GelSight is a vision-based optical tactile sensor developed by researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL). Unlike traditional tactile sensors that measure contact force, GelSight primarily measures geometry with high spatial resolution. The sensor consists of a soft elastomer surface that deforms when contacted, and an embedded camera captures the deformation to reconstruct the 3D geometry of the object. The sensor can also infer contact force and slip from the deformation. This paper discusses the development of GelSight, focusing on its sensing principle, sensor design, and hardware and software support for its application in robot hands. The authors introduce the optical system design, algorithms for shape, force, and slip measurement, and the fabrication processes for different sensor versions. Experimental results demonstrate the sensor's performance in measuring geometry and force, and its successful application in various robotic tasks, including material perception, recognition, and in-hand localization. The paper also highlights the challenges and advancements in tactile sensing for robotics, emphasizing the importance of geometry sensing alongside force sensing.GelSight is a vision-based optical tactile sensor developed by researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL). Unlike traditional tactile sensors that measure contact force, GelSight primarily measures geometry with high spatial resolution. The sensor consists of a soft elastomer surface that deforms when contacted, and an embedded camera captures the deformation to reconstruct the 3D geometry of the object. The sensor can also infer contact force and slip from the deformation. This paper discusses the development of GelSight, focusing on its sensing principle, sensor design, and hardware and software support for its application in robot hands. The authors introduce the optical system design, algorithms for shape, force, and slip measurement, and the fabrication processes for different sensor versions. Experimental results demonstrate the sensor's performance in measuring geometry and force, and its successful application in various robotic tasks, including material perception, recognition, and in-hand localization. The paper also highlights the challenges and advancements in tactile sensing for robotics, emphasizing the importance of geometry sensing alongside force sensing.