A few months ago I had two problems that looked unrelated:
- My AI coding sessions felt “cheap” in the moment, then expensive at the end of the week.
- I kept opening one social feed “for 2 minutes” and losing 45.
Turns out both problems came from the same thing: invisible drift.
Problem #1: Token costs are sneaky
When you’re shipping fast, you don’t feel token spend in real time.
You feel it later, when a random bill hits and you realize half of it came from messy retries and context bloat.
So I built TokenBar — a tiny $5 Mac menu bar app that keeps AI token and cost tracking visible while I work.
What changed for me:
- I stopped “YOLO prompting” giant blocks of context.
- I started running smaller passes with clearer constraints.
- I caught expensive workflows the same day instead of at month-end.
The weird part? Seeing the number live made me write better prompts.
Problem #2: Feeds destroy deep work in micro-doses
I don’t need to block the whole internet.
I need to stop getting trapped in algorithmic rabbit holes during work blocks.
So I built Monk Mode — a $15 Mac app that blocks distracting feeds at the feed level.
Not all-or-nothing website blocking. Just removing the part that eats your attention.
What changed for me:
- I still open the sites I need.
- I don’t get sucked into endless scroll loops.
- I can actually return to the task I sat down to do.
The combo effect
Using both apps together gave me a simple system:
- TokenBar protects money.
- Monk Mode protects attention.
Attention and spend are the two fastest leaks for solo builders. Plug both and your output changes fast.
If you’re building with AI right now
Try this for one week:
- Track token/cost usage during work, not after.
- Set one clear scope for each prompt.
- Remove feed-style distractions during coding blocks.
- Review where your money + focus went at the end of each day.
You don’t need perfect discipline.
You need fewer invisible leaks.
That’s the whole thesis behind what I built.
Top comments (0)