I spent ~44 hours routing little men in rooms for ICFPc 2026. I had a good time.
See my repo for more technical report: https://github.com/uguu-org/icfpc2026
Or read on for more random thoughts related to my contest experience.
This year's contest was reminiscent of the 3D problem from ICFPc 2024, and that was my favorite part of ICFPc 2024, so this year was entirely my favorite bit from beginning to end. Not only that, the organizers kindly provided a debugger for running the programs, and it's the best kind of debugger with time rewinding capabilities. It was just a really great experience.
I was honestly really looking forward to this year's contest, much more so than previous years, once I found out that Nolen Royalty was an organizer. Nolen had many game development related articles made their way to Hacker News, so I was quite hyped to experience another one of his creations. The fact that it felt similar to ICFPc 2024 was even better, although this appears to be just a coincidence.
Regarding the time spent, ~44 hours is roughly the same range as the amount of time I spent on previous contests, although this year I put in a bit more hours in the lightning round (~17 hours) compared to previous years. For the individual problems:
The numbers don't add up to 44 since I spent a bit of time reading contest specs and writing infrastructure scripts, but I didn't do much coding in conventional languages since most of the solutions were implemented directly in the quirky contest language. I did write a submission script and a few prototypes, all in Ruby. This is one of the rare ICFPc years where I didn't write any C++ at all.
Regarding routing the little men... It was a challenge to fit all the desired functionalities in minimal area, and it was a headache to keep track of all the overlapping routes. If I had more time (like, a few weeks), I would implement a conventional-language-to-little-men compiler. Within the limited time available, I was mostly doing that compilation manually, with VIM.
Despite all the layout headaches, it was obvious that there are endless optimization opportunities for this kind of problems, and that was just something that was really fun to grind. I was overjoyed when I finally found the optimal solution for Triangle. It's the kind of fun that can only be had after taking the effort to do things manually.
After the standings are frozen, uguu.org appears to have landed in 109th place (out of 268), which is somewhere above the middle, and that's where I always strive to be.
It was a great experience, many thanks to the organizers for another great year!
Previous (2026-07-03): Kadokawa Megumin DX