I had to plan a bachelor party in Atlanta for a group of 17, and I kept not doing it.
The task I kept avoiding
The planning had a lot of moving parts: restaurants that seat 17 or 18, one standout activity worth flying in for, somewhere outside to throw a ball around, a dinner and a plan for each night, transportation for a big group to a stadium with no train to it, and the one fixed point on the schedule, a Saturday night game. Every version of it in my head was a Google Doc or a spreadsheet. I opened one a few times, wrote a couple of lines, and closed it. The work was not hard, but sitting down to start it was.
Building it as a Game Boy instead of a doc
Instead of a document, I had Claude build the whole thing as an interactive web page styled like an original Game Boy: green screen, pixel font, a working D-pad and A/B buttons.
The itinerary lives inside the screen. Each day is a set of time slots, and each slot has a few options I tap to pick. The picks save and build into a plan. Saturday morning has a blocked-out recovery window before the night game. The finished page is one file I can send to the group as a link.
What Claude did
The research came first. Claude ran five searches at once, each on its own part of the trip: group restaurants, the standout activity, the game and stadium logistics, lodging, and nightlife. Each came back with real names, capacities for a party of 17 or 18, prices, and booking notes. I dropped those into the day options.
Then it built the page: the Game Boy shell, the tabbed screens, the pick-and-save logic, and the styling to match my site. It pulled the details of the Airbnb we had already booked off the listing and wrote them into the page.
It also built a second thing I did not plan on. The boot screen has two buttons. One opens the itinerary. The other opens a playable baseball game, a pinball version like the old wooden tabletop toy, where you pull back a plunger and the ball loops around the field into a hit or a home run. The first version had a bug where the ball never reached the field. I had Claude open the game in a real browser with Playwright, watch it play, find the bug, and fix it.
The finished page runs on my own site behind a password, unlisted, so only the group can open it.
What I took from it
This was a fun way to approach something that felt daunting, and I finished it.
What stuck with me is that I can do the same with other tasks like it. The task does not have to be serious. If turning it into a video game is what gets me to sit down and do it, that is a fine trade. AI is what makes the trade cheap. Building a custom Game Boy around a chore used to cost more time than the chore itself. Now it costs a few prompts, so it is worth doing any time a task is boring enough that I would otherwise avoid it.
I have a list of things I keep putting off. The next boring one, I will try building into a game before I force myself through it.