paint-brush
How I Built an AI Tool To Craft the Perfect Tweetby@hackercm36m8r2v00003b7kvspfyd4g
128 reads

How I Built an AI Tool To Craft the Perfect Tweet

by November 14th, 2024
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

Bad at writing tweets, worse at coding. Built an AI writing tool to fix my own terrible tweets first (example included!). Accidentally found others had the same problem - from physiotherapists to NASA folks. Hit $266 MRR in 7 days with zero following. Key lesson: Build for yourself, others will follow.
featured image - How I Built an AI Tool To Craft the Perfect Tweet
undefined HackerNoon profile picture

Here's my original tweet draft from two weeks ago:

one thing i've been trying to do more, is to simplify my notification, allowing me to focus more, havign less, discord messages, less dms, less things to think about, and just focus on what's actually important, and do a digital detox


Painful to read, right? That's what I was producing when I decided to start building in public. My thoughts were clear in my head, but they came out as rambling messes. Grammar errors, no structure, zero hook. I was always afraid my tweets were getting the engagement they deserved: none.

Scratching My Own Itch

I could have hired a copywriter. Could have spent months studying viral tweets. Instead, I did something that matched my circumstances:

  • I couldn't write engaging tweets
  • I couldn't code


But I understood my problem deeply.


So, I decided to build https://twtfast.com - initially just to fix my own terrible tweets.


The irony of a bad writer building a writing tool wasn't lost on me.

1. The Build Process (With Zero Coding Background)

Here's how someone who couldn't code or write ended up building a writing tool:


First, I collected 10,000+ high-performing tweets and studied their patterns. Used Claude to help me understand and write the code for:

Tweet analysis, Pattern recognition, and Framework implementation.


Cursor helped me refactor and improve code I barely understood at first. V0 handled the design implementation because my CSS looked like it was from 1995.


Some problems I hit:

  • Spent 6 hours figuring out why my API calls weren't working (async/await concepts were completely foreign).
  • Too many Claude Accounts to solve 1 simple issue.


-Had to rebuild the tweet analysis engine twice because my first attempts couldn't handle thread generation.

-Lost half a day learning what "state management" meant after my UI kept breaking.

-Till this day, I hate databases.


However, because I pushed through...

2. The Tool That Fixed My Own Tweets First

Here's what the tool did to my original tweet:


Digital minimalism experiment:


Before:

  • 147 daily notifications
  • Constant Discord pings
  • Endless DM distractions


After turning off notifications:

  • 2 hours of deep work daily
  • Zero anxiety checking phone
  • Actually finishing projects


A simple rule that changed everything: If it's not urgent, it can wait 24 hours.


Your mind will thank you.


The transformation in my own tweets' was immediate. What started as a tool to fix my writing became something others wanted to use.


It's important to notice that I don't have an audience or following, and did very little marketing, and here's the current app reality: $266 MRR in 7 Day.s


The numbers aren't life-changing, but they're validating.


But here's the interesting part:

The majority of users are not even like me - founders and makers who are good at building but struggle with consistent, engaging tweets, I have people writing about physiotherapy, NASA, and some even write amazing tweets already, and just use the platform to optimize theirs, or get new perspectives.


Through this process, I found that you can be selfish and build stuff for yourself, and most likely, other people will have the same problems as you.

3. The Technical Learning Curve For Those Curious About The Actual Learning Process:

Started with basic JavaScript concepts (took 2 days to understand promises properly). Learned React as I built (lots of "why is my component not re-rendering?" moments). Actually understood API design by building one badly first; I ended up using Python for the backend since it was easier to train the model and used fastapi.


Figured out user authentication the hard way (after accidentally logging everyone out twice).


What's Next: From Writing Tool to Growth Engine Right now, TweetFast is solving one problem: helping people write better tweets. But the vision is larger:


Building an AI that learns from your existing tweets to maintain your voice. Adding engagement analytics to understand what works for your specific audience. Developing an automated scheduler to help people tweet... well... faster ;).


Would love to hear from others who built tools to solve their own limitations. What did you learn in the process?


If you're curious, give it a visit at https://twtfast.com