Read the classification before you read the title. NASA's grant US12649233B1, "System and methods for robotic grasping" (issued June 9, 2026; inventors William A. Watson, Orlando E. Gordillo, Brady T. Campbell, assigned to the United States as represented by the NASA Administrator), carries exactly two CPC codes — and that economy is itself informative. B25J 9/1664 covers program-controlled manipulators that use sensory feedback; B25J 9/1697 covers manipulator control using vision. The fence is drawn around sensor-and-vision-guided grasping, not grasping in general.
The way to think about this: a grasp is a control problem, not a gripper problem. The hardware that closes is commodity; the hard, protectable part is the loop that perceives the object, plans the contact, and adjusts as the fingers close. The two B25J subclasses on this grant point straight at that loop. When the claim says grasping, it means the closed-loop decision process — which is where the engineering value, and the defensible scope, actually sit.
Note the assignee. This is federally funded IP, held by the U.S. government. That changes the strategic read versus a Figure or a Boston Dynamics filing: government-held manipulation patents tend to be licensable and are less about exclusion than about establishing a public record of a technique. For a portfolio analyst, a NASA grasping grant is less a competitive weapon and more a prior-art anchor — a dated, examined description of vision-and-sensor grasp control that any later commercial claim has to reckon with.
That prior-art function is worth dwelling on. Humanoid and warehouse players are filing aggressively on manipulation right now. A granted government patent in B25J 9/1664 and B25J 9/1697, dated June 2026, becomes a reference point: a later applicant claiming sensor-feedback or vision-guided grasping has to distinguish over it. The narrower the NASA claims, the more room competitors have; the broader they are, the more this grant constrains the field. Either way it's now on the board.
The caveats are the usual ones, applied honestly. A two-CPC grant is not automatically narrow — classification breadth and claim breadth are different things, and you have to read the independent claim to know the metes and bounds. And a granted method is not a deployed system; NASA describing a grasp-control technique is not NASA shipping a warehouse robot. But the dating and the classification are exactly the kind of verifiable anchor this desk trusts over a press release.
The takeaway for the manipulation beat: when you map who owns grasp control, government and university grants belong on the map alongside the startups, because they shape what the startups can claim. This NASA grant, deep-linked on PatentBear, is one of those shaping references — small in CPC footprint, potentially large in its effect on everyone else's claims.