It all boils down to a simple resolution: Publish More.
That’s it. Nothing fancy, not even physically difficult. Continue to tinker and build things, and then, one small tweak: release them to the wilds of the internets.
No one has to see them, but they can be found.
No one has to like them, but they can be judged.
No one needs to know, but they can be known.
And for me that’s terrifying. It’s a far shift from perfectionism. From doing the extra homework to “get it right” and have “the answer” like I do at work. No boss to judge, to please, to provide feedback. Just me seeing where I can let my own hand take me; with the secret, undying, scary-but-exhilarating-and-probably-conceited hope that it might just go somewhere amazing.
At barely 30 days in, there are no results yet, no indicators of direction, no feedback, nothing.
But I am having fun.
Here’s the latest things I’ve built. First up, a simple Truth or Dare python app. I wanted to build a local library of prompts and calls to action that will help me work my emotional and mental resiliency. I used Codex to build out a v1, and then upgraded it with an ability to pull down further questions from online sources.
It’s small, it’s lightweight, and it punches far above anything I could have built on my own in a weekend. I love how, with some thinking, a few prompts, and a solid test plan, I can build tools to extend my self. I can improve my self-improvement.
And that wasn’t the only thing I built. I wanted to test out a real use of an MCP, and decided to use Perplexity’s Search API. I have a more than passing curiosity about Formula 1 and with the latest news about the upcoming 2026 season, I wanted to see if I could use Codex + Perplexity MCP to build my own 2026 F1 Season Tracker. Setting up the MCP was easy enough, updated the config.toml file (See screenshot below)
…And that was it. I had to re-generate and update my Perplexity key once due to some finickiness on my end (I had no chill and didn’t follow the ridiculously easy instructions in order). And then restarted Codex. I first prompted Codex to search for the full 2026 F1 lineup: teams, race dates, locations, drivers. Codex chose to use the Perplexity MCP, pulled a bunch of information together, and then formatted it into a nice table.
Then, here’s the fun part, I told it to build me an F1 tracker as a webpage with racing themed styling, and some modern animations. Nothing else. And then this beauty showed up.
A few of the dates are wrong; and I think that’s from a timezone / daylight savings time conversion of sorts.
But look at it! Track outlines with animation. Clean formatting, a world map with pins you can hover on. A date countdown. To me, the seamless handoff between research prompt and visualization is amazing. With this proof of concept, I can see quite a few use cases centered around compiling deep research data and visualizations. In my running Notion page titled “Idea List, Brain Dump, Thoughts, Musings, Future Explorations” I keep a tally of future “experiments”. One of them is to get a sense of the shift in social norms through popular songs and lyrics. I have barely any idea of how to go about it other than to pull a bunch of song lyrics, analyze them for recurring patterns, and then show them in a visualization. There’s a ton of skillsets required, probably a few advanced degrees, and definitely some career experience required to do this for fun. But now, I think I might be able to dabble and indulge a whim. Or at the very least, understand how out of my depth I might be.
All in a few hours of work and one half-pint of Ben & Jerry’s!