.jpeg)
IntuiCell is building a new kind of software brain that lets any machine (physical robots or digital systems) learn in real time, directly in the physical world. It’s a biologically inspired “digital nervous system”, that doesn’t require pretraining, massive datasets, or millions of simulations to learn and adapt. CEO and co-founder Viktor Luthman explains why starting from a spinal‑cord‑like learning core can unlock the next paradigm of intelligent machines, and why Europe should bet on bolder visions.
What is IntuiCell and what are you building?
IntuiCell builds the first software that allows machines to learn continuously in real time, directly in the real world. We’re the brain people, not a robotics company, the robots are merely vehicles to intuitively demonstrate what our brain can do.
What drew you from tech commercialization into deep science ventures?
Personally, I’ve always been passionate about scientific breakthroughs that can transform how we live and work. In my previous venture, I worked with top bacteriology and immunology researchers who had fascinating discoveries but limited understanding how to build a company around it. Being “the stupidest person in the room” at the science–market intersection, building a company to commercialise truly groundbreaking science, is addictive.
How did you come in contact with the team behind IntuiCell?
The spark came when a former colleague-who had become the head of the tech portfolio of Lund University holding company called about “the most interesting case he had ever seen on his desk”: a group of neurophysiologists with new discoveries about the mechanisms of how the brain predicts the world, and a mission to build “AI like the human mind”. The company launched in late 2020, the first hire (Linus Mårtensson, tech founder) joined the scientific founders to translate research into C++. I joined in January 2021. Intelligence was never our end-goal, something we hoped would magically emerge by stacking more GPUs on existing blueprints. From the very first neuron we built, we made intelligence the starting point. Because it was necessary. And because finally, we knew how.
What was the early “it works” milestone?
We wanted proof of real‑time learning from scratch in the real world. Using an inverted pendulum robot, we watched it fail to balance-until it suddenly learned, using the same principles biological creatures use. It was a pivotal moment. The system adapted online, not from a static, pretrained model.
How is IntuiCell different from contemporary AI?
Modern AI trains on massive datasets or simulations with backpropagation-learning by being told the correct answer to every input. That’s brilliant for static pattern analysis, like protein folding. But backprop is fundamentally incompatible with self‑driven, real‑time learning. In nature, there’s no external teacher at every step. For decades, the AI industry has been busy building static intelligence for an extremely dynamic world. IntuiCell changes that, and opens the door to machines that learn as they act, without pretraining.
Why start with a “spinal cord”?
Our first functioning module replicates the functional role of a biological spinal cord: low‑level reflexive control for movement and body understanding, adapting to environmental change, and handling unforeseen problems as they occur-without a big predictive “cortex” on top yet. In and of itself, it’s a game changer for physical robotics (in essence, the missing link for enabling reinforcement learning in the real world) and has strong relevance in digital systems as well, for example anomaly detection, as it can learn the “normal behaviour” and adapt as that normal shifts over time.
What did Luna, the robot dog, prove?
Luna is an off‑the‑shelf quadruped, a Unitree Go2. We didn’t touch the hardware. Inside however, she runs the first working module of our digital nervous system (our reflexive, low-level control system) and learns from scratch in her environment. Before Luna, it was harder for partners to frame what our new kind of brain could do. After Luna, we now have visionary industrial partners call with their own use‑cases: “Could your brain do this on our robot?”. That’s the point-we’re building a universal learning architecture, an omni-brain, just like in biology. The same principles scale across bodies and tasks.
So IntuiCell is infrastructure, not an app?
Yes. Think of deep learning 20 years ago. We see ourselves as analogous infrastructure for the next technological revolution. But our tech needs to mature, and we have to earn our right to be a platform. For that reason we will initially do bespoke, flagship deployments, however, long-term, we aim for a community of smart people and companies building solutions on top of our platform. Our vision is the original aspiration of intelligent machines from the 1950s-realized in a decentralized, highly efficient intelligence that lets agents fulfill their purpose in the world.
Why is this crucial now for robotics and industry?
AI is largely stuck behind screens-performing narrow tasks in rigid contexts-while the world is (extremely) dynamic. If we want to work with intelligent machines, those machines must operate and thrive in real environments. Think of training a service dog to understand intent: different tasks need different levels of embodied intelligence. We’re not building an all‑powerful cloud brain; we’re building on‑agent intelligence tuned to each form factor and purpose.
Europe needs bolder visions-how does Intuicell embody that?
If this had come out of Stanford or MIT, the starting line with investors would be different. Many early conversations said, “Cool tech, but you’ll never make it from Lund.” We embraced the underdog spirit. Our edge is a unique understanding of how learning really happens in biology, rooted in a cluster of exceptional researchers at Lund University-and the ambition to light a new tech paradigm from here.
What’s contrarian about your approach-and is that a risk or an asset?
Our entire DNA is contrarian. Professor Henrik Jörntell and his neurophysiology lab spent 35 years overturning common views of how the brain works. We’re not tweaking deep learning or reinforcement learning. We’re rebooting the very foundation of how machine intelligence is built. Coming from neuroscience, not from AI, makes us the oddbird in the mix. It’s an asset, but we also have to educate. We’re not “a better GPT.” We’re a new kind of intelligence, and we’re still early, so problem complexity today doesn’t match LLM expectations.
Recent breakthroughs you can share?
About six months ago we showed a robot that learned from scratch to stand and balance, then adapted online to ice and rocks. Now in the office, our system is learning to take its first steps-proving autonomous, adaptive locomotion, and also working on solving dexterous manipulations, one of the critical bottlenecks for robotics today. It’s surreal to see “science fiction” turning into functioning software you can watch learn in front of you.
Where is this headed-humanoids or beyond?
Physical AI will find the right brains for the extraordinary bodies we already know how to build. Today’s demos are impressive, but still the trailer is a lot cooler than the three‑hour movie - because the real world is hard. In the future, intelligence will appear in many form factors, including ones we haven’t imagined. Initially, humanoids will slot into factories designed for humans; later, we’ll design factories for intelligent machines-unlocking new efficiencies and value.
What excites you most as CEO right now?
The team’s energy tackling problems without an answer around the corner-turning our understanding of brains into software that works. Balancing the cart‑pole in 2023 was pivotal; since then, we’ve placed our brain into new robot forms, including humanoids and a learning quadruped in our office. At the moment, progress feels slow; looking back, the steps are staggering.
If you’re a CTO at a global robotics firm-why Intuicell?
Because we bring machine intelligence into the real world. Our software lets your machines learn from scratch, adapt online, and handle unforeseen problems-without pretraining or brittle assumptions. It’s decentralized, efficient, and morphology agnostic, from robot dogs to humanoids to industrial systems.
One line that captures IntuiCell’s ambition?
We build adaptive systems inspired by biology that learn, move, and act with the autonomy once imagined, bridging the gap between artificial and intelligent.
Thank you to Viktor Luthman for sharing Intuicell’s journey and bold vision with us. For more Founder Series interviews and insights from the Nordic startup ecosystem, visit Luminar Ventures’ News & Insights.