I stopped leaking money to AI coding and doomscrolling by fixing one workflow

Source: Dev.to

Problem Overview

• My AI‑coding costs kept jumping.
• My focus was constantly wrecked by endless feeds.

I treated these as separate issues, but the root cause was the same bug: distraction led to sloppy prompts, more reruns, and exploding token usage.

Typical Bad Night

1. Start a coding task.
2. Alt‑tab to another app for “2 minutes.”
3. Return with broken context.
4. Fire off another vague prompt.
5. Get mediocre output.
6. Retry three times.

That loop is expensive.

Solution

Spend Cap & TokenBar

I now set a rough spend cap for each coding block. This forces better scoping from the first prompt. I built TokenBar because invoices arrive too late to change behavior.

• Product: https://tokenbar.site
• Price: $5

Monk Mode

I stopped deleting social apps; instead, I block algorithmic feeds and keep only utility surfaces (DMs, search). I built Monk Mode for this exact use case.

• Product: https://mac.monk-mode.lifestyle
• Price: $15

Pre‑run Checklist

Before each run I ask myself:

• What output do I want?
• What files/context are actually needed?
• Which model tier is enough?

This tiny check saves far more than any “best prompting hack.” I also leave a short note for future‑me, reducing context rebuilds and expensive reruns.

Results

After one week of running this system:

• Fewer reruns
• Cleaner prompts
• Lower token burn
• More shipped features
• Less doom‑scrolling relapse during coding blocks

If your AI bill and your focus both feel chaotic, treat them as one system problem. That framing did more for me than any tool switch.