Back Original

[RC] Week #4: Ziglings are done; Google AI Pro fiasco; bricked my new car

25 Apr, 2026 #rc, #coding-retreat, #snippets

Context about what “Recursers” do day-to-day

See my Week #2 snippet for details. Here’s a LinkedIn Short video, where Nicholas (RC CEO) explains it really clearly.

Week 4 Progress

Completed all Ziglings. Was it worth it? Maybe; it was a good intro to different parts of the language and ecosystem, but actual learning happened when I overcame some challenges:

I’ve collected some interesting “quirks” about the language below.

Challenges

I had a day of firmware doom. I got a car with a fancy OS (based on AOSP) and managed to brick it during the over-the-air upgrade. The same day, the laundry machine got corrupted firmware and locked my clothes inside…

MINI Firmware Upgrade Washing Machine Fail

That reminded me how integral software is to our day-to-day chores.

Procrastination

I procrastinated on a personal finance tracker project. In the US, I used the Monarch app, and it was great. Worked 95%. In the EU, it doesn’t really work.

When I actually cornered myself to get it done, I had an idea to purchase a “Google AI Pro” subscription to get this task completed quickly, since with “AI Pro” all Google apps get those magic “Gemini” buttons. Google Sheets, in particular, gets the magic formula, AI("your prompt").

Observation 1:
:point_right: Having a financial commitment forced me to start making progress.

When I actually began to work on this task, the “thinking” Gemini model built a plan with a series of steps. At this point, I was able to make incremental progress.

Observation 2:
:point_right: This very specific type of procrastination—where a “task” is not an atomic simple task like taking out the trash or paying a bill, but a whole “project” with multi-step planning and readjustment…

In the end, I was really frustrated with the “AI Pro” service; it’s really poor quality. I ended up using Gemini in the form of “Stack Overflow that knows my context.” After a couple of evenings, I was able to squeeze out the three formulas I needed to use—(VLOOKUP, IMPORTRANGE, and QUERY).

In particular, a task that should be “easy” for LLMs—given a list of credit card transactions, generate a regular expression to match them to categories in @file-with-categories. Use the web-search tool to find details about transactions—Gemini really struggled to produce output for this.

Quirks in Zig that I found interesting / worth exploring more

(P.S. KEXP music performance, I listened to this as a soundtrack while doing Ziglings).

:eyes: :eyes: :eyes: