Losing My Edge
It’s been one of my deepest fears for a while that someone who has no real technical background will eventually make a better app than any of the stuff that I’ve created and I’ll just be sitting there with my 5+ years of professional experience developing software like, “Yeah, that’s really cool. I guess.”
It’s probably why Losing My Edge is one of my favorite songs. To a certain kind of flawed personality type, this is a primal fear, the way better-adjusted people are afraid of poverty, being alone, or becoming senile. I’m afraid that a guy that used to be in a band or a DJ will make a better app with these new tools and my hard-earned skills won’t matter. That guy will get to be on the All-In podcast talking about how he only drinks water and sleeps 22 hours a day and spends the other 2 vibecoding and contributing to offshore retirement accounts with cryptocurrency and airline miles. I will work at the mall food court talking about how I used to write code.
Yeah, I'm losing my edge.
I'm losing my edge.
The kids are coming up from behind.
I'm losing my edge.
I'm losing my edge to the kids from France and from London.
But I was there.
It took longer than I thought it would, to be honest. I’ve used AI really aggressively over the past year to develop code, and because it’s what I do all day every day I have basically been operating under the assumption that it’s really easy for everyone else too. When I go to a coffee place in LA, I’m not thinking, oh man, these actors are all so handsome. I’m thinking: what if one of these guys with 100,000 Instagram followers develops an app, and it’s actually really good? Ryan Reynolds did that Mint Mobile thing, so it could happen. If these guys get into vibecoding, it’s game over for me.
But I'm losing my edge
To better-looking people
With better ideas and more talent
And they're actually really, really nice
Then my most famous friend, Joe Weisenthal, with 400,000 Twitter followers, and one of the world’s biggest podcasts, started to post about messing around with Claude and vibecoding. His comments sounded suspiciously well-informed. I got up early, ready to pounce on the latest news, and refreshed the page on Spotify for the Odd Lots podcast, expecting an episode on interest rates or the surging prices for gold and silver. The collapse of the U.S. dollar! Real problems.
No! Odd Lots had released an episode on Claude, the arch-rival AI to ChatGPT, as I understood the ecosystem. I need to start using Claude, I thought. Everybody else was using Claude. My most eager colleague brought up his relationship with Claude approximately every 30 seconds. Every time anyone had an issue, he reminded us he paid for a license to Claude with his own money, and offered to go ask Claude.
I hear that you and your band have sold your guitars and bought turntables
I hear that you and your band have sold your turntables and bought guitars
I hear everybody that you know is more relevant than everybody that I know
I had not sent Claude a friend request just yet because I already had a year of conversation history stored with ChatGPT. We had months of back-and-forth refining certain bits and pieces of code. I couldn’t possibly stand to lose that and start all over with a new person weird robot that lives in my computer and never sees my most vulnerable side! Or could I?
Whipping out my debit card to go spin up a Claude account because I saw Joe Weisenthal post about an app he made for analyzing text samples on Twitter made me feel like the proverbial bad boyfriend. It’s okay! I thought to myself. I’m just doing what’s right for me, and my career. I’ve outgrown ChatGPT , and it’s time to move on! Not all relationships are meant to last forever!
I asked Claude to take care of some tasks I had already completed once with a boost from ChatGPT. Claude handled those chores in a way that appeared straight out of science fiction. I felt oddly like I was cheating on ChatGPT, trying to prove there was something better out there. The end was coming for me. These bros were right, and I had almost missed out. But it’s not too late, it’s never too late, to go on a date with Claude.
Then I asked Claude a systems architecture question for a part of the project we were still working on and it spit out two thinly-sourced sentences that amounted to something like, “Don’t worry about it! That’s what MQTT is for!” and I realized the Claude UI was better for my particular use case of writing code, but the architect-level problems where a lot of the complexity of a real engineering project lives were still a few iterations away.
The appearance of magic had come about for that one specific problem because I already knew what to ask to get the correct answers from an AI for it. I didn’t already have that knowledge for the parts that were still in progress or yet to come, so Claude’s talents seemed less magical when faced with a true unknown instead of a problem where I already had the answers.
You don't know what you really want
You don't know what you really want
You don't know what you really want
You don't know what you really want
In some ways it’s a little bit like reading a horoscope or a fortune cookie message. When you talk to an AI, you’re going to find whatever is already on your mind, not something truly novel. The AI can’t take your bad day and reveal a winning lottery number, but Vibecoding could help you create an app that generates lottery numbers. For a lot of people and businesses, AI is giving them the ability to solve problems with software that weren’t cost effective to build tools to work on if it had to be done manually. At least at this phase of things, I would say AI is a bit like having an $20 toolbox from Target or IKEA on hand. You can solve a lot more problems with a tape measure and screwdrivers than you can without those tools, but that wouldn’t be an efficient starting point to get to work building a skyscraper either.