Robot hardware is getting crazy cheap: service bots run about $11.5k now (down from $21k just two years back), and you can grab printable humanoid kits for under $6k (min $16k last year).
At the same time, vision models like GPT-4V, Gemini Robotics, and ViLa are becoming really
Moyai
186 posts
Joined March 2024
- github.com/codecflow/optr Introducing optr: the Operator Kit (Early Alpha) A year ago, I started working on Codec to help @_lilkm_ with some quality of life improvements while he worked on the @LeRobotHF project. What began as simple tooling grew into something bigger. We
- Thoughts after reading "Robotics & The Age of Physical AI" by @0xPrismatic LLMs thrive on decades of internet text and images, enabling human-level language and visual understanding. Robotics, however, lacks such data. Robots need labeled visual data and poses, but current
- With optr, you can now run simulations remotely and interact with them in the browser, no need to worry about having the right GPU or installing anything locally. Simulations run in cloud containers and stream to the 3D viewer through WebRTC.
00:00 - Great read, @veradittakit, LLMs changed the world a few months ago, and VLAs represent the logical next evolution. Given the ongoing trend toward cheaper, commoditized hardware, robotics applications seem positioned to capture the greatest value from these advances. I'd like to
- Replying to @CryptotrissyI consider myself really lucky to have met @0xdetweiler and @Cryptotrissy when we started building CodecFlow. Having them believe in our team and get our vision from the start gave us the fuel to get to where we are. I'm genuinely thankful for all the work and guidance they've
- Best way to learn is to just build, I have been tinkering with a few recent papers in motion control and imitation learning in order to integrate them into optr. some of the papers I explored: - github.com/LeCAR-Lab/ASAP - github.com/YanjieZe/GMR - github.com/zixuan417/humaโฆ Still
- they donโt know ๐
- We had an incredible time at the @Hadronfc Fireside Chat on AI x Robotics in Web3. Appreciated the insightful questions, thoughtful discussion, and the energy from everyone there. Grateful to meet so many great minds pushing this space forward.
- Replying to @cryptoswag123 @FarzaTV and @codecopenflowThis can be built as an Operator on our marketplace Developer: Focus on training the coaching model. SDK handles camera access, processing, and rendering. CodecFlow: Manages infrastructure, billing, and delivery. User: Install, grant camera permission, get coaching feedback.
- congrats @openmind_agi on the raise! It's exciting to see this validation for what we've been building toward these past 7 months. definitely confirms we're on the right track, and the momentum in this space keeps growing
- Excited to be on the panel at tomorrowโs @Hadronfc Fireside chat speaking on AI x Robotics in Web3.Two AI visionaries. One evening in Dubai. Join @sandeepnailwal & @aakrit at our office on Oct 21st. Fireside chats on Sovereign AI, open vs closed models & more.
- after sharing the GTA demo few people asked if it's possible to make it play and farm web3 games i spent sometime today experimenting with Shogun Curse from @OhBabyGames. the video is a replay of the operator thoughts and actions
00:00 - github.com/mit-han-lab/stโฆ Real-time agents and operators are emerging faster than most realize. The gap between perception and action is closing, blurring the line between software and the physical world. This advancement moves us closer to systems that interpret not just video,







