Yes, that surprised me, too. Cliff detection isn't that hard. They seem to have cameras in the nose area that are positioned to see whether there's a walkable surface ahead. The basic vision system is cameras with IR projection of a pattern, like the original Kinect. How can they do foot placement on stairs while unable to detect cliffs?
There's an optional LIDAR vaguely mentioned in the instructions. No specs on field of view or range, though.
"Spot’s joints can pinch fingers and other body parts and entangle loose clothing, long hair, and jewelry. Only handle Spot when the motors are locked out or robot power is off. Fingers may break or get amputated if caught in joints while Spot’s motors are active." Were pinch points not considered during design?
Jogging/trotting is only available in "demo mode", and is listed as "less stable".
Boston Dynamics can do better than this. I was expecting full Big Dog capability in a smaller package. Maybe this is just the minimum viable product for a market test. If someone finds a use case for this robot, the next model should have those problems fixed.