At MJBizCon 2025, Sorting Robotics unveiled Stardust 2.0, the latest version of the company's automated kief-coating system. According to Nohtal Partansky, company founder and CEO, the newest iteration has a lowered design to make it more accessible, runs about 20% faster than its predecessor and is much more serviceable and easier to clean.
The Stardust 2.0, which is scheduled to start shipping by the end of Q1 2026, is a much more rigid system. Partansky said the new machine is a full weldment, or has a fully welded assembly, which enables the robotic arm to move much faster. Sorting also has customers running several machines across multiple shifts, which made cleanability a priority for the latest iteration.
The new pre-roll coater doesn't have any intrusions or grooves, making it easier to clean and preventing entrained or entrapped particles. Sorting also simplified the internal wiring.
While materials like stainless steel and aluminum have remained readily available, Partansky did note that costs have increased dramatically, doubling in some cases.
Sorting's showstopper in Vegas was Telti, a foundational robotics model system built for unstructured assembly. Telti is a robot that can take random inputs, like a pile of joints in its field of vision, infer and understand them, and organize them into finalized outputs. For example, Telti's first application is picking up joints and putting them into tubes. The robot knows how to identify the joint, pick it up by the crutch (filter or tip) and put it into a tube.
Next, Telti will be trained to assemble five-pack jars and tins, and to put doob tubes and vape cartridges into mylar bags.
"The application for unstructured assembly in cannabis, and frankly outside of cannabis, is really high, especially when you run into these high-SKU, low-to-medium volume scenarios, which is all of cannabis," Partansky said.
Telti isn't just a robot using computer vision, but a foundational robotics model that can execute inference, recognizing and deciding what to do with an object based on training and instruction.
"You can't solve unstructured assembly problems with classic automation; you can only solve them with robotics, because it requires decision making," Partansky said.
