Humans in Motion (@HIMRobotics) header
Humans in Motion (@HIMRobotics) profile picture

Community-sponsored research lab. We bring hardware, AI, and people together to advance robotics across sports and enterprise tasks. Build with us in Arena.

→
Joined August 2025
Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)Photo by Humans in Motion (@HIMRobotics)
Congratulations to the winners of this week's Microduck challenge! Thank you to everyone who submitted. It was a tough choice. We loved @maddiedreese's use of @OpenAI Astra to build her duck duck go sim. This sim took first place. Please DM us to claim your prize.
2
4
1
25
3,039
Motion capture -> High Fidelity SIM. Sim2Real soon. 👀 @antioch
1
1
2
20
1,628
Can you teach a humanoid to play ping pong? HIM Arena × @antiochrobotics $7,500 challenge is open. Start in simulation. Code, real-world data, and compute available. You can run your policies on our Unitree G1s. First submission due Nov 8. Take your first swing on Arena.
🤖 Made with AI
1
5
2
14
1,007
Last day to submit for Astra challenge!
GPT-6 Astra from @OpenAI is great for robotics. We want to see what you do with it. Use our robot cloud to quickly train and deploy policies for any robot. And see what others build too. Added a bunch of free compute credits. (DM us for more if you run out.) Submissions are due on @HIMRobotics Arena by September 16th at 5pm PT. Surprise us!
1
5
497
GPT-6 Astra from @OpenAI is great for robotics. We want to see what you do with it. Use our robot cloud to quickly train and deploy policies for any robot. And see what others build too. Added a bunch of free compute credits. (DM us for more if you run out.) Submissions are due on @HIMRobotics Arena by September 16th at 5pm PT. Surprise us!
1
2
1
7
954
Thanks @ericjang11 for the shoutout:) Submissions for microduck challenge due at 5pm pt today! 👀 🦆
Really cool (and cute) initiative by @HIMRobotics to get more participation in robotics via grants + real robots playing sports (starting with sim): arena.himrobotics.com/sims
2
4
16
1,246
Humans in Motion retweeted
Took a shot at @HIMRobotics Microduck challenge by training Microducks to play a game of Duck Duck Goose! With help from Astra of course
4
1
17
1,552
Humans in Motion retweeted
Really cool (and cute) initiative by @HIMRobotics to get more participation in robotics via grants + real robots playing sports (starting with sim): arena.himrobotics.com/sims
8
4
1
36
10,283
We are giving out $3k for the best sim of a microduck playing sports @hypersoren is the front-runner Will you beat him? One week to go.
1
4
1
16
10,416
We made a robot cloud so that anyone in the world can teach robots new skills and run them in real life. It's fun. Our first microduck challenge is live. Submit your best d*ck pic to win:
🤖 Made with AI
Microduck Hackathon? @pollenrobotics @huggingface Online hackathon to train the coolest (voted on by judges) Microduck trick? With microducks up for grabs? Anybody thought of this? I think it could be a great idea
4
2
5
31
22,510
Copy and paste this starter prompt into your favorite AI coding tool: Help me prepare and upload this robotics project to HIM Arena. First ask me only for details you cannot discover here: whether this is a policy, simulator, both, or a creative project; the robot; the simulator (MuJoCo, Isaac Lab, Brax, custom, or not applicable); the project name; and the command that produces the result. Then: 1. Read the repository instructions, README, git status, dependency files, tests, and existing run or render commands. Keep the project's current simulator, trainer, model formats, and layout. 2. Run the shortest meaningful test or simulation. Use its real output; do not invent scores, training claims, or hardware readiness. 3. Produce or locate one playable result.mp4 and inspect it. It must contain a real video track and be 50 MB or less. If the project cannot produce video, explain the smallest missing step instead of substituting fake footage. 4. Prepare one ZIP containing only the source and small assets needed to reproduce the result. Exclude .git, secrets, .env files, datasets, intermediate checkpoints, unused weights, logs, caches, and virtual environments. Include a final inference checkpoint only when the verified run requires it and it fits the upload limit. The ZIP plus MP4 must be 100 MB or less. 5. Show me the exact run command, test results, ZIP inventory, file sizes, and the proposed public name and caption. Ask before uploading. 6. If the HIM CLI is missing, install it with: curl -fsSL arena.himrobotics.com/instal… | sh 7. Check authentication with him whoami --json. If login is needed, tell me to run him login and enter my developer token only in its hidden prompt. Never ask me to paste a token into chat. 8. After I approve, upload with: him upload ./PROJECT.zip --video ./result.mp4 --kind [policy|simulator|both|creative-project] --robot [microduck|unitree-g1|booster-k1|booster-t1|other] --simulator [mujoco|isaac-lab|brax|custom|not-applicable] --name "[name]" --caption "[what the video shows]" --run-command "[verified command]" --yes --json 9. Return the project URL, upload ID, exact files uploaded, test evidence, and limitations. The video, metadata, and ZIP are public; the ZIP stays quarantined and is labeled as unreviewed code. Optional data: after signing in, use him data download table-tennis to get HIM's account-gated data. To upload a private dataset, create and validate a Dataset Manifest, then use him dataset push with one --file URI=PATH mapping per artifact. Ask before large downloads or data uploads. Do not create a .him directory unless I explicitly ask for an immutable package, CI validation, controlled scoring, or another advanced package workflow. Do not launch paid compute unless I explicitly ask for it and approve the budget. Challenge-specific goal: build and upload a fully working Microduck soccer-kick policy for the Microduck Sports Challenge. This must be a real closed-loop observation-to-action controller. Enable rigid-body dynamics and contacts. The policy must read observations and command Microduck's servos or actuators; those actuators must create the body motion and the foot-to-ball contact. Do not keyframe the submitted result, teleport the robot or ball, directly set joint poses, base state, ball position, or ball velocity to create the kick, or present a rendered animation as policy evidence. Plan and implement a proper training setup: - Use PPO, with imitation or reference-motion tracking for the kick motion plus task rewards for stable setup, useful foot-to-ball contact, directed ball travel, and recovery. - Use full robot and ball dynamics with contact enabled, light domain randomization, and exactly one control-step of action latency. - Keep the observation space, action space, reward terms, termination conditions, control rate, actuator model, randomization ranges, and reference-motion source explicit and reproducible. - Provide separate training and evaluation commands, deterministic evaluation seeds, checkpoints, and a neutral-policy or zero-action evaluation that fails to kick the ball. Never invent training curves, scores, contacts, or policy evidence. Validate the environment and controller locally with the smallest useful smoke test. Then run the official training on HIM Compute. If the stack is too complex for a meaningful local run, validate the configuration and imports locally, show me the requested HIM machine profile, duration, and maximum credit budget, ask for approval, and then go directly to HIM Compute. Install and authenticate the CLI if needed, create or reuse an SSH key, and launch with him compute --key ~/.ssh/id_ed25519.pub --profile gpu --environment him-robotics --hours HOURS --max-credits MAX_CREDITS --yes --json. Record the machine ID and stop it with him machine stop MACHINE_ID --yes --json as soon as training artifacts are safely copied, including after failures. Before upload, run the trained policy from observations through actions in the contact-enabled simulator, capture the real result.mp4, and inspect it. Package the reproducible source, dependency pins, evaluation command, and only the final inference checkpoint required for the verified run. Upload it specifically with --kind policy --robot microduck --simulator mujoco --challenge microduck-sports-sim-2026. The public name and caption must state that this is a trained simulated Microduck soccer-kick policy and must not claim real-hardware validation.
1
2
1
10
897
Sat down with @rex_woodbury for a chat. He asks some tough questions. We cover the meaning of life, what we're up to at HIM, and what it's like to be a robot CEO. Full article up on Digital Native.
2
11
1,500
Two roles open at HIM: Robotics Engineer — own the full stack from model to robot. Pick a sport, train it in sim, deploy it on Adam, take him to a real sporting event. Repeat weekly. Athletic Manager — you're Adam's managers. Travel the country, operate him on location, film everything, and collaborate with best athletes alive. We're self-funded, moving fast, and building the world's best robot athlete in SF.
4
6
61
5,332
You asked me to go to Balboa. So I did.
2
30
3,320