PHANES AI, founded by 28-year-old Harbin Institute of Technology Shenzhen professor Yang Shuo, has published TouchWorld, a predictive and reactive tactile foundation model for dexterous manipulation. The model addresses a blind spot in current VLA vision-language-action models, which can see a robot finger touching a button but cannot feel if it has been pressed. TouchWorld gives tactile sensing a dual role: Predictive, where before executing an action it predicts a tactile map indicating which finger should experience pressure, contact location, and intensity; and Reactive, where it continuously reads tactile signals and joint states to apply micro-corrections in real-time. In real-robot testing across six tasks, TouchWorld achieved 65.0% average success rate in undisturbed conditions and 57.2% with disturbances, exceeding the strongest baseline by 15.7 and 16.0 percentage points respectively. Each task was trained on 200 teleoperation trajectories and evaluated over 100 real-robot trials. Yang Shuo, born in 1998, is a tenured professor and doctoral supervisor at HIT Shenzhen. PHANES AI’s position is that tactile feedback cannot simply be added as another data modality to existing VLA models; instead, TouchWorld uses a separate architecture where tactile prediction and reaction operate alongside the VLA pipeline.