Blog
Results, methods and what we learn.
We write up every result we publish, including the ones that don't go our way.
All posts
Why training data for robots has to be manufactured
Language models had the internet to learn from. Drones and robots don't. This is the argument behind our company, and why we think simulation is the only way to cover the cases that matter.
Company24 Jul 2026How we checked our flight controller against real Betaflight
The method behind our first proof point, step by step.
EngineeringComing upOur physics against NeuroBEM's real flight data
Where our model beats the published one, and where it still doesn't.
ResultsComing upSynthetic-to-real drone detection on DUT Anti-UAV
Our first public benchmark: method, code and numbers.
ResultsJan 2027
Get new results by email.
About one email a month. Only when we publish something.