Human skill for helper robots

Skill has a shape.

We capture how skilled people work and teach robots to help them.

From human experience to machine intelligence

The best teacher is the real world.

Every movement has a story. Every skilled task holds a lesson. We help robotics teams learn from the people who make complex work look effortless.

We learn from real work.

What we build and who it is for.

For robotics teams

Demonstrations of skilled manual work, recorded where the work really happens. Off the shelf, or a campaign built around your task: we bring in professionals of the right trade, record them and deliver data ready for training.

Discuss your use case

For factory automation

Automating a station starts with knowing how your best operators work at it. We record them on site and hand the team building your automation the data it needs.

Discuss your use case

For homes and service businesses

We are developing a small robot model that will run on the robot itself: video is processed on the robot, and data comes back to us only with the owner’s consent. Its first jobs are the chores around skilled work: tidying, fetching and laying out tools, cleaning, carrying. For salons, tailors, workshops and homes.

Discuss your use case
The best teacher
is the real world.
Capture
Deliver, then train
Illustrative image

From a moment to a breakthrough.

01Capture
01

Capture

Skilled people doing real work, recorded where the work happens.

02

Process and label

Raw recordings become structured, richly annotated datasets.

03

Deliver, then train

Training-ready data for your robots. Next we train our own small models on it.

Robots learn best from skilled people.

Where we are.

  • V2 prototype finished

    Our capture system is built and recording.

  • Two letters of intent

    One from a robotics startup, one from a manufacturer.

  • Alien Bazaar 2026

    We co-organised a three-day robotics hackathon in Warsaw and sat on its jury. Event partners included NVIDIA, Hugging Face and Spectacles.

Built by curious minds.

Based in Warsaw. Building for a world that moves.

Bartosz Kostarczyk

Founder

Physics at the University of Warsaw and management with AI at Kozminski University. Laureate of the Polish national stage of EUCYS (2nd-degree prize in the Odkrycia competition). Co-founder of ExtremeBox and APPI; background in AI safety testing and agentic systems.

LinkedIn

Let’s teach robots something remarkable.

Tell us what you’re working on.

This opens your email app with a draft. Nothing is sent until you choose to send it.

CogniCap

Teach the machine.

Watch the sequence. Repeat it. One more action each round.

Actions learned0