RoboStrategy published our first Shareholder Letter, and in it we discuss our outlook on Robotics industry. While we've already reached the GPT-3 era, there won't be a ChatGPT moment. The adoption curve will be somewhere between AI chatbots and autonomous vehicles. There will be no single day where it clicks for the global consciousness, but nonetheless a rapid permeation of new autonomous machines of all shapes and sizes. Physical AI is already getting really good for a variety of use cases, but we can't instantly supply millions of robots to customers like we can with AI instances. By next year, intelligence will no longer be the bottleneck, it will be the actual robots. robostrategy.co/news/shareho…

Just realizing that robots might develop their own caste systems

I love humanoids but it’s awesome to see the step change evolution of the robot arm from essentially a 1970s PC to something you can now vibe code Programmable robots -> instructable robots

Robotics research output is accelerating More accessible and cheaper hardware, tooling, talent and capital waves, better general purpose models are all major contributors here

Recent developments from frontier Physical AI companies are showing that robots are capable of moving from single task automation to complete workflow automation. This unlocks a market an order of magnitude larger than single task autonomy. Similar to the jump from L3 to L4 driving automation that allowed Waymo to start scaling its operations or the jump from AI chatbots to AI agents. Less than 5% of physical work in the world is actually single task. Machine tending, packaging, simple assembly. A lot of work, even if seemingly repetitive, require more higher level decision making than we think. Take for example Dyna's recent demonstration for their laundry physical agent. Addressable annual laundry labor itself is ~$9.5B. 85% of North America's 59,000 hotels have 200 rooms or fewer. These have on average one full time laundry attendant. That laundry attendant is continuously navigating multiple decision trees - when to load the washer, when to transfer from washer to dryer, when to fold laundry, when to move folded stacks to the shelves, etc. Correctly making these decisions requires real scene understanding, memory, and the ability to reliably execute a large number of subtasks. A laundry attendant that has a single task folding robot still needs to be near the laundry room for most of their shift since they need to support the robot in loading, unloading, transfering stacks and moving laundry baskets. They free up some time, but they can't effectively reallocate it to housekeeping due to the constant support required. When a robot is able to complete entire laundry workflows unattended, now the hotel manager can quickly fill a real laundry attendent hiring need with a robot or reallocate an attendant to housekeeping work. The ability to complete entire workflows makes robots much more useful to businesses. This is the value of physical agents.

Anyone waiting for the ChatGPT moment for robotics is going to miss the first wave of adoption @Standardbots are already deployed to hundreds of customers across almost every state @DynaRobotics has robots deployed across restuarant chains, hotels, logistics, data centers @PathRobotics has robots welding across defense, oil & gas, energy companies and signed a $600m deal with HII, one of the largest shipbuilding companies We will eventually build a big single model that works out of the box for anything, but specialized models are starting to tackle massive labor markets today