I'm trying a different format for this check-in. Instead of writing this as a formal blog post, I used GPT-5.5 to generate questions for me and then answered them for the purposes of this update.
For reference, the goals I set in January were:
Yes, they still feel like the right goals.
I feel like I've made the most progress in producing more and consuming less, I've taken an active interest in building my own games using Claude Code and Codex. I submit them to game jams on Itch.io. One particular game that's worth checking out is GRB, which was a NES game I made with Claude Code.
GRB felt especially worth mentioning because I used a trick I didn't expect to work. Namely, that I would tell Claude Code to leverage a NES emulator to do end-to-end tests. While doing so, instead of loading the NES emulator in the browser and playing the game (taking screenshots while doing so), it instead just loaded the game into the emulator and monitored the memory addresses to confirm that the game was operating as expected.
To me, that was super cool! I wasn't expecting it to use a trick like that at all. It taught me that building games for emulators is a really interesting way to help agentic coding tools like Claude build games since the restricted environment makes testing and writing code easier for them in some ways.
Pebble was very cool, it was the first time I discovered that browsers can run LLMs and run LLMs that are actually okay at generating little snippets of text.
I believe that's self-explanatory, they connect back because I'm building something. That something is admittedly not particularly productive but it's really cool to make these games and they remind me of what I originally enjoyed about programming when I was a kid. Making cool things.
I have a Marimo notebook I made for doing coast FIRE projections. It was a good exercise running through that with Claude Opus 4.8. I think there's some more room for work around pulling my investment account information and getting the numbers checked by a financial planner but I'm glad I finally have a tool to help me with financial projections and that I'm building out a process for doing this regularly.
With regard to bespoke software outside of this domain, I have had less luck although I have been playing around with making Telegram chatbots for managing my website and managing my emails. I think this is pretty rich ground although I haven't gotten results I'm proud of yet.
I haven't made much progress since last time, I started Vimy by Pierre Berton since I haven't read any Canadian history in a while. I figure I should do at least a bit.
I was in Vienna last month, it was lovely! I found it especially moving to be in the crypt that appears at the end of Emperor's Tomb (the sequel to Radetzky March).
No, it was just a personal/ literary connection.
I have made very little progress on reading up on LLMs, there's more to do there. Same with improving my climbing, I remain on the plateau there.
It slipped my mind.
Change into something lighter like this Q & A. I've noticed that for all their faults, I enjoy talking to the bots like you. I find it pleasant. Since clearly writing doesn't come naturally to me, perhaps answering questions from you is sufficient for our purposes here.
That AIs make things much easier to do but that doesn't mean that I don't have to take the initiative to do things and that doesn't mean that I shouldn't think about how to make things more efficient and natural for myself. Writing a formal blog post feels like a chore. Perhaps this Q & A format won't? We shall see though.