Multiple robotics companies are wrestling with one of AI's hardest problems: building foundation models that can do things they were never explicitly trained for. Approaches range from watching millions of web videos to running computer simulations to strapping sensors on humans doing chores. Enigma, a research lab emerging from stealth on Monday, is taking a different tack: instead of just making robots smarter, it wants to figure out how humans actually want to boss them around.

The less-than-one-year-old startup has raised a $70 million seed round led by Index Ventures and Ribbit Capital, with help from Sarah Guo of Conviction Partners. To test human-robot communication, Enigma is launching a large-scale online experiment where anyone can interact with over 100 of its proprietary AI robots. These bots, housed in hangars in Israel and California, can paint, sword-fight, and mix chemicals. Enigma claims it built both the robot arms and their AI models from scratch.

Co-founders Jonathan Jacobi (Microsoft's youngest-ever employee, recruited by Wiz founder Asaf Rappaport) and Gal Niv met as teens in hacking competitions and later served together in Israel's elite Unit 8200. They admit they have zero direct robotics experience but figured that's the best way to think outside the box. Their team includes “smartest friends” from Israel’s tech scene, math Olympiad winners, and people who dropped out of PhD programs.

“There are a lot of robotics industry insiders, but Jonathan and Gal are outsiders - they’re not roboticists. It affords them more room for originality,” said Index Ventures partner Shardul Shah. Jacobi says Enigma aims to make human-robot interaction completely effortless: “If you had to spend 15 minutes explaining to a robot where to put dishes, everyone reaches ‘Forget it, I’ll just do it myself.’” He wants robot control to be as intuitive as adjusting a car’s volume knob - not like setting percentages and hoping for the best.

The startup’s public test will evaluate different interfaces: text, audio, video examples, or tap/drag/drop. Jacobi admits the experiment is open-ended, but hopes it will reveal both better interfaces and better ways to train its AI model. For now, Enigma’s business use case remains an enigma itself - though it’s already partnering with companies in healthcare, logistics, and entertainment.