The full catalog — from underwater autonomy and flight avionics to research tools, web utilities, music tech, and games.
A search problem is a search problem whether it's routing an underwater robot or picking piano fingerings. I look for the structure under the vocabulary, then reuse what already works in another field.
Invention alone rarely changes anything — electricity mattered once people lit their homes with it. I'd rather build tools people can actually pick up, explore, and fold into how they work.
As AI makes building cheap, the hard part is deciding what's worth building. Clear problems, surfaced assumptions, and a sharp question are worth more than raw output.
A linear scan beats a clever index on small data; a spreadsheet beats a database for a few dozen rows. I start with the simplest thing that could work and add complexity only when reality demands it.