Ralf Gulde's Recent LinkedIn Posts

DeltiLog GmbH is going dual arm. Double digit single arm systems already live across the site. Now dual arm. Plus kitting, returns, and value added services. Returns tells you everything about a system. Every item comes back different. Damaged boxes. Missing components. Products that don't match the manifest. You can't pre-program your way through that. Kitting is the same kind of hard. High mix. Constant changeover. Nothing repeats. All running in production. Next phase already in planning. A partner doesn't go from one workflow to six because a demo impressed them. They go because it held up when the work got ugly.

A robot picks a damaged box from a cluttered bin under bad lighting. The clip does well online. Three months later, the same company installs at a second site. New products. Different bin geometry. Different light conditions. The integration takes almost as long as the first. The robot didn't get smarter. It just performed. For a long time, that was fine. Progress in robotics meant a better machine. Faster arms. More accurate grippers. Cleaner demos. And that framing made sense when hardware was the bottleneck. Hardware isn't the bottleneck anymore. The teams that struggle most right now have impressive robots that need to be re-taught at every new location. New SKUs. New packaging. New edge cases. Each deployment starts close to zero. The unit of progress used to be the robot. Now it's the loop behind it. When a fix deployed at one warehouse propagates to every warehouse. When a new SKU learned at one site generalizes across all sites. When the tenth deployment converges faster than the third because the system already carries the exposure. That's not a better robot. That's a different kind of progress entirely. The robot becomes a sensor and an actuator. Important. But not what compounds. What compounds is the intelligence underneath. Execution data flowing back from every pick, every retry, every recovery. Corrections accumulating into a system that doesn't reset when the next deployment starts. Most architectures reset. New site, new tuning, new effort. The robot doesn't carry what it learned last month at the other facility. The ones that don't reset are the ones where progress compounds. Once the loop is live, every deployment makes the next one faster. Every edge case encountered anywhere becomes knowledge available everywhere. The system gets denser. Not just wider. Not a better robot. A system that improves every time it runs. And once that loop is turning, it doesn't stop.

Our office is 20 minutes from Messe Stuttgart. LogiMAT is the one week a year when 65,000 logistics professionals come to our backyard. Most robotics companies fly across the world to show their work at trade shows. We drive. That's not a coincidence. Stuttgart shaped this company. The warehouses we first deployed in were within driving distance. The operators who taught us what actually breaks in production were local. The problems we solved first were German logistics problems, which are some of the most demanding in the world. Next week we're showing Cortex across multiple robot form factors on the floor. Single arm. Dual arm. Same AI. No teleoperation. No pre-recorded clips. Live picks on real items. Most booths at logistics trade shows play videos. We run robots. There's a difference between showing what your system did once in a controlled environment and letting 65,000 people walk up and watch it work in real time. One is marketing. The other is a statement. Hall and booth number in the comments. Come see it break or come see it work. Either way, you'll know what's real.