published 6/7/2025
Your Job Now: Be The Outlier
I’ve been working on a startup for the last month. I’ll have an MVP very soon, but until then, I thought you could use a short LinkedIn cringe post.
Median Pilot
Humans have been operating machines for a while now. We use them for all sorts of things, from mass producing bobbleheads to flinging us through the air in a metal tube at a speed that I’m still very uncomfortable with. They’re fantastic, and we’re going to keep utilizing machines more and more until we meet some fiery end. And that’s my job as a software engineer.
We’ve always aimed for the median. You want a car that drives straight. You want 2+2 to equal 4. You want the metal tube in the sky to stay level in the sky. Outliers are identified and eliminated.
LLMs continue that trend, doing an amazing job slurping up the internet and finding a (rough) median. We don’t want hallucinations. They’re the outliers that models are getting better at filtering out. Every day models get more and more accurate, and more and more homogeneous.
Outlier Pilot
Our job now is to introduce outliers.
AI is getting extraordinarily good at generating boring code. There’s nothing wrong with boring code. Boring code flies a plane. Boring code collects a paycheck. But boring pilots are the first to be automated. They don’t add value, and a pilot that doesn’t add value is often just a liability.
If you want to make something special in current_year
, then you need to learn when to introduce outliers.
You have to ignore the AI sometimes not because it’s wrong, but because it’s too boring.
In a sea of fish, you want to be a narwhal.
Sure, the horn doesn’t make a whole lot of sense and they’re a boring shade of gray, but they are undeniably unicorns.
My humble advice is to vibe code…whatever that means. Absolutely hit tab on the boilerplate and non-consequential stuff. Automate the boring parts so you can spend more of your time on other things. Choose some components to be the outliers so your application is different. It doesn’t matter if your outliers are “worse” than the median, they’ll still stand out.
For example, I used ChatGPT to spit out a pretty good controls UI.
I was going to do the same thing for my new project, but I realized, “fuck that”.
I don’t even like emojis, why am I copying that style?
Is it because every README.md
does it, or worse, because ChatGPT loves emojis?
- 📌 A bullet point list.
- 😄 With an emoji in front of each line.
- 🔥 Is super easy to do.
- 👴 But gets old fast.
I’m not an artist, but the least I can do is draw a narwhal instead. That’s my outlier, and you’ll remember it.
The Skeptics
One last note for the software engineers who don’t want to vibe. I know you exist, I see the complaining on Reddit.
Look, I’m not trying to sell you anything. I’m not founding a startup marketing itself as “Cursor for X” and collecting my riches. I’ve been doing live video for a decade now, I like it, and it’s nice being an expert at something.
Change is hard. Case and point, I’ve been using vanilla vim for most of my career. I only recently switched over to VSCode (vim bindings ofc) and was blown away by rust-analyzer. Same thing happened when I started using Copilot. Same thing happened when I started using Cursor. Now I can barely code offline.
You should use AI tools in your day-to-day work. They’re the new Google, the new Stack Overflow, etc. Your job as a software engineer is to copy-paste boring code from the internet and fix it until it works. In many regards, the only thing that’s changed is the source.
Conclusion
What does this have to do with Media over QUIC?
Absolutely nothing.
I would set up a personal blog but I’m too busy right now. Hope you enjoyed something different like comment subscribe.