I’ve been toying with a challenge: 30 apps in 30 days. So I started with the obvious question. What do I actually use? Notion. Slack. Linear. Miro. Amplitude. Looker. Granola. Wispr Flow.
I started cloning them. Some as web apps, some as desktop apps, every one of them with AI and agents built in. By the end of that first session I had about fifteen of them working.
The prompt for the Notion clone was, more or less, “make a Notion clone.” I set a few ground rules, went off to do something else, and came back to a Notion clone.
They aren’t perfect. Some are rough. I’m still iterating on all of them, and who knows how long it’ll take to polish them. But they work.
That’s a strange thing to sit with when you work at a SaaS company. I sell software for a living. And I’m saying software is dead? Is my company going out of business?
No. Not at all. What’s dead is software as we know it. The thing we’ve all come to know and love over the past ten years. Here’s what I actually mean:
Code is free now, so you’re paying for something else.
I used to explain the code. Now I describe the outcome.
About three years ago, when Cursor first came out, I had a newborn. I remember being up late in the nursing chair we’d bought, laptop open, while my kid slept.
I was building a scraper to automate logging into an account. I’d done scraping at scale a few jobs back, and that old system never had a DSL. So I told Cursor: let’s make a DSL. It came back with “what about this?” and wrote one. In Python, which I’d barely touched, because I’d been a TypeScript guy.
There were no MCPs. I have no idea what model was under the hood. But it wrote code, it tested itself, and I sat there thinking: wow.
Even then, I was the one designing it. I’d say I want a file, and a class, that does this. I was explaining the code and it was typing it.
Today I don’t care what’s under the hood. I care about outcomes. I say “make a Notion clone” and it goes and researches Notion, figures out what to build, and builds it. I set the high-level ground rules. That’s my whole part in the loop.
In three years we went from describing classes to describing products.
Software for software’s sake is basically free
On a single $200 a month Claude plan, you can build dozens of apps. I built three in an hour, and I had a great time doing it.
Look at websites. Businesses used to pay thousands of dollars for a basic one. Today you can ask pretty much any agent, even on a $20 ChatGPT plan, and get something decent looking. It’s hard to picture anyone paying thousands for that now.
This isn’t hypothetical for me either. I’m going through everything I pay for right now and asking what I can replace. I pay for Notion today. I’m going to stop. Some of these tools are going to become things I built myself and run locally.
If the only thing your product offers is code, that’s a problem. The code is the cheap part now.
It was always possible to do it yourself. People still hired a guy.
When I was a kid, I could get a template on HostGator for $8 a month. And I did. You could install WordPress and have a turnkey website. Add WooCommerce, or go Magento, and you had a store. The tools to do it yourself have been around for a long time.
Most businesses still didn’t. A plumbing company, a real estate brokerage, a bank: technology isn’t their thing. So they have a guy. Someone who does their marketing, their website, their IT. “I’ve got a guy for that” is almost a joke because it’s so universal.
That isn’t going away. You’ll need fewer of them, and one guy will get way more done. But the plumber still isn’t going to run his own stack, even though now he could.
He was never paying for the code. He was paying for someone who owns the outcome.
So what are people paying for?
If code is free, here’s what I think is left to sell:
- Someone accountable. That’s the guy. At some point you want a person who does the thing for you and answers for it.
- Breadth and trust. When anyone can build anything, products will do more and more. You’ll see more all-in-ones. And you’ll pick one because you like the UX, you like the people, you trust them.
- Someone to call when it goes down. Liability is worth paying for. When your system is down at 2am, “I built it myself” is not a comforting sentence.
- Not having everything go down at once. This is the flip side of the all-in-one. If one vendor runs everything and they go down, you’re totally messed up. Spreading your services across vendors spreads your risk. In some cases your contracts demand it.
- Time. Industries move at different speeds. A big bank isn’t adopting agentic coding overnight. There are security implications, red tape, bureaucracy. A 1,000+ person company has its own problems to solve, and it would much rather pay someone who knows a thing than build it.
Will that last forever? In five years, when this is even easier, probably not. At least the economics won’t look the same. But the good software companies aren’t going to sit around selling the same product either. Their products will evolve too.
There will always be something to buy and something to sell. It’s just not the code.
Nobody pays thousands for a website anymore. They will for an agent.
Spinning up an agent is easy now. Configuring one to drive real value is not. That still takes an understanding of the technology, the tools, and the business problem you’re pointing it at.
Your average non-technical business owner isn’t going to pick that up just because the tools exist. Same as they didn’t build their own WordPress site.
That’s where the specialist goes next. Customized, specialized agents for a specific business. Making them faster and more effective. Tying them to real business problems. And then teams of agents, which is where I think all of this is headed.
The work didn’t disappear. It moved up a level.
It’s never been more fun to be a builder
I used to spend late nights, hours and hours, on passion projects. It was always addictive. Since agentic coding took off, and especially since Claude Code came out, it’s a different thing entirely. If you have an idea, you can just go with it.
I’d never built a desktop app before this. Now that’s the part I’m having the most fun with.
Software isn’t dead. Software as a moat is. The code was never really what people were paying for. Now that it’s free, it’s just obvious.
So if you build software for a living, the question isn’t whether people will still pay. It’s what they’re paying you for once the code is free.
Discover more from zach wills
Subscribe to get the latest posts sent to your email.