🐴 What to do if you hit the usage limit of AI tools

Tips and tricks with token-saving strategies

A few month ago, I was using Claude to generate user journey maps.

My work-in-progress experiment

I was getting close to something exciting. I could specify the style, customize the diagram, and more. My train of thoughts was running.

But suddenly, I hit a wall. This error message showed up:

It was so frustrating.

So if you’ve hit your usage limit before too, I feel your frustrations.

Today, I’ll share what you can do if you hit the usage limit for some common AI tools—V0, Bolt, ChatGPT, and Claude.

And yes, there are other options besides just waiting for the limit to reset or paying to upgrade your plan.

I’ll also share my token-saving strategies.

Let’s dive in!

---

Bolt & V0

If you are on their free plans, you can hit the usage limit in roughly as many as 15-20 prompts or as few as 5-10 prompts (if your tasks are complex).

Their usage caps are much lower than those of tools like ChatGPT, Claude, and Cursor.

Option 1: Wait for limit reset

For Bolt and V0, their limit reset routine have not always been consistent.

Roughly speaking, you need to wait for a day.

Option 2: Pay to upgrade the plan

You can pay $20 a month to upgrade to their next level.

You’ll get roughly 10-20x more messages than the free plan.

Option 3: Use the code editor

Both Bolt and V0 have a code editor, which is great.

The code editor section in V0

Since editing code isn’t considered “generating,” it doesn’t consume tokens.

In other words, you can continue using the code editor for coding even after you hit the limit, and the Preview mode will still work.

This is helpful because it allows you to bypass the limit.

But you may wonder: what if I don’t know how to code?

I often take the steps below. They're especially effective when you're close to the finish line:

1. Go find the corresponding code files.

   1. Knowing where to find code in your editor based on your needs is important:

      1. Update the UI of a page → Find page.tsx

      2. Change the layout shared across all pages → Find app/layout.tsx

      3. Adjust global styles → Find app/globals.css or tailwind.config.ts

      4. Modify shared components → Find files inside the “components” folder

2. Copy the code and paste it into Claude or ChatGPT, describe what you want to achieve, and ask it to revise the code for you.

3. Copy and paste the revised code back into Bolt or V0.

Claude helped me revise the code

Option 4: Transfer to Cursor.

You can easily download the code from Bolt or V0. It will come as an organized folder with files.

Attach all the files in Cursor to continue building. You’ll be able to pick up where you left off, which is great.

That said, if you have minimal programming knowledge, working in Cursor can feel overwhelming.

Option 5: Create a new account

Warning: Although the idea may sound interesting, I don’t encourage abusing it. Only create a new account when absolutely necessary. Creating many free accounts just to bypass the limit may violate the tool’s terms of service and I don’t recommend that.

1. Take a snapshot of the file structure of the code editor.

The file structure shown in V0’s code editor

2. Create a new account, attach the snapshot, and ask it to create the same files as placeholders with this prompt → “Create a blank file structure with placeholder files that match the attached structure”.

New account with the same file structure and placeholder files

3. Then copy and paste the code, file by file, from the old account to the new one.

This can save you a lot of time by letting you pick up where you left off in the old account, without having to start over from scratch.

---

ChatGPT & Claude

Option 1: Wait for limit reset

ChatGPT’s advanced model limit resets every 7 days, while Claude resets daily for free users and every 5 hours for Pro users.

Option 2: Pay to upgrade the plan

You can upgrade the plan to immediately reset the limit.

Option 3: Switch to a basic model

If you are doing a less resource-intensive task, you can use one of their basic AI models instead. (Not great, but a temporary compromise.)

Option 4: Switch to other similar tools

If you hit a limit on ChatGPT, you can continue your task on Claude, Gemini, or Grok.

Likewise, if you hit a limit on Claude, switch to ChatGPT, Gemini, or Grok.

Option 5: Use API

Take Claude for example, you can get an account at Anthropic Console, get an API key, and use Claude API. It’s billed based on the tokens you use, so there’s no usage cap. However, the API can get expensive depending on your usage, and it also requires technical knowledge to set up.

---

Token-saving Strategies

As I’ve mentioned in my newsletters before, token-saving strategies are an underrated topic that not many people talk about. But they’re becoming increasingly important.

No matter which option you choose from what I shared today, you still shouldn’t abuse token usage, because you’ll eventually hit a limit.

Here are some of my strategies to save tokens:

1. Plan strategically when to use each tool

“Don't put all your eggs in one basket."

For tasks like clarifying your thinking, understanding concepts, creating design specs, and writing prompts, you can use ChatGPT first.

Then, you can test your prompt in Claude as a proof of concept before feeding it into app builders like Bolt, V0, or Lovable. This is because Claude has a higher usage cap and can generate a quick interactive prototype.

If the result in Claude makes sense directionally, you can move forward with the same prompt in Bolt, V0, or Lovable.

If not, revise the prompt and then try it in those tools.

An example of my token-saving strategy

The benefit of this strategy is that you allocate tasks to different AI tools based on their usage limits and strengths, rather than overusing a single tool.

For example, I know Claude is good at coding and has a higher usage limit than Bolt, V0, or Lovable (I’m also on Claude’s Pro plan), so I often use it to test the effectiveness of my prompts and to help debug code.

2. Avoid unnecessary information

For Bolt and V0, I always make sure the prompt includes only the information that’s absolutely necessary, so tokens aren’t wasted.

This applies to ChatGPT and Claude as well, even though they have higher usage limits.

Additionally, a long chat history can burn through tokens fast, since ChatGPT or Claude reprocess the entire conversation each time you send a message.

So be mindful when the chat gets very long.

Start a new “chat” when switching to a different topic.

3. Use basic modes when appropriate

For Claude, the Concise mode uses fewer tokens.

Claude’s style setting

Likewise, for ChatGPT, it’s not necessary to use token-intensive models like o1 for simple tasks.

---

Thanks for reading.

The usage limit can be frustrating, but it also encourages more mindful use of AI tools rather than overusing them.

Limitations can spark creativity.

See you next week!

Xinran

-

P.S. My daughter: “Dad, why do you always talk about AI?!”

Me: “Do you know what AI is?"

“No.”

I couldn’t help wondering what the world would be like when she grows up.