Like a lot of people into credit card rewards, I started with a spreadsheet. Which card earns what at which store. Rotating quarterly categories. Annual fee offset calculations. The works.
I updated it diligently for about two months. Then I stopped. Real life doesn't wait for you to open a spreadsheet before you pay at a restaurant.
I started building a prototype on weekends — just a simple lookup tool. Type a merchant, see which of your cards earns the most. It worked well enough that I started using it myself every day.
Then I added AI. Not as a gimmick, but because "which card for Costco gas?" is genuinely the kind of question you want to ask in plain English and get an immediate answer. The AI layer (Cleo) knows your wallet and can reason across your cards, categories, and offers in a way no spreadsheet can.
Today RewardSmart supports 600+ cards, has a full MCP server so you can ask reward questions from Claude or ChatGPT, and earns users an average of $300+ more in rewards per year. It's still just me building it — and I plan to keep it that way as long as it stays focused and useful.