Tacta Systems

There is a peculiar embarrassment at the center of the artificial intelligence boom, and Tacta Systems has built a company around it. The embarrassment is this: a modern AI model can pass the bar exam, write a sonnet, and summarize a legal brief, but ask it to pick up a strawberry without turning it into jam and it has no idea where to begin. Text is solved. Video is more or less solved. The physical world - the part where you have to actually touch something soft and not crush it - remains, in the words of Tacta's co-founder and CEO Andreas Bibl, largely "incomprehensible" to machines.

Tacta Systems, based in Palo Alto, thinks this gap is not a footnote but the whole ballgame. The company is building what it calls Dextrous Intelligence, which it describes - and this is the phrase that appears in every one of its announcements, so we may as well use it - as a "smart nervous system" for robots. The pitch is that intelligence, in the way most robotics companies have been chasing it, has been all brain and no hand. You can bolt a genius language model onto a robot arm and the arm will still fumble a bolt if the bolt is a millimeter out of place. Tacta wants to supply the missing part: the sensing, the reflexes, the fine motor adjustments that a human does without thinking and a robot, so far, cannot do at all.

If you want to understand why anyone would give this idea $75 million, it helps to know who is asking. Bibl is not a first-time founder with a slide deck and a dream. His previous company, LuxVue Technology, built microLED displays and was acquired by Apple in 2014 - the kind of exit that buys a founder the benefit of the doubt on his next hard, expensive, physics-heavy idea. Robotics is nothing if not hard, expensive, and physics-heavy.

The millimeter problem

Here is the thing about factory robots that nobody outside of a factory tends to appreciate: they are astonishingly good at doing exactly the same motion, in exactly the same spot, ten million times, and astonishingly bad at anything else. Move the part slightly. Change its shape. Ask the robot to adapt. The whole illusion of competence collapses. This is why so much of the physical economy - warehouses, assembly lines, the delicate and repetitive labor of making things - is still done by people, often doing work that is dull, physically punishing, or both.

Tacta's bet is that the way out is not a bigger brain but a better body - specifically, a sense of touch and the reflexes to use it. A robot that can feel how much force it is applying, sense when an object shifts, and adjust in real time is a robot that can work in the messy, variable physical world rather than the pristine, pre-arranged one. The company frames its mission in refreshingly unglamorous terms: automating "much of the drudgery of factory work and grueling physical labor." Not humanoid butlers. Not robot uprising. Just the tedious stuff nobody wants to do, done reliably.

Why touch, and why now

There is a broader current in AI research that makes Tacta's timing interesting. The field has become obsessed with "world models" - systems, pursued by the likes of DeepMind and Meta, that try to give machines an internal understanding of how the physical world behaves. Tacta's argument, essentially, is that a world model without a sensory layer is a map without a compass. You can model the world all you like, but if the robot can't feel what it's touching, it can't act on the model. Tacta positions its tactile intelligence not as a nice-to-have accessory but as a prerequisite - the thing that has to exist before robots can be deployed at scale anywhere that isn't perfectly controlled.

Whether that framing is exactly right or merely mostly right is the kind of question that $75 million is meant to answer. What is clear is that the investors buy it. The syndicate is unusually global for an early-stage hardware company, which tends to signal that strategic players - the ones who actually build cars and move goods - see something they want a seat next to.