9 free fixes to stop burning Claude credits

I just stumbled across a post that hit me right between the eyes. I burn through my Claude credits by mid-afternoon all the time, and I always assumed it was just the cost of doing real work. Turns out, I’ve been using the tool wrong this whole time.

This savvy professional broke down exactly why your $20/month plan evaporates so fast, and shared 9 completely free fixes to stretch every credit. I was blown away by how simple some of these are. We’re talking changes you can make in the next 30 seconds that cut your token usage by a factor of 20.

Here’s the core problem the creator laid out, in plain English:

• Every follow-up message re-reads your entire conversation from the top.
• Message 30 costs roughly 31x more than message 1.
• You upload the same PDF to five different chats.
• You use Opus for a grammar check. That’s a Ferrari run to pick up groceries.

Sound familiar? Yeah, me too. Now here’s the fix list the original poster put together.

The 9 free fixes

1. Shrink your files. Open a Google Doc, paste the text from your PDF, then download it as a .md file and upload that instead. A 15-page PDF eats around 45,000 tokens. The exact same text as Markdown? About 2,000 tokens. That’s a 22x reduction for thirty seconds of work.
2. Plan first, build last. Open Claude Chat and prompt: “Help me plan a [financial model]. Ask me questions first.” Then paste the final plan into Cowork and prompt, “Build exactly this.” The author says this alone saves you about 1/10th of the cost.
3. Let Claude ask you. Stop writing 500-word prompts. Try this instead: “I want to [TASK] to [SUCCESS CRITERIA]. Ask me questions using AskUserQuestion before you start.” Then just click answers. Clicking costs almost zero tokens, while your massive prompt costs you for every single word.
4. Edit your messages, don’t resend. Click the “Edit” button on your last message, fix the mistake there, hit Save. Claude regenerates without stacking a new message on top of the old one. Every “no wait, I meant…” message you send doubles your cost.
5. Summarize every 15 messages. After a long chat, prompt: “Summarize this entire conversation into a brief.” Copy that summary and paste it as the first message in a fresh chat. The expert points out this can compress 105,000 tokens down to about 500.
6. Switch models before you start. Click the model dropdown. If the task takes under 30 seconds to answer, pick Haiku or Sonnet. Only reach for Opus on deep, multi-step work. This saves 3 to 5x per message.
7. Use Projects, not uploads. Go to Projects, create one, upload your files there once. Every new chat inside that Project reads those files without re-tokenizing them. Stop uploading the same contract to five separate chats.
8. Turn off extra features. Open the tools panel. Turn off Web Search. Turn off connectors. Turn off Extended Thinking. Only flip on what this specific task actually needs. Idle features burn credits silently in the background.
9. Batch 3 tasks into 1 message. Instead of sending “Summarize this,” then “List the key points,” then “Write a headline” as separate messages, send all three in one. Three messages mean three full context reloads. One message means one reload.

Why this matters more than it looks

Most of these fixes aren’t about being cheap. They’re about working with how the tool actually charges you instead of fighting it. Conversations are cumulative. Every message you send pulls the entire history along with it.

The mind behind this post nailed something most people miss: you’re not paying per question, you’re paying per token, and tokens stack. Once you internalize that, the fixes feel obvious.

My favorite three from the list

If you only do three things from the post’s author this week, I’d pick these:

• The Markdown swap. Going from 45,000 to 2,000 tokens is the kind of math that makes you angry you didn’t know it sooner.
• The summary reset. Long chats turn into tax bills. Compressing them into a fresh chat keeps your sessions fast and cheap.
• The model switch. So many of us default to the heaviest model out of habit. Most daily tasks don’t need that horsepower.

A practical workflow you can copy today

Here’s a quick way to put it all together, based on what the curator suggests:

1. Start a Project for any recurring topic (client work, research, writing).
2. Upload your reference files as Markdown, not PDFs.
3. Turn off every tool you don’t need before you start.
4. Pick the lightest model that fits the task.
5. Let Claude ask you questions instead of typing a wall of text.
6. Batch related asks into a single message.
7. Edit, don’t resend, when you make a mistake.
8. Summarize and reset around message 15.

That’s it. No new plan, no new tool, no upgrade. Just better hygiene around how you talk to the model.

Check out the full LinkedIn post for the original breakdown and every detail straight from the source.