In short:

What are we building?  A frontier robotics system for repetitive manual work, starting with electronics assembly tasks. It includes a robotic foundation model, inference infrastructure, and necessary hardware.

Go-to-market: Step 1: Build a SOTA (best) robotic model that allows users to deploy robots themselves - Done ****Step 2: Start the revenue flywheel by selling inference. Step 3: Start the data flywheel by collecting data from deployments.

Demo:

https://x.com/ihorbeaver/status/2054253758406910457

Team

Igor and Viktor ran an electronics manufacturing business for 9 years. We know the domain from the ground up because we assembled products with our own hands, then hired and trained workers to do it instead of us.

Igor (CEO)

Before: Built an app that reached 1.3 million daily users.

Viktor (CTO)

Before: Founding engineer at a telepresence robot company.

Denys (Founding engineer)

Before: A founding engineer at K-Scale (YC W24 humanoid startup)

Yevgen Chebotar (investor on standard terms/advisor)

A scientist behind the Helix model at Figure humanoid (see mention), co-creator of robotic VLA models (see authorship)

Assembly work we did

Telepresence robot built by Viktor

Progress:

• Apr 2026: Breakthrough. Created Loop Action Model (LAM) a new type of model that is 20× more efficient than the SOTA model from a leading robotics lab, measured by throughput per unit of data.
• Oct 2025: Made a pilot with our customer, where the robot autonomously assembled a real product end-to-end.
  • Main learning: We cannot spend significant time on each customer task, because it does not scale. Customers need a way to collect data and push the model to a 100% success rate without relying on us.
  • Details: https://x.com/ihorbeaver/status/1984781192647024670
• May 2025: The first release went viral without any marketing effort on our side.
  • It got
    • 6.1k likes on Twitter
    • 1.5k likes on Reddit
    • 150 paid deposits.

Market:

We started with electronics assembly because it is a large industry, comparable in scale to software development (17.4 million people in electronics assembly vs. 27.7 million people in software development in 2023). It aligns with the major trend of US reindustrialization. And it allows us to build our own hardware autonomously by dogfooding our technology.

What is your Moat?

Our moat comes from two self-reinforcing loops: the revenue flywheel and the data flywheel, both of which compound over time.

Because our data-collection approach is deployment-heavy, hardware becomes an additional moat against research labs that lack manufacturing expertise and smaller competitors.

ASK

We are raising a seed round for: