Most people use a fraction of what their AI can do. Tick off what you have already done, and use the gaps as your to-do list. It saves as you go.
Two different dials, often confused. Choosing a model sets the underlying brain, bigger models are more capable but slower and pricier. Turning on thinking (also called reasoning or extended thinking) tells the model to work through a problem step by step before answering. A fast model with thinking on will often beat a bigger model with it off for tricky, multi-step tasks.
For quick drafting, rewriting and everyday questions, a fast model with thinking off is snappy and cheap. For analysis, planning, maths, code and anything where the reasoning matters, reach for a stronger model or switch thinking on. If an answer feels shallow, that is usually the dial to change.
A skill (also a custom GPT, a Gem, a project instruction, depending on your tool) is a saved setup for a job you do often, so you do not rebuild it each time. Your custom instructions are the simplest version of this. PromptPro is the layer that writes the per-task prompt on top.