What are people building with Jev?

Intelligent features in the background instead of a chat: Jev fills in forms from a button, fast enough to feel native.

Jev can't reason, so this interpreter reasons for it: facts and rules are written in English and Jev evaluates each step.

"Pick up the cans", then "the bottles": Jev picks the target and the MOSS robot collects it. Real decisions, replayed in simulation.

OCR plus Jev: a little classifier that sorts a folder of images into categories.

GPT-6 Astra designed the world, Jev chose the actions and H3 Max Turbo turned each round into video: 264 clips in 5 minutes.

Jev looks at a PDF page by page and decides which pages need OCR; the rest is extracted locally. Cheaper and faster.

Jev reviews your row-level security policies and flags the risky ones.

Jev went through thousands of outreach messages and answered in 40 seconds, for under $0.20. Coming to Gojiberry.

Feed it the news and Jev finds the stocks worth buying. Walkthrough in the thread.

Paste a token address and Jev re-reads it every 5 seconds: the call, dump risk, who's in control. Each call is graded 15 minutes later.

Jev as an improv judge.

When the template says "name" and Airtable says "full_name", Jev matches the fields. Live in Bannerbear.

Write what you want to compute in plain words; Jev works out the calculation.

GLM won in 29 moves, but at ~5.8 s and $0.008 per move against Jev's ~0.3 s and under $0.0001. A lesson in using each for its strengths.

Jev reads the prompt first and picks the skill, tool and parameters before the agent starts, halving its response time.

Each hand it sees the cards, jokers and blind, then plays or discards in 200–500 ms, and shops after each ante. A full run costs under a cent.

Jev acts as the computer itself and runs very short programs, sometimes wrong against a reference emulator.

From a company's homepage, Jev finds Careers, picks the links to follow and scores each role against your profile. Was 5 minutes with an LLM.

Helps systematic reviewers find and extract information from research articles quickly. Open source.

An SO-101 arm with Jev choosing the next move about twice a second. After a slip, it went back for the ball and got it in the bowl.

Tempo, instruments, chords and rhythm are all structured choices, so Jev makes every one of them.

Talk and Jev turns it into Figma actions. Total cost from first line of code to the video: $0.01.

Jev chooses the join order; after some tuning, queries on the Join Order Benchmark ran 12% faster.

Jev watches the macOS clipboard; if it holds a terminal command, a quick action offers to run it.