiently large reductions in the marginal cost of usable machine evaluation may increase its organizational consumption where latent demand is substantial and complementary costs do not dominate.
Jev Practices & PatternsArticle#community#research#arxiv
Community
0GitHub stars
A systematic study of Jev's security and privacy implications using the official API and NanoJev, a local model with controllable training data. It adapts prompt injection, adversarial suffixes, and poisoned training data, runs inference attacks against typed outputs, and maps dual uses.
Jev Practices & PatternsArticle#community#research#arxiv
Community
shihe98/Security_Privacy_JevPython
shihe98/Security_Privacy_JevPythonA pre-registered evaluation of Jev as a one-second, training-free screen for SCADA alarm triage in water distribution networks. On a four-class EPANET C-Town benchmark it matched a hand-written rule tree, and a gated cascade spared an LLM reviewer 35-38% of windows.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A position and case-study paper on using System One decision models like Jev and the open-source Laya as decision layers in LLM-driven penetration-testing harnesses. It names four decision points, runs an exploratory 13-vulnerability NeuroSploit comparison, and proposes Rave, an unevaluated model.
Jev Practices & PatternsArticle#community#research#arxiv
Community
An evaluation of prompt injection against Jev's typed probabilistic decisions using 510 reconstructed InjecAgent cases. Malicious content shifts action probabilities but rarely flips Jev to the attacker's target; score-feedback attacks raise fresh-run success from 1.8% to 3.5%.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Documents a failure mode in letter-bound remote sensing classification: 97.71% OA on UCM in vocabulary but 7.35%–96.40% cross-dataset accuracy and inverted confidence (AUROC 0.38). RSAJEV replaces the letter readout with training-free prototype retrieval, reaching 89.43% on the UCM cross-dataset gate.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A training-free method that studies redundancy in candidate evaluation for JEV-style multimodal decisions. Shared context anchoring, candidate prefix sharing, and decision-guided pruning cut candidate depth by 43.75%–45.83% while retaining 93.66%–97.52% of task scores.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Studies a fast learned actor in closed-loop Doom that hands control to a reasoning vision-language model only when a gate opens. On 900 held-out questions, deferring the least confident 30% of decisions gains +0.13 with doomLaya's option order, but no closed-loop variant reaches the exit.
Jev Practices & PatternsArticle#community#research#arxiv
Community
dexmac221/doomgemmaPython
dexmac221/doomgemmaPythonA study of whether Jev can serve inside reinforcement learning training rather than as a single-domain black box. Frozen Jev fills reference-policy, exploration-judge, and replay-rater roles, and training with it beats a standard RL learner across nine MiniGrid tasks and three Atari games.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Reference-bound Visual Jev rewards verify whether requested attributes, actions, and relations hold for specified reference subjects in multi-subject image generation. Training a Qwen3.5-4B verifier as a GRPO reward raises a GPT-5.4 composite score from 41.78 to 52.50 on a 897-task MICo-Bench subset.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A study of using Jev inside Stockfish search and multi-teacher distillation. A bot with Jev's judgment distilled into a compact evaluator climbs above a 2200 Lichess bullet rating, and averaging Jev with Qwen3-32B labels beats spending the whole labeling budget on Qwen alone by 9.6 Elo.
Jev Practices & PatternsArticle#community#research#arxiv
Community
JevCascade allocates inference-time compute for prefill-only decision models using readout stability: when only the candidate menu changes, post-intervention accuracy is already determined by the cached first-pass distribution. Menu-only interventions are predicted to within 4.2 points across six model families.
Jev Practices & PatternsArticle#community#research#arxiv
Community
An analysis of open-weight typed decision models that finds their probabilities mostly follow option labels rather than written definitions. Deleting every definition leaves laya-td accuracy unchanged (0.8559 versus 0.8487), while renaming options to A and B raises accuracy by 0.1511.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A student-guided distillation pipeline for routing across 60 LLM task categories: a ModernBERT student retrieves top-k candidates and a DeBERTa-v3 zero-shot NLI teacher reranks only those. The best checkpoint reaches 77.5% teacher agreement, with Coverage@16 of 91-100%.
Jev Practices & PatternsArticle#community#research#arxiv
Community
LLM-as-Jev asks how far off-the-shelf LLMs already go as Jev-style decision models and when fine-tuning is needed. Frozen Qwen3.5-4B and Qwen3-0.6B match or exceed existing training-free readouts on the same 4B backbone, and Thinking Jev adds budgeted test-time reasoning.
Jev Practices & PatternsArticle#community#research#arxiv
Community
4GitHub stars
An analysis of ordinal-scale bias in JEV 1.13 and three open KEV models: across 36 ordinal datasets, final decisions use only 67-76% of the effective gold support versus 87-102% on four nominal tasks. BA-LoRA post-training raises utilization from about 47% to 86%.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Glax147/jev_ordinal_scale_biaPython
Glax147/jev_ordinal_scale_biaPythonDyad augments a pretrained LLM with an environment-conditioned action encoder that scores candidate action descriptions to yield a distribution over typed actions. Training only the encoder helps across four unseen environments; joint training adds a 3.80% absolute gain on ALFWorld with a 9B model.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Koa-decision: Fast and Consistent Structured Decision Making with Generative LLMs (arXiv 2609.36115)
Koa-decision reduces single-step decisions such as classification, endpointing, Boolean checks, and scoring to constrained generation with single-token outputs. On a production intent-routing benchmark it reaches 85.5% accuracy in about half a second, up to roughly 7.5x faster than frontier models.
Jev Practices & PatternsArticle#community#research#arxiv
Community
3GitHub stars
PACT turns contrastive training pairs into four loss terms for single-token typed-decision models. On a frozen 324-item holdout it matches the published recipe's accuracy (84.6% versus 85.2%) while cutting answer flips under relabelling from 13.8% to 9.8%.
Jev Practices & PatternsArticle#community#research#arxiv
An evaluation of Jev and two open-weight typed decision models when option names and definitions conflict. With yes/no names, decision-flip rates are up to 70.4 percentage points higher than the 0/1 control, and mean AUC drops from 93.8% to a below-chance 23.2%.
Jev Practices & PatternsArticle#community#research#arxiv
Community
JevOut shows short background context additions can flip correct decisions in four systems including Jev across seven datasets. Within 64 accepted evaluations per decision, additions redirect 61.4%-73.2% of initially correct decisions toward a wrong option fixed in advance.
Jev Practices & PatternsArticle#community#research#arxiv
Community
xzx34/JevOutPython
xzx34/JevOutPythonAn exploratory study of JevForest, a budgeted feature-acquisition policy that aggregates path-dependent proposals from bootstrapped trees and queries Jev for semantic answers. On small held-out samples it beats static and random orderings on AG News but not on TREC.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A training-free framework with two implementations that turn emotion classification into Jev judgments: one decomposes tasks into atomic judgments, the other aggregates complementary judgment paths. Across eight datasets standard Jev averages 62.93% macro-F1 versus 67.28% for the strongest LLM baseline.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A report identifying an arithmetic-dependent rejection bottleneck in Jev: on paired arithmetic problems answer-present accuracy reaches 99% while correct rejection falls to 7%. A development-selected decision threshold lifts rejection from 7% to 79% while retaining 97% answer-present accuracy.
Jev Practices & PatternsArticle#community#research#arxiv
Community
7GitHub stars
A system that decomposes all three tasks of SemEval-2026 Task III Track A into typed Jev decisions aligned with 488 CPU-fitted coefficients. It reaches 1.0645 RMSE on valence-arousal regression and 52.09/44.06 continuous F1 on extraction, above fine-tuned Llama-3.3-70B and GPT-OSS-120B baselines.
Jev Practices & PatternsArticle#community#research#arxiv
Community
ZhangYiqun018/jev-dimabsaPython
ZhangYiqun018/jev-dimabsaPythonA system that codes Texas crash narratives as gated, typed Jev decisions returning probabilities and no text: a screen covered 499,500 narratives and 195,857 were coded with a 27-question schema. The typed model attains F1 of 0.908 against human labels, and recalibration cuts calibration error 3.3-fold.
Jev Practices & PatternsArticle#community#research#arxiv
Community
A behavioral diagnosis of why System One decision models choose wrong actions in agent loops. On a frozen benchmark of agent decision states, most wrong actions follow a correct reading of the state, and the more accurate the system, the more of its errors take this form.
Jev Practices & PatternsArticle#community#research#arxiv
Community
Index of all 296 entries
- Jev + Claude Code case study
- SystemOne in SeerScore: workflow review
- SystemOneAffordance
- Jev isn't tricked by ordering, and the others are.
- Jev Does Not Play Dice — kantahayashiai.github.io
- Images and Assets (Dreamer-Ready)
- Box Questions (Dreamer-Ready)
- JevCoder: When Does Decision Delegation Pay Off in Code Repair? (arXiv 2610.19596)
- Jev-Guided Strategy Bandit: Multi-Dimensional Judge Feedback for Query-Efficient Jailbreak Attacks (arXiv 2610.19971)
- Jev Engineering: Typed Decision Systems for Reliable Agent Workflows
- Jev Lab — pasblinn
- Using Jev in Domain Abuse Review: A Workflow Design
- A Thermometer, Not an Analyst: A Text Judge in a Trading Lab
- Laya in shadow, next to Jev on every request
- Jev handed our LLM the wrong page a tenth as often as our embeddings
- Jev review: tested against Fable on 520 posts
- Jev in practice: typed decisions, scoped authority
- Routing Agents with Jev and Laya: Adding System One Decisions to Quarkus LangChain4j
- Recalibrating Many Questions with Few Labels: Minimax Rates and Identification Limits (arXiv 2610.18488)
- Consider, Then Choose: Order-Invariant Decisions in System One Models (arXiv 2610.18233)
- Code Owns the Simulation, Jev Owns the Evaluation (arXiv 2610.01834)
- Judgement in the Age of Jev: From Evaluation Scarcity to Evaluation Abundance (arXiv 2610.01231)
- Hidden Risks of Jev: An Empirical Study of Security, Privacy, and Dual Use (arXiv 2610.04985)
- HydroJEV: A one-second, training-free screen for cyber-attack and fault attribution in water distribution networks (arXiv 2610.02048)
- Calibrated Decision Models for Autonomous Penetration-Testing Harnesses: JEV and Laya as System One Decision Layers for LLM-Driven Pentest Agents (arXiv 2609.28940)
- Decision Hijacking: Prompt Injection Attacks on Jev's Typed Probabilistic Decisions (arXiv 2609.28613)
- RSAJEV: Training-Free Abstaining Prototype Memory for Eliminating Silent Hallucination in JEV-Based Remote Sensing Classification (arXiv 2610.15742)
- FastJEV: Understanding Redundancy for Compact JEV Inference (arXiv 2610.11379)
- System Switch: When Should a Fast Decision Model Stop and Think? (arXiv 2610.09683)
- Can Jev be Your Q or Policy in Reinforcement Learning? (arXiv 2610.11692)
- Visual Jev Rewards: Reference-Bound Verification for Multi-Subject Image Generation (arXiv 2610.09328)
- From Probabilities to Decisions: Search and Multi-Teacher Distillation with Jev (arXiv 2610.09188)
- JevCascade: Readout Stability and Inference-Time Compute Allocation for Prefill-Only Decision Models (arXiv 2610.07716)
- Labels Override Definitions in Jev-Style Typed Decision Models (arXiv 2610.02586)
- Student-Guided Teacher Distillation for Efficient LLM Task Routing: Positioning Against Jev-Style System-1 Classifiers (arXiv 2610.02516)
- LLM-as-Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them (arXiv 2610.02076)
- More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models (arXiv 2609.38827)
- Dyad: Extending Large Language Models with Native Typed Decision-Making (arXiv 2609.36116)
- Koa-decision: Fast and Consistent Structured Decision Making with Generative LLMs (arXiv 2609.36115)
- PACT: Pairwise-Anchored Calibrated Tuning for Single-Token Typed Decisions (arXiv 2609.35865)
- Type-Safe Is Not Error-Free: Typed Decision Models Follow the Option Name, Not the Definition Bound to It (arXiv 2609.26758)
- JevOut: Natural Context Can Flip Decision Models (arXiv 2609.30243)
- JevForest: Path Voting for Budgeted Feature Acquisition (arXiv 2610.10615)
- Emo-Jev: Probabilistic Reasoning for Emotion Classification with Jev (arXiv 2610.08829)
- When the Right Answer Is Missing: An Arithmetic-Dependent Rejection Bottleneck in Jev (arXiv 2609.39496)
- Decide, Don't Generate: Competitive Dimensional ABSA with Jev's Typed Decisions (arXiv 2609.35293)
- Calibrated Decisions at Scale: Converting Police Crash Narratives into Probabilistic Crash Variables with a System One Model (Jev) (arXiv 2609.24052)
- Reading Right, Acting Wrong: Diagnosing Decision Failures of System One Models in Agent Loops (arXiv 2610.18579)
- Orcha Jev Judgment Spike
- Jev in the Wild (arXiv 2609.30216)
- Jev-Driven SRE Diagnosis: What Worked and What Failed
- Auto-optimizing Jev: Half the Errors, 1/7 the Cost
- Jev Local — Argos1111
- Jev Calibrate
- Jev On A Laptop
- Jev Column Race
- Jeval — rlaope
- Jev RAG Benchmark
- System One — sgoedecke
- System One Open
- TypeAR
- Jev_apps
- Dejevu
- Openvons
- Jev Stock
- Jev Project Context
- AnyDecisionModel
- Jevmory
- Jev.nu
- Hermes And Jev Play Minecraft
- Local Jev — amithgc
- Eikos Arena
- Nanojev Arena
- Laya Browser Agent
- Kev — arjun988
- TypeSafe Local
- Jev Search — larguesa
- Learn Jev End To End
- Decision Index
- LegalForecastBench
- Jev Korean Benchmark
- Jev Lm — y0usaf
- Claude Jev — buchmark
- Jevchess — choxos
- Jevbench — dhruvmehra
- Jev Agent Authorization
- Daf Jev
- Jev Align — caiovicentino
- Jev Codes
- OpenJev — GPT-AGI
- Jevarena — chenmingtang830
- Jev Reward Model Evaluation
- Jev Realtime Trading
- Jev Bot — nssmd
- Jevloop — parkavenue9639
- Sysone Bench
- Jev Model Tokengate
- Reflexbench
- Jev Chat — adhyaay-karnwal
- Openjev — DECRUX9812
- Jev Behavior Study
- Tenbin
- Jev — cobusgreyling
- Jevtest — realZachi
- Rh Guard
- ORIGIN-CIVILIZATION
- cu-Jev
- Jev As Quant
- Ego Jev Ultrafast
- Jev Storyboard Lab
- Jet
- Jev Benchmark — YidiDev
- Cairn Jev Lab
- Jev Guardrails — deepansh-saxena
- System One Gemma
- Jevgpt — Bewinxed
- Trade Jev
- Jev Does Not Play Dice — KantaHayashiAI
- Mcts Agent
- Jev Gate — MongLong0214
- Jev Exploration
- Jev Benchmark — wondertwins
- Siftr
- Jev Mode — ddfeyes
- Research Desk
- Jev Chess — hemanth
- TypeSafe Chess — Dimesio
- Toolgate — RiskAverseTech
- Jev Starter — hamakyo
- Jevbus
- Watfile
- Jevshield
- Polymarket Btc 5M Agent
- TypeSafe Jev MCP
- Ruby LLM Providers TypeSafe
- Jev MCP Server — wangkuangkuang
- Metask Jev
- Bes Kelime Jev
- Dsh Jev Decide
- Jevalyzer
- Jev Deepresearch
- Jev X Kit
- Jev Bot — bl888m
- Openpave Jev
- Jev Reviews
- Open System One
- Jev System One Reference
- Jev Dllm
- Paper Package
- Decisionbridge
- RISC-jeV
- Jev Freeform
- Typed Decisions
- Jev Synergy Screening
- Stuntdouble
- Qwen Rlcd
- Kyotsu AI Bench
- Jev Routing Experiment
- Openjev Experiments
- Padflow Jev Evals
- Jev Carryforward
- Modelsystem
- TypeSafe Vs DeepSeek
- Jev Labs
- Openpoke Meets Jev
- What Is Jev
- Jev Benchmark — themsquared
- System One Security Triage
- Openjev Multimodal
- Yc Jev Bench
- Rerank Bench Jev
- System One — mpuig
- Can Jev Play
- Jev-Research-Index
- LetJevDecide
- Decide — gopaljigaur
- System One Router
- MEDJEV — PAI-CUHK
- Jev Filter
- Jev Zig CLI
- Typesafe_chess_eval
- Jev Shadcn Lint Eval
- System One Adapter Rust
- Jev Deferred Crispification
- Jev Alpha Bench
- Jev Headline Bench
- Jev Report
- Jev Finance Benchmark
- Jevals Data
- Jev Calibration Audit — jujumilk3
- Labs
- Jev Anotacao Sentencas
- Jev Vs Luna
- Jev Lab — Menny1337
- Jev Skillbench
- Misereru Slide Jev
- Shade Arena Jev Monitor
- Jev Certify
- Jev Eval — onlyoneaman
- Jev No Enem
- Jev Jp Address
- Jev Trace Classifier
- FinancialPredictionJev
- Jev Pick And Place Study
- Thaiexam Jev Charts
- Jev Evaluation
- Minojev
- Snbt Jev Bench
- WindTunnel
- Jev Dataops
- Jev is the fish at the poker table
- Jev AI Model in Action: 5 Practical QA Use Cases + Hands-On Demo
- Jev's Architecture Unmasked
- Jev: The Language Model That Won't Talk
- Jev in Search: Three Practical Evaluations
- Jev Engineering: turning an agent stack into a control system
- Local Laya vs cloud Jev on Tetris
- What is Jev? (and is laya better?)
- Is Jev actually good at computer use?
- Run a Jev-Style Model on Your Own GPU
- Is Jev Really Better, Faster and Cheaper? I Put Jev To Test vs. Luna
- Jev vs Jev in Gomoku
- Can LLMs Do What Jev Does? A Replication Attempt
- fx auto mode benchmarked with Jev
- A working note on Jev harnesses for coding agents
- Jev within 5 points of a fine-tune
- Shrinking gomoku from 225 moves to 40
- Workflow evals
- TypeSafe AI Benchmark
- Jevcal
- Jev Rerank Bench
- Jev Eval Agent
- Jev Sec Bench
- Zero-shot email classification with TypeSafe's Jev
- Decider
- SemIf
- Jev Agent Failure Benchmark
- Kev — jaredpalmer
- Jev Search Rerank Eval
- Does Jev know when it doesn't know?
- Jev vs LLM: a phishing decision benchmark with a calibration audit
- Jev Secret Detection
- Jev Benchmarks
- Little Airways
- Janus
- Jev Research Eval
- Jev Playground — hegargarcia
- openJev-verdict-2.0
- Jev DSPy Lab
- Jev Orderby Bench
- JevBench — fstandhartinger
- ASSAY-001
- Jevbetter
- Openjev — daseinlabs
- OpenJev (Verdict)
- Reflex — kshetrajna12
- Jev as a judge for agent evals
- Jev Arena
- Jeff — logan-markewich
- Jev Capability Atlas
- JEVfire
- Poorjev
- Mini-Vibe Check: TypeSafe's Jev Judged Everything I've Written in 0.7 Seconds
- Jev: one judge call, or twelve dimension scores? I measured both on three tasks
- Testing Jev on public and private data: classifier or filter?
- An early-access test of TypeSafe's Jev: calibrated judgments for half a cent
- Testing TypeSafe Jev, Mistral and Gemini for local event validation
- How Much Does a Technical SEO Audit Cost in 2026?
- Jevals — Independent Jev Benchmark for Models That Decide
- A 706K-parameter form filler
- Security decisions that fit Jev
- Reranking 33,047 catalog entries
- Inferring Jev's internals from 1,000 calls
- State
- Jev 1.13 Jaggedness
- Choice
- Score
- Noul
- Advanced: Structure
- Pattern: Composite Scoring
- Pattern: Confidence-Gated Routing
- Pattern: Speculative Fan-Out
- Pattern: Intent Routing
- Building with Jev
- What Can You Build With Jev? Documented Use Cases
- Jev in the Agent Runtime
- Typed Decision Models: Jev and Laya in Agentic AI
- Jev as an LLM Judge
- Jev Beats Gemini Flash Lite on a Classifier Eval
- On Jev-Scored Compaction: A Critique
- Jev + WebMCP Breaks a Benchmark
- Primitives: Choice, Score & Noul
- Confidence
- Patterns
- How to Build with TypeSafe
- Example Use Case Map
Start here with the Jev AI model
Featured entries in practices & patterns.
The box question type: one bounding box per question, returned as four edges that are fractions of the image, plus confidence, with no criteria field and no not-found value.
Jev Practices & PatternsDocs#official#docs#vision
Official
Published evaluations of four automation workflows, security incidents, agent trace observability, invoice processing, and customer service, comparing Jev and frontier LLMs as structured workflows versus single prompts.
Jev Practices & PatternsDocs#official#benchmarks#evaluation
Official
Apache-licensed, locally runnable Jev-style decision models (0.8B, 4B, 9B on Qwen3.5) with released weights, training code, a System One-compatible server, frozen eval suites, and a playground.
Jev Practices & Patterns#community#python#open-models
Community
jaredpalmer/kevPython
jaredpalmer/kevPythonJev AI model articles
Original guides that go deeper on Jev practices & patterns.
Jev's compaction tools keep or drop tool calls instead of summarizing them — the thresholds they fire at, the savings they report, and why 1.14's opt-in 262k context doesn't settle the choice.
Jev Practices & Patterns#compaction#long-context#coding-agents
Read article
Per-tick buy/sell calls on Monad, a five-sleeve Hyperliquid desk, Nifty-50 rankers, and Polymarket confidence gates: how the catalog's trading repos wire Jev in without asking it to predict prices.
Jev Practices & Patterns#trading#markets#gates
Read article
How the box question type turns a crop into a decision: four fractional edges, a presence Noul for missing targets, the cut-off problem, and two patterns worth building.
Jev Practices & Patterns#vision#box#patterns
Read article
From the Jev community
Watch walkthroughs, read the threads, and see what builders are posting.
YouTube: A podcast segment on TypeSafe AI, its $40M raise, and the Jevons paradox behind the name.
YouTube: Gary Explains the catch: Jev understands natural language but answers with structured values and confidence.
YouTube: Sam Witteveen on System 1 thinking, then Choice, Score, and Noul demos, and chained actions.
Hacker News: Pipeable onesie CLI with shell-safety exit codes and a calibrate mode over labelled commands.
frodikarlsson3 pointsShow HN: onesie – An expressive Unix-pipeable CLI for System One models like JevRead the thread on Hacker NewsHacker News: Spoken-language learning with video scenes branching on TypeSafe Jev.
j_rcht1 pointShow HN: Worklaude – learn by speaking; video scenes branch on TypeSafe JEVRead the thread on Hacker NewsHacker News: OpenAI's Luna Decisions API framed as a fast follow to TypeSafe's Jev wave.
matt_d6 pointsOpenAI Answers TypeSafe's Jev with a Decision API Built on LunaRead the thread on Hacker NewsHacker News: Nikola Balić on what made the Jev launch easy to try and easy to share.
nkko2 pointsWhat TypeSafe Got Right with the Jev LaunchRead the thread on Hacker News