A systematic fragment-based de novo design procedure was developed and applied to discover new potent and selective inhibitors of glycogen synthase kinase-3 beta (GSK3β). Candidate inhibitors were generated to simultaneously maximize the biochemical potency and the specificity for GSK3β through three design steps: identification of the optimal molecular fragments for the three sub-binding regions, design of proper linking moieties to connect the fragmental building blocks, and final scoring of the generated molecules. By virtue of modifying the ligand hydration free energy term in the scoring function using hybrid scaled particle theory and the extended solvent-contact model, we identified several GSK3β inhibitors with biochemical potencies ranging from low nanomolar to picomolar levels. Among them, the two most potent inhibitors (12 and 27) are anticipated to serve as promising starting points of drug discovery for various diseases caused by GSK3β because of the high specificity for the inhibition of GSK3β.