API공식 가격보다 최대 28% 저렴API 사용해 보기 →
Hermes
← Back to case library
Jev코딩

Roman Slack

@RomanSlack1

Roman Slack: @typesafeai의 Jev는 드론 애플리케이션에 매우 적합합니다. 15분만에 만들었는데 가격은 10센트입니다. 저장소: https://t.co/

Localized reading

@typesafeai의 Jev는 드론 애플리케이션에 매우 적합합니다. 15분만에 만들었는데 가격은 10센트입니다. 저장소: https://t.co/b2ytFvehTV https://t.co/rC36lr4iqK

Original post / EN

Jev by @typesafeai works quite well for drone applications. Made this in 15 minutes and only cost 10 cents. Repo: https://t.co/b2ytFvehTV https://t.co/rC36lr4iqK
views
13만
likes
351
saves
296
reposts
28

Links and projects

Related posts

Zyvex ↗

This drone flies itself for 10 cents. pip install typesafe-sdk Camera only, no scripted path, no waypoints. A judgment model sits in the loop at 2.5 Hz. Jev does not generate text. It returns typed decisions with probabilities. Here is what it actually does: → reads a JSON scene built from depth and segmentation → answers three questions per call: go over, go around, or brake → every answer carries a probability, not a sentence → code keeps the veto at 50 Hz and overrides it near obstacles → 80 calls, 65 seconds of flight, 0.11 s median latency Built on the Skydio X2 airframe from MuJoCo Menagerie, flown by a geometric controller. And here is the part that got me: With the model disabled the same stack stops dead at station 2, every run. It cannot express "fly over it". With Jev it clears all 77.5 m. 49 stars. MIT. Input tokens cost $42 per billion. Repo below

This independent archive preserves a public post with attribution. Text, media, account details and trademarks belong to their respective owners.