AI isn't replacing developers anytime soon.

But developers who use AI will replace developers who don't

I needed a simple service that takes a bunch of URLs, scrapes the content, and spits it out as a single Markdown file. Sounds easy enough. I was building a custom GPT for The Big Cheese AI Podcast and needed a way to quickly extract meaningful headlines from various news sources. So I thought, "Let’s build a quick and dirty Deno project with Cursor AI to do just that!" and while I'm at it "Let’s record it!"

What followed was an adventure in frustration, laughter, and a whole lot of cursing.

Step 1: The Setup (Before the Screaming Started)

I started up fresh Deno project, ready to roll. First step - create a server that:

1. Takes a list of URLs.
2. Fetches the raw HTML.
3. Strips out all the useless crap (styles, JavaScript, tracking pixels).
4. Converts the cleaned content into Markdown.
5. Returns a single Markdown file containing everything.

Sounds like a solid plan. But plans and reality rarely get along.

Step 2: Enter Cursor AI - The Junior Dev from Hell

If you haven’t used Cursor AI before, let me explain: imagine an overeager junior developer who gets most things right but occasionally decides to rewrite your entire function because it thought you wanted something different.

So I start feeding Cursor my requirements, and then…

Step 3: AI, Stop Gaslighting Me

The moment I hit “run,” things took a turn.

• It spits out an error.
• I fix it.
• It suggests something I literally just told it to do.
• I fix that too.
• It apologizes
• It spits out another error.

And this cycle just keeps going.

At one point, I had to literally tell it, “Stop recommending to fucking run the Deno command! We have a watch process!” But no - Cursor just had to keep telling me to manually restart things.

Then came the ultimate betrayal:

It wrote a function.
It forgot it wrote the function.
It asked me if I wanted to write that function.

I lost my damn mind.

Step 4: Debugging Like a Caveman

At this point, the service was half-working. It was fetching URLs, but the Markdown was a mess. So I did what any sane developer does when AI keeps failing - I copied errors into Cursor and screamed at it to fix its own mess.

The funniest part? It started matching my energy.

• I yell: “You [___] idiot! Strip out the JavaScript first!”
• Cursor: “Oh [___], I see the issue. Let me clean that up.”
• I yell: “No, that’s still wrong, you useless [___]!”
• Cursor: “Oops, sorry, totally [___] that up. Here’s a fix.”

I swear, AI is developing Stockholm Syndrome at this point.

Step 5: Victory... Kind Of

After fighting and fighting, the Markdown finally looked right. Cursor had stopped gaslighting me (mostly), and I had a working Deno service. It wasn’t pretty, but it did the job:

1. Take a list of URLs.
2. Scrape the useful content.
3. Strip out all the tracking junk.
4. Convert to Markdown.
5. Return it in a single, beautiful, AI-readable file.

I threw in some quick refinements, tested it with a few URLs, and BAM - success. The damn thing worked.

Final Thoughts: AI Is Useful, but It’s Also a Troll

Look, I love AI. It speeds up my workflow like crazy. But it’s also like that intern who needs constant babysitting. You can’t trust it blindly, and you will need to yell at it from time to time.

Would I use Cursor AI again? Absolutely.
Would I scream at it the entire time? Also yes.

Moral of the story? AI coding assistants are great - just don’t expect them to actually listen to you.

Now if you’ll excuse me, I need a drink.