We present a new molecular geometry optimization package which is especially robust for transition-state optimization. Key innovations include a new gradient-based trust-radius update, a robust way to define dihedral angles that avoids problems with near-linear bond angles, a way to selectively enhance the accuracy of the Hessian for key internal coordinates, and a robust way to select internal updates and update the Hessian expressed therein. By prioritizing software modularity, it became easy to support new features, allowing the \(\mathrm{G}\)eometry \(\mathrm{Opt}\)imizer package to evolve into a \(\mathrm{G}\)eneral \(\mathrm{Opt}\)imizer. \(\mathrm{Gopt}\) now supports an arbitrary number of coordinate transformations (allowing problems to be expressed in the coordinates for which the optimization/constraints are most efficiently expressed), myriad quasi-Newton methods, and alternative convergence criteria. This allows us, for example, to use \(\mathrm{Gopt}\) in wavefunction optimization applications. In preparation for As \(\mathrm{Gopt}\) 's imminent release, this talk will discuss key scientific innovations and software architecture decisions, while providing examples of use cases, especially those where \(\mathrm{Gopt}\) is preferable to competing software.