Defining the problem, scope, and requirements are my first steps to starting a project. Without knowing those three basic elements, there's no foundation to move forward on.
From there, I'd move into an ideation phase. Concept exploration sketches on napkins, whiteboards, sticky notes - whatever is available. Low fidelity is a personal must at this point. I believe that diving into more polished end-stage visuals at this stage is a fantastic, beautiful waste of time. Motion, transitions, and state changes are good to keep in mind throughout this stage, but the real purpose is to come up with a set of structured states that can be tested and iterated upon. Culling some of the more "out there" designs, I'd select a few front-runners to proceed with.
Visual design follows ideation. Exploring different visual executions of a structured UI is, in all honesty, a passion of mine. I love tinkering with visual layouts and exploring new ways to display information. Prototyping motion and state transitions can help fine-tune or rapidly stress-test designs at this point.
As a final part of the iteration loop, designs get tested against one another with people who have never seen them before. If no unbiased eyed are around to test a prototype, or if time doesn't allow for it, in-team testing and open critiques can help kill proverbial darlings and ultimately validate or invalidate design assumptions.