Spranic builds spatial memory for physical AI

Munich-based Spranic is betting that robots need more than a script. The company is building a spatial memory layer for physical AI, one that records places and how those places change over time. Jordi Spranger, Spranic’s founder and CEO, says that context is what lets machines move beyond fixed instructions and into something closer to real adaptation.

His pitch is simple enough to hear once and then keep hearing everywhere: robots need the right context at the right time. Spranger told theCUBE that the same idea already works for digital agents, and that the missing piece is bringing it into the physical world. In his framing, maps are central. Not nice-to-have central. More like the thing standing between a robot that can only follow a plan and one that can adjust when the room changes.

That is where knowledge graphs enter the picture. Spranger says they can tie together different sources of information while preserving provenance, so the robot can know where a piece of data came from before acting on it. He argues that matters because robots operate across many scenarios, and those scenarios pull in many domains. Graphs, in that view, are the glue that lets useful sources in and keeps irrelevant ones out.

There’s also a practical angle here that sounds a lot less flashy than “AI for the physical world” and a lot more useful. If a robot can reuse spatial context across tasks, it needs less task-specific planning each time it deploys. Spranger even compares the goal to “Google Maps on steroids,” a system that helps a humanoid replan across many tasks instead of being trapped in one statically planned routine.

The other promise is accountability. Spranger says robots need to work in an auditable and safe way, and that provenance helps make that possible. That is the real story here: not just smarter robots, but robots whose decisions can be traced back to the information that shaped them.