💸 Lower Your AI Bill
My cheat sheet for smarter token usage
Ngl, I used to burn through my tokens so fast I'd just sit there waiting for the reset. These are the exact habits that fixed it. None of them cost a thing.
First, the one thing to understand
Tokens are the small pieces of text an AI processes. Long chats and large files can use more of your allowance. The exact cost or usage limit depends on your tool and plan, but keeping your context focused is a useful habit.
The 9 habits
1. New project, new session. Don't cram everything into one chat. Start a fresh one for each new thing so it isn't dragging along (and charging you for) unrelated history.
2. Only stay in a chat to reference today's work. Moving on to something new? Start fresh. That old context follows you and you pay for it on every message.
3. Convert PDFs to Markdown (.md) before uploading. PDFs are heavy and bloated. The same document as a clean MD file uses a fraction of the tokens, and the AI reads it better too.
4. Be specific in your prompts. "Fix the headline in this one file" is cheap. "Improve my whole project" makes it scan everything. The more precise you are, the less it wastes.
5. Stop it the second it goes wrong. Hit escape. Don't pay for it to keep building something you're just going to scrap.
6. Pick the right model for the job. Use the lighter, cheaper model (Haiku / Sonnet) for everyday stuff. Save the expensive one (Opus) for genuinely hard problems. You don't need a Ferrari to buy groceries.
7. Plan before you build. For anything big, let it map out the approach first and approve it. Stops it going down the wrong path and making you pay to redo everything.
8. Keep your standing instructions lean.
Your CLAUDE.md (the instructions that load every session) costs tokens all day if it's bloated. Keep it tight, move niche stuff into skills that only load when needed.
9. Check what's eating your tokens.
Type /context to see what's hogging space, and turn off any connected tools you're not using. Type /usage to watch your spend.
The 10-second version
Short chats. One project per session. Light files. Specific asks. Right-size the model.
That's it. Small habits, big difference on your next reset. ☺️
Made by @chloesinyin