I tried Cursor AI for creating a chrome extension

3 key mistakes that I made

In partnership with

Before we talk about Cursor, a quick note about what to expect from the newsletter, in case you haven’t noticed it yet:

  • The Weekday Edition: Mostly about AI, Product Development, or a process that we went through at Ertiqah and I share our learnings about it.

  • The Weekend Edition: Either about marketing your product, or transparent statistics from our experiments in the past week (Build in public).

In this one, I am going to break the pattern a little bit:

  • Process I used for building a Chrome Extension with Cursor AI

  • 3 key takeaways from using Cursor for building a Chrome Extension

  • Experimented with reels on YouTube Shorts & Twitter | Results, Observations & the tool that I used for it

Building a Chrome Extension with Cursor AI

I’ll jump straight into it, here’s the process I used:

  1. (5 minutes) Document the features that I wanted the extension to have.

    1. I used AudioAI to record my brainstorming notes,

    2. Fed it to Claude to turn it into an organized instructions set that I can feed to my team for building the app.

  2. (5-10 minutes) Fed that info to Composer within Cursor to create the necessary files and initial version of the extension for me.

  3. (2 minutes) Asked Cursor to give me the steps to test the extension locally.

  4. (3 hours) Used the Composer and kept feeding it the errors that I got, until it took care of all of it for me.

    1. Incrementally kept asking it for the changes I needed.

    2. Copy/pasted the errors I got and fed those back to Cursor to fix.

Total time taken: 3 to 4 hours.

Scale yourself and your business for less

  • Get 25% off all annual HoneyBook plans

  • Manage proposals, contracts, invoices, and payments

  • Get the tools to build stronger client relationships

3 key takeaways from the mistakes I made

Here’s what I would do differently the next time (for a video that I am planning to make for building a basic version of a fully functional web app with AI), based on my learnings from this experiment:

  1. Document the deployment environment in the instructions.

    1. Ideally, start with something stupidly simple.

    2. Do not feed all the features/requirements in one go. Best to ask it for one feature at a time, test, and then ask it to add the next.

  2. Only use Composer mode for the initial files creation.
    Debugging must be done using Chat Mode because “web search” is not available within Composer, and the models (GPT o1-mini, Claude 3.5, etc.) information can be outdated.

  3. (Specific to Chrome extensions) Do not aim for building something that can be deployed to the Chrome WebStore straight away. Make it clear to AI that:

    1. This will be deployed locally and only used by you/one user.

    2. The backend should be on the local machine.

    3. Ask it to NOT care about security issues in the code.

Treat AI like an intern, not a product manager. If you give it the full feature roadmap, it will set your whole project on fire and write a LinkedIn post bashing you for the workload. One step at a time.

Reels experiment on YouTube Shorts & Twitter | Results, Observations & the tool I used

About 2-3 weeks ago, I told you about the 100 day YouTube challenge. I honestly don’t know what that was about, because it for sure was not about “posting on youtube everyday for a 100 days” as these challenges are usually supposed to be.

But anyways, I had published about 4 long videos on YouTube since then. And that is when I recalled that there’s a tool called OpusClip (no, not sponsored) that I had trialed about 6 months ago.

So here’s what I did:

  • Copied the links to those “long” videos into Opus Clip.

  • It gave me about 4 “reels” from each of those videos.

  • I liked a few of them.

  • Posted them directly to YouTube, TikTok & Twitter (not all of them though).

This is what their interface internally looks like:

it’s very handy

And these are the results:

The common reel that got the most traction (relatively) that I posted on all of the above:

OpusClip took this clip from the 1 hr+ long video where I was creating a custom landing page for LiGo. I then added an “AI voice over” for the starting part (built-in feature) and then pushed it to socials.

Observations:

  • this was a very low quality reel,

  • Twitter gave it approximately 8x more views than my normal tweets
    (shows that X is encouraging video content just like LinkedIn).

  • you can easily repurpose a 15 to 20 minutes long video you make once into 4 to 5 reels that you can post over the next 2 to 6 weeks

  • it’s free distribution of content that you’ve already created

That’s all for today.

Super excited for Monday! Here’s a sponsored ad before you go:

AI & ChatGPT Mini Crash Course - Eliminate workplace burnout & save 16+ hours every week. Learn 20+ AI tools, prompting techniques & hacks for free.

Rate this newsletter! (or drop an email)

How did you like it?

Login or Subscribe to participate in polls.

Reply

or to participate.