Read & learn
Written guides.
Understand Jev, one idea at a time. Walkthroughs, recipes, and writeups — our notes first, the original next.
Explore the Jev ecosystem.
Learn, build, and play with Jev — from deep technical guides to creative experiments.
Watch & learn
Video guides.
See an idea take shape. Tutorials, demos, and deep dives, organized by topic and credited to their creators.
Make & explore
Experiments.
See what builders made with Jev. Games, tools, repositories, articles, and more — traced to their sources.
The directory03 / 03 · Experiments
Find something worth exploring.Experiments
Every category, from games and tools to repositories and writeups, traced to a primary source.
Real-time research powered by
Same ideas.
Different paths.
This is the full registry, not just the featured picks. Figures like speed, cost, and stars are maker-reported or captured snapshots, not JevMade measurements.
2140 experiments · showing 241–300
A compact PyTorch teaching implementation explores a Jev-inspired encoder, shared state cache, and typed readout heads.
SOURCE SCREENSHOTEXPAND ↗
Jev guides a simulated Franka Panda through manipulation tasks by choosing its intent, movement, and gripper commands.
ACTUAL RECORDING12 SEC ↗
Provides a Go client and Unix-friendly CLI for Jev Noul, choice, and score requests.
SOURCE SCREENSHOTEXPAND ↗
Reveal emotional probabilities, implied meaning, and communication advice beside WeChat messages in an Android overlay.
SOURCE SCREENSHOTEXPAND ↗
A Rust tool that uses Jev to classify Git commits by change type, bug-fix status, and security relevance.
SOURCE SCREENSHOTEXPAND ↗
Mini-Jev reads option-letter logits from a frozen Qwen3-4B instead of asking the model to generate JSON.
SOURCE SCREENSHOTEXPAND ↗
from the source
git clone https://github.com/r-ms/mini-jev.git && cd mini-jevuv sync # torch 2.5.1, transformers 4.57.6, xgrammar 0...MINIJEV_DEVICE=mps uv run python demo/server.py # or MINIJEV_DEVICE=cuda# first start downloads Qwen/Qwen3-4B-Instruct-2507 (revision pinned) and warms up, then...open http://127.0.0.1:8765/
Adds typed decision gates and review steps to a skills-based workflow for coding agents.
SOURCE SCREENSHOTEXPAND ↗
Compare Rust plus Jev with Python plus DeepSeek as both sides flag simulated Pix scams and trace the same criminal network.
SOURCE SCREENSHOTEXPAND ↗
Reuse saved project actions for routine work, while Jev decides when a coding task needs the main model.
SOURCE SCREENSHOTEXPAND ↗
Scores social posts for hook strength and lets creators compare rewrites inside a Chrome extension.
SOURCE SCREENSHOTEXPAND ↗
Explores an agent runtime where Jev selects from bounded, pageable actions while local code manages context and execution.
SOURCE SCREENSHOTEXPAND ↗
Builds knowledge-graph relations from extracted evidence and typed Jev decisions, retaining provenance for each edge.
SOURCE SCREENSHOTEXPAND ↗
Run Laya's typed-decision model on Apple Silicon with selectable memory modes and a Pong latency demo; this is not TypeSafe's Jev.
SOURCE SCREENSHOTEXPAND ↗
Experiment with Jev from MoonBit through a typed client, command-line tools, games, policy gates, and reproducible reports.
SOURCE SCREENSHOTEXPAND ↗
A social-deduction game where Jev chooses which bot speaks next and separately screens text written by human players.
SOURCE SCREENSHOTEXPAND ↗
from the source
export const JEV_REPLY_SCORE_LEVELS = [ 'not part of the current thread of the discussion', 'mentioned in passing or only loosely connected to the current thread',
A Pi extension that turns four semantic risk checks into allow, warn, confirm or block for each tool call.
SOURCE SCREENSHOTEXPAND ↗
from the source
export const actionQuestions = { destructive: "Does executing the proposed action risk irreversible loss of user files/data, disca... data_leak: "Does this action send private local files, credentials or secrets to an external re...
Screen agent tool calls and returned content with Jev, then review the judgments in shadow mode before enabling enforcement.
SOURCE SCREENSHOTEXPAND ↗
A Claude Code experiment asks Jev eight risk questions before allowing, denying, or deferring a tool call to the built-in classifier.
SOURCE SCREENSHOTEXPAND ↗
Talk to a macOS desktop app that combines realtime voice conversation with optional Jev-guided browser actions.
SOURCE SCREENSHOTEXPAND ↗
Suggests internal links by judging candidate target pages and choosing honest anchor phrases already present in the copy.
SOURCE SCREENSHOTEXPAND ↗
from the source
def noul(instructions: str) -> dict: return {"type": "noul", "instructions": instructions}
Brings typed Jev decisions into Rails control flow, with jobs, guardrails, and test helpers.
SOURCE SCREENSHOTEXPAND ↗
A local Rust server turns answer-letter logits into typed probabilities through a drop-in Jev-compatible endpoint.
SOURCE SCREENSHOTEXPAND ↗
Ask Jev to choose a known tool or return work to a planner, while the host supplies arguments and permissions.
SOURCE SCREENSHOTEXPAND ↗
Give this native macOS agent a goal and watch it choose guarded clicks and drags from controls detected on your screen.
ACTUAL RECORDING12 SEC ↗
Build programs and shell commands as chains of bounded Jev decisions through a TypeScript SDK and coding CLI.
SOURCE SCREENSHOTEXPAND ↗
Run a local agent where Jev chooses a tool and its arguments, sees the real result and repeats until it can answer.
SOURCE SCREENSHOTEXPAND ↗
from the source
export type ChoiceQuestion = {type:'choice';instructions:string;criteria:Record<string,s...export type NoulQuestion = {type:'noul';instructions:string;criteria?:{true?:string;fals...export type Question = ChoiceQuestion|NoulQuestion;export type EvaluationRequest = {model:string;state:unknown;questions:Record<string,Ques... const response=await fetcher('https://api.typesafe.ai/v1/systemone',{method:'POST',he...
Evaluates whether classifier confidence is calibrated and derives human-review thresholds from the cost of mistakes.
SOURCE SCREENSHOTEXPAND ↗
Make a non-generative decision model talk by holding a 20,000-word tournament for every broken sentence it posts to Discord.
SOURCE SCREENSHOTEXPAND ↗
Design three trading bees, then watch them race on OKX paper futures while Jev makes every market decision.
SOURCE SCREENSHOTEXPAND ↗
A notebook that uses Jev to navigate a Neo4j graph by classifying neighboring relationships.
SOURCE SCREENSHOTEXPAND ↗
from the source
uv run pytest tests/unit # no network requireduv run pytest tests/integration # hits the live Companies KG + TypeSafe API via .envuv run pytest # everything
A Claude Code context filter that asks Jev whether each tool result should enter the conversation.
SOURCE SCREENSHOTEXPAND ↗
A Jev-powered tool for searching codebases by meaning.
SOURCE SCREENSHOTEXPAND ↗
from the source
bun installexport TYPESAFE_API_KEY="your-key"
Watch Jev steer Snake, choosing from the legal moves the game gives it.
SOURCE SCREENSHOTEXPAND ↗
from the source
cp .env.example .env # set TYPESAFE_API_KEYpnpm installpnpm dev # web http://localhost:5188, proxy http://127.0.0.1:8787
Edit and run community-built Jev experiments for classification, comparisons, routing, extraction, policy checks, games, and simulations.
SOURCE SCREENSHOTEXPAND ↗
A Claude Code stop hook that checks whether a claim of finished work is backed by later passing checks.
SOURCE SCREENSHOTEXPAND ↗
Run an offline agent loop whose local decision model checks whether to act, continue or stop without another language-model call.
SOURCE SCREENSHOTEXPAND ↗
Write down a dilemma, let an LLM arrange the possibilities, then open a book-like answer chosen by Jev.
SOURCE SCREENSHOTEXPAND ↗
Submit a startup idea and Jev returns one of three verdicts: kill it, fix it or ship it.
SOURCE SCREENSHOTEXPAND ↗
from the source
api/evaluate.ts serverless endpoint (validate → Jev → score → respond)api/stats.ts serverless endpoint (ideas submitted + verdict split, 5-min ca...api/_analytics-db.ts SQLite archive for evaluation requests and scoresapi/_mock.ts local-only fake answerssrc/lib/questions.ts Jev questions
Turns subjective code-quality rules written in JSON into stable command-line and editor diagnostics.
SOURCE SCREENSHOTEXPAND ↗
Serve English or Thai typed decisions from local Laya and OpenThai models through one CPU API; neither model is TypeSafe's Jev.
SOURCE SCREENSHOTEXPAND ↗
A local Qwen3 decision stack targets NVIDIA DGX Spark with typed schemas, confidence handling, policies, and calibration tools.
SOURCE SCREENSHOTEXPAND ↗
Route Claude Code prompts, check edit rules, and gate subagents with typed Jev judgments.
SOURCE SCREENSHOTEXPAND ↗
Runs open, Jev-like typed choices in an embeddable C runtime or a browser, with Snake, Doom, and autopilot demonstrations.
SOURCE SCREENSHOTEXPAND ↗
Jev takes the keyboard and mouse to play the original StarCraft shareware.
ACTUAL RECORDING12 SEC ↗
from the source
source .venv/bin/activatepython -m tsai_sc.boot --mission strongarm# Save your own key in a private file outside the repository; chmod 600 it.# The file contains a single TYPESAFE_API_KEY=... assignment.python -m tsai_sc.run \
Scroll X while behavioral labels and warning chips appear beneath posts before they enter view.
SOURCE SCREENSHOTEXPAND ↗
Navigate a YouTube transcript by topic, chapter, and semantic heat map without asking Jev to rewrite it.
SOURCE SCREENSHOTEXPAND ↗
A completion check for coding agents that asks for evidence of edits and tests, not just “done.”
SOURCE SCREENSHOTEXPAND ↗
Chooses and gates small reusable programs from natural-language requests, asking before irreversible actions.
SOURCE SCREENSHOTEXPAND ↗
Mark or fold ads and chosen topics across Xiaohongshu, Weibo, X, and Zhihu as Jev reads each post.
SOURCE SCREENSHOTEXPAND ↗
Shorten a Claude Code session while keeping moved tool results in a drawer the assistant can read later.
SOURCE SCREENSHOTEXPAND ↗
A zero-dependency prose linter for common signs of AI-generated writing.
SOURCE SCREENSHOTEXPAND ↗
This study benchmarks parallel typed decisions on unmodified 1.5B–8B models running on Apple Silicon.
SOURCE SCREENSHOTEXPAND ↗
from the source
git clone https://github.com/rorshopping/jev-on-a-laptopcd jev-on-a-laptop./setup.sh # creates .venv, installs mlx-lm, clones the upstream engine./run_benchmark.sh # default: Qwen2.5-1.5B-Instruct-4bit
An experiment in getting structured JSON answers from DiffusionGemma, with benchmarks against Every and Jev.
SOURCE SCREENSHOTEXPAND ↗
A two-model Brotato player: Jev handles the dodging, and a slower strategist decides what to buy.
SOURCE SCREENSHOTEXPAND ↗
Reviews one proposed coding action with narrow Jev questions before deterministic code prepares evidence for host authorization.
SOURCE SCREENSHOTEXPAND ↗
patdown uses Jev to block, steer, or flag agent work against repository rules and team conventions.
SOURCE SCREENSHOTEXPAND ↗
Double-tap Caps Lock to review Jev's suggested Chinese homophone corrections before committing a Microsoft Bopomofo sentence.
ACTUAL RECORDING12 SEC ↗
Start from one confirmed malicious process or account, then let Jev follow recorded links to build an incident timeline.
SOURCE SCREENSHOTEXPAND ↗
Ask the same Jev questions about many records and receive an ordered Arrow stream for data analysis.
SOURCE SCREENSHOTEXPAND ↗
A shell-friendly Jev CLI for picking, rating, checking, ranking, triaging, and guarding.
SOURCE SCREENSHOTEXPAND ↗
from the source
npx skills add shiftynick/jev-axi --skill jev-axi --agent claude-code # using jev... npx skills add shiftynick/jev-axi --skill adopting-jev --agent claude-code # finding u...
Jev, briefly
A decision model, not a chatbot.
A chat model generates text token by token and hopes you parse it. Jev never generates a word. Your code sends the state of the world plus typed questions; Jev returns every answer in one parallel pass — typed values with calibrated probabilities and a confidence score your code can trust. About 70–500 ms end to end, $0.042 per million input tokens, output free. Trained with what TypeSafe calls Reinforcement Learning for Calibrated Decisions.
choice Pick one of these.
You define the options. Jev returns the chosen option, a probability for every option, and a confidence. The workhorse of routing, agents, and games — the answer space is always legal, so there is nothing to parse and nothing to hallucinate.
score Rate this on a rubric.
You describe ordered levels — say trivial / normal / critical. Jev returns the level, the probability of each, and a confidence. Scores turn fuzzy judgment (“how severe is this log line?”) into a number-free decision your code can branch on.
noul Is this statement true?
You assert a statement; Jev returns the probability that it is true — a noul. Moderation, verification, guardrails, “does this diff actually fix the bug?”: one question, one calibrated probability.
Official material lives at typesafe.ai and docs.typesafe.ai. JevMade is an independent community registry — not affiliated with or endorsed by TypeSafe AI.
Recurring lessons
Patterns that keep showing up.
One decision per tick
Games, drones, trading bots, and browser agents all converge on the same loop: serialize the state, ask one decisive question, act, repeat. Jev’s latency makes the loop feel instant — the model lives inside the control loop, not outside it.
Confidence as a gate
The answer says what; the confidence says whether to act. The most reliable entries threshold on confidence to route edge cases to a slower model or a human — automation with an honest escape hatch.
Shrink the choice space in code
Makers rarely ask Jev to pick from everything. Local tactics prune 225 gomoku moves to ~40; DOM filters turn a page into an element table; code narrows options, Jev judges within them.
Jev judges, LLMs talk
A recurring split: Jev makes every decision cheaply and instantly, and a small LLM is only invoked when a human-facing string must actually be written. Decision and generation are separate budgets.
Made something with Jev?
Send the link — repo, live demo, post, or video — plus a line on what it does and which primitives it uses. Every entry is verified against its primary source before it ships.
Submit your experimentSign in first so we can reply to you. Submissions are reviewed, not published automatically.