AI Update: Work has fully embraced AI and is encouraging us to use it as much as possible — which honestly works in my favor. This project has a ton of new moving parts, and without AI helping me troubleshoot and crank out code, I’d be buried. Even naming variables slows me down; I swear every name I pick makes me cringe later when I revisit my code.
The nice part is I find myself adding features I would normally push to a mythical ‘future update’ that never actually happens because other new features keep coming down the pipe. I’m almost out of Claude tokens, so I’m trying to ration them as much as possible. If I have any left at the end of the month, I’m unleashing Claude on all my ‘any’ types or telling it to generate any missing unit tests.
GPT‑4.1 has been surprisingly helpful — it even solved a couple of things Claude got stuck on — but sometimes it just can’t see the path forward. That’s when spending a few Claude tokens saves me from explaining the same thing to GPT for three hours straight.
On a side, I’m working on a game with FU and Gemini. It’s coming along, but it still needs sound effects and some cleanup on the intro and outro screens. Progress was slowed by Gemini's default ‘clean and optimize’ mode. I’d ask it to change a color, and suddenly my file goes from 800 lines to 273. Apparently you have to start your request with ‘zero-refactor’ if you want it to behave. I haven’t done major updates since learning that, so I’ll test it this week — assuming I’m not still working until 9:17 at night again. Also, Gemini seems to get really slow when you keep cutting and pasting 800+ lines of code into it. I tried adding Gemini to Visual Studio Code but that was even slower.