Today’s send is sponsored by distll.ai, be the most informed person in the room by keeping a daily pulse on literally any topic you can think of. Use code ROBOTS90 for 90% off your first three months of distll!
Top headlines
Genesis AI launches with $105M seed to build robotics foundation model
Amazon deploys one millionth robot, launches DeepFleet robotics model
Unitree Robotics valuation surpasses $1.7 billion, eyes IPO
MIT increases robot jump height 41% by using AI assisted designs
Researches develop TWIST, an AI framework for humanoids
Neural Jacobian Fields (NJF) allows robots to model themselves using a single camera
Genesis AI launches with $105M seed to build robotics foundation model
Emerging from stealth mode, Genesis AI, founded by Dr. Zhou Xian and Théophile Gervet, aims to tackle the $30-40 trillion physical labor market. Their $105 million seed round, co-led by Eclipse Ventures and Khosla Ventures with participation from Bpifrance, HSG, Eric Schmidt and Xavier Niel, will fuel the development of a universal robotics foundation model (RFM) and horizontal platform, slated for release by the end of 2025, targeting applications across manufacturing, logistics, and healthcare.
Why it matters
Genesis AI is architecting the future of work, poised to disrupt industries ripe for automation and finally address those pesky labor shortages. If they succeed, we could see a Cambrian explosion of robotic capabilities that redefine what's possible in the physical world.
Sources:
TechCrunch.com
Amazon deploys one millionth robot, launches DeepFleet robotics model
Amazon Robotics reached this milestone with the deployment of a robot in a Japan fulfillment center. The newly launched DeepFleet model leverages AWS infrastructure to optimize the navigation of robots like Hercules, Pegasus, and Proteus, promising a 10% improvement in travel efficiency across Amazon's 300+ facilities. VP Scott Dresser explained that DeepFleet analyzes logistics data in real-time to reduce congestion and accelerate deliveries.
Why it matters
This milestone signals Amazon's relentless pursuit of automation, turning warehouses into hyper-efficient robot armies. DeepFleet isn’t just about faster shipping; it’s a glimpse into a future where AI orchestrates complex systems, leaving competitors scrambling to keep up or be relegated to the slow lane.
Sources:
aboutamazon.com
Unitree Robotics valuation surpasses $1.7 billion, eyes IPO
The Chinese firm, specializing in legged and humanoid robots, finalized a Series C round in May 2025, spearheaded by investments from Tencent, China Mobile, and Alibaba, valuing the company between $1.7 and $2.95 billion. CEO Wang Xingxing is now eyeing a potential IPO as early as Q1 2026, signaling a major push into public markets amid rising demand for affordable robots in education and industry. Unitree's rapid growth underscores the intensifying competition between China and the US in the robotics arena.
Why it matters
Unitree's surge to unicorn status isn't just about valuation; it's a potent symbol of China's ambition to dominate the global robotics market, challenging US leadership. Their aggressive pricing strategy could force competitors to innovate faster or face obsolescence, ultimately accelerating the arrival of robots in our daily lives.
Sources:
therobotreport.com
Quick Hits
MIT increases robot jump height 41% by using AI assisted designs
MIT CSAIL's new generative AI framework uses diffusion models to let users optimize robot designs, leading to impressive results. Applying this approach to a jumping robot yielded a 41% increase in jump height, achieved through innovative curved linkages that store more energy. Optimized designs are production-ready, eliminating manual adjustments and dramatically accelerating the build process.
Sources:
mit.edu
Researches develop TWIST, an AI framework for humanoids
TWIST emerged from a collaboration between Stanford and Simon Fraser, leveraging H2L's Capsule Interface to map muscle activity instead of just positions. This allows a Unitree Robotics H1 robot to perform nuanced tasks with greater realism, capturing intent and force behind movements. Demonstrations in late June showcased TWIST's ability to enhance human-robot interactions beyond simple motion replication.
TWIST marks a pivotal shift from clunky teleoperation to genuine embodiment for robots, threatening to disrupt fields from remote surgery to elder care with more intuitive control. Imagine robots that not only move like us, but react like us – that's the paradigm shift TWIST unlocks, and it's going to be a race for dominance in human-compatible AI.
Sources:
rudebaguette.com
Neural Jacobian Fields (NJF) allows robots to model themselves using a single camera
MIT CSAIL's Neural Jacobian Fields (NJF) allows robots to model themselves using a single camera, eschewing traditional sensors and pre-programmed digital twins. Published in Nature in June 2025, the system, spearheaded by PhD student Sizhe Lester Li, lets robots learn body dynamics visually, critical for adaptable and affordable robotic solutions. This approach tackles limitations of rigid, sensor-dependent systems, opening doors for robots that can self-learn and operate even with deformable or hidden components.
NJF flips the script on robotic control, moving from meticulously engineered systems to adaptable machines that learn by seeing. This spells trouble for legacy hardware vendors, as the future belongs to robots that can teach themselves, dramatically lowering costs and expanding their utility in unpredictable environments.
Sources:
mit.edu
👋 Thanks for reading! I read all replies, so don’t hesitate to respond to this email with feedback, questions, and suggestions.
- Andrew