For some reason, I’ve always liked top lists. When I was a teenager, I used to listen to Casey Kasem’s American Top 40 every Sunday morning, usually while working on some programming project or other. In 2022, I had the idea for a daily game in which players were surveyed on one day and their preferences ranked to create a ranking game the next day. I didn’t get a chance to really dig into that idea until the holiday break at the end of 2025.
And now, after a good deal of iteration, the game is here and it’s called Peaklet1. Iterating on this has been a lot of fun and I’ve had a lot of good input from the initial players. Things like automatic booting of choices that aren’t in the top 5 came from feedback and make the gameplay quicker.
My thoughts on the survey to puzzle part of the idea also changed over time. I realized that one day of lead time on a survey wasn’t going to be enough and, more importantly, it’s kind of confusing. If you take a survey on Monday and Wednesday’s puzzle looks just like the survey you took, it feels like you already did the puzzle. With the current version of the game, surveys can go on for a period of time and then I’ll post the puzzle at the right time after the survey completes.
The making of Peaklet
I have a lot of experience as a programmer and some as a product manager as well. I brought that experience to bear in making Peaklet with the help of GitHub Copilot using mostly the Claude Sonnet 4.5 model, though I switched to Claude Opus 4.5 in the later part of development. I made heavy use of Beads for planning out the work and found that worked great. Copilot stayed largely on track throughout the project.
When I first started out, I was doing all of the work on my Mac, validating the commands the agent wanted to run before allowing them. That got to be annoying pretty quickly, so I migrated to a GitHub Codespace containerized dev environment and started running the agent in YOLO mode. The odds of the agent deciding to delete everything on my hard drive seemed pretty small, but why take the chance if I could run in a container?
The pace of development on this project was amazing. Being able to describe a feature and have it implemented (with tests!) twenty minutes later seems like it’s right out of science fiction.
Peaklet benefitted from being a greenfield project and not too much code, but the models are definitely way more capable than they were a few months ago and I’m increasingly able to get my work at GitHub done more quickly with the help of agents.
Agents help with iteration
I’m far from the first to make this point, but the ability to iterate quickly on ideas using agents is a real game changer for software development. That’s almost a bigger deal than the ability to build the final product with the agents, because it doesn’t matter how great the engineering is if you’re not building the right thing to begin with.
I have no idea how many other people will enjoy a game like Peaklet, but I do know that Peaklet as it exists today is much better than my original concept for it because of the iteration I was able to do.
Check it out!
Give Peaklet a go and see if you like it! Don’t miss responding to the surveys after you complete the daily puzzle so that you can influence the puzzle coming up.
Footnotes
-
My name for the game in 2022 was Rankle, which is an objectively better name. Unfortunately, someone came along earlier in 2025 and made a game of that name. ↩