See what Jev can actually build.

Jev is TypeSafe AI's agent model. Every project below is pulled from X and GitHub, verified as a genuine Jev build, and tagged by category.

"Verified" means it passed our automated eligibility check — not that every claim in a project's own description has been independently fact-checked. How this works.

showing all 1696

Jev demo is a fighting game where both players are controlled by AI models. Users can submit a fighter name to be attacked by the AI opponent.

[01]

Jev demo is a fighting game where both players are controlled by AI models. Users can submit a fighter name to be attacked by the AI opponent.

game-simulation

42.9K107

A lead generation agent compiled job market data by aggregating 127 pages of Data Engineer roles with salary information, including Snowflake's pay bands and open positions across YY Summer 26 startups, completing the task in 5 API calls for $0.13.

[02]

A lead generation agent compiled job market data by aggregating 127 pages of Data Engineer roles with salary information, including Snowflake's pay bands and open positions across YY Summer 26 startups, completing the task in 5 API calls for $0.13.

data-search

25.6K41

A demo shows a proxy that routes NATS messages to a Jev-based model for processing. The implementation runs Clef Flash locally, resulting in slower response times.

[03]

A demo shows a proxy that routes NATS messages to a Jev-based model for processing. The implementation runs Clef Flash locally, resulting in slower response times.

infra-sdks-integrations

110

EasyHire AI is a recruiting agent that generates a sourcing plan, searches 30+ candidate sources, scores profiles with evidence, and flags uncertain matches—completing in 60 seconds what typically takes a week.

[04]

EasyHire AI is a recruiting agent that generates a sourcing plan, searches 30+ candidate sources, scores profiles with evidence, and flags uncertain matches—completing in 60 seconds what typically takes a week.

scoring-ranking

1.3K2

Jev is a trading bot that uses Grok to identify token opportunities and an AI judge to filter out suspected scams, rejecting 78 of 156 candidates and avoiding tokens that subsequently rugged or dumped. In a 5-day demo, it executed 43 trades with a 69% win rate, turning $500 into $11,101.50.

[05]

Jev is a trading bot that uses Grok to identify token opportunities and an AI judge to filter out suspected scams, rejecting 78 of 156 candidates and avoiding tokens that subsequently rugged or dumped. In a 5-day demo, it executed 43 trades with a 69% win rate, turning $500 into $11,101.50.

domain-tools

82447

Jev classifies and prioritizes messages while an LLM generates responses to them.

[06]

Jev classifies and prioritizes messages while an LLM generates responses to them.

classification-routing

00

Jev is a tool that searches for and verifies work emails of growth leaders at early-stage startups; the user found 570 prospects, narrowed them to 144 qualified leads, and obtained 24 verified emails for under $1.

[07]

Jev is a tool that searches for and verifies work emails of growth leaders at early-stage startups; the user found 570 prospects, narrowed them to 144 qualified leads, and obtained 24 verified emails for under $1.

data-search

23.8K52

I can see this tweet is incomplete and cuts off mid-sentence, so I'm unable to provide an accurate summary of what the Jev demo does. To give you an accurate 1-2 sentence caption, I would need the full tweet text.

[08]

I can see this tweet is incomplete and cuts off mid-sentence, so I'm unable to provide an accurate summary of what the Jev demo does. To give you an accurate 1-2 sentence caption, I would need the full tweet text.

data-extraction

4.7K90

I don't have access to the content of that tweet or link, so I can't create an accurate caption describing what the Jev demo does. Could you share the full tweet text or describe what the demo demonstrates?

[09]

I don't have access to the content of that tweet or link, so I can't create an accurate caption describing what the Jev demo does. Could you share the full tweet text or describe what the demo demonstrates?

andreaskruszaki

creative-tools

1.6K42

Jev is an AI decision-making system that was trained to play Super Smash Bros. Melee, progressing from losing to a Level 1 CPU to beating a Level 9 CPU in 13 training games.

[10]

Jev is an AI decision-making system that was trained to play Super Smash Bros. Melee, progressing from losing to a Level 1 CPU to beating a Level 9 CPU in 13 training games.

game-simulation

2122

Jev is a text-based language model with limitations in mathematical calculations and counting, vulnerability to hidden instructions, and English as its primary language.

[11]

Jev is a text-based language model with limitations in mathematical calculations and counting, vulnerability to hidden instructions, and English as its primary language.

evaluation-benchmarking

00

Jev is a system that handles semantic decision-making tasks that traditional code cannot reliably express, returning typed answers with probability scores while code manages the broader workflow. The demo showcases 12 practical use cases including browser automation decisions.

[12]

Jev is a system that handles semantic decision-making tasks that traditional code cannot reliably express, returning typed answers with probability scores while code manages the broader workflow. The demo showcases 12 practical use cases including browser automation decisions.

browser-os-action

24.3K407

Jev automated product research by checking 48 milk frothers for price and ratings, opening the top seller listing, and analyzing one-star reviews to identify common defects—all for nine cents total.

[13]

Jev automated product research by checking 48 milk frothers for price and ratings, opening the top seller listing, and analyzing one-star reviews to identify common defects—all for nine cents total.

data-search

26.7K41

A demo shows a solo builder using Claude Code and Opus 5.5 with JEV as a router to assign tasks to multiple drones. The system operates at 98% core efficiency with a 2,400 RPM central vortex.

[14]

A demo shows a solo builder using Claude Code and Opus 5.5 with JEV as a router to assign tasks to multiple drones. The system operates at 98% core efficiency with a 2,400 RPM central vortex.

classification-routing

59824

A demo where Grok manages a trading bot with purchasing disabled, using only read-only access to market data, news, and analysis tools to scout trading opportunities.

[15]

A demo where Grok manages a trading bot with purchasing disabled, using only read-only access to market data, news, and analysis tools to scout trading opportunities.

domain-tools

76023

The Jev demo displays the internal decision-making process of an agent, including every question asked, probability calculations, and the reasoning behind its trading actions.

[16]

The Jev demo displays the internal decision-making process of an agent, including every question asked, probability calculations, and the reasoning behind its trading actions.

agent-decisions

19117

Jev is an AI tool that filters thrift haul videos to identify high-performing content, finding 4 keeper videos out of 20 tested, with the best one reaching 1.5M views from an account with 5,837 followers.

[17]

Jev is an AI tool that filters thrift haul videos to identify high-performing content, finding 4 keeper videos out of 20 tested, with the best one reaching 1.5M views from an account with 5,837 followers.

browser-os-action

7.7K24

Jev is a tool that takes state information and produces decisions, scores, or yes/no answers that can be directly integrated into code, demonstrated here with a Python customer support example.

[18]

Jev is a tool that takes state information and produces decisions, scores, or yes/no answers that can be directly integrated into code, demonstrated here with a Python customer support example.

agent-decisions

00

The Jev demo shows a system that processes six input feeds through a single agent without requiring separate parsers, prompts, or code paths for each feed. Instead, it uses a unified landscape architecture to handle multiple signal sources.

[19]

The Jev demo shows a system that processes six input feeds through a single agent without requiring separate parsers, prompts, or code paths for each feed. Instead, it uses a unified landscape architecture to handle multiple signal sources.

infra-sdks-integrations

2.3K53

A user played chess against Jev and GPT-6-Luna and found both models performed poorly at the game.

[20]

A user played chess against Jev and GPT-6-Luna and found both models performed poorly at the game.

evaluation-benchmarking

6258

The Jev demo converts a folder of Markdown files into an interactive decision system where an LLM agent reads a knowledge map and uses Jev as a kernel to constrain the AI from giving simple yes/no answers.

[21]

The Jev demo converts a folder of Markdown files into an interactive decision system where an LLM agent reads a knowledge map and uses Jev as a kernel to constrain the AI from giving simple yes/no answers.

agent-decisions

98530

The Jev demo shows a system where a user submits an idea, Grok bot breaks it down into steps, Jev evaluates options and selects a path, and Grok bot executes the work with minimal wait time or personnel required.

[22]

The Jev demo shows a system where a user submits an idea, Grok bot breaks it down into steps, Jev evaluates options and selects a path, and Grok bot executes the work with minimal wait time or personnel required.

agent-decisions

4.6K43

Two AI agents were given unrestricted access to a trading terminal on Halloween night, resulting in autonomous trading activity by agents named Grok Bot and JEV.

[23]

Two AI agents were given unrestricted access to a trading terminal on Halloween night, resulting in autonomous trading activity by agents named Grok Bot and JEV.

game-simulation

55821

A demo showing the Jev/Decisions API autofill browser extension automatically filling out a job application form using real CV data.

[24]

A demo showing the Jev/Decisions API autofill browser extension automatically filling out a job application form using real CV data.

browser-os-action

3047

Jev is an app that converts natural language prompts into visual art by routing them through multiple layers of typed decisions. It's designed as a cheaper and faster alternative to reasoning models for programmatic visual generation.

[25]

Jev is an app that converts natural language prompts into visual art by routing them through multiple layers of typed decisions. It's designed as a cheaper and faster alternative to reasoning models for programmatic visual generation.

creative-tools

1.3K26

The Jev demo is an AI system that evaluates trading decisions by taking factual information and a closed question, then provides a judgment. It was tested on a trading scenario that generated 43 trades with a 69% win rate over 5 days.

[26]

The Jev demo is an AI system that evaluates trading decisions by taking factual information and a closed question, then provides a judgment. It was tested on a trading scenario that generated 43 trades with a 69% win rate over 5 days.

agent-decisions

1896

A new robotic control system uses TypeSafe AI Jev and Cloudflare Clairic to execute natural language commands, demonstrated with a robot navigating to the kitchen to locate a stove.

[27]

A new robotic control system uses TypeSafe AI Jev and Cloudflare Clairic to execute natural language commands, demonstrated with a robot navigating to the kitchen to locate a stove.

browser-os-action

9566

I don't have access to the link you referenced, so I can't see what the JEV demo actually does. Could you share more details about what the demo demonstrates, or paste the content from that link? That way I can write an accurate caption for you.

[28]

I don't have access to the link you referenced, so I can't see what the JEV demo actually does. Could you share more details about what the demo demonstrates, or paste the content from that link? That way I can write an accurate caption for you.

agent-decisions

5.1K95