Directory
Projects, tools, demos, and claims from launch week onward. Each carries a verdict on what was claimed, and a test status for whether we've reproduced it.
upweight: a personal Hacker News front page
Project by Vishesh Baghel, added 16 Sep 2026
VerifiedNot yet tested by us
Six sliders (technical depth, drama, practical utility, AI slop, novelty, career relevance) re-rank the HN front page instantly, with each story's scores visible.
JevScoring and ranking
Jev Ultrafast for Browser Use
Project by browser-use, added 16 Sep 2026
PlausibleNot yet tested by us
Jev picks both the browser operation and the page element in one request; a small LLM is only called when text needs typing.
JevAgents and real-time
Model router: Jev picks which model answers
Demo by X demo (948 likes), added 17 Sep 2026
PlausibleNot yet tested by us
Jev decides which LLM should serve each request before it's forwarded.
JevRouting and triage
Jev as an agent safety monitor
Claim by X test report, added 18 Sep 2026
UnverifiedNot yet tested by us
Checking each agent action with Jev first catches most attacks with almost no false blocks, at much lower latency.
JevGuardrails and safety
One 50 ms pass vs. 23 LLM turns
Claim by X post, added 18 Sep 2026
UnverifiedNot yet tested by us
A 706K-parameter model fills a whole form in one 50 ms pass while an LLM agent needs 23 turns and 39.6 seconds.
Other modelAgents and real-time
jev-trader: a trade decision every block
Project by jarrodwatts, added 16 Sep 2026
Hype riskNot yet tested by us
Asks Jev for a buy or sell decision on every Monad block on a live market.
JevAgents and real-time
Jev + fallback vs. GPT-5.6 Luna on passage support
Benchmark by Community benchmark, added 20 Sep 2026
PlausibleNot yet tested by us
Same accuracy as GPT-5.6 Luna and faster, but the fallback route cost more overall.
JevRetrieval checks
TypeSafe: 193.6× faster, 444.6× cheaper
Vendor claim by TypeSafe AI, added 15 Sep 2026
MisleadingNot yet tested by us
Jev is 40–200× faster and 40–400× cheaper than frontier LLMs, peaking at 193.6× and 444.6× on TypeSafe's workflows.
JevClaim vetting
“Jev can't hallucinate”
Vendor claim by Launch marketing, added 15 Sep 2026
MisleadingNot yet tested by us
Because outputs are typed, Jev cannot hallucinate.
JevClaim vetting
Laya: “6–8× faster than Jev”
Vendor claim by Convai Innovations, added 18 Sep 2026
MisleadingNot yet tested by us
At 32.8 ms per question on one GPU, Laya is 6 to 8 times faster than Jev.
LayaJevClaim vetting
Laya's 0.766 benchmark accuracy
Vendor claim by Convai Innovations, added 18 Sep 2026
MisleadingNot yet tested by us
Laya reaches 0.766 on the typed-decisions benchmark.
LayaClaim vetting
typesafe-ai/skills: official agent skill
Tool by TypeSafe AI, added 17 Sep 2026
VerifiedNot yet tested by us
An MIT-licensed skill that teaches Claude Code, Codex, and other agents to call Jev.
JevTooling and open replicas
jev-mcp: an MCP server for Jev
Tool by jkudish, added 17 Sep 2026
VerifiedNot yet tested by us
Lets Claude Code, Claude Desktop, and Codex call Jev and branch on the probabilities.
JevTooling and open replicas
llm-typesafe plugin for the LLM CLI
Tool by Simon Willison, added 22 Sep 2026
VerifiedNot yet tested by us
Adds Jev to the llm command-line tool and Python library.
JevTooling and open replicas
openjev (now SemIf): a Jev-style model on one RTX 3090
Open replica by TheoLeeCJ, added 16 Sep 2026
VerifiedNot yet tested by us
Runs a Jev-style decision model locally on a single consumer GPU.
Other modelTooling and open replicas
jevlike: an open reimplementation
Open replica by vinnylarouge, added 16 Sep 2026
VerifiedNot yet tested by us
Trains a small model to score a changing list of options in one forward pass, with Doom, chess, and Wikispeedia demos.
Other modelTooling and open replicas
Kev: Jev-like models on Qwen
Open replica by Community, added 21 Sep 2026
PlausibleNot yet tested by us
Open Jev-like decision models built on Qwen, with a TypeSafe-compatible API.
Other modelTooling and open replicas
jevchat, jev-leftpad, and jev-2048
Demo by Kyle Pena, Fatih Kadir Akın, Andy Gayton, added 20 Sep 2026
ToyNot yet tested by us
A chat model that picks one symbol at a time, left-pad as a Choice question, and Jev playing 2048.
JevAgents and real-time