Watch Later Extension
How it works?
Watch Later Extension
How it works?
Watch Later Extension
How it works?

Click here to read complete article on Medium

Sometimes, success starts in the most unexpected places — like an IPL innings break.

3 min read article by Girish Shedge

The Spark: A Match, a Moment, and a Messy Playlist

It was April 3, 2025. I was unwinding with an intense IPL match between MI vs. LSG when I opened my laptop during the innings break. For just a casual scroll, nothing serious. And then it hit me. Those tabs overload.

In the chaos of tabs, I opened YouTube, and my eye just went to the newly uploaded video of my favourite YouTube channel, Mark Rober. But unfortunately, the video was 25:00 minutes long, and the match was just about to resume. So, without fail, I added that video to my YouTube “watch later playlist” and just to be curious, I just opened it & had a look at my playlist, and what I saw next was a nightmare.

A black hole of intent full of saved tutorials, podcasts, documentaries, and startup stories, all bookmarked for a “someday” that never came.

I kept scrolling. The list felt infinite.

Crazy ideas don’t come when you’re thinking — they sneak in when you’re not.

Ironically, just as I was toying with the idea, a calendar notification popped up on my mobile phone — a reminder for a late-night design review meeting. An idea sparked.

Why not build something that checks your calendar for free slots and automatically fills them with your youtube videos which you wish to save to your watch later playlist, based on your time preferences?

Something smart. Something useful.

I’m not an avid developer, but…

I had some understanding of the product development process and the courage to start.

So, I turned to my co-pilot: ChatGPT

I explained my idea and asked for help getting started. It responded with a pretty solid end-to-end blueprint:

  • Authenticate via Google OAuth.

  • Access calendar data via Google Calendar API.

  • Store & schedule everything with a Supabase backend.

Sounds clean, right?

The real lessons came after 1000+ lines of code.

Building this wasn’t a linear process. Here’s what I learned the hard way:

  1. “Prompt Engineering” is a superpower.
    The quality of your questions determines the quality of the LLM models that help you get. Be specific, be clear, and refine constantly.


  2. LLMs can hallucinate
    ChatGPT and other LLM Models are brilliant — and may occasionally be wrong. Always read, test, and understand the code before you just copy/paste it.


  3. Build the base first.
    Login, storage, API connection — before you even think of any other fancy features or UI, get the core rock solid. Mark me on this, this will save you a lot of time further in the process.


  4. After 1000 lines = Chaos without clarity.
    As the codebase grows, the real work becomes staying organized. Mental clarity is your best debugging tool.


  5. Non-linear is the only path
    The product you end up with won’t look like what you started with. That’s not a failure — it’s the process. Trust it.

Action breeds ideas

One thing I’ve truly internalized from this build:

Stop waiting for perfect ideas. Start building, and the ideas will evolve along the way.

What started as a simple pain — “too many saved videos, no time to watch them” — turned into a functional tool that solves the problem with smart automation.

And yes, it still reminds me of that cricket match.

Because sometimes, all it takes is a regular night, a half-formed thought, and the willingness to chase it.

Final Thought

If you're sitting on a half-baked idea, don't wait for the stars to align.

Start small. Ask good questions. Break things. Fix them.

And celebrate the heck out of every little “Success” in the console.

That’s how things get built.

Policy

|

T&C

|

Idea

|

Stats
Policy

|

T&C

|

Idea

|

Stats
Policy

|

T&C

|

Idea

|

Stats