User repository on GitHub
Jev (TypeSafe System One) vs Claude Opus 5 driving a simulated robot arm in MuJoCo
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev Robotics Demo
A MuJoCo Franka arm with an Allegro hand stacking a blue cube on a red one, driven two ways: Claude Opus 5 reasoning through low-level tools, and Jev picking each simulated move while code owns the physics.
- Stars
- 7
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- FazalAAli
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- One recorded run each: Jev stacked in 19.1 s for $0.0006; Claude Opus 5 took 158.8 s and $0.75 with one retry.
- The Jev agent simulates each small move in a scratch physics copy and asks Jev to pick, grasp, release, and finish.
- The split follows jev_fsd: code owns the math and simulation while Jev owns the judgment.
- README pricing: Jev at $0.042 per million input tokens with output free, against Opus 5 list prices of $5/$25 per million.
- Headless runs are faster because arm motion is not paced to real time, and replay.py renders recorded runs to MP4.
Watch out
MIT, but a demo: robot models come from MuJoCo Menagerie under their own licenses and are not included. Setup targets macOS or Linux with Python 3.12 and uv, macOS viewer runs need mjpython, and API keys are read from environment variables only.
Questions about this entry
Short answers drawn from the linked project or page.
Two agents in the same MuJoCo scene with a Franka arm and a four-finger Allegro hand: Claude Opus 5 with move_hand_to, set_hand, and get_state tools, and a Jev agent where code proposes small moves and Jev judges them.
Jev stacked the cubes in 19.1 s for $0.0006, while the recorded Opus 5 run took 158.8 s and $0.75 and needed one retry; clean first-try Opus runs took about 55 s and $0.19.
Code proposes small moves and simulates each one ahead in a scratch copy of the physics; Jev picks the move and decides when to grasp, release, and call the task done, following the jev_fsd split.
A macOS or Linux machine, Python 3.12 with uv or nix develop, sparse checkouts of franka_emika_panda and wonik_allegro from MuJoCo Menagerie, and keys in environment variables.
Yes, pass --headless; the README notes it is much faster because arm motion is not paced to real time, and headless recording is how the comparison videos are produced.
TYPESAFE_API_KEY for Jev, and either OPENROUTER_API_KEY or ANTHROPIC_API_KEY for the LLM arm, with Anthropic used when there is no OpenRouter key; they are read from environment variables only.
Runs are recorded headless to .npz files, replayed to 720x1440 portrait MP4 with a live clock and cost line, then joined side by side with ffmpeg; record.py is the older screen-recording approach needing macOS Screen Recording permission.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker News: Pick Verify: a single decision at the warehouse pick station catches confusable picks.
Show HN: Verifying robot picks with a decision modelRead the thread on Hacker NewsX: A dual-arm robot with Jev in the middle layer
YouTube: Dan Makes Things drives an ESP32 robot with Jev decisions.
More Jev resources like this
A MuJoCo city simulation where one, two or four camera-observed drones pursue a target: each role asks Jev role-specific typed questions while deterministic mission navigation, safety guards and motor control run independently at 50 Hz.
Jev Demos & Experiments#community#github#python
Community
YongLD/Multi-UAV-JEVPython
YongLD/Multi-UAV-JEVPythonA Clash Royale bot that turns a phone screenshot into JSON once a second and asks Jev for a strategy, a card, and a square.
Jev Demos & Experiments#community#python#gaming
Community
bytelabs-oss/clash-jevPython
bytelabs-oss/clash-jevPythonResearch prototype that tests where Jev can help a driving policy choose continue, observe, replan or defer: Jev receives structured text, a separate vision model handles camera images, and a local controller owns vehicle execution.
Jev Demos & Experiments#community#catalog-import#logicrw
An agricultural advisory demo that turns sensor readings, a forecast, and a crop stage into a next-action choice plus an urgency score, routing marginal readings to an agronomist instead of triggering an application.
Jev Demos & Experiments#community#python#agriculture
Community
A simulated high-frequency crypto trading loop that makes its buy and sell calls with Jev: recent BTC and ETH prices and volumes go in, a typed confidence-aware decision comes out, and a cost-aware execution simulator turns it into P&L. Not live trading.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
waxmell114514/jev-tradePython
waxmell114514/jev-tradePythonMuJoCo simulation where a local decider-2b model picks the next skill for an AgileX PiPER arm through a Jev-style typed-choice interface while the red cube target is pushed to new positions mid-run. Deterministic checks may reject a chosen skill but never substitute another.
Jev Demos & Experiments#community#robotics#simulation
Community
Hu-xiao-max/jev_robotPython
Hu-xiao-max/jev_robotPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.