Five real projects built with Jev in one week: debate meter, command gate, trading, speaking, clone
A roundup of five open-source projects using Jev for binary decisions: debate rhetoric classification, ALLOW/ASK/DENY gating for Claude Code commands, pre-trade approval, spoken English correction, and a 0.6B NanoJev clone that beat the original on a Doom test.
Jev came out and within days people were doing the weirdest sh*t with it debates, Claude Code, a live trading account, English homework, and one guy rebuilt the whole thing himself here's what you missed: > jevmeter Jev fact-checks the vibe of a debate. one call per sentence: dodge? attack? straight answer? 1,191 calls. 5,955 yes/no answers. total cost: $0.05. then it cuts the video with live meters on screen. > http://github.com/ChetasLua/jevmeter > bouncer a literal bouncer for Claude Code. before every command runs, Jev checks it against your rules: ALLOW / ASK / DENY ~190ms per check. ~4 cents a day. > http://github.com/clownware/bouncer > QuantDinger 12K-star trading OS. before a trade hits the exchange, Jev looks at the order, positions and budget and says yes or no. exits and stop-losses skip the line. > http://github.com/OpenByteInc/QuantDinger > dasheng read English out loud. Jev catches every word you mess up while you're still talking. 80–200ms per judgment. > http://github.com/wquguru/dasheng > NanoJev someone built a 0.6B copy of Jev. it went 128/128 on a Doom test. the original Jev went 56/128. the copy beat the teacher. > http://github.com/TianyuCodings/NanoJev and here's the pattern: none of these need Jev to write anything. they need it to look and pick: DODGE / ATTACK / ANSWER ALLOW / ASK / DENY ENTER / SKIP RIGHT / WRONG LEFT / RIGHT / SHOOT that's the whole thing. the big model plans. Jev picks. code does the rest.
This post quotes another post: x.com/thegreatest_sv/status/2102802378378621355