What are we building? Two-arm robotic cells for automating manual work, starting with electronics assembly.
Problem Today, 5–10% of electronics manufacturing operations, primarily in final assembly, remain manual. This often forces U.S. companies to offshore the entire production to stay competitive. We automate that final layer so our customers can build U.S. based autonomous factories.
Insight Electronics assembly is a surprisingly good “first” domain for AI driven robots. The tasks are simple, short, and repeatable. The parts are small, so we can put the entire workspace inside a controlled, portable box. It’s easier to make a robot do precise operations than to make an AI model work in other domains.
Goal: One billion robots assembling one trillion products per day.
Business model? $10k initial payment + $1/hour per robot Marketplace for tuned robot setups
https://x.com/ihorbeaver/status/1986859432165405179
We — Igor and Viktor — have owned an electronics manufacturing business for 9 years, so we understand the problem of manual labor and know how to build electronics.
https://x.com/ihorbeaver/status/1984781192647024670
https://x.com/ihorbeaver/status/1986859432165405179
We found a way to make existing VLA robotic models reliable enough for real-world use.
Unlike large language models, robotics tasks do not require massive datasets to have reliable performance. Good hardware + task-specific data coverage + recovery logic is enough to reach ~99% success rate.