Last month I had two problems that looked unrelated:
1) my AI coding spend kept drifting up every week
2) my attention was getting shredded by feed scrolling between builds
Turns out they were the same problem: context collapse.
When I was distracted, my prompts got sloppier, retries went up, and expensive runs stacked quietly.
The weekly reset that fixed both
Every Sunday night I run a 30-minute reset:
- review where token spend actually went
- mark 3 tasks that deserve heavyweight models
- route everything else to cheaper/default models
- block feed surfaces on social apps for the week
- define 2 no-scroll deep-work windows for weekdays
That single ritual cut my wasted retries and made coding sessions calmer.
What changed in practice
- fewer "one more rerun" loops after midnight
- cleaner prompts because I start with scope, not vibes
- less doomscrolling during compile/test wait time
- lower variance in weekly cost (the big one)
Tools I built because I needed this
I built TokenBar (https://tokenbar.site) as a $5 Mac menu bar tracker so I can see AI token/cost drift in real time instead of discovering it at billing time.
I built Monk Mode (https://mac.monk-mode.lifestyle) as a $15 Mac feed-level blocker so I can keep useful app functions while removing the infinite feed trap.
Using both together changed my workflow more than any motivation hack.
If you’re a solo builder
Don’t optimize for "maximum AI usage." Optimize for clean context + intentional runs.
That’s where the biggest savings and best output quality come from.
If you want, I can share the exact Sunday reset checklist I use in plain text.
Top comments (0)