Nikhil Kumar

Drift v1.0.24 is live! for robotics engineers tired of re-explaining their project to agents every command. Shipped: 🧠 Project context - drop a DRIFT.md or SKILLS.md, it's used on every step 🤖 MuJoCo + Gazebo, end-to-end - URDF, MJCF, plugins, launch, debug, run, train ⚙️ Multi-step builds that hold context - robot -> world -> sim -> tune, one prompt

@midjourney Excited for this!

Midjourney will be announcing its first hardware project tomorrow (Wednesday 6/17) at 6pm PT. Stay tuned for a livestream of our in-person launch event in San Francisco. If you're in town and want an invite, reply below, we have just a few slots left.

This is so accurate! Let me tell you how we learned it the hard way 👇 We had a walking policy for the lower body only. So we tried to run it by jamming the upper body joints and figured the arms could just be dead weight. But they couldn't! The arm motion turned out to affect the gait cycle directly. Locking the upper body degrades the walk. Try it yourself: run with your arms pinned to your sides, you'll notice that you hit a speed ceiling fast. Turns out the body is one coupled system. There's no "just the legs", funny how the robot taught us what evolution already knew :)

We actually made their robot walk on our end @GoDrift_ai as well after training it for few hours in simulations and it was so cool! Crazy good work @asimovinc 🚀

Humanoid companies aren't rushing into factories for the labor but for the data. Every robot on a real floor generates the exact data that makes the next one way more general and the factory looks next gen. The quiet part here is that many of these factories aren't selling robots. They're selling that data to the teams building world models and VLAs. The robot is still the collection device and the data is the product.

@krandiash cartesia

We’ve seen incredible traction this year, with enterprises transforming every part of their operations with Voice AI. We’re sharing what we’ve learned and are giving away a guide to the top 10 enterprise use cases for Voice AI. Retweet and comment "CARTESIA" on the main video and we'll send you the guide and $100 in credits so you can get started today.

We released Sonic-3.5 and Ink-2, the #1 streaming models for text to speech and speech to text you can use in your voice agents today. New architectures enable new frontiers for speed and quality. We're now the only provider to have #1 models for both speaking and listening.

@CyberRobooo Curr-0 feels like a glimpse of what happens when movement and intelligence finally become one system

It’s truly impressive--humanoid performing dexterous tasks while on the move. Shanghai-based Current Robotics released Curr-0: a humanoid foundation system for loco-dexterous manipulation. It runs on a single unified policy trained with over 21k hours of real human whole-body data on ~70 DoF hardware. The whole-body coordination in dynamic tasks looks remarkably natural. Yeah,that could be quite relevant for caregiving, home-use, office assistant,complex task scenarios…

@DominiqueCAPaul So wish I could be in Zurich

Over the last few years, I’ve met so many people working on robot learning across Zurich’s academic labs, start-ups, and big tech. What keeps surprising me is how little exchange happens between these groups. To change that, I’m starting a Robotics AI Paper Club for researchers, engineers, and anyone looking to get deeper into the field. Wrote a paper or method yourself? → Present it! Read a paper others should know about? → Show us what you liked and why it’s worth reading! First edition: this Thursday. Drinks, pasta & pesto are on me. Luma link below 👇🏼

@evan_wineland @kaandogrusoz @weaverobotics Big congrats on your launch, Evan🚀

The showing we got for Isaac 1 yesterday was everything we could've hoped for Roboticists, engineers, designers, students, kids, grandparents all in one place Seeing how home robots will change the way we live for the better

@roboindexio @weaverobotics @evan_wineland that's the cutest robo i've ever seen

Isaac 1 just stepped into the home robot conversation. YC-backed Weave Robotics opened its San Francisco space for a first public look, with Isaac 1 pitched as a robot for the chores nobody wants to do: laundry, tidying, bed-making, and checking on the house while you’re away. The company is keeping signups tight, with a refundable $1,000 reservation model and a limited first batch approach that looks built for early demand testing.

Everyone thinks the robotics problem is building smarter robots. It's not. We've already taught machines how to see, reason, move, write code and even drive cars. The harder challenge is dealing with reality and it's messy but how? > a box is slightly crushed > a tool isn't where it's supposed to be > a package arrives upside down > a human does something unexpected The list goes on and honestly, the world is an endless stream of edge cases which humans barely even notice. But robots do and they work exactly on these edge cases I believe that the companies that win the next decade won't necessarily have the smartest AI but they will build systems that keep working when the environment stops being predictable. Because the real breakthrough isn't automation but it's adaptability.

@asimovinc what an excellent breakdown of the work👏

@brianzhan1 Big congrats Brian!

Excited to be on the Business Insider list of top robotics investors. The data wall is falling. For years, the bottleneck was teleoperation: slow, expensive, narrow. Now robots can learn from human video, practice in simulation before touching reality, and there is enough funding to scale data deployments. Robotics is finally running the playbook that took LLMs from autocomplete to reasoning. There are still many open frontiers. General manipulation. The right API layer for the robotics models. Application robotics. That's exactly why I'm bullish. We've just found a recipe that works. Thanks to @ryajetha for compiling the list.

@techniahq now that's replacing humans, I believe

Humanoid robots were supposed to replace workers. This one skipped straight to replacing my cousin at 2AM watching TikTok with one leg in the air.

@shiri_shh don't scare already😭

"one day it won't rebalance. It'll respond 😭"

@stevencheng love the speed!

A parallel SCARA robot that combines both speed and precision. It’s truly a genius design.

@spaceandtech_ This is incredible!

Researchers at ETH Zurich have developed ANYmal, a quadruped robot that can play badminton with humans. The robot tracks the shuttlecock, predicts its path, and moves into position to return shots. Its control system coordinates vision, movement, and racket swings to play the game effectively.

Just attended Saturday Robotics Reading Club Session with a fantastic talk by @DanielXieee from Origami Robotics. We’re moving beyond binary success rates and the old taxonomy to rich multi-axis dexterity metrics: spatial precision, temporal composition, contact richness, coordination & tool use. Dexterity for robots is accelerating to reality 🚀

@anieasyy Damn😂

Perry stops Dr Doofenshmirtz from misusing Fable Major Monogram and Carl were the ones who found the jailbreak, have it on good authority

@weaverobotics It's very cute!

Isaac 1 Today, 4.30pm - 6.30pm 285 7th St, San Francisco

@VaderResearch Final boss for humanoids!🙌

Figure 03 makes it look easy to walk down the stairs Yet this a difficult engineering problem for humanoids 🔵 When you walk downstairs, your foot swings out over empty air before finding the next step 🔵 The robot can't fully see where it's landing; cameras & depth sensors struggle with stair edges & shadows 🟢 The entire upper body is already tilting forward from gravity so if the foot lands even a few cm too far forward/back, the humanoid will fall 🔵 Humans fix this with a lifetime of balance training the body does unconsciously 🔵 Humanoids have rigid flat feet and milliseconds of latency between sensing and correcting