@rubenhassid: You pay $20/month for a Claude subscription. But you still hit the limit by 2 pm. I did too, until I understood how Cla…

X AI KOLs Timeline News

Summary

The article shares nine tips for optimizing Claude subscription usage to avoid hitting token limits, such as reducing file sizes, planning in chat, and batching requests.

You pay $20/month for a Claude subscription. But you still hit the limit by 2 pm. I did too, until I understood how Claude calculates credits: It re-reads the entire chat every time you send a prompt. Prompt 30 costs about 31x more than message 1. Upload the same PDF to 5 chats & you pay for it 5 times. Run Opus on a grammar check & you pay Opus prices for a Haiku job. Here are the stupid simple 9 fixes to use every day: FIX 1. Shrink your files. Open a Google Doc. Paste the text from your PDF. Download it as a .md file and upload that instead. A 15-page PDF costs ~45,000 tokens. The same text as .md costs ~2,000. FIX 2. Plan in Chat, then build in Cowork. In Claude Chat, prompt: "Help me plan a [financial model]. Ask me questions first." Paste the finished plan into Cowork & say: "Build exactly this." FIX 3. Let Claude prompt you instead. A 500-word prompt costs 500 tokens. A click costs almost zero. So prompt: "I want to [TASK] to [SUCCESS CRITERIA]. Ask me questions using AskUserQuestion before you start." Full guide here: http://claude.rubenhassid.ai FIX 4. Edit your last message. If you want to correct a mistake, go to your sent prompt, click "Edit" on your message, retype your prompt, and hit Save. Claude regenerates from that point. FIX 5. Summarize every 15 prompts chat. Prompt: "Summarize this entire conversation into a brief." Copy the brief. Paste it as message # 1 in a new chat. Turning 105,000 tokens into 500. FIX 6. Pick the model before you start. Open the model dropdown in the chatbox. Under 30 seconds of thinking? Sonnet. Execution, multi-step work? Opus. Brainstorming or solving a complex problem? Fable FIX 7. Use Projects for repeated files. Create a Project. Upload your files there once. Every chat inside that Project reads them without re-tokenizing. FIX 8. Turn off the settings that a task doesn't need. Go to the tools panel. Turn off Web Search, connectors unless this specific task needs them. Every idle feature adds tokens. FIX 9. Batch your asks. Send "Summarize this, list the key points, and write a headline" as ONE message. Three separate messages make Claude reload the full context three times. Grab the free guides to copy my full system: ✦ The 27-fixes breakdown: http://claude.rubenhassid.ai ✦ The infographic cheat sheet: http://how-to-ai.guide
Original Article
View Cached Full Text

Cached at: 09/01/26, 07:47 PM

You pay $20/month for a Claude subscription. But you still hit the limit by 2 pm. I did too, until I understood how Claude calculates credits:

It re-reads the entire chat every time you send a prompt. Prompt 30 costs about 31x more than message 1.

Upload the same PDF to 5 chats & you pay for it 5 times. Run Opus on a grammar check & you pay Opus prices for a Haiku job.

Here are the stupid simple 9 fixes to use every day:

FIX 1. Shrink your files.

Open a Google Doc. Paste the text from your PDF. Download it as a .md file and upload that instead. A 15-page PDF costs ~45,000 tokens. The same text as .md costs ~2,000.

FIX 2. Plan in Chat, then build in Cowork.

In Claude Chat, prompt: “Help me plan a [financial model]. Ask me questions first.” Paste the finished plan into Cowork & say: “Build exactly this.”

FIX 3. Let Claude prompt you instead.

A 500-word prompt costs 500 tokens. A click costs almost zero. So prompt: “I want to [TASK] to [SUCCESS CRITERIA]. Ask me questions using AskUserQuestion before you start.”

Full guide here: http://claude.rubenhassid.ai

FIX 4. Edit your last message.

If you want to correct a mistake, go to your sent prompt, click “Edit” on your message, retype your prompt, and hit Save. Claude regenerates from that point.

FIX 5. Summarize every 15 prompts chat.

Prompt: “Summarize this entire conversation into a brief.” Copy the brief. Paste it as message # 1 in a new chat. Turning 105,000 tokens into 500.

FIX 6. Pick the model before you start.

Open the model dropdown in the chatbox. Under 30 seconds of thinking? Sonnet. Execution, multi-step work? Opus. Brainstorming or solving a complex problem? Fable

FIX 7. Use Projects for repeated files.

Create a Project. Upload your files there once. Every chat inside that Project reads them without re-tokenizing.

FIX 8. Turn off the settings that a task doesn’t need.

Go to the tools panel. Turn off Web Search, connectors unless this specific task needs them. Every idle feature adds tokens.

FIX 9. Batch your asks.

Send “Summarize this, list the key points, and write a headline” as ONE message. Three separate messages make Claude reload the full context three times.

Grab the free guides to copy my full system: ✦ The 27-fixes breakdown: http://claude.rubenhassid.ai ✦ The infographic cheat sheet: http://how-to-ai.guide


27 Claude Tips (Anthropic should have given you)

Source: https://claude.rubenhassid.ai/ (Anthropic should have given you)

I’ve used Claude every day since March 2023. That’s about 1,800 hours.

Here are the 27 thingsI had to work out on my own.

TIP 1. Two words fix Claude’s writing.

Add this to the end of any prompt:

Use ASD-STE100.

ASD-STE100 is the writing standard for aerospace manuals. Short sentences, one idea in each, no decoration. Claude already knows it, so two words switch the whole answer over.

Two Claude chats side by side: the left one prompted with Use ASD-STE100 answers in short, plain sentences; the right one, without it, answers in longer, more decorated prose.Don’t use it for poems or LinkedIn posts. It goes cold fast.

Use it for emails, explanations, instructions and summaries.

27 Claude Tips Anthropic Should Have Given You2 of 30

Want the other 26 tips?

Put your email in and we’ll send the full 30-page PDF straight to your inbox.

Similar Articles