Building with an AI Buddy

The only limit is your imagination is finally true. Here's the caveat nobody puts on the bumper sticker, plus how friends are quietly buying back hours of their week.

The Ghostwriter Loophole

I'd been carrying a story idea around since I was a kid. One of those things that lives in your head fully formed and then dies the second you try to put it on a page, because I write like someone who builds software, not like a novelist. I can see the scene perfectly. I just can't make the sentences do what the scene does in my head.

So I rationalized my way in. Authors use ghostwriters, the work still counts as theirs, and this was just a ghostwriter that happened to be a deeply trained autocomplete on steroids. (You tell yourself what you need to tell yourself to get started. We all do it.) And it worked. It got me to an outline faster than I'd ever gotten to one alone. That was the moment it stopped being a toy.

Then I Stopped Asking It for Summaries

Quick context, because it matters for what comes next. I've worked in technology for fifteen years. Started as a web designer in 2008 and somehow turned into a user-experience designer capable of many impressive design feats. I'm kidding. I'm a product design director now, and I still pull off the occasional impressive feat, except these days I've got much smarter people doing it with me.

Anyway. Today I don't just use these tools for summaries and the odd creative brainstorm. I've built agents that summarize my week of meetings and feed me updates on my projects and my team's projects. I point it at mountains of information and have it build plans and outline outcomes. I've been coding prototypes and full applications for all sorts of use cases, which is not a sentence a design director was supposed to be able to write a few years ago.

And my design team doesn't make designs the old way anymore. Nobody fires up Figma to hand-build UI or pull components from a system anymore. We sketched out the bones of a design system, pointed our favorite model at those rules, and vibe-coded our way to a working prototype. Functional. On brand, mostly. Ready to put in front of users and then hand to development for the real implementation push. It's a genuinely different way to work, and I don't say that lightly.

It Stopped Being About Me

The part that surprised me wasn't my own output. It was watching the people around me catch the same bug, and helping them when they got stuck. Once you've seen what's possible you can't un-see it, and you start spotting the exact moment a friend is one workflow away from getting their life back.

A buddy of mine has a daughter in competitive soccer, and at some point he noticed something obvious that nobody was doing. Nobody was recording the games and playing them back to the kids. Competitive league, real stakes, and the film just didn't exist. So he got annoyed enough to fix it himself. He set up a camera and, at first, all he did was stream a link to the parents so they could watch. Simple. Then the idea did what good ideas do and refused to sit still. He realized he could layer coaching mechanisms on top of the footage, map out plays, break down positioning. What started as a webcam pointed at a field rolled into something close to an AI coaching tool. For a kids' soccer league. That's the part that gets me.

Then there's my girlfriend, who's a psychologist. She was drowning in the same task every week: analyzing reports and synthesizing them into something she could actually submit. She was already using a tool to capture all the notes, so the raw material was there. The bottleneck was turning the pile into a finished thing. I showed her a workflow that let her synthesize the notes, review the draft, make her edits and then aid in reviews. It chopped hours off her day. Not minutes. Hours. And it kept paying out, because now she runs the same loop to review notes from previous team meetings and build the agenda before she walks in. She got time back, which is the only currency any of us actually care about.

That's the through-line in all of this. People aren't using these tools to do something flashy. They're using them to make one annoying thing simpler and buy back a piece of their week.

The Part Nobody Warns You About

Here's what I actually wanted to talk about, because the coding has been the best and most frustrating part of all of it.

I remember marveling at certain websites and applications. The interactions, the transitions, the design. They were beautiful and a little exciting, and they pushed the envelope on what felt possible in a browser. I tried to build that stuff myself. Pulled examples off Stack Overflow, cornered developers, copied and pasted and prayed. I never once reached the moment where I genuinely believed anything was possible.

Then I did. Now I sit down and I actually believe it.

But belief is not the same as a free lunch. The caveat I mentioned at the top is this: you still have to understand the underlying structure. You need enough front-end literacy to know what you're manipulating, what you're trying to achieve, and where the constraints are. The model will happily build you a beautiful thing that collapses the moment you ask it the wrong question, and if you can't read the room, you won't know it collapsed until a user shows you. Token burn is real. Watching an agent confidently dig a hole for twenty minutes because you didn't define the outcome clearly is real. The tool removed the barrier of execution. It did not remove the barrier of knowing what good looks like.

That's the trade, and honestly it's a great trade. The skill that used to matter most, the manual ability to produce the thing, matters less. The skill that matters now is taste plus enough structural understanding to steer. Which, conveniently, is the part fifteen years was actually teaching me the whole time.

So here's the thing to do. Pick the one idea you've been carrying around because you assumed you didn't have the skills to build it. Open a blank session tonight and describe the outcome you want in plain language, including the constraints you already know about. Don't ask it to impress you. Ask it for the specific thing. Then watch how far you get before you hit the wall that only your own understanding can climb. That wall is the new job. The rest is the most fun I've had in years.