4Stars
GitHubOpen source
Jev-Compatible
Make an existing SGLang or vLLM deployment answer as a decision service.
David-Lolly
Updated Sep 25, 2026
Projects, posts and guides about Jev, the System One model from TypeSafe AI. Each entry links to its source and shows the cost and speed its author reported.

Diogo Almeida
@CompleteSkeptic
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
4Stars
GitHubOpen source
Make an existing SGLang or vLLM deployment answer as a decision service.
David-Lolly
3Stars
yangyu666
1,000 papersVolume
GitHubDocuments and OCR
A thousand arXiv papers, one Jev decision each, checked by an LLM judge.
stas4000
43Stars
sumleo
~3Stars
Iron-LYK
3Stars
HaoxiangXia
4Stars
GitHubBenchmarks and evals
Research on which LLM-decision tasks Jev actually speeds up.
Tongyun1
3Stars
Elue
2Stars
anpicasso
2Stars
Metask AI
11Stars
bytelabs-oss

GitHubAgents and browsers
One ~0.4 s typed decision per DOM step, instead of an LLM turn.
ZephyrDeng
4~0.4 s2Stars
Rohit Mehta
19Stars
GitHubOpen source
Typed decisions in the browser: one state, any number of options.
Nico Martin
19Stars
zhangcy122
13Stars
zjunlp
9Stars
GitHubOpen source
Knowledge graphs built from typed relation decisions, with the evidence kept.
chenmingtang830
8Stars
Qew7
5Stars
Mawfyy
agent-chaperone
2Josh Long
3<15 msLatency
wfzyx
411Alex Volkov
@altryne
This is actually insane. This uses @typesafeai Jev model, as a plugin in Claude to review all the un-nesseasary tool calls, and it takes 1s to run! Like, literally, 1 second to take my Claude session from nearly 1M to ... 86K tokens! 😮 Ask your claude to install it and be amazed Use this prompt ``` Install, and configure : https://github.com/tamaratran/fast-jev-compaction ```
18kStars
Vercel Labs
1.5kStars
GitHubAgents and browsers
Desktop control from the accessibility tree, with Jev picking the operation.
Lahfir
201Stars · ≈ −60% vs largest model backtest
GitHubRouting and model choice
Jev picks the model and the thinking effort for every Codex turn.
Natoshi
57Stars
GitHubCoding and code review
Codebase search by sending a hundred walkers through the file tree.
Ellipsis
33Stars
Sharvil Saxena
32Stars · ~0.25 s judgment
Qkal
83Stars
monteduro
344Stars · every 8 steps decision
GitHubGames and real time
Jev picks NES controller inputs from telemetry, never from pixels.
fhshaik
20Stars · 60 Hz sim, 20 Hz snapshots server
GitHubGames and real time
A browser shooter where the opponent is Jev, first to five kills.
Garrett Emrick
71Stars · 1 call / 2 s pacing
GitHubTrading and markets
An LP rebalancing agent that routes every Jev call through one gate.
Irfandi M
85Stars · 1 request cost per hop
Michael Hunger
23Stars · 24.0 rows/s, one node measured
GitHubDocuments and OCR
A Rust dataset sifter that keeps clean rows verbatim and drops the rest.
Akash Priyadarshi
Akshay 🚀
@akshay_pachaar


SkillRouting and model choice
PickModel routing, memory, compaction and skill selection for Hermes agents.
kerpopule


GitHubOpen source
A source-backed field guide to TypeSafe’s System One model.
AbdelStark
438
GitHubOpen source
Examples indexed by the decision each one makes.
kydlikebtc

GitHubOpen source
Curated Jev resources and runnable examples for typed decisions.
AppitStudio


GitHubSDKs and integrations
A provider-neutral System One runtime for TypeScript and Pi.
iamaamir


GitHubRouting and model choice
Typed IntentSpec contracts as the input layer for routers.
angel291592

GitHubOpen source
Four decision-quality tools on Jev, with code keeping the final say.
klauswg

GitHubOpen source
A self-hosted API for Laya, compatible with the Jev API format.
1Panel-dev





GitHubAgents and browsers
A research design for a Jev-native agent system.
6Mikao9






GitHubSDKs and integrations
80+ TypeScript recipes for agents, retrieval and verification.
agencyenterprise


GitHubAgents and browsers
PickAn LLM proposes, Jev answers, code decides, every step leaves a receipt.
TypeSafe AI

GitHubBenchmarks and evals
Calibration-aware reinforcement learning for decision systems.
weitianxin

GitHubOpen source
Natural-language constraints for JEPA planning, judged by a decision model.
AbdelStark

GitHubOpen source
A Jev-like Rust port using Laya, fully local and with no Python.
Trystan-SA



GitHubOpen source
Windows and Apple Silicon builds, with local OCR and human control.
Aimark-dai



GitHubDocuments and OCR
Jev reads every new arXiv paper each morning and keeps the few you need.
Eliot5566


GitHubBenchmarks and evals
Tests, calibration audits and failure-mode studies of Jev.
Yifan-Lan

GitHubAgents and browsers
A free course: 13 agent use cases with a fast brain and a slow brain.
harshithsunku


GitHubOpen source
An evidence survey of Jev and Jev-like typed decision models.
Eurekaleo



GitHubAgents and browsers
AX-first computer use on macOS with optional Jev semantic guards.
Sur-Cai

GitHubTrading and markets
Typed System One decisions as the judgment layer of a quant stack.
jiayylu



GitHubContext and memory
Test what your AI should remember, with Jev deciding admission.
Cairn-ink

a_c


GitHubSearch
Search an Obsidian vault with Jev deciding relevance.
MikeLembo

GitHubOpen source
An attempt to beat Jev’s accuracy, speed and cost with open models.
rob313


Omid Sayfun
@OmiidSayfun
Couldn't find my important unread emails, so I built a Chrome extension. Jev finds them and labels them by what they're about. You can check it out here: github.com/iamomiid/jev-i…
GitHubInbox and support
Daniel Raeder
@DanRaeder
While Jev is down for new invites, I worked with Claude last night to build Vej (a jev clone that runs locally.) It's weakest at score types, but almost 1:1 for noul and choice. You can use it to build and play with jev-based apps until you can get in. github.com/draeder/Vej
GitHubOpen source
Muskan Paliwal
@PaliwalMuskan19
built this skill-picker for myself because apparently having 30+ agent skills also means remembering which one does what 😭 jev (@typesafeai), being the god that it is, ranks them for a given task and tells me and my agents which skills are actually worth using phew. one less thing for my brain to cache. https://github.com/MuskanPaliwal/skill-picker
GitHubRouting and model choice
Code you can read, fork and run: jev-ultrafast, OpenJev, nanojev, jevlike, simplejev, jev-trader, jev-mcp and the rest. Each repo calls Jev for a decision — which link to follow, which bucket a ticket goes in, which move to make — and the figures on a card are the ones its author published.
318 builds and 12 guides