Discover public tools, agents, experiments, and research built around typed decisions.
Browse 433 projects→Browser Use's ultrafast agent: Jev picks the operation and DOM element in one request; a small LLM only writes text when typing is needed.
Browser Use 的超高速 Agent：Jev 在一次请求中选择操作和 DOM 元素，仅在需要输入文字时调用小型 LLM。
Open repository ↗Official agent skill for Claude Code, Codex, and compatible agents: primitives, patterns, and how to structure evaluations.
面向 Claude Code、Codex 及兼容 Agent 的官方技能，涵盖原语、模式和评测结构设计。
Official TypeScript/JavaScript client with inferred answer types. npm install @typesafe-ai/sdk.
官方 TypeScript/JavaScript 客户端，可推断答案类型；安装：npm install @typesafe-ai/sdk。
Official drop-in TypeSafeClient replacement backed by OpenAI, Anthropic, and compatible LLM APIs, for comparing Jev against chat models.
官方即插即用 TypeSafeClient 替代实现，支持 OpenAI、Anthropic 及兼容 LLM API，用于对比 Jev 与聊天模型。
Official sync and async Python client. pip install typesafe-sdk.
官方同步及异步 Python 客户端；安装：pip install typesafe-sdk。
A small, type-safe client for asking AI questions about your data, powered by TypeSafe Jev.
基于 TypeSafe Jev 的轻量类型安全客户端，可针对数据提出 AI 问题。
TypeSafe Jev for OTP: reply to Jev from a GenServer and pattern match on its answer.
面向 OTP 的 TypeSafe Jev 客户端：从 GenServer 响应 Jev，并对答案进行模式匹配。
Typed TypeSafe AI clients for Rust, with async and blocking backends and observable retries.
Rust 类型化 TypeSafe AI 客户端，提供异步和阻塞后端以及可观测重试。
Go SDK for the TypeSafe AI API — typed questions in, probability distributions out.
TypeSafe AI API 的 Go SDK：输入类型化问题，返回概率分布。
Zod validates the shape, Jev validates the meaning: semantic checks on request bodies become calibrated probabilities you threshold in code.
Zod 校验结构，Jev 校验语义：将请求体的语义检查转为可由代码设定阈值的校准概率。
A small, extensible decision-to-action harness for TypeSafe Jev.
小型、可扩展的 TypeSafe Jev“决策到动作”运行框架。
Community .NET SDK for the TypeSafe AI System One API — typed noul, choice, and score questions with structured, confidence-scored answers. Not affiliated with TypeSafe AI.
社区版 .NET SDK，支持 Noul、Choice 和 Score 类型问题并返回带置信度的结构化答案；非 TypeSafe AI 官方项目。
Agent-first Haskell DSL for TypeSafe's Jev judgment model: typed packets, inferred types, answers under the same labels.
面向 Agent 的 Haskell DSL，为 Jev 判断模型提供类型化数据包、类型推断和同标签答案。
Ruby client for typesafe.ai.
typesafe.ai 的 Ruby 客户端。
Idiomatic Java SDK for TypeSafe AI Jev System One decision engine.
符合 Java 习惯的 TypeSafe AI Jev System One 决策引擎 SDK。
Semantic schemas over TypeSafe's Jev — validate the state locally, then project typed answers.
构建在 Jev 之上的语义 Schema：先在本地验证状态，再映射类型化答案。
Independent async and blocking Rust SDK for the TypeSafe AI System One API.
TypeSafe AI System One API 的独立 Rust SDK，支持异步和阻塞调用。
Unofficial go SDK for typesafe AI, with typed answers, retries, and context support.
非官方 TypeSafe AI Go SDK，支持类型化答案、重试和上下文。
Community Java client for the TypeSafe System One API (unofficial).
TypeSafe System One API 的社区 Java 客户端（非官方）。
Dependency-free Swift 6 client for Jev Choice, Score, and Noul questions, with strict concurrency, retries, and offline transport tests.
无依赖的 Swift 6 客户端，支持 Jev Choice、Score、Noul、严格并发、重试和离线传输测试。
Swift SDK for TypeSafe AI.
TypeSafe AI 的 Swift SDK。
An idiomatic, type-safe Elixir port of the official TypeScript AI SDK (ai / ai-sdk) providing unified LLM integrations, streaming text and structured outputs, tool calling, and agentic workflows. Jev is their current flagship model and is the first System One model.
官方 TypeScript AI SDK（ai / ai-sdk）的惯用、类型安全 Elixir 移植版，提供统一 LLM 集成、流式文本、结构化输出、工具调用和 Agent 工作流；Jev 是其首个 System One 模型。
.NET SDK for the TypeSafe AI platform.
TypeSafe AI 平台的 .NET SDK。
Scala 3 / ZIO client for the System One API: typed end-to-end, several questions per round-trip via NamedTuple.
System One API 的 Scala 3 / ZIO 客户端，通过 NamedTuple 实现端到端类型化和单次往返多问题。
Go client for TypeSafe's System One API and its model Jev: typed judgments and calibrated probabilities instead of generated text.
TypeSafe System One API 及 Jev 模型的 Go 客户端，以类型化判断和校准概率替代文本生成。
Unofficial Go SDK for TypeSafe AI's Jev API.
TypeSafe AI Jev API 的非官方 Go SDK。
Async Python client for TypeSafe Jev. Typed questions in, probabilities and choices out, no prose to parse.
TypeSafe Jev 的异步 Python 客户端：输入类型化问题，直接获得概率与选择，无需解析文本。
Go client for TypeSafe AI's System One API (Jev), with optional Langfuse instrumentation.
TypeSafe AI System One API（Jev）的 Go 客户端，可选集成 Langfuse 观测。
Unofficial typed async Rust client for the TypeSafe System One API.
TypeSafe System One API 的非官方类型化异步 Rust 客户端。
Rust SDK for the TypeSafe AI API.
TypeSafe AI API 的 Rust SDK。
Go SDK for TypeSafe AI API https://docs.typesafe.ai/api.
TypeSafe AI API 的 Go SDK。
Probabilistic decisions for Python. Use Jev or bring your own provider; crawl with Playwright.
Python 概率决策库，可使用 Jev 或自带提供商，并通过 Playwright 抓取网页。
Unofficial Go client for TypeSafe's System One API and its model, Jev.
TypeSafe System One API 及 Jev 模型的非官方 Go 客户端。
Rust client for the TypeSafe System One API (Jev).
TypeSafe System One API（Jev）的 Rust 客户端。
Typed System One layer for Rust (Choice/Score/Noul).
Rust 的类型化 System One 层，支持 Choice、Score 和 Noul。
An integration with jev by typesafe.ai in Rust.
以 Rust 实现的 typesafe.ai Jev 集成。
An idiomatic Elixir client for the TypeSafe AI API.
符合 Elixir 习惯的 TypeSafe AI API 客户端。
Idiomatic Go SDK for the TypeSafe AI API.
符合 Go 习惯的 TypeSafe AI API SDK。
Latency-first Rust SDK for TypeSafe System One.
以低延迟为优先的 Rust TypeSafe System One SDK。
PHP & Laravel SDK for TypeSafe AI's JEV Model series.
面向 TypeSafe AI JEV 模型系列的 PHP 与 Laravel SDK。
Unofficial Go SDK for the TypeSafe AI System One API — 1:1 parity with the official JS and Python SDKs. Not affiliated with TypeSafe AI.
非官方 Go SDK，与 TypeSafe AI System One 官方 JS 和 Python SDK 保持 1:1 功能对应；非 TypeSafe AI 官方项目。
Unofficial PHP SDK for the TypeSafe AI System One API — 1:1 parity with the official JS and Python SDKs. Not affiliated with TypeSafe AI.
非官方 PHP SDK，与 TypeSafe AI System One 官方 JS 和 Python SDK 保持 1:1 功能对应；非 TypeSafe AI 官方项目。
An idiomatic Zig client for the TypeSafe AI API.
符合 Zig 习惯的 TypeSafe AI API 客户端。
A supervised Mint client for the TypeSafe AI System One API.
TypeSafe AI System One API 的受监督 Mint 客户端。
Unofficial Elixir SDK for the TypeSafe AI API.
TypeSafe AI API 的非官方 Elixir SDK。
Typesafe AI SDK in Elixir using Req.
基于 Req 的 Elixir TypeSafe AI SDK。
Vercel's open agent framework, which ships Jev as the default evaluation model in its experimental evaluate path.
Vercel 开源 Agent 框架，在实验性 evaluate 流程中将 Jev 作为默认评测模型。
Vercel Labs terminal CLI that can run Jev as the evaluation model for its evaluate command.
Vercel Labs 终端 CLI，可在 evaluate 命令中使用 Jev 作为评测模型。
Agentic TypeScript workflow framework with a Jev session checker wired into its workflows.
Agent 化 TypeScript 工作流框架，内置 Jev 会话检查器。
Ask your Postgres tables questions in plain language. A PostgreSQL extension powered by TypeSafe's Jev.
由 TypeSafe Jev 驱动的 PostgreSQL 扩展，可用自然语言查询 Postgres 表。
Self-hosted, versioned skills library for AI agents with optional Jev recommendations through TypeSafe or an AI gateway.
面向 AI Agent 的自托管、版本化技能库，可通过 TypeSafe 或 AI 网关提供 Jev 推荐。
Fish-style zsh history autosuggestions ranked by Jev (TypeSafe).
由 Jev 排序的 Fish 风格 zsh 历史自动建议。
Self-improving agent harness with an optional TypeSafe Jev companion for typed Choice, Score, and Noul judgments.
自我改进 Agent 框架，可选配 TypeSafe Jev，以执行 Choice、Score 和 Noul 类型化判断。
Typesafe.ai System One Model Jev navigating a Neo4j graph by using a classifier over neighbouring relationships.
使用邻接关系分类器，让 typesafe.ai 的 Jev System One 模型在 Neo4j 图中导航。
TypeSafe structured-output provider for RubyLLM 2.
RubyLLM 2 的 TypeSafe 结构化输出提供商。
Home Assistant integration for TypeSafe Jev. Ask a question about your house and get a probability, a choice or a score as an entity.
TypeSafe Jev 的 Home Assistant 集成，可针对家庭状态提问，并将概率、选择或评分作为实体返回。
Pre-alpha PostgreSQL extension for TypeSafe AI (Jev) categorical classification.
用于 TypeSafe AI（Jev）分类的早期预览版 PostgreSQL 扩展。
TypeSafe AI Jev judgments for Agent Zero, with typed tools and probability cards.
为 Agent Zero 提供 TypeSafe AI Jev 判断，包括类型化工具和概率卡片。
Async LangGraph workflow that gets a typed Jev Choice (invoice or general) and routes each inbound email to the matching handler.
异步 LangGraph 工作流，通过 Jev Choice 判断邮件属于发票还是一般邮件，并路由到对应处理器。
Neon Function proxy for the Neon AI Gateway with TypeSafe Jev routing.
面向 Neon AI Gateway 的 Neon Function 代理，使用 TypeSafe Jev 进行路由。
Unofficial Laravel integration for TypeSafe Jev AI with typed responses, async requests, scoped dependency injection, and testing fakes.
TypeSafe Jev AI 的非官方 Laravel 集成，支持类型化响应、异步请求、作用域依赖注入和测试替身。
LlamaIndex reranker + router powered by TypeSafe Jev — typed scores/choices, cheaper than LLM-as-judge.
由 TypeSafe Jev 驱动的 LlamaIndex 重排序器与路由器，提供类型化评分和选择，成本低于 LLM-as-judge。
Скилл для агентов Letta: суждения по критериям через TypeSafe System One (Jev).
面向 Letta Agent 的技能，通过 TypeSafe System One（Jev）按标准执行判断。
Pydantic AI capabilities made stronger with Jev: small runnable demos, one file each.
使用 Jev 增强 Pydantic AI 能力的一组单文件、可运行小型示例。
Community Rails integration on the typesafe-sdk gem: configuration, persisted usage and cost telemetry, and opt-in confidence policies.
基于 typesafe-sdk gem 的社区 Rails 集成，支持配置、用量与成本持久化，以及可选置信度策略。
Home Assistant Assist conversation agent powered by TypeSafe's Jev (System One) model.
由 TypeSafe Jev（System One）驱动的 Home Assistant Assist 对话 Agent。
Jev adapter to Mellea.
连接 Jev 与 Mellea 的适配器。
N8n community node for the TypeSafe AI System One API — typed yes/no, choice and score questions with calibrated probabilities.
TypeSafe AI System One API 的 n8n 社区节点，支持带校准概率的是非、选择和评分问题。
TypeSafe AI System One task plugin for New API with a native /v1/systemone endpoint, routing, and token billing.
面向 New API 的 TypeSafe AI System One 任务插件，提供原生 /v1/systemone 端点、路由和 Token 计费。
Shadcn-style reusable components and blocks for using TypeSafe AI.
用于 TypeSafe AI 的 Shadcn 风格可复用组件与区块。
Open-source Claude Cowork alternative whose eval testkit can use Jev as a typed verification judge for agent-produced work.
开源 Claude Cowork 替代方案，其评测工具包可使用 Jev 作为类型化裁判，验证 Agent 产出的工作。
Claude Code plugin that replaces the compaction summary with Jev decisions: every tool call and result is scored in one fast request, stale ones are dropped or truncated, everything kept stays verbatim.
Claude Code 插件，用 Jev 决策替代压缩摘要：一次快速请求评估所有工具调用与结果，丢弃或截断过期内容，其余内容原样保留。
Software Factory Foreman: an agent supervisor that uses Jev decisions to keep coding agents on task.
软件工厂 Foreman：使用 Jev 决策让编程 Agent 保持任务方向的监督器。
A staged code-review workflow and local dashboard built with TypeSafe Jev.
基于 TypeSafe Jev 的分阶段代码审查工作流与本地仪表板。
TypeSafe Jev as a decision layer for the Pi coding agent: a measured tool-call gate plus jev_ask for typed, calibrated answers.
Pi 编程 Agent 的 TypeSafe Jev 决策层，提供可度量的工具调用门控和用于类型化校准答案的 jev_ask。
Route to the cheapest model in claude code for your task using jev-router.
使用 jev-router 为 Claude Code 的任务选择成本最低的模型。
Local-first MCP plugin for continuous software-quality review by AI coding agents, powered by Jev.
本地优先的 MCP 插件，由 Jev 驱动，为 AI 编程 Agent 持续审查软件质量。
A skill for writing and improving programs that call Jev, TypeSafe's System One model.
用于编写和改进 Jev 调用程序的 Agent 技能。
Proof of concept MCP for Typesafe's new Jev AI model.
TypeSafe 新 Jev AI 模型的 MCP 概念验证。
Go CLI and single-binary MCP server exposing TypeSafe judgments to Claude Desktop, Claude Code, and Codex.
Go CLI 与单二进制 MCP 服务器，向 Claude Desktop、Claude Code 和 Codex 提供 TypeSafe 判断。
Guardrails for Pi built on pi-typesafe that steer the agent instead of interrupting you: Jev judges irreversible and off-task tool calls, detects stuck loops, checks unverified done claims, flags slop.
基于 pi-typesafe 的 Pi 护栏：由 Jev 判断不可逆和偏离任务的工具调用、检测死循环、检查未验证的完成声明并标记低质输出。
Search the web with TypeSafe's Jev: source selection, query understanding and relevance ranking. Built with Search1API.
使用 TypeSafe Jev 搜索网页，覆盖来源选择、查询理解和相关性排序；基于 Search1API。
Rust CLI powered by Jev from TypeSafe.ai that ranks agent skills for the next step using live session context. Includes Claude Code hooks, structured JSON, abstention, and local feedback. Requires a TypeSafe API key.
由 TypeSafe.ai Jev 驱动的 Rust CLI，依据实时会话上下文为下一步排序 Agent 技能；支持 Claude Code hooks、结构化 JSON、拒答和本地反馈。
Full-stack agent harness with memories, long-running tasks, multi-agent relay, and a Jev-backed model-routing table.
全栈 Agent 框架，支持记忆、长任务、多 Agent 中继和基于 Jev 的模型路由表。
Per-turn model & reasoning routing for Codex, driven by Jev (TypeSafe System One): picks the model, thinking depth and speed mode for every turn.
由 Jev 驱动的 Codex 单轮模型与推理路由器，为每轮选择模型、思考深度和速度模式。
Skill for Hermes, and other agents, to ask typesafe's jev.
让 Hermes 及其他 Agent 调用 TypeSafe Jev 的技能。
Route HTTP requests by meaning. A semantic router for Hono powered by Jev.
按语义路由 HTTP 请求的 Hono 路由器，由 Jev 驱动。
A calibrated context sieve for Claude Code: every tool result is judged by a System One model before it enters context.
Claude Code 的校准上下文筛选器，每个工具结果进入上下文前都由 System One 模型判断。
Agent-ergonomic CLI for TypeSafe's Jev: fast calibrated judgments (pick, rate, check, rank, triage, guard) from the shell.
面向 Agent 的 Jev CLI，可在 Shell 中快速执行带校准的选择、评分、检查、排序、分诊和护栏判断。
Context-pruning proxy for Claude Code and Codex: Jev judges which history is still needed, measured not claimed. POC here now, heading soon into Compozy.
Claude Code 与 Codex 的上下文裁剪代理，由 Jev 判断哪些历史仍有价值；概念验证将迁移至 Compozy。
Jev (TypeSafe System One) backed auto mode for the Pi coding agent: semantically auto-approves bash, write, and edit tool calls and fails closed when a decision cannot be made.
由 Jev 驱动的 Pi 编程 Agent 自动模式，按语义自动批准 Bash、写入和编辑调用，无法决策时默认拒绝。
MCP server for TypeSafe Jev: typed classify, score, check, match and screen for any agent, with confidence on every answer.
TypeSafe Jev MCP 服务器，为任意 Agent 提供类型化分类、评分、检查、匹配和筛查，并为每个答案附带置信度。
Configurable semantic linting powered by Jev, with file-level NOUL judgments and a magic-strings plugin.
由 Jev 驱动的可配置语义 Linter，支持文件级 Noul 判断和魔法字符串插件。
Bounded TypeSafe Jev workflows for coding agents.
面向编程 Agent 的有边界 TypeSafe Jev 工作流。
AI workspace whose approval review uses Jev to gate workspace actions with typed decisions.
AI 工作空间，通过 Jev 审批审查以类型化决策约束工作空间操作。
Semantic tool routing and typed System One decisions for the Pi coding agent using TypeSafe Jev.
使用 TypeSafe Jev 为 Pi 编程 Agent 提供语义工具路由和类型化 System One 决策。
Automatic model routing for Pi using TypeSafe's Jev through Vercel AI Gateway.
通过 Vercel AI Gateway 使用 TypeSafe Jev，为 Pi 自动选择模型。
A prose linter that sniffs out AI writing tells. Zero dependencies, countable rules plus one judgment model.
检测 AI 写作痕迹的文本 Linter，零依赖，结合可计数规则与一个判断模型。
TypeSafe (Jev) skill routing for Hermes Agent: names the one skill worth loading, before the model call. Opt-in, stdlib only, ~$0.001 per routed turn.
Hermes Agent 的 TypeSafe（Jev）技能路由器，在模型调用前选出唯一值得加载的技能；按需启用且仅使用标准库。
Typed, confidence-aware agent skill routing with TypeSafe Jev.
使用 TypeSafe Jev 实现类型化、感知置信度的 Agent 技能路由。
Connect JEV to MCP clients and compare its judgments against general-purpose LLMs using shared datasets and measurable accuracy.
将 Jev 接入 MCP 客户端，并用共享数据集和可量化准确率对比其与通用 LLM 的判断。
A lightweight Jev-powered router for models, tools, and subagents.
由 Jev 驱动的轻量模型、工具及子 Agent 路由器。
Jev decision layer for agents: MCP server, embeddable DecisionModel library, and an escalate-only Claude Code plugin (TypeSafe AI's Jev).
Agent 的 Jev 决策层，包含 MCP 服务器、可嵌入 DecisionModel 库和仅在升级时介入的 Claude Code 插件。
Every code file in a pull request, judged against Uncle Bob's Clean Code by TypeSafe's Jev, then reviewed by Luna. Built on eve and Next.js.
逐个使用 Jev 按《代码整洁之道》判断 Pull Request 文件，再由 Luna 审查；基于 eve 和 Next.js。
DiffJury — TypeSafe Jev PR risk router + code review coach.
使用 TypeSafe Jev 进行 Pull Request 风险路由并提供代码审查指导。
Typed System One decisions, ranking, verification, and an opt-in Hermes tool gate using TypeSafe Jev.
使用 TypeSafe Jev 提供类型化 System One 决策、排序、验证及可选 Hermes 工具门控。
PoC: TypeSafe Jev as the reviewer for Hermes Agent smart command approvals. 8.7x faster, 4.4x fewer prompts, measured on 153 real commands. Approvals only.
概念验证：让 TypeSafe Jev 审查 Hermes Agent 的智能命令审批；在 153 条真实命令上测得更快且提示更少，仅处理审批。
A claude code plugin for jev.
面向 Claude Code 的 Jev 插件。
Minimal agent loop where Jev directs control flow and a LangChain chat model writes argument values and the final response.
最小化 Agent 循环，由 Jev 控制流程，LangChain 聊天模型生成参数值和最终回复。
TypeSafe Jev as the pi coding agent's quiet decision layer.
作为 Pi 编程 Agent 静默决策层的 TypeSafe Jev。
A small, fast prose linter: ruff-style rule codes for writing, backed by TypeSafe's Jev model.
小巧快速的文本 Linter，使用类似 Ruff 的写作规则代码，并由 TypeSafe Jev 支持。
Voice-driven semantic auto-advance for Slidev, powered by Cloudflare Agents and TypeSafe AI Jev.
Slidev 的语音驱动语义自动翻页工具，由 Cloudflare Agents 和 TypeSafe AI Jev 驱动。
CLI and agent skill for TypeSafe System One (Jev): typed Choice, Score, and Noul judgments.
TypeSafe System One（Jev）的 CLI 与 Agent 技能，支持 Choice、Score 和 Noul 类型化判断。
Type-safe model router. Jev (System One) banks each request to a typed catalog route.
类型安全的模型路由器，由 Jev（System One）将每个请求分配到类型化目录路由。
Hybrid coding harness: System 2 writes, System 1 (Jev) runs reflexes.
混合编程框架：System 2 负责写作，System 1（Jev）执行反射式判断。
Codex plugin: verbatim Jev-guided context restoration around session compaction. Port of tamaratran/fast-jev-compaction to Codex lifecycle hooks.
Codex 插件，在会话压缩前后按 Jev 指引原样恢复上下文；由 fast-jev-compaction 移植至 Codex 生命周期 hooks。
Local proxy that picks the Claude model and effort per message using TypeSafe Jev. Routes subagents, leaves your cached main chat alone.
本地代理，使用 TypeSafe Jev 为每条消息选择 Claude 模型和推理强度；路由子 Agent，不影响主对话缓存。
Prompt-injection and dangerous-action guard for coding agents (Claude Code, Codex, pi, ACP), powered by Jev.
由 Jev 驱动的编程 Agent 提示词注入与危险操作护栏，支持 Claude Code、Codex、Pi 和 ACP。
Agent Skill: send closed coding-agent judgments to TypeSafe Jev.
Agent 技能：将编程 Agent 的封闭式判断交给 TypeSafe Jev。
MCP server exposing TypeSafe Jev as typed, calibrated judgment tools: classify, score, check, batched ask. Ships as a Claude Code plugin.
MCP 服务器，将 TypeSafe Jev 暴露为类型化、校准的判断工具，支持分类、评分、检查和批量提问，并提供 Claude Code 插件。
Semantic MCP firewall powered by Jev — screens every tool call, tool result, and tool description with calibrated System One verification. 94% block recall, 0 false positives, ~$0.00002/check.
由 Jev 驱动的语义 MCP 防火墙，以校准的 System One 验证筛查每个工具调用、结果和说明；报告阻止召回率 94%、零误报，单次检查成本约 0.00002 美元。
Agent-first SEO and GEO CLI suite and MCP server using DuckDuckGo evidence and Jev scoring.
面向 Agent 的 SEO 与 GEO CLI 套件和 MCP 服务器，使用 DuckDuckGo 证据与 Jev 评分。
Cut Claude Code's skill manifest by ~75% with TypeSafe Jev. Scores every installed skill for relevance and hides the rest via skillOverrides — 12,750 → 3,185 tokens on a 217-skill install, for $0.0009 a session.
使用 TypeSafe Jev 将 Claude Code 技能清单缩减约 75%，按相关性评分并通过 skillOverrides 隐藏其余技能。
System-architecture skill for TypeSafe AI Jev/System One — find fuzzy semantic judgment and turn it into small Choice/Score/Noul primitives.
面向 TypeSafe AI Jev/System One 的系统架构技能，将模糊语义判断拆为小型 Choice、Score 和 Noul 原语。
Build versioned judgment functions on TypeSafe's Jev once, then call the same published version from your backend over HTTP and from coding agents over MCP. The vendor key stays on your machine.
在 TypeSafe Jev 上构建版本化判断函数，并从后端经 HTTP、从编程 Agent 经 MCP 调用同一发布版本；供应商密钥保留在本机。
Grep, but the pattern is a description. Filters lines by meaning with TypeSafe's Jev decision model: ~200 ms and a thousandth of a cent per line.
以自然语言描述作为匹配模式的 Grep，使用 Jev 决策模型按语义过滤行。
Verbatim Jev-scored context reduction for omp, over TypeSafe or OpenRouter.
通过 TypeSafe 或 OpenRouter 为 OMP 提供由 Jev 评分、原文保留的上下文缩减。
Pi extension: verbatim context compaction with TypeSafe Jev decisions.
Pi 扩展：使用 TypeSafe Jev 决策进行原文保留式上下文压缩。
Runtime constraints for the pi coding agent: checks every side-effecting tool call against what you said, before it runs. Powered by TypeSafe Jev.
Pi 编程 Agent 的运行时约束，在有副作用的工具调用执行前，根据用户要求逐项检查；由 TypeSafe Jev 驱动。
Tell Claude Code which installed skill a session needs, using Jev (TypeSafe AI) for the decision and skills.sh for discovery.
使用 Jev 判断 Claude Code 会话需要哪个已安装技能，并通过 skills.sh 发现技能。
Shift every LLM call to the cheapest model that can handle it. Routing decided by TypeSafe Jev in ~180 ms. No training data. Policy in plain YAML. TypeScript and Python.
将每次 LLM 调用路由到能够完成任务的最低成本模型；约 180 毫秒内由 TypeSafe Jev 决策，策略使用 YAML，支持 TypeScript 和 Python。
⚡ Ultra-fast, low-cost intelligent task classifier and 3-tier routing engine powered by TypeSafe Jev (System One).
由 TypeSafe Jev（System One）驱动的高速低成本任务分类器与三级路由引擎。
Automated database migration safety reviewer powered by TypeSafe AI (Jev System One model).
由 TypeSafe AI（Jev System One）驱动的自动化数据库迁移安全审查器。
Claude Code mod that routes decisions to TypeSafe's Jev model: ranks installed skills per prompt, and answers the agent's own this-or-that questions when confident.
Claude Code 模组，将决策路由到 TypeSafe Jev：按提示词排序已安装技能，并在置信度足够时回答 Agent 的封闭式问题。
Supervises coding-agent edits and uses Jev to flag violations of project rules.
监督编程 Agent 的编辑，并使用 Jev 标记违反项目规则的修改。
Fast semantic code search & diff sanity auditor for AI coding assistants (Antigravity, Cursor, Claude Code) powered by TypeSafe System One.
由 TypeSafe System One 驱动的快速语义代码搜索与 Diff 合理性审计器，面向 Antigravity、Cursor 和 Claude Code 等 AI 编程助手。
Utilizing Jev, the RLCD-type model provided by TypeSafe AI, to independently and cheaply judge agentic coding sessions.
使用 TypeSafe AI 提供的 RLCD 类模型 Jev，独立且低成本地评判 Agent 编程会话。
Scans a codebase for covert, deceptive, or data-stealing behavior with Jev, then reports suspicious files and line ranges before the user runs it.
使用 Jev 扫描代码库中的隐蔽、欺骗或窃取数据行为，在运行前报告可疑文件与行范围。
Command-line tool for TypeSafe's Jev AI model.
TypeSafe Jev AI 模型的命令行工具。
MCP server that puts TypeSafe Jev on the coding loop in Cursor, Codex, and any MCP client.
将 TypeSafe Jev 接入 Cursor、Codex 及任意 MCP 客户端编程循环的 MCP 服务器。
Predict another skill's next closed decision with TypeSafe Jev — without running that skill.
无需运行目标技能，即可使用 TypeSafe Jev 预测其下一项封闭式决策。
Open-source LLM router that uses TypeSafe's Jev to pick a model, on top of LiteLLM.
构建在 LiteLLM 之上的开源 LLM 路由器，使用 TypeSafe Jev 选择模型。
Grounded GitHub and crates.io discovery CLI and MCP server that uses Jev to score architectural fit, licensing, and maintenance.
有依据的 GitHub 与 crates.io 发现 CLI 和 MCP 服务器，使用 Jev 评估架构适配、许可证与维护状态。
Per-prompt capability router for coding agents: resolves installed skills, MCP servers, agents and commands against your prompt via TypeSafe Jev, and measures whether the injection actually helps.
面向编程 Agent 的逐提示能力路由器，用 TypeSafe Jev 将提示词匹配到已安装技能、MCP 服务器、Agent 和命令，并衡量注入是否真正有效。
A Stop hook that stops your coding agent from stopping too early. Plain-language rules, judged by jev.
阻止编程 Agent 过早结束的 Stop hook，使用自然语言规则并由 Jev 判断。
TypeSafe AI (Jev) adversarial reviewer and typesafe_ask tool for the omp coding agent.
为 OMP 编程 Agent 提供 TypeSafe AI（Jev）对抗式审查器和 typesafe_ask 工具。
Agent skill for designing Jev-assisted systems with decision theory, composition patterns, question diagnostics, and validation gates.
用于设计 Jev 辅助系统的 Agent 技能，涵盖决策理论、组合模式、问题诊断和验证门控。
Reusable GitHub Action: agent fix loop gated by checks, an AI reviewer, and TypeSafe Jev.
可复用 GitHub Action：由检查、AI 审查器和 TypeSafe Jev 门控的 Agent 修复循环。
Experimental protocol for evidence-aware agent handoffs, with Jev-assisted review before results reach the lead agent.
面向证据感知 Agent 交接的实验协议，在结果到达主 Agent 前使用 Jev 辅助审查。
A CLI and GitHub Action that assesses code-change risk using deterministic rules and TypeSafe Jev, recommending checks and reviewers before merge.
结合确定性规则和 TypeSafe Jev 评估代码变更风险的 CLI 与 GitHub Action，在合并前推荐检查项和审查者。
Configurable decision MCP server with AI SDK, Jev, percentage scores, and bias-profile routing.
可配置的决策 MCP 服务器，支持 AI SDK、Jev、百分比评分和偏差配置路由。
A flexible and configurable CLI model router using TypeSafe Jev.
使用 TypeSafe Jev 的灵活可配置 CLI 模型路由器。
Experimental Hermes plugin: Jev-assisted model routing plans with budget and capability constraints. API access pending.
实验性 Hermes 插件：在预算与能力约束下生成 Jev 辅助的模型路由计划；API 访问尚待开放。
Grok skill: Jev as a judgment sensor in a builder-agent loop (priors × probabilities → next act).
Grok 技能：在构建型 Agent 循环中将 Jev 用作判断传感器，以先验概率乘以观测概率决定下一步动作。
Sub-second pre-commit and pre-push semantic gate that screens staged diffs with Jev.
亚秒级 pre-commit 和 pre-push 语义门控，使用 Jev 筛查暂存 Diff。
Autonomous Jev pull-request review with typed decisions, calibrated approval gates, and trusted-owner escalation.
自主 Jev Pull Request 审查，提供类型化决策、校准审批门控和可信所有者升级机制。
Cross-agent software-development skills that add Jev decisions to planning, execution, debugging, and completion gates.
跨 Agent 软件开发技能，将 Jev 决策加入规划、执行、调试和完成门控。
Uses typeful jev, zero sync to pull and sync large repositories for issue triage.
使用类型化 Jev，并以零同步方式拉取和同步大型仓库以进行 Issue 分诊。
Semantic code review with Jev, plain-English rules, and installable agent skills.
使用 Jev、自然语言规则和可安装 Agent 技能进行语义代码审查。
Mastra input processor that uses Jev for a typed block decision and moderation category.
Mastra 输入处理器，使用 Jev 作出类型化阻止决策并判定内容审核类别。
Jev-powered model-judged permission gate for OMP (TypeSafe System One).
由 Jev 驱动、模型判断的 OMP 权限门控。
Send Pi agents back to work when they stop before the job is done.
当 Pi Agent 在任务完成前停止时，将其送回继续工作。
Single-agent Pi coding coprocessor with Jev semantic gates, baseline-to-current diff review, and append-only observability telemetry.
单 Agent Pi 编程协处理器，提供 Jev 语义门控、基线到当前状态的 Diff 审查及仅追加式可观测遥测。
Pi coding-agent extension built on the TypeSafe AI System One API (Jev).
基于 TypeSafe AI System One API（Jev）的 Pi 编程 Agent 扩展。
A pi extension that exposes TypeSafe (Jev, System One) judgments as five pi tools, so a model can make narrow semantic judgments while your code and your users keep control of thresholds, weights, and actions.
将 TypeSafe（Jev、System One）判断暴露为五个 Pi 工具，使模型执行狭窄语义判断，同时由代码和用户控制阈值、权重与动作。
Guardrail + model router for LLM gateways on TypeSafe's Jev (System One model), with an independent accuracy/calibration/latency evaluation. Stdlib Python.
基于 TypeSafe Jev 的 LLM 网关护栏与模型路由器，附带独立准确率、校准度和延迟评测；使用 Python 标准库。
MCP server exposing TypeSafe System One judgments (noul, choice, score) as agent tools.
将 TypeSafe System One 判断（Noul、Choice、Score）暴露为 Agent 工具的 MCP 服务器。
Using Typesafe.AI to generate diff reviews.
使用 Typesafe.AI 生成 Diff 审查。
Typed judgment layer for coding agents — gates from PRD to ship. Jev-ready, provider-agnostic.
面向编程 Agent 的类型化判断层，从 PRD 到发布全程设置门控；支持 Jev 且不绑定提供商。
Browser Use's ultrafast agent: Jev picks the operation and DOM element in one request; a small LLM only writes text when typing is needed.
Browser Use 的超高速 Agent：Jev 在一次请求中选择操作和 DOM 元素，仅在需要输入文字时调用小型 LLM。
Computer use for about $0.0002 a step: OCR the screen, classify the next action with TypeSafe, click. macOS.
单步成本约 0.0002 美元的 macOS 计算机操作方案：OCR 识别屏幕、TypeSafe 分类下一动作并执行点击。
Standalone Android agent for Mobilerun where Jev makes every decision, with a live React studio and an Uber demo.
面向 Mobilerun 的独立 Android Agent，每项决策均由 Jev 完成，附带实时 React Studio 和 Uber 演示。
Browser use using Typesafe's Jev model.
使用 TypeSafe Jev 模型的浏览器操作工具。
WXT browser extension: Jev-powered page clutter removal with reusable template rules.
WXT 浏览器扩展，使用 Jev 和可复用模板规则清除页面干扰内容。
🧹 Fun project: a Chrome extension that asks a tiny AI decision model (TypeSafe Jev) "is this DOM element an ad?" and pops it off the page. BYOK, no backend, not a real ad blocker.
实验性 Chrome 扩展，让 Jev 判断 DOM 元素是否为广告并将其移除；BYOK、无后端，并非完整广告拦截器。
5–10x faster browser operations: Jev clicks, Codex thinks and verifies. Built at EZCollegeApp.
速度提升 5–10 倍的浏览器操作方案：Jev 点击，Codex 思考并验证；由 EZCollegeApp 构建。
Control a real browser by voice. Jev (TypeSafe System One) decides intent + target in ~300 ms per spoken word; Playwright acts — often before you finish the sentence.
通过语音控制真实浏览器，Jev 在约 300 毫秒内判断意图与目标，Playwright 随即执行。
ACP and MCP adapter that bridges TypeSafe Jev with any LLM — computer use and typed decisions alongside Codex, Claude, Grok, and OpenCode.
ACP 与 MCP 适配器，将 TypeSafe Jev 接入任意 LLM，为 Codex、Claude、Grok 和 OpenCode 提供计算机操作与类型化决策。
Chrome extension: vibe-check your X posts with TypeSafe's Jev before you hit Post.
Chrome 扩展，在发布 X 帖子前使用 TypeSafe Jev 检查内容观感。
Browser automation where an LLM plans and Jev (Typesafe System One) decides. Library, CLI and MCP server.
由 LLM 规划、Jev 决策的浏览器自动化方案，提供库、CLI 和 MCP 服务器。
在 X 的时间线上，给每条帖子标出它想让你干什么。判断来自 Jev，一个只返回概率、不生成文本的模型。
在 X 时间线上标注每条帖子的行为意图，判断由仅返回概率而不生成文本的 Jev 完成。
Hybrid browser harness: an LLM turns goals into verifiable subgoals, Jev chooses each action and DOM field, Playwright acts.
混合浏览器框架：LLM 将目标拆为可验证子目标，Jev 选择动作和 DOM 字段，Playwright 负责执行。
2D autonomous car simulation in the browser, driven by TypeSafe's Jev decision model.
浏览器中的二维自动驾驶模拟，由 TypeSafe Jev 决策模型驱动。
AskJev — Jev autopilot for any website + guard on irreversible clicks (TypeSafe System One, not Claude).
适用于任意网站的 Jev 自动驾驶工具，并为不可逆点击提供护栏。
MacOS computer use driven by Jev (TypeSafe System One) as the decision maker.
以 Jev（TypeSafe System One）作为决策器的 macOS 计算机操作方案。
One grounded Jev/Playwright core: typed SDK, persistent CLI, and MCP server with native browser operations and deterministic assertions.
统一的 Jev/Playwright 核心，提供类型化 SDK、持久化 CLI、MCP 服务器、原生浏览器操作和确定性断言。
TypeScript browser agent for ego lite: Jev Ultrafast indexed actions, TypeSafe Jev decisions, persistent observe/act CLI. No Chrome or Playwright.
面向 ego lite 的 TypeScript 浏览器 Agent，使用 Jev Ultrafast 索引动作、TypeSafe Jev 决策及持久 observe/act CLI，无需 Chrome 或 Playwright。
Evidence-driven frontend QA built on Jev Ultrafast and Browser Harness, with a synthetic todo demo.
基于 Jev Ultrafast 与 Browser Harness、以证据为导向的前端 QA，附带合成 Todo 演示。
Privacy-first Chrome extension that semantically blocks native ads, sponsored feed cards, and video ads using TypeSafe Jev.
隐私优先的 Chrome 扩展，使用 TypeSafe Jev 按语义阻止原生广告、赞助信息流卡片和视频广告。
Almond-fastloop: Almond's browser computer-use rig (Chrome DevTools + TypeSafe Jev), and the Browser Use Olympics benchmark it is measured on.
Almond 的浏览器计算机操作框架，结合 Chrome DevTools 与 TypeSafe Jev，并使用 Browser Use Olympics 基准测试。
A Chrome extension that brings TypeSafe's Jev to X.com to analyze posts as you browse.
将 TypeSafe Jev 接入 X.com、在浏览过程中分析帖子的 Chrome 扩展。
Drives a real browser with Jev making every decision and Vercel's agent-browser performing every action, with a benchmark.
由 Jev 完成所有决策、Vercel agent-browser 执行动作的真实浏览器 Agent，附带基准测试。
Open-source native computer use for macOS and Windows: TypeSafe Jev, local OCR, and selective planning.
适用于 macOS 和 Windows 的开源原生计算机操作方案，结合 TypeSafe Jev、本地 OCR 和选择性规划。
Parallel web search for terminals and agents, with local Chromium and Jev-guided exploration.
面向终端与 Agent 的并行网页搜索工具，使用本地 Chromium 和 Jev 引导探索。
Browser Use Olympics by Almond: one prompt, five events, one clock. Plus fast loop, a ~200-line browser computer-use agent (Chrome DevTools + TypeSafe Jev).
Almond 的 Browser Use Olympics：一个提示、五项任务、统一计时；另含约 200 行、结合 Chrome DevTools 与 Jev 的快速浏览器 Agent。
Local browser automation UI that uses TypeSafe Jev to choose bounded page actions and a text model only for field values.
本地浏览器自动化 UI，使用 TypeSafe Jev 选择有边界的页面动作，仅由文本模型生成字段值。
Jev-powered browser MCP for LLM agents — ~300ms decisions, no LLM tokens in the loop. Benchmark vs Playwright MCP included.
由 Jev 驱动的浏览器 MCP，面向 LLM Agent 提供约 300 毫秒决策，循环内不消耗 LLM Token，并附 Playwright MCP 对比基准。
Fast structured Android control loops with TypeSafe Jev and Mobile MCP.
结合 TypeSafe Jev 与 Mobile MCP 的高速结构化 Android 控制循环。
Jev-augmented Playwright MCP proxy — page-state triage, prompt-injection shielding, goal-based snapshot pruning, risky-action gating. Drop-in wrapper around @playwright/mcp for any coding agent.
Jev 增强的 Playwright MCP 代理，提供页面状态分诊、提示词注入防护、按目标裁剪快照和高风险动作门控。
Bounded exploratory browser testing with Jev, deterministic assertions, and replayable evidence.
使用 Jev 的有边界探索式浏览器测试，结合确定性断言和可重放证据。
Chrome extension that re-ranks Google results with TypeSafe Jev and folds away sales pages and SEO filler.
Chrome 扩展，使用 TypeSafe Jev 重排 Google 搜索结果并折叠销售页面与 SEO 填充内容。
Open-source Chrome extension that filters AI-generated prose and ads with Jev.
开源 Chrome 扩展，使用 Jev 过滤 AI 生成文本和广告。
Jev-powered semantic browser use.
由 Jev 驱动的语义浏览器操作工具。
Browser agent where Jev selects actions, an LLM reads and plans, and answers cite page evidence.
浏览器 Agent：Jev 选择动作，LLM 阅读并规划，最终答案引用页面证据。
One AI trade decision every Monad block. Jev on Kuru MON-USDC.
在 Monad 每个区块执行一次 AI 交易决策，在 Kuru 的 MON-USDC 市场上运行 Jev。
Marketing analytics platform whose feature flag routes brand-visibility classifiers off an LLM and onto Jev boolean decisions.
营销分析平台，通过功能开关将品牌可见度分类器从 LLM 切换为 Jev 布尔决策。
Put a live Jev (TypeSafe) meter on any video: every sentence scored, rendered as a 16:9 edit.
为任意视频叠加实时 Jev 仪表：逐句评分并渲染为 16:9 成片。
Real-time Discord moderation bot: Jev evaluates messages and metadata in parallel to catch phishing, spam, and social engineering with a progressive escalation ladder.
实时 Discord 审核机器人，Jev 并行评估消息与元数据，通过渐进式升级机制识别钓鱼、垃圾信息和社会工程。
Classify Git commit diffs and messages with Jev. Bug fixes, security fixes/CWEs, and change types.
使用 Jev 分类 Git Commit Diff 与消息，识别 Bug 修复、安全修复/CWE 和变更类型。
Codebase search powered by Jev from @typesafe-ai.
由 @typesafe-ai Jev 驱动的代码库搜索。
Trading bot with the all new TypeSafe AI's first system one model named as Jev.
使用 TypeSafe AI 首个 System One 模型 Jev 的交易机器人。
Open-source Jev log triage for OpenTelemetry. Score the signal before expensive LLM analysis.
面向 OpenTelemetry 的开源 Jev 日志分诊，在昂贵的 LLM 分析前先评估信号。
Self-hosted Stacks data service whose Slack gate and fault-triage paths use Jev decisions.
自托管 Stacks 数据服务，其 Slack 门控和故障分诊路径使用 Jev 决策。
High-speed recursive AI Elo tournament engine powered by Jev and Swiss matchmaking.
由 Jev 和瑞士制匹配驱动的高速递归 AI Elo 锦标赛引擎。
Typed semantic decisions for Unix pipelines and CI, powered by TypeSafe AI Jev.
面向 Unix 管道与 CI 的类型化语义决策工具，由 TypeSafe AI Jev 驱动。
A spatial reference explorer for creators. Local Jev query choices, metadata highlights and source-linked collections.
面向创作者的空间化参考资料探索器，提供本地 Jev 查询选择、元数据高亮和带来源链接的收藏。
Ask Jev typed questions from the shell: noul, choice, and score answers as numbers, not prose.
从 Shell 向 Jev 提出类型化问题，以数字而非文本返回 Noul、Choice 和 Score 答案。
Helping JEV speak <3.
帮助 Jev 进行自然语言表达的实验项目。
Live Jev trader on Hyperliquid.
运行于 Hyperliquid 的实时 Jev 交易工具。
Ask a yes/no question of every function in a codebase. Ranked answers in seconds, for cents. Grep whose pattern is a question, powered by TypeSafe Jev.
针对代码库中的每个函数提出是非问题，并在数秒内低成本返回排序答案；由 Jev 驱动的“问题式 Grep”。
Small dependency-free CLI for TypeSafe Jev.
TypeSafe Jev 的小型无依赖 CLI。
Recursive Jev choice over a taxonomy. Select from more than 255 options without breaking TypeSafe Jev's choice cap.
在分类树上递归执行 Jev Choice，绕过单次最多 255 个选项的限制。
A live tone labeler for Bluesky posts and drafts, using TypeSafe's Jev API.
使用 TypeSafe Jev API 为 Bluesky 帖子和草稿实时标注语气。
Everyday Stocks Status with Jev.
使用 Jev 展示日常股票状态。
SQL with natural-language predicates, powered by TypeSafe's Jev. Filter, rank, classify and score rows by meaning — batched, cached and cost-guarded.
由 Jev 驱动、支持自然语言谓词的 SQL，可按语义过滤、排序、分类和评分，并提供批处理、缓存与成本保护。
Reimagined window switcher for macOS using frontier artificial intelligence. Predicted by TypeSafe's Jev model.
使用前沿 AI 重构的 macOS 窗口切换器，由 TypeSafe Jev 模型预测目标窗口。
Experimental multi-horizon BTC signal generator using TypeSafe Jev probabilities and Binance market data.
实验性多时间尺度 BTC 信号生成器，使用 TypeSafe Jev 概率和 Binance 市场数据。
Put your code comments on trial. Powered by Jev.
使用 Jev 审判代码注释的质量。
Desktop GUI for agentic coding that routes tasks across local agents and accounts, with Jev picking the best agent per task, running basic code-review checks, and supplying context.
面向 Agent 编程的桌面 GUI，在本地 Agent 和账户间路由任务，由 Jev 为每项任务选择最佳 Agent、执行基础代码审查并提供上下文。
CLI for TypeSafe AI's Jev evaluation model — typed questions in, structured JSON answers out.
TypeSafe AI Jev 评测模型的 CLI：输入类型化问题，输出结构化 JSON 答案。
Use Jev (TypeSafe's System One model) as a calibrated reranker: one call, up to 30 documents, a probability per document. Apache-2.0.
将 Jev 用作校准重排序器：单次调用最多处理 30 个文档，并为每个文档返回概率。
Browser extension that grades page sections for clarity, writing quality, and on-page SEO with Jev.
浏览器扩展，使用 Jev 对页面区块的清晰度、写作质量和站内 SEO 进行评分。
Typed decisions from a System One model. An uncertain answer is a different type from a confident one — and the compiler makes you handle it.
System One 模型的类型化决策库，将不确定答案与确定答案表示为不同类型，并由编译器强制处理。
Screen a folder of CVs with the TypeSafe Jev decision model: typed judgments, an editable policy, free re-scoring.
使用 Jev 决策模型筛选简历文件夹，提供类型化判断、可编辑策略和免费重新评分。
Check whether each cited paper supports the sentence citing it. Claude proves the quote, TypeSafe's Jev scores it, a human decides.
检查被引用论文是否支持引用它的句子：Claude 验证引文，Jev 评分，最终由人工决定。
Experimental: live X draft viral scorer powered by TypeSafe Jev.
实验性实时 X 草稿传播潜力评分器，由 TypeSafe Jev 驱动。
Emoji autocomplete at the speed of typing. TypeSafe AI Jev on a Whop-hosted TanStack Start app.
打字速度级别的 Emoji 自动补全，在 Whop 托管的 TanStack Start 应用中使用 TypeSafe Jev。
Low-latency audio censorship POC using Jev typed decisions and ffmpeg.
使用 Jev 类型化决策与 ffmpeg 的低延迟音频消音概念验证。
JEV Document Classification enables the rapid and cost-effective classification of text-based documents using AI, leveraging TypeSafe's "System One" model.
利用 TypeSafe System One 模型快速、低成本地分类文本型文档。
CV diagnostics and job alignment with TypeSafe Jev, React and FastAPI.
结合 TypeSafe Jev、React 和 FastAPI 的简历诊断与岗位匹配工具。
Scores public note.com articles across eight Jev dimensions and produces an inspectable AI-slop score.
从八个 Jev 维度评估公开 note.com 文章，并生成可检查的 AI 低质内容评分。
Guardrails for LLM apps in one API call. Prompt injection, jailbreaks, leaks, unsafe content. Built on TypeSafe Jev. MIT.
一次 API 调用为 LLM 应用提供护栏，检测提示词注入、越狱、信息泄露和不安全内容；基于 TypeSafe Jev。
A .NET 10 and React 19 application for fast, structured AI-powered ticket triage using TypeSafe Jev.
基于 .NET 10 与 React 19 的快速结构化 AI 工单分诊应用，使用 TypeSafe Jev。
Rewrite AI drafts in your own voice with a bounded TypeSafe feedback loop.
通过有边界的 TypeSafe 反馈循环，将 AI 草稿改写为用户自己的表达风格。
Pre-install security gate for npm lifecycle scripts using TypeSafe System One.
使用 TypeSafe System One 检查 npm 生命周期脚本的安装前安全门控。
System One Search: navigate and trace code with TypeSafe judgments and repository evidence.
System One Search：结合 TypeSafe 判断和仓库证据导航并追踪代码。
Scam text message filter powered by TypeSafe's Jev model.
由 TypeSafe Jev 模型驱动的诈骗短信过滤器。
ACME live support-call scoring demo with TypeSafe AI, Effect, SQLite, React, Vite, and Turborepo.
ACME 实时客服通话评分演示，使用 TypeSafe AI、Effect、SQLite、React、Vite 和 Turborepo。
Small Python package that uses typesafe.ai to evaluate code comments on certain heuristics.
使用 typesafe.ai 按特定启发式规则评估代码注释的小型 Python 包。
Three composable judgment pipelines on TypeSafe's Jev: support-ticket triage, observability alert triage, and a deploy-risk gate.
构建在 TypeSafe Jev 上的三个可组合判断管道：客服工单分诊、可观测性告警分诊和部署风险门控。
A TypeSafe/Jev agent that plays Super Mario Bros. from structured emulator state.
从结构化模拟器状态出发、使用 TypeSafe Jev 游玩《超级马力欧兄弟》的 Agent。
Camera-only autonomous drone in MuJoCo with a small judgment model (TypeSafe Jev) in the loop at 2.5Hz.
MuJoCo 中仅依赖摄像头的自主无人机，控制循环以 2.5Hz 调用 TypeSafe Jev 小型判断模型。
Snake auto-played by TypeSafe's Jev model: one System One choice per tick, legal moves and facts generated in code.
由 TypeSafe Jev 自动游玩的贪吃蛇：每个 Tick 执行一次 System One 选择，合法动作和事实由代码生成。
TypeSafe Jev controls original StarCraft shareware through keyboard and mouse with recorded action probabilities.
TypeSafe Jev 通过键盘和鼠标控制原版《星际争霸》共享版，并记录动作概率。
Python prototype that plays NES Super Mario Bros. from structured RAM observations, with Jev answering focused movement and jump questions.
Python 原型：从结构化 RAM 观测游玩 NES《超级马力欧兄弟》，由 Jev 回答聚焦的移动和跳跃问题。
Interactive experiments from support routing to a 3D driving simulation: structured sensor state in, typed steer, brake, and overtake decisions out.
从客服路由到三维驾驶模拟的交互实验：输入结构化传感器状态，输出类型化转向、制动和超车决策。
1v1 Jev quickscope arena — Three.js + TypeSafe System One.
一对一 Jev 快速狙击竞技场，使用 Three.js 与 TypeSafe System One。
Observable browser stealth game: Jev makes typed guard judgments while deterministic code owns the world.
可观测的浏览器潜行游戏：Jev 作出类型化守卫判断，确定性代码负责世界规则。
Zero-shot English goals on a sim Franka. Jev chains hardcoded primitives.
让模拟 Franka 机械臂执行零样本英文目标，由 Jev 串联硬编码动作原语。
A browser-native Doom agent experiment with structured spatial state, composable AI controls, live decision telemetry, and a Chocolate Doom WebAssembly runtime.
浏览器原生 Doom Agent 实验，包含结构化空间状态、可组合 AI 控制、实时决策遥测和 Chocolate Doom WebAssembly 运行时。
弈瞬：双 Jev 五子棋九宫格输入实验台，逐手查看模型决策，支持真实对局回放与实时对战。
弈瞬：双 Jev 五子棋九宫格输入实验台，可逐手查看模型决策，支持真实对局回放与实时对战。
Challenge the Jev's intelligence in Rubik Cube puzzles.
用魔方谜题挑战 Jev 的推理能力。
TypeSafe Jev plays original Civilization II in a browser, with live action probabilities. Experimental full-game harness.
TypeSafe Jev 在浏览器中游玩原版《文明 II》，实时展示动作概率；实验性完整游戏框架。
Using the new model JEV to play the game tetris.
使用新模型 Jev 游玩俄罗斯方块。
Pong benchmark where every ball step is a model decision, comparing Jev with LLMs through Vercel AI Gateway.
每次球移动都由模型决策的 Pong 基准，通过 Vercel AI Gateway 比较 Jev 与 LLM。
Pokémon Red agent that turns structured game state into Jev decisions and validated moves.
《宝可梦 红》Agent，将结构化游戏状态转为 Jev 决策和经过验证的动作。
SO-101 robot-arm workbench where Jev selects bounded joint actions under a spend budget.
SO-101 机械臂工作台，由 Jev 在预算约束下选择有限关节动作。
Three.js drone simulator with a Python backend and live Jev navigation decisions.
Three.js 无人机模拟器，配有 Python 后端和实时 Jev 导航决策。
Browser game: try to beat Jev at spotting a spam message.
浏览器游戏：尝试在识别垃圾信息方面击败 Jev。
A Next.js brick breaker whose paddle is controlled in real time by TypeSafe AI's Jev model. Built with Claude Code.
Next.js 打砖块游戏，挡板由 TypeSafe Jev 实时控制；使用 Claude Code 构建。
Cyber-Breach: The Jev Protocol - A tactical cyberpunk arena combat game powered by TypeSafe AI Jev System One decision model.
《Cyber-Breach: The Jev Protocol》：由 TypeSafe Jev System One 决策模型驱动的赛博朋克战术竞技游戏。
100 AI NPCs live in a tiny town. Jev chooses the next action; the world writes the story.
100 个 AI NPC 生活在小镇中，Jev 选择下一动作，世界系统书写故事。
Wikipedia link races with direct Jev ranking and a live terminal display.
维基百科链接竞速，使用 Jev 直接排序并在终端实时展示。
Visual Jev lab for multiple games and emulator platforms.
面向多种游戏和模拟器平台的可视化 Jev 实验室。
MoonBit client for Jev plus a Jev-vs-Jev gomoku match, with timing logs.
Jev 的 MoonBit 客户端，以及带计时日志的 Jev 对 Jev 五子棋对局。
Jev plays browser table tennis in real time: structured telemetry, typed decisions, ordinary Chrome inputs, and auditable evidence.
Jev 实时游玩浏览器乒乓球，使用结构化遥测、类型化决策、普通 Chrome 输入和可审计证据。
An experimental Snake environment where the game engine owns deterministic rules and TypeSafe AI's Jev makes the movement decision from structured state on every tick.
实验性贪吃蛇环境：游戏引擎负责确定性规则，Jev 每个 Tick 根据结构化状态决定移动方向。
Chrome dino game played by Typesafe AI Jev model.
由 TypeSafe Jev 模型游玩的 Chrome 恐龙游戏。
A visual TypeSafe demo where Jev chooses verified Tetris placements.
可视化 TypeSafe 演示，由 Jev 选择经过验证的俄罗斯方块落点。
Minecraft mod where Jev tries to finish the game from scratch without a scripted route.
Minecraft 模组，让 Jev 在没有预设路线的情况下从零尝试通关。
Let Jev (TypeSafeAI) solve 2048.
让 Jev（TypeSafe AI）解决 2048。
A cyberpunk border encounter powered by TypeSafe Jev. Bluff the guard. Inspect the receipts.
由 TypeSafe Jev 驱动的赛博朋克边境遭遇游戏：欺骗守卫并检查决策记录。
USER + JEV: P(DOOM) PROTOCOL — co-op platform shooter where TypeSafe Jev plays alongside you.
USER + JEV: P(DOOM) PROTOCOL：TypeSafe Jev 与玩家并肩作战的合作平台射击游戏。
TypeSafe's Jev plays Atari Pong. One typed Choice question per frame, no coordinates sent to the model.
TypeSafe Jev 游玩 Atari Pong，每帧一个类型化 Choice 问题，不向模型发送坐标。
Autonomous PlayStation 2 AI Agent with real-time visual telemetry HUD powered by TypeSafe Jev System One.
由 TypeSafe Jev System One 驱动、带实时视觉遥测 HUD 的 PlayStation 2 自主 AI Agent。
NPCs of River Oaks Houston, Texas using Jev to power NPCs.
以休斯敦 River Oaks 为背景、使用 Jev 驱动 NPC 的模拟项目。
River shooter game in Python, inspired by Atari's River Raid, played by a TypeSafe AI pilot.
受 Atari《River Raid》启发的 Python 河流射击游戏，由 TypeSafe AI 飞行员游玩。
A 3D planetary rover sandbox for experimenting with autonomous decisions using TypeSafe AI.
三维行星车沙盒，用于试验 TypeSafe AI 的自主决策。
Shady Town: social-deduction party game for the living room TV, moderated by TypeSafe Jev.
适合客厅电视的社交推理派对游戏，由 TypeSafe Jev 主持。
SIEGE: 200 people vs one agent. A typed action gate (TypeSafe System One) that learns from every breach, evaluated by W&B Weave, hardened by a defender loop. Built at CoreWeave Hacks: Agent Loops 2026.
SIEGE：200 人对抗一个 Agent；TypeSafe System One 类型化动作门控会从每次突破中学习，并通过 W&B Weave 评测及防守循环强化。
Snake controlled by parallel Jev assessments, with one API call per game tick.
由并行 Jev 评估控制的贪吃蛇，每个游戏 Tick 仅调用一次 API。
A sandbox where a TypeSafe System One model presses the controls of a small creature. Code runs the world.
让 TypeSafe System One 模型操控小型生物的沙盒，世界逻辑由代码运行。
3D chess powered by TypeSafe AI (Jev). AI vs AI by default, or play either side. Multiple difficulty levels.
由 TypeSafe Jev 驱动的三维国际象棋，默认 AI 对 AI，也可选择任一方游玩，并提供多种难度。
Chess where both players are TypeSafe's Jev model: every move is a typed Choice decision.
双方均由 TypeSafe Jev 控制的国际象棋，每一步都是类型化 Choice 决策。
A Minecraft Java player controlled by TypeSafe AI, with live decisions, Canadian flag building, and a side-by-side dashboard.
由 TypeSafe AI 控制的 Minecraft Java 玩家，实时展示决策、搭建加拿大国旗并提供并排仪表板。
Describe your startup idea. Jev decides: kill it, fix it or ship it.
描述你的创业想法，由 Jev 决定：放弃、修改或发布。
A grill-me style interrogation of your idea, with Jev doing the grilling.
以连续追问方式审视创意，由 Jev 负责“拷问”。
Community TypeSafe AI playground: 110 use cases, games, dilemmas and model challenges, with editable prompts, A/B comparisons and a mobile-friendly UI.
社区 TypeSafe AI 游乐场，包含 110 个用例、游戏、两难问题和模型挑战，支持编辑提示词、A/B 对比和移动端界面。
This demo uses Jev from TypeSafe AI to autonomously fly a drone in a random city from point A to point B, avoiding obstacles along the way. A trip costs $0.01.
使用 TypeSafe Jev 让无人机在随机城市中从 A 点自主飞到 B 点并避障，单次行程成本约 0.01 美元。
A polished OpenAI + TypeSafe Jev terminal interface for answers with transparent decision reports.
精致的 OpenAI + TypeSafe Jev 终端界面，提供答案和透明决策报告。
Live verdict page: feeds Jev the day’s Florida Man, odd-news, politics and world headlines and asks all three primitives whether AI should kill us all, refreshed every ten minutes.
实时判决页面，每十分钟向 Jev 输入当日奇闻、政治和国际新闻，并用三种原语判断 AI 是否应该消灭人类。
Community TypeSafe AI playground: 110 use cases, games, dilemmas and model challenges, with editable prompts, A/B comparisons and a mobile-friendly UI.
社区 TypeSafe AI 游乐场，包含 110 个用例、游戏、两难问题和模型挑战，支持编辑提示词、A/B 对比和移动端界面。
Multi-axis writing quality checker powered by TypeSafe AI's Jev model. Separate named checks, each with its own verdict and confidence.
由 Jev 驱动的多维写作质量检查器，每个命名检查项独立返回结论与置信度。
Jev (TypeSafe AI) PoC through Game of Thrones.
通过《权力的游戏》场景演示 TypeSafe Jev 的概念验证。
Compare GPT generated language with JEV structured Noul decisions on the same input.
在相同输入上比较 GPT 生成语言与 Jev 的结构化 Noul 决策。
Comment-moderation playground: paste a comment, Jev decides what to do with it.
评论审核游乐场：粘贴评论，由 Jev 决定处理方式。
JEV picks which stream idea becomes the live MVP. TypeSafe System One decision board.
由 Jev 选择哪个直播创意成为现场 MVP 的 TypeSafe System One 决策面板。
Instinct: describe a case in free text and Jev picks the UI from a fixed catalog without generating a line of code or copy.
输入自然语言案例，由 Jev 从固定目录中选择 UI，不生成任何代码或文案。
A small Next.js app for experimenting with TypeSafe AI's Jev model (System One).
用于试验 TypeSafe Jev System One 模型的小型 Next.js 应用。
Agent tool/MCP call gate — allow / ask_human / deny via TypeSafe Jev.
Agent 工具/MCP 调用门控，通过 TypeSafe Jev 决定允许、询问人工或拒绝。
A playground for experiments around Jev, TypeSafe's System One model.
用于试验 TypeSafe System One 模型 Jev 的游乐场。
A playground for TypeSafeAI's Jev Model.
TypeSafe AI Jev 模型的实验游乐场。
CartShield — SMB checkout fraud disposition via TypeSafe Jev.
CartShield：使用 TypeSafe Jev 为中小企业结账交易判定欺诈处置方式。
Twelve members. Zero qualifications. A playful TypeSafe AI council with animated votes, inspectable decisions, and shareable verdicts.
由 12 位毫无资质的成员组成的趣味 TypeSafe AI 委员会，提供动画投票、可检查决策和可分享结论。
Judge agent steps — ok / retry / escalate / stop via TypeSafe Jev.
使用 TypeSafe Jev 判断 Agent 步骤应继续、重试、升级还是停止。
HireSignal — resume first-pass fit+interview via TypeSafe Jev.
HireSignal：使用 TypeSafe Jev 进行简历首轮匹配和面试筛选。
A compiler for human language. Paste text, get diagnostics. Measured by TypeSafe Jev.
人类语言编译器：粘贴文本即可获得由 TypeSafe Jev 衡量的诊断结果。
Jev Gamecast: replay-first React app that asks Jev typed questions about live sports data.
以回放为先的 React 应用，针对实时体育数据向 Jev 提出类型化问题。
Typesafe/Jev public X demo.
TypeSafe/Jev 的公开 X 广告预检演示。
Interactive explorer and Jev question workspace for Jev Board datasets.
用于 Jev Board 数据集的交互探索器和 Jev 问题工作区。
TypeSafe の Jev を TypeScript SDK で使ってみる最初の 1 歩.
使用 TypeScript SDK 上手 TypeSafe Jev 的第一步示例。
Demos to test the effectiveness of TypeSafe's "Jev" System One Model.
用于测试 TypeSafe Jev System One 模型效果的演示集合。
同じ発言を jev と LLM の両方に判定させ、感情の変動値のズレと応答速度を1画面で見比べるデモ（affectus + Vercel AI Gateway）.
让 Jev 与 LLM 判断同一发言，在单一界面比较情感变化偏差和响应速度。
Feedback on your paper in seconds.
在数秒内获得对论文的反馈。
Can a System One model steer music? Jev picks the plan (enums only); code renders sheet, audio and MIDI.
探索 System One 模型能否指挥音乐：Jev 只用枚举选择方案，代码生成乐谱、音频和 MIDI。
Typesafe/Jev public X demo.
TypeSafe/Jev 的公开 X 求职申请判断演示。
Typesafe/Jev public X demo.
TypeSafe/Jev 的公开 X 用户陪审团演示。
TypeSafe Jev playground — custom Choice/Score/Noul builder with live distributions.
TypeSafe Jev 游乐场，可自定义构建 Choice、Score 和 Noul 并实时查看概率分布。
Job Risk Analyzer: CLI and REST API that uses Jev to score an occupation's exposure to AI-driven layoffs and its resilience.
职业风险分析器：通过 CLI 和 REST API 使用 Jev 评估职业受 AI 裁员影响的程度及韧性。
LaneBreak — support ticket priority+routing via TypeSafe Jev.
LaneBreak：使用 TypeSafe Jev 判断客服工单优先级并路由。
Match user goals to MCP catalog (two-stage) via TypeSafe Jev.
使用 TypeSafe Jev 分两阶段将用户目标匹配到 MCP 目录。
Probably: live BTC, ETH, and XRP prices with a shared TypeSafe buy-or-wait demonstration. No trades placed.
展示 BTC、ETH 和 XRP 实时价格及 TypeSafe“买入或等待”共享决策的演示，不执行真实交易。
PulseLane — clinic triage decisions via TypeSafe Jev.
PulseLane：使用 TypeSafe Jev 进行诊所分诊决策。
A test project based on Jev AI, the goal is to build a search function for a blog/article website that has 100s of articles to search from, So the user can actually use the search as chat to question anything and find related answers/articles.
基于 Jev AI 的搜索功能测试项目，让用户以聊天方式检索拥有数百篇文章的网站并找到相关答案。
Agent budget brake — continue / downgrade_model / stop via TypeSafe Jev.
Agent 预算刹车，通过 TypeSafe Jev 决定继续、降级模型或停止。
Route tasks to research/code/browser/support/writer agents via TypeSafe Jev.
使用 TypeSafe Jev 将任务路由到研究、编程、浏览器、客服或写作 Agent。
TrustGate — indie media T&S gate via TypeSafe Jev.
TrustGate：使用 TypeSafe Jev 为独立媒体提供信任与安全门控。
Diffusion-style pixel art out of a classifier: 256 parallel per-pixel Jev questions plus refinement passes.
从分类器生成扩散风格像素画：并行提出 256 个逐像素 Jev 问题并执行多轮细化。
Worked ticket-triage and reranking examples for Jev, runnable through OpenRouter with sample data and a Makefile.
Jev 的工单分诊与重排序完整示例，可通过 OpenRouter、样例数据和 Makefile 直接运行。
Semantic ifs from open models, on a 3090 at home. Independent; not affiliated with Jev or TypeSafe.
在家用 RTX 3090 上通过开源模型实现语义 If；独立项目，与 Jev 或 TypeSafe 无隶属关系。
Train a small model that chooses among a changing list of text options, one probability per option in a single pass. Includes Doom, chess, and Wikispeedia demos.
训练小模型，从动态文本选项列表中单次前向输出每个选项的概率；包含 Doom、国际象棋和 Wikispeedia 演示。
Open 0.6B Jev replica with parallel decisions, complete probability distributions, training pipeline, weights, dataset, and live demos.
开放的 0.6B 参数 Jev 复刻，支持并行决策、完整概率分布，并提供训练管道、权重、数据集和在线演示。
An educational Jev-like visual inference experiment on Apple Silicon: shared context, direct candidate scoring, and local visual demos.
Apple Silicon 上用于教学的 Jev 风格视觉推理实验，采用共享上下文、直接候选评分和本地视觉演示。
One-pass typed decisions with calibrated probabilities (System One style model), fine-tuned from Qwen3.5-2B.
从 Qwen3.5-2B 微调的 System One 风格模型，单次完成类型化决策并返回校准概率。
Personal-assistant agent built on Vercel's eve with 100 mocked tools, measuring how many steps it takes when Jev picks the tool versus the LLM.
基于 Vercel eve 的个人助理 Agent，配置 100 个模拟工具，对比由 Jev 和 LLM 选工具时所需步骤。
A small open decision model: state + typed questions -> calibrated probabilities. A Jev / System One re-creation on Qwen3.5.
小型开源决策模型：输入状态与类型化问题，输出校准概率；基于 Qwen3.5 复刻 Jev/System One。
WebMCP benchmark comparing browser-agent interfaces, with Jev included as one of the evaluated configurations.
用于比较浏览器 Agent 接口的 WebMCP 基准，将 Jev 作为被评测配置之一。
This is a LLM Gateway that mimics typesafe ai structured output. Like an imposter Jev.
模拟 TypeSafe AI 结构化输出的 LLM Gateway，相当于一个“仿冒 Jev”。
One-pass option scoring with a local Gemma 3 4B on Apple silicon via MLX, inspired by jevlike, with a Doom demo.
受 jevlike 启发，在 Apple Silicon 上通过 MLX 使用本地 Gemma 3 4B 单次评分选项，并附 Doom 演示。
Jev-style parallel constrained decisions for any MLX model on Apple Silicon. Typed, schema-valid JSON in one forward pass.
让任意 MLX 模型在 Apple Silicon 上执行 Jev 风格并行约束决策，单次前向输出类型化、符合 Schema 的 JSON。
Mini-Jev: what a Jev-style typed-decision interface looks like on a frozen Qwen3-4B — read the option letter's logits instead of generating JSON. Preregistered experiment, results, teaching bench.
展示冻结 Qwen3-4B 上 Jev 风格类型化决策接口的 Mini-Jev：读取选项字母 Logits 而非生成 JSON，提供预注册实验、结果和教学基准。
Open, Jev-compatible System One decision server on DiffusionGemma.
基于 DiffusionGemma 的开放 Jev 兼容 System One 决策服务器。
Unofficial study: Jev-style parallel typed decisions on stock 1.5B-8B models on an Apple Silicon laptop. Benchmarks, research notes, and a Hugging Face Space demo.
非官方研究：在 Apple Silicon 笔记本的原生 1.5B–8B 模型上执行 Jev 风格并行类型化决策，附基准、研究笔记和 Hugging Face Space 演示。
Typed JSON inference with DiffusionGemma, with Every and Jev benchmark results.
使用 DiffusionGemma 进行类型化 JSON 推理，并提供 Every 与 Jev 基准结果。
A stronger one-pass scorer over a variable list of text options: hashed n-gram encoder, rival-aware attention, gated head, temperature scaling, benchmarked against jevlike.
面向动态文本选项列表的更强单次评分器，采用哈希 n-gram 编码器、对手感知注意力、门控头和温度缩放，并与 jevlike 比较。
Type-safe one-decision-per-token decoding engine for autoregressive LLMs, inspired by Jev.
受 Jev 启发、面向自回归 LLM 的类型安全单 Token 单决策解码引擎。
Probability-aware evaluation for typed decision models: calibration, selective risk, latency, and reproducible benchmarks.
面向类型化决策模型的概率感知评测，覆盖校准、选择性风险、延迟和可复现基准。
Jev vs Gemini 3.8 Flash: labelling 1,000 app reviews, 4.1× faster and 7× cheaper.
Jev 与 Gemini 3.8 Flash 对比：标注 1,000 条应用评论，报告速度快 4.1 倍、成本低 7 倍。
Openvons (open-Jev): 有限選択肢に確率で答える判断層 — テキスト / 画像 / 日本語音声コマンド.
开放 Jev 风格判断层，以概率回答有限选项，支持文本、图像和日语语音命令。
看看 Jev 能做什么：用中英文讲清热门应用、工作原理和各自优缺点。Explore Jev apps with plain-language examples, explanations, and comparisons.
用中英文介绍 Jev 热门应用、工作原理及各自优缺点。
JEV-inspired parallel decisions for CUDA LLMs. One context, many decisions. vLLM API, game-agent examples, and reproducible benchmarks.
面向 CUDA LLM、受 Jev 启发的并行决策：一次上下文完成多个判断，提供 vLLM API、游戏 Agent 示例和可复现基准。
LegalForecast-MTD benchmark alpha and official evaluation workflows.
LegalForecast-MTD 法律预测基准 Alpha 版及官方评测工作流。
Local bilingual probability decisions from context, questions, and candidate answers. Independent research preview inspired by TypeSafe Jev.
根据上下文、问题和候选答案执行本地双语概率决策；受 Jev 启发的独立研究预览。
Reproducible early-access evaluation of Jev on Korean understanding and medical text, with runtime and cost evidence.
针对 Jev 韩语理解和医疗文本能力的可复现早期访问评测，包含运行时间与成本证据。
A word-level language model whose output layer is Jev: n-gram drafter, Noul chunk verification, bits-per-token eval.
输出层采用 Jev 的词级语言模型，包含 n-gram 草稿器、Noul 分块验证和每 Token 比特评测。
Open replica of TypeSafe's Jev: typed calibrated decisions in one forward pass, on Gemma 4 E2B / Gemma 3 270M (Modal).
TypeSafe Jev 的开放复刻，在 Gemma 4 E2B / Gemma 3 270M 上单次前向完成类型化校准决策。
Inspired by TypeSafe Ai, Ask a local LLM typed questions, get calibrated probabilities instead of text. Structured output without generation or parsing. MLX / Apple Silicon.
受 TypeSafe AI 启发，让本地 LLM 回答类型化问题并返回校准概率，而非生成文本；无需生成或解析结构化输出，适用于 MLX/Apple Silicon。
Daf-jev: composable Python toolkit for TypeSafe's Jev (System One) decision API — question builders, confidence gates, evaluator, calibration, CLI, MCP server, agent skill.
面向 TypeSafe Jev API 的可组合 Python 工具包，包含问题构建器、置信度门控、评测器、校准、CLI、MCP 服务器和 Agent 技能。
Independent Jev 1.13.0 behavior study: report, controlled prompt experiments, raw results, and offline verification.
独立 Jev 1.13.0 行为研究，提供报告、受控提示实验、原始结果和离线验证。
A chatbot from typed Jev decisions: hierarchical speculative decoding over System One probabilities.
从类型化 Jev 决策构建聊天机器人，通过 System One 概率进行分层推测解码。
Does a TypeSafe Jev rerank beat embedding search? Graded relevance eval (9,831 pairs, 164 zh/en queries) over the Agent Skills Hub catalog, with the judge-circularity bias measured.
评测 Jev 重排序是否优于向量搜索：在 Agent Skills Hub 目录上使用 9,831 个相关性对和 164 个中英文查询，并测量裁判循环偏差。
A reproduction of Jev that turns any Qwen model into a fast decision model, serving the same /v1/systemone schema (Choice, Score, Noul) with no training and no generated answer text.
Jev 复现方案，无需训练即可将任意 Qwen 模型变为快速决策模型，提供相同 /v1/systemone Schema 且不生成答案文本。
Batched single-token choice inference for open language models, compatible with TypeSafe.
面向开放语言模型、兼容 TypeSafe 的批量单 Token Choice 推理。
Non-autoregressive decision engine on ModernBERT (151M) with calibrated uncertainty (RLCD), TypeSafe AI Jev benchmark audit, and in-browser WebGPU playground.
基于 ModernBERT（151M）的非自回归决策引擎，提供校准不确定性、TypeSafe Jev 基准审计和浏览器 WebGPU 游乐场。
Calibration and confidence-based routing measured on Banking77: 80.2% accuracy at $0.103 per 500 decisions.
在 Banking77 上测量校准与基于置信度的路由：每 500 次决策成本 0.103 美元，准确率 80.2%。
Benchmarks and a playground for TypeSafe's Jev (System One) model: chess, and who-is-the-player-talking-to for speech-to-text game NPCs.
TypeSafe Jev System One 模型的基准与游乐场，包含国际象棋和语音转文字游戏 NPC 对话对象识别。
Eight minimal working examples of TypeSafe's Jev (a System One model) applied to mechanical and electrical engineering: CAD/CAE/CAM routing, FEM result triage, DFM screening, BOM alignment, hallucination-proof extraction. Zero dependencies.
八个 TypeSafe Jev 在机械与电气工程中的最小可运行示例，覆盖 CAD/CAE/CAM 路由、FEM 结果分诊、DFM 筛查、BOM 对齐和防幻觉提取，零依赖。
Not every coding task needs your best model. Experimental Jev-powered model routing for Claude Code — V3 prototype runs today, V4 routes at the task boundary.
并非每项编程任务都需要最强模型：面向 Claude Code 的实验性 Jev 模型路由，V3 可运行，V4 在任务边界路由。
Decision harness for TypeSafe Jev — confidence gates, shadow mode, recipes, and evals. Claude CLI 48.9s → Jev 1.3s on the same row-filter job.
TypeSafe Jev 决策框架，提供置信度门控、影子模式、配方和评测；同一行过滤任务从 Claude CLI 48.9 秒降至 Jev 1.3 秒。
An experimental JEV-powered framework for forecasting short-term stock price direction from structured market data.
实验性 Jev 股票框架，根据结构化市场数据预测短期股价方向。
Stop guessing confidence thresholds: calibrate, threshold, and drift-check typed decision models (TypeSafe Jev) against an LLM teacher.
用于校准、设定阈值并检测类型化决策模型漂移的工具，以 LLM 教师评估 TypeSafe Jev，避免凭感觉设置置信度阈值。
A chatbot built on a model that cannot generate text (TypeSafe AI's Jev, driven autoregressively).
构建在无法生成文本的 TypeSafe Jev 上、以自回归方式驱动的聊天机器人。
Open alternative to Jev: typed, calibrated decisions from any open-weights LLM in one forward pass (HF + vLLM), with benchmarks.
Jev 的开放替代方案，让任意开放权重 LLM 单次前向完成类型化校准决策，支持 Hugging Face、vLLM 和基准测试。
Backtest Jev (TypeSafe) as a BUY/SELL/HOLD trader on NQ L10 order-book data.
在 NQ L10 订单簿数据上回测 Jev 作为买入、卖出或持有交易器的表现。
A Jev-inspired decision interface for existing LLMs. Explicit choices, scores, calibration, and review thresholds.
受 Jev 启发、面向现有 LLM 的决策接口，提供显式选择、评分、校准和审查阈值。
Benchmarking Jev (Typesafe.ai) against a strong LLM on the Who&When Pro agent-failure-attribution benchmark (text subset).
在 Who&When Pro Agent 失败归因基准的文本子集上，对比 Jev 与强力 LLM。
High-throughput synthetic & pretraining dataset sifter powered by TypeSafe AI Jev (api.typesafe.ai). Stream, filter, and score Parquet & JSONL datasets at 1,500+ rows/sec using System One typed decisions (Choice, Score, Noul).
由 TypeSafe Jev 驱动的高吞吐合成与预训练数据集筛选器，以每秒 1,500 行以上的速度流式过滤和评分 Parquet、JSONL 数据。
Jev (TypeSafe) exploratory thread: claim audit, live demos, and runnable code.
Jev 探索记录，包含声明审计、在线演示和可运行代码。
An observable raw-character chat experiment powered entirely by TypeSafe Jev Choice.
完全由 TypeSafe Jev Choice 驱动、可观测的原始字符聊天实验。
A show-and-tell capability study for Jev, TypeSafe's System One decision model.
TypeSafe System One 决策模型 Jev 的展示型能力研究。
Can a decision model beat dedicated rerankers? TypeSafe Jev vs Cohere Rerank 4 vs ZeroEntropy zerank-2 vs a chat-model baseline: 14 datasets, every raw API response, bootstrap ranges on every gap.
评测决策模型能否击败专用重排序器：比较 Jev、Cohere Rerank 4、ZeroEntropy zerank-2 和聊天模型基线，覆盖 14 个数据集及全部原始响应。
Reproducible Jev Ultrafast research-browser eval harness + field note (QC’d cases, suite runner, report generator). Not investment advice.
可复现的 Jev Ultrafast 研究浏览器评测框架与实地笔记，包含质检案例、套件运行器和报告生成器；不构成投资建议。
Blind security benchmarks for Jev, TypeSafe's System One model: prompt injection and vulnerable code detection, built on jev-go.
Jev 的盲测安全基准，覆盖提示词注入和脆弱代码检测，基于 jev-go。
Jev (TypeSafe System One) × ASReview SYNERGY abstract screening demo — Choice/Noul vs gold labels.
Jev 与 ASReview SYNERGY 摘要筛选演示，将 Choice/Noul 结果与金标准标签对比。
An agent skill to discover TypeSafe Jev opportunities, design typed questions, and learn from recent community experiments.
Agent 技能，用于发现 TypeSafe Jev 应用机会、设计类型化问题并借鉴近期社区实验。
AI benchmark on Japan's 2026 Common Test: Jev vs luna-none vs luna-low (static dashboard).
日本 2026 年大学入学共通测试 AI 基准，对比 Jev、Luna-none 和 Luna-low，并提供静态仪表板。
Discriminative Monte Carlo Tree Search using TypeSafe Jev System One Primitives and Gemini.
使用 TypeSafe Jev System One 原语和 Gemini 的判别式蒙特卡洛树搜索。
Typed-decision benchmark from PadFlow (land development SaaS): schemas, anonymized labeled rows, and a runner for confidence-calibrated models like TypeSafe Jev.
来自 PadFlow 土地开发 SaaS 的类型化决策基准，提供 Schema、匿名标注样本和置信度校准模型运行器。
Jev-style calibrated decision model (Choice/Score/Noul) on Qwen3.5-0.8B.
基于 Qwen3.5-0.8B 的 Jev 风格校准决策模型，支持 Choice、Score 和 Noul。
I tortured Jev into being a RISC-V CPU.
将 Jev 强行构造成一颗 RISC-V CPU 的实验。
Using Jev to test how well it predicts financial markets(just like most llms as of september 2026, it doesnt do that good).
测试 Jev 预测金融市场的能力；结果与多数 2026 年 9 月的 LLM 类似，表现并不理想。
Does Jev predict stock returns from news? It reads the news well; there is no tradeable alpha. Three arms separate reading from recall.
测试 Jev 能否从新闻预测股票收益：阅读新闻表现良好，但没有可交易 Alpha；通过三组实验区分阅读与记忆。
Jev (TypeSafe) vs. Gemini 3.8 Flash vs. GPT-5.6 Luna na anotação estruturada de sentenças do TJSP: qualidade, tempo e custo.
在巴西圣保罗州法院判决结构化标注任务上，对比 Jev、Gemini 3.8 Flash 和 GPT-5.6 Luna 的质量、时间与成本。
Using Jev as an evaluator.
将 Jev 用作评测器的实验。
Position paper: the Hidden-Markov and fuzzy primitives missing from TypeSafe AI's Jev and System-One decision models. Two lemmas, one principle (Deferred Crispification), one architecture (BSF-S1).
立场论文：讨论 TypeSafe Jev 与 System One 决策模型缺失的隐马尔可夫及模糊原语，提出两个引理、延迟清晰化原则和 BSF-S1 架构。
Typesafe.ai model jev finance benchmark.
typesafe.ai Jev 的金融基准。
Can Jev pick the winner of a real headline A/B test? 64.5% across 10,984 Upworthy randomized experiments, 74.7% when the difference was decisive.
测试 Jev 能否选出真实标题 A/B 测试赢家：在 10,984 项 Upworthy 随机实验中准确率 64.5%，差异明显时为 74.7%。
Jev (TypeSafe) 性能評価プロジェクト — 日本郵便 KEN_ALL をマスタに、AI SDK 経由の Jev が住所のあいまい一致にどこまで使えるかを検証.
以日本邮政 KEN_ALL 为主数据，评估通过 AI SDK 调用 Jev 进行日文地址模糊匹配的性能。
TypeScript experiments, evaluations, and latency benchmarks for TypeSafe's Jev model.
TypeSafe Jev 的 TypeScript 实验、评测和延迟基准。
Jev (TypeSafe) vs Claude Haiku 4.5 on 2 000 phishing emails: accuracy, calibration, latency, cost. Reproducible benchmark.
在 2,000 封钓鱼邮件上比较 Jev 与 Claude Haiku 4.5 的准确率、校准度、延迟和成本，结果可复现。
A small reproducible MuJoCo pilot comparing Jev, Claude Haiku, and reactive rules for pick-and-place.
可复现的小型 MuJoCo 先导研究，对比 Jev、Claude Haiku 和反应式规则在抓取放置任务中的表现。
Benchmarks Jev against other evaluation models in games with explicit states, legal actions, and measurable outcomes.
在具有显式状态、合法动作和可量化结果的游戏中，将 Jev 与其他评测模型进行基准比较。
发明 RLHF 的人，这次做了个不会说话的模型：Jev 独立研究报告。52 页 PDF + 50 条中文实测复现包 + 143 条可回溯数据表.
关于 Jev 的独立中文研究报告，包含 52 页 PDF、50 条中文实测复现包和 143 条可追溯数据表。
Benchmarking TypeSafe's Jev decision model as a cost-efficient LLM router on RouterArena.
在 RouterArena 上评测 TypeSafe Jev 作为低成本 LLM 路由器的表现。
Measures how well TypeSafe's RLCD-Jev model spots real secret credentials in file snippets.
测量 TypeSafe RLCD-Jev 模型识别文件片段中真实密钥凭据的能力。
A small second eval for shadcn-ui/lint that uses TypeSafe's Jev to judge the linter's own output.
针对 shadcn-ui/lint 的小型二次评测，使用 Jev 判断 Linter 自身输出。
Zero-shot spam filtering with TypeSafe Jev Noul questions, compared with TF-IDF baselines.
使用 TypeSafe Jev Noul 问题进行零样本垃圾信息过滤，并与 TF-IDF 基线比较。
Application of TypeSafe Jev (noul judgment primitive) on the collusion.wiki corpus: agent vs human page authorship, head-to-head vs local Qwen3.8-Flash-Next.
在 collusion.wiki 语料上应用 Jev Noul 判断，识别页面由 Agent 还是人类创作，并与本地 Qwen3.8-Flash-Next 正面对比。
Ongoing Japanese research deck on Jev and System One models, maintained as Markdown slides.
持续维护的日文 Jev 与 System One 模型研究幻灯片，使用 Markdown 编写。
Experiments with openjev, an open Jev-style option-logit runner, on local models.
在本地模型上试验 openjev 开放 Jev 风格选项 Logit 运行器。
On-device iPhone visual decision tool using MLX and Qwen3-VL direct option logits.
设备端 iPhone 视觉决策工具，使用 MLX 和 Qwen3-VL 直接选项 Logit。
Evaluating TypeSafe's Jev as a fast monitor and action gate for agent sabotage in SHADE-Arena, compared with Gemini 2.5 Flash/Pro.
在 SHADE-Arena 中评估 Jev 作为 Agent 破坏行为快速监控器和动作门控的效果，并与 Gemini 2.5 Flash/Pro 比较。
Rust port of TypeSafe system-one-adapter (LLM-backed system_one evaluations).
TypeSafe system-one-adapter 的 Rust 移植版，用 LLM 支持 System One 评测。
Open-source Jev-style System One decision model. Gemma 3 270M with a scoring head — fast, calibrated decisions in a single forward pass. No text generation. Inspired by TypeSafe.ai's Jev.
开源 Jev 风格 System One 决策模型，基于 Gemma 3 270M 与评分头，单次前向完成快速校准决策，不生成文本。
Charts: TypeSafe Jev evaluated on Thai standardized exams vs 110 other models.
图表：在泰国标准化考试上评估 TypeSafe Jev，并与其他 110 个模型比较。
Evaluating TypeSafe's System One primitives (Choice/Score/Noul) — where a typed oracle beats an LLM call.
评估 TypeSafe System One 原语 Choice、Score 和 Noul，研究类型化 Oracle 优于 LLM 调用的场景。
An evaluation of typesafe AI chess. As it turns out, the AI isn't doing really well even though chess is not a particularly open-ended game. Still, it's only a prototype and this probably wasn't optimzied for games.
TypeSafe AI 国际象棋能力评测；结果显示原型表现较弱，且可能未针对游戏进行优化。
Curated list of official resources and community projects for TypeSafe, System One models, and Jev, with a GitHub Pages site.
TypeSafe、System One 模型和 Jev 的官方资源与社区项目精选列表，并提供 GitHub Pages 站点。
A curated list of public projects, integrations, and discussions built on Jev — TypeSafe AI's System One model for typed decisions.
围绕 TypeSafe AI 类型化决策 System One 模型 Jev 构建的公开项目、集成和讨论精选列表。
A curated, source-backed list of projects built with Jev, TypeSafe AI's System One model for typed decisions.
有来源依据的 Jev 项目精选列表；Jev 是 TypeSafe AI 面向类型化决策的 System One 模型。
A curated list of awesome Jev / TypeSafe System One applications, libraries, and resources.
Jev / TypeSafe System One 应用、库和资源精选列表。
A source-backed Jev project directory with a reusable Jev-only GitHub review workflow.
有来源依据的 Jev 项目目录，提供可复用的 Jev 专用 GitHub 审查工作流。
Awesome Jev: source-backed open-source ecosystem radar, plain-language project discovery, and automatic GitHub sync.
Awesome Jev：有来源依据的开源生态雷达、通俗项目发现和 GitHub 自动同步。
Papers, open reproductions and independent evaluations behind System One models and Jev.
汇集 System One 模型与 Jev 背后的论文、开放复现和独立评测。
Jev is TypeSafe AI's System One model for decisions that software needs to consume. Instead of generating free-form prose, it accepts a state and one or more typed questions, then returns structured answers with calibrated probabilities.
Jev 是 TypeSafe AI 的 System One 模型，专门处理软件中的决策问题。它不会生成自由文本，而是读取当前状态和一个或多个类型化问题，返回结构化答案及经过校准的概率。
It fits workflows where the option space or scoring rubric is known in advance: routing an agent, validating a tool call, ranking candidates, checking a policy, or deciding whether an action should proceed.
当候选范围或评分标准已经明确时，Jev 很适合用来做 Agent 路由、工具调用校验、候选项排序、策略检查，或判断某项操作是否应该继续。
Applications can set deterministic thresholds, inspect uncertainty, and route low-confidence cases to a person or a larger model.
开发者可以自行设定阈值，查看结果的不确定性，再把低置信度的情况交给人工或更大的模型处理。
Select from a known set and return probabilities with confidence.
Place an input on an ordered scale with a full distribution.
Return the probability that a statement is true, from zero to one.
Each entry is checked against public evidence: the upstream repository, project-authored documentation, and visible Jev usage. Stars are a dated discovery snapshot, not a quality ranking.
每个条目都会根据公开证据核验，包括上游仓库、项目作者文档和可见的 Jev 使用方式。Star 数只是带日期的发现快照，不代表质量排名。
“Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out.”— TypeSafe AI, introducing Jev