Enigma Raised $71M on a Thesis: Robot Interfaces, Not Raw AI, Are the Adoption Bottleneck
Most of the $18.8 billion that flowed into robotics startups in the first half of 2026 went to companies working on making robots smarter. Enigma, which emerged from stealth on July 27, took a different bet with its $71 million seed round: the problem isn’t that robots aren’t smart enough — it’s that no one has figured out how to make them easy to use.
That distinction matters for anyone building, buying, or deploying robotic systems. The company’s thesis — that the interface between human and machine is the primary bottleneck, not raw capability — is backed by an unusual kind of evidence: a live public experiment at robots.online where anyone in the world can right now interact with more than 100 of Enigma’s own AI-powered robots in real time. What makes that experiment worth paying attention to is not just what it will teach Enigma about usability. It is also how the company intends to close the data gap that every serious physical AI lab is racing to fill.
Robotics Has an Interface Problem Every Capable Robot Makes Worse
“If you had to do your dishes and spent 15 minutes explaining to a robot where to put everything, everyone reaches the point of ‘Forget it, I’ll just do it myself,'” said Jonathan Jacobi, Enigma’s co-founder and CEO, speaking to TechCrunch at the company’s stealth emergence. “Right now, everyone is at that point — even with the most capable models.”
Jacobi’s analogy is deliberate. Adjusting a car’s volume is intuitive because the interface — a physical dial — gives proportional, immediate feedback. You turn it, something changes, and you understand what just happened. Robots today have no equivalent. The more sophisticated the underlying model, the larger the vocabulary of potential actions, and the harder it becomes for an ordinary person to communicate what they actually want done.
This is not a new problem in human-robot interaction research, which has been a formal academic field since the early 2000s, tracing conceptual roots back to Isaac Asimov’s Three Laws of Robotics. Researchers have documented that during initial interactions, people are more uncertain, expect less, and prefer to communicate with humans rather than robots. They have also found that when a robot has no clear use or produces unpredictable responses, negative feelings increase and willingness to try again decreases. What Enigma argues is that these documented user-experience failures are not fixed by building smarter models — they are fixed by building better interfaces, which then feed better data back into the models.
robots.online Is Also a Data Engine
The technical underpinning of Enigma’s strategy is more significant than the press release framing suggests. Physical AI foundation models — the AI systems that give robots general-purpose intelligence — are severely limited by the scarcity of real-world human-robot interaction data. Industry estimates in 2026 put the total volume of high-quality robotic interaction data at roughly half a million hours, against a requirement measured in billions of hours for anything approaching general-purpose robot competence. Every hour of quality teleoperation data requires a human, a robot, and a facility — a cost curve that does not scale.
Enigma’s public experiment directly addresses this. The 100-plus robots housed in hangars in Israel and California are performing real physical tasks — drawing pictures with a paintbrush, sparring with swords, picking up and mixing flasks in simple chemistry experiments — while real humans worldwide interact with them in real time. Every interaction is a labeled training sample: a record of what a human communicated, how they communicated it, and what behavior the robot produced in response. At scale, that generates the kind of multimodal human-robot interaction data that no simulation can replicate and that teleoperation labs collect only slowly and expensively.
Jacobi confirmed this explicitly. “We’re going to learn a lot about what is the right way to interact with robots,” he said. “Do we want to just talk to them over text or audio? Do we want to show them an example as a video? Or maybe do we want to tap, drag, and drop?”
The answer to those questions is not just a UX insight. It determines which input modalities the company’s foundation model learns to process, how reliably it can interpret human intent from natural communications rather than expert-coded instructions, and whether Enigma’s models can eventually generalize across the diverse hardware types and environments that real-world robotics deployment requires.
Founders: Outsiders With a Specific Pedigree
The company was co-founded by Jacobi and Gal Niv, both veterans of Israel’s Unit 8200, the Israel Defense Forces’ signals intelligence and cyberwarfare unit, which is regarded as the Israeli equivalent of the NSA and has produced the founding teams behind Wiz, CyberArk, and Palo Alto Networks, among others.
Jacobi began a computer science degree at 13, finished it while still in high school, and went on to become the youngest-ever employee at both Microsoft and Check Point — a position at Microsoft he reached through a direct recruitment by Asaf Rappaport, who later founded Wiz. He subsequently joined cybersecurity startup Dazz, which Wiz later acquired, before leaving last year to start Enigma.
Niv started hardware hacking at 10, joined a cybersecurity startup at 17, completed a four-year degree in a single year, and became Unit 8200’s youngest cyber-operations manager. The two met as teenagers in hacking competitions, then served together in Unit 8200. When they decided to build a company, they chose a field where neither had direct experience — deliberately.
Shardul Shah of Index Ventures, who led Enigma’s seed round, explained the appeal of that outsider status. “Someone who’s an insider may start with the capability of teleoperation or dexterity,” Shah told TechCrunch, “but Enigma is starting from a very different place: ‘What’s the ultimate experience?'”
The team Jacobi and Niv assembled includes alumni from leading AI labs, math olympiad winners, and researchers who left PhD programs to join the startup.
Where Enigma Sits in a $14 Billion Competitive Field
The hardware-agnostic intelligence layer is exactly where several of 2026’s largest robotics bets are concentrated. Skild AI raised $1.4 billion in January 2026 at a $14 billion valuation, building what it calls an “omni-bodied brain” — a single foundation model that can control any robot on any hardware without extensive retraining. Physical Intelligence, valued at approximately $5.6 billion, has trained its π0 model across seven robot platforms and 68 tasks. Both companies are approaching the intelligence layer from the capability side: they build AI that makes robots competent at diverse tasks.
Enigma’s differentiation is structural. Rather than starting from “how do we make robots more capable,” the company is starting from “how do we make capable robots usable” — and using the resulting human-interaction data to train its own foundation models. The press release states that Enigma’s proprietary AI models reduce reliance on massive manually labeled datasets while improving robot performance across physical settings. That claim, if borne out, would mean that Enigma’s public experiment is not just a user research project but a training data flywheel: more interactions produce better models, better models enable more useful robots, more useful robots attract more interactions.
The company has confirmed partnerships in healthcare, logistics, and entertainment, and plans to use the $71 million to grow its research and engineering teams, scale compute, and expand real-world deployments.
Travis Kalanick’s robotics company Atoms closed a $1.7 billion round led by Andreessen Horowitz just days before Enigma’s announcement, underscoring how quickly capital is concentrating in the physical AI space. The robotics venture capital market has set records at every measurement point in 2026, with $55.8 billion tracked globally through midyear by Dealroom — nearly double 2025’s full-year total.
How Does the robots.online Experiment Actually Work?
The robots housed in Enigma’s hangars in Israel and California are proprietary robotic arms built entirely by the company, including both the hardware and the underlying models, according to Enigma’s claims. The experiment allows any user, anywhere in the world, to connect to one of more than 100 of those robots in real time, direct it to complete physical tasks, and observe the results.
The specific tasks currently available — drawing with a paintbrush, sword combat with another robot, and simple chemistry (picking up and mixing flasks of liquid) — are chosen to require fine motor control and some degree of adaptive response to environmental conditions. They are not trivial demonstrations of pre-programmed behavior but tests of how well the robot’s AI model can interpret human direction and convert it into reliable physical action.
Enigma is testing four interaction modalities: text commands, voice instructions, video demonstrations, and tap-and-drag gestures. The goal is to determine which modality, or which combination, produces the most reliable and intuitive robot behavior — and, critically, which modality yields training data that is most useful for improving the underlying model. The experiment is accessible at robots.online.
Can You Visit Without a Robotics Degree?
The explicit design goal of the interface is accessibility. Enigma’s software is described as “robot-agnostic” — meaning it is intended to work across different hardware platforms, not just Enigma’s own robotic arms. The company’s thesis is that if the interface is right, users with no technical background should be able to direct a robot effectively. The online experiment is live evidence of how far that goal has been achieved.
Whether the interface-first thesis proves correct — whether a better volume knob can be found, and whether the data that experiment generates can train foundation models that are both capable and broadly accessible — is the open question that $71 million and 100 live robots are currently being asked to answer.
Frequently Asked Questions
What does Enigma mean by “robot interface,” and why does it matter more than capability?
A robot’s capability is how much it can physically do. Its interface is how a human communicates what it should do. Enigma’s argument is that even fully capable robots go unused when the communication between human and machine is clunky, slow, or unpredictable. A robot that can fold laundry but requires a 15-minute technical setup to receive new instructions will be abandoned for human labor. Enigma is specifically researching which input modes — text, voice, video demonstration, or tap-and-drag — make that communication natural enough that ordinary users don’t give up.
How is Enigma’s public robot experiment different from a product demo?
Most product demos are controlled to show a robot succeeding at a prepared task. Enigma’s robots.online is specifically designed to gather unscripted human-robot interaction data at scale. The modalities people choose, the instructions they give, and the points at which robots fail or users abandon an attempt are all training signals for improving the foundation model. The physical AI industry estimates that only around half a million hours of high-quality real-world robotic interaction data currently exists — far less than what general-purpose robot AI requires. Enigma’s public experiment is one of the few mechanisms that can generate that data without requiring a dedicated facility or trained operator for every session.
How is Enigma different from Skild AI or Physical Intelligence, which are also building hardware-agnostic robot AI?
Skild AI and Physical Intelligence approach the intelligence layer from the capability side: they train AI models that can execute a wide range of tasks across different hardware, focusing on what the robot can do. Enigma approaches the same layer from the human side: it researches how people naturally want to communicate with robots, then builds both the interface that enables that communication and the model that interprets it. The two approaches are complementary research directions, but Enigma’s public experiment gives it a specific data collection mechanism — real-time interactions from users worldwide — that capability-first competitors are not currently running.
Is Enigma’s technology available to buy or license?
As of July 28, 2026, Enigma is a pre-commercial research company that has confirmed partnerships in healthcare, logistics, and entertainment but has not announced a commercial product. The robots.online experiment is publicly accessible for free. The company has said it will use its $71 million seed round to expand research, engineering, and compute capacity and to accelerate real-world deployments with existing partners.
