I Coded an App in English

Welcome to AI 2.0. Build something or get out of the way.

Something shifted for me recently. And I think anyone who’s been deep in this space for the last year is starting to feel it too.

The writing phase is over

The first phase of AI, for most non-programmer people, was basically writing and research. Write this email. Summarize that doc. Rewrite this in a better tone. Turn my rambling into a LinkedIn post. That was the whole game for a while, and it was fine. Useful, even. But it had a pretty low ceiling.

Because if AI is just a writing tool, I already have a pen. I have Google Docs. I have fifteen years of muscle memory and a decent vocabulary. And honestly, most of these models aren’t great writers out of the box. It takes real effort to get output that sounds like a human being actually wrote it. So if the main use case requires that much prompting just to get to “good enough,” something’s off.

That’s where the shift started.

From chatbot to co-worker

I moved into Claude.

The draw was how it works. Specifically, a feature called Cowork, which is essentially a semi-autonomous way to actually do things on your computer. It suggests, sure. But more importantly, it does.

I started small. I had Claude pay bills. Fill out forms. Handle the kind of tedious, badly designed internet experiences that we all deal with and quietly hate. The broken UX. The weird interfaces. The moments where you’d normally open a new tab, dig through Google for an answer, and come back ten minutes later having forgotten what you were doing.

Claude just kept going. It navigated the bad design. It solved the problems without me having to context-switch. That loop where you hit a wall and go searching for answers? It disappears. The thing just gets done.

That was the first moment where I thought: okay, this is something else entirely.

So I gave it a bigger task.

I had a bunch of webpages that had been sitting on my to-do list for months. The kind of thing you keep meaning to get to on a Saturday that never comes. I told Claude to build them.

It did it overnight.

They weren’t perfect. I still had to guide it, clean up some design choices, do some hand-holding. But it got me 80 percent of the way there, and in some cases it made better decisions than I would have. Partly because I’d been working inside a Squarespace template, constrained by what the template allowed me to think was possible.

Which led to an uncomfortable but exciting realization: I probably don’t need Squarespace anymore. And if I don’t need Squarespace, what else don’t I need? What else can I build that I previously assumed was out of reach because I didn’t have the budget or the dev team?

I coded my first app in English

That question led me to Near Me Now.

The idea came from something incredibly mundane. My wife asks me every week: what’s going on this weekend in Austin? It should be a simple question. It is a simple question. But Google gives you a pile of links. Eventbrite and Luma and Meetup show you closed ecosystems and ignore everything outside of it. Bars and restaurants hide calendars on their websites and Instagrams/ There’s no clean, complete picture of what’s actually happening around you.

So I built something to solve that.

Someone asked me what language I coded it in. I said English. Claude handled the backend. The DNS. Railway. JSON. I still have no idea what any of it means. I described what I wanted, and Claude Code built it. That’s the world we’re in now.

The front end is one thing. What sits behind it is where it gets interesting.

Near Me Now aggregates information that already exists but has never been pulled together in a useful way. And then it does something I think is actually new: it lets you subscribe to that information based on what you actually care about. If you want to make sure you never miss a jazz event in your city, you can get a personalized stream of exactly that. No checkboxes. No advertisers shaping the feed. Just a clean signal based on an AI-powered search that actually understands what you’re looking for.

The accidental Google Killer

I joked with someone that my little app is a Google Killer. They laughed. But here’s why I’m half serious: it does something that you’ve come to rely on Google to do. You just didn’t realize it was a shitty experience until you saw something better. That’s how disruption actually works. It just quietly makes the old thing feel broken.

Once the first version was up, the pattern became obvious. Events have the exact same problem everywhere. Official events, unofficial events, pop-ups. There’s no single place to find everything. So I dropped two more versions: one for Climate Action Week Sydney and one for SXSW. After a brief marketing push, the SXSW version picked up over 500 users, including a small handful who actually paid for a feature that automatically syncs events with their calendar. Which, by the way, why has nobody built that yet? Calendar experiences are just as broken as search. You find an event you want to go to and then you’re copy-pasting details into a calendar app like it’s 2009. The whole chain is held together with duct tape and downloads.

And now I’m building one for the FIFA World Cup. (Sponsors, give me a call.)

Prototype, refine, deploy, repeat.

AI 2.0 is an experience layer

This is the part that feels different from everything else happening in AI right now. There’s a whole category of problems that AI solves better than existing platforms, but people have been trained to think in terms of big destinations. Google for search. Eventbrite for events. Yelp for restaurants. These platforms have been coasting for years on the assumption that they’re the only game in town.

AI breaks that model. It lets you build smaller, more specific experiences that do one thing really well. And in doing that, it starts to chip away at the platforms we’ve all just accepted as good enough.

This is where AI 2.0 shows up for me. The move from AI as a writing tool, to AI as an operating layer, to AI as user experience, to AI as a creative multiplier. The question becomes: how do we fix all the broken experiences we’ve just accepted for the last 20 years?

Because if you step back and look honestly at what we use every day, most of it is kind of a mess. Clunky interfaces. Fragmented information. Workflows that require five different tools to do one simple thing. The way we present information on the internet, broadly speaking, is pretty bad. A few companies are starting to catch on, but the opportunity is wide open.

I know there’s a lot of noise around “agents” and fully automated companies and all of that. Most of it sounds like a pitch deck, to be honest. There’s a strange end state in that vision that nobody really explains. If everything is automated, what’s left for the company to do? What’s the point? It’s the latest version of “give us all your money and we’ll figure it out,” which Silicon Valley has been running since before AI was cool.

But underneath the noise, something real is happening. The gap between idea and execution has collapsed. The gap between the promise of the internet and the value it can bring is collapsing. A clear idea and the willingness to actually try is all you need to build something useful.

And here’s the thing that should make a lot of companies uncomfortable: building AI-first will disrupt entire categories of businesses that still think AI is a feature they can bolt on later. It’s an experience layer. The companies that figure that out will eat the ones that treat it like a productivity hack.

Create now

That’s why I keep coming back to the same point: create now.

This window is open. The tools will get more polished. The space will get more crowded. The excuses will get easier. Right now, it’s still a little messy and chaotic, which is exactly what makes it interesting. The people who build things in this moment, even imperfect things, are the ones who’ll have a head start when the dust settles.

So if you’re sitting on an idea, even a small one, this is the moment to test it. Build the thing that’s been bugging you. Answer the question that everyone answers poorly.

I coded my first app in English. You can figure the rest out as you go.