Notebook.
Short notes from the architect's bench. Decisions, reversals, and lessons paid for the hard way — partly so I don't repeat them, mostly because the list is genuinely funny.
-
The list I write when I'm done
A small evening ritual — I plan tomorrow's work while today's context is still hot. Not a TODO list, not the team's tracker, just a private planning surface written from inside the day, when I still know what's worth doing first.
-
Thick walls, thin walls
Two recent personal projects, two different stacks, one decision rule. Framework choice is downstream of the workload's interactivity profile, not personal taste.
-
I stopped reaching for a global state library
After ten years and three production stores (Redux, NgRx, then TanStack Query), most of what I used to call "state" turned out to be server data with a bad caching strategy. Here's the decision tree that replaced the default reach.
-
Lighthouse is a compass, not a destination
Synthetic benchmarks are useful right up to the moment they start setting the priority. A real example from CyclingHero — a 69 on mobile Lighthouse, well into the yellow zone, blazing fast in production, zero user complaints — and why the score-chasers were wrong.
-
Capacity planning fails the same way at any scale
Customer Engineering Manager at spaceOS, post-acquisition, three timezones, a relocation to Japan. I said yes to too many roles individually. The architectural rule for systems applies to people; I just hadn't applied it.
Decisions, in writing.
Case studies show what I shipped. The notebook is for how I decided — context, trade-off, hindsight. Updated when the world changes; reversed when I do.
get in touch