A two-code CPC is a gift: it tells you precisely where the inventor thinks the novelty lives. Mitsubishi's grant US11104008B2 ("Dexterous gripper for robotic end-effector applications," issued August 31, 2021; inventors William Yerazunis, Parthasarathi Ainampudi, Nakul Gopalan) classifies only into B25J 15/0233 and 15/0266 — both finger/gripping-head mechanism codes. There is no perception code, no learning code, no control code. This is a mechanical claim about how the gripper's fingers are constructed.
That purity matters because dexterity in grasping is, at the limit, a mechanism problem before it's a software one. A gripper that can hold varied objects securely with simple control is doing the work in its kinematics and compliance, not in a neural net. The B25J 15/02xx classification — gripping heads with movable fingers — points the reader straight at the finger geometry and actuation as the inventive core.
For the manipulation beat, the reading discipline here is to ask: broad foundational finger mechanism, or narrow improvement on an existing design? A claim classified this tightly is almost always the latter — a specific finger arrangement or compliance feature — because the B25J 15 space is old and dense. The word "dexterous" in the title is aspiration; the granted scope is whatever the two finger codes' worth of structure claim 1 actually recites.
From a portfolio angle, MERL files research-grade mechanism patents that often seed later product or licensing activity. A dexterous-gripper mechanism grant is the kind of asset that matters if Mitsubishi pushes into manipulation products or licenses to integrators. It's narrow, but mechanism patents are also among the hardest to design around — you can't software your way past a claimed finger geometry.
Caveats. Gripper mechanics is a centuries-deep field once you count general mechanical grippers; the grant must clear that, so claim 1 is specific. The two-code classification tells you the claim is mechanical and finger-centric — it does not tell you the exact structure. Read the independent claim for the recited finger arrangement.
For the file: a tight, mechanism-centric dexterous-gripper grant from a research lab. Pull US11104008B2 on PatentBear and read claim 1 for the specific finger structure — in a mechanism claim, that geometry is the whole asset and the whole defense.