It's Hard To Compete With Free

The Real Estate Espresso Podcast

It’s Hard To Compete With Free

Spotify Google Podcast Apple Podcast Subscribe via RSS Leave a Review

It’s Hard To Compete With Free

Welcome to the Real Estate Espresso podcast, your morning shot of what’s new in the world of real estate investing. I’m your host Victor Menasce.

On today’s show we’re talking about the acceleration of AI capability. Now, what I’m about to share is my own first-hand experience with writing a complete software application for the Mac in less than 15 minutes.

Now I know what you’re thinking, that’s all good Victor, you’re an engineer, and you’re used to doing super complicated things. While that may be true, what I did yesterday was nothing amazing at all. I used the simplest of prompts that you could possibly imagine to generate a program from scratch.

So let me be clear, I’ve never written a software program for macOS and I’ve not read any of the documentation on where to start or how to do it. Yes, I’ve designed software in the past, I’ve designed microchips, but I have not written macOS applications.

I installed the Claude application from Anthropic, which includes their chat interface, the Cowork interface, and the Claude Code interface. And here’s what I asked the chat interface. I said, using already existing open-source tools, create an Apple Macintosh application to play tic-tac-toe. Confirm your plan with me before executing, including the cost to develop it using Claude Code.

Here’s what it came back with. It proposed an approach where it would create a Mac native application using Swift and SwiftUI. These are Apple’s modern user interface frameworks, and it’s the most straightforward approach for Mac development, requiring no third-party dependencies.

It proposed a technology stack consisting of the Swift application and Xcode, which is their development environment, and it’s available for free from the Mac App Store. It provided a justification for that particular approach and said there’s no external dependencies or open-source libraries required. It would provide native performance and native look and feel. By using these libraries, the user interface is simple and it would be declarative. And then, finally, it would work with all of the newest versions of macOS.

It estimated the development time to be somewhere between one to two hours and then it provided a breakdown, told how much would be required for the project setup, for the logic of the game, the user interface, the win-lose-draw detection software, and then, finally, some testing and bug fixes.

It estimated the cost somewhere between two to five dollars to write this application. And then it asked me a bunch of questions. Do I have Xcode installed? Is there a particular preferred version of macOS that I am targeting? Are there any specific user interface requirements that I have? Would I want to play the game against an AI opponent or just keep it as a two-player game for now?

So, I told it that I had indeed installed Xcode, that I’d give it the current version of the operating system, and I told it that I would play just human-to-human instead of human-to-computer.

It gave me the choice of building the application two different ways. One is using Claude Code, where it would write the software and then provide me that software to paste manually into the software development environment so I could build it myself, and the second approach would be absolutely free. There would be no cost to doing that.

Now, you’ll notice that this kind of interaction might take place between myself and an employee, is what would happen if I was walking through the thought process with one of my staff. Except in this particular case, I’m talking to a language model, not a human.

I chose neither option. I chose to use Claude Cowork instead of Claude Code, and I pasted the entire prompt that Claude generated for me directly into Cowork just to see what would happen. After only a couple of minutes, it said all of the files are ready in the folder you selected. The project is fully functional and ready to build in Xcode.

So I said, wow, that was fast. It was nowhere near the one to two hours that were being predicted. So I opened Xcode and hit the required command, and immediately got three error messages. So I went back to Cowork and told it I got three error messages.

Now, to nobody’s surprise, I pasted the three error messages into Cowork and it responded immediately that it had fixed the three bugs in the program and that it was now working correctly.

So I set about to test the program and, lo and behold, I now have a functioning game of tic-tac-toe on my computer that an AI tool wrote for me custom in only a few minutes. This is a custom program just for me, the specification was mine. I told it to use a blue banner in the window, and my game has a blue banner.

I realize there’s nothing particularly earth-shattering about a game of tic-tac-toe. But let me put this in context for you: I have a pretty good idea of what it takes to write a program of this complexity. I’ve been programming computers since the 1970s. And I’ve been designing computer chips since the 1980s, including microprocessors that formed the heart of computation and communication systems for more than a few decades.

I know what it takes to design a software program. It starts with a specification, then into design, then coding, then testing.

Now I didn’t need to tell the AI what the rules were for a game of tic-tac-toe. It could look up the rules much faster than I could describe them. In actual fact, I didn’t need to tell it what kind of human interface to choose. It chose a pretty sensible balance of window size, proportion, size of the characters, spaces, and so on.

In actual fact, the program was created much faster than the original estimate of one to two hours. I did lose a few minutes trying to figure out what to do when the first attempt failed, when there were three error messages. At first I put on my engineering hat and tried to debug the error messages myself, but pretty quickly realized that I should just send the work back to my employee, in this case my AI tool, and tell it to fix the problems that crept into the software somewhere along the way. And it fixed them readily in only a few seconds.

The true power of the process is that I got AI to write the instructions for writing the software from a pretty vague set of specifications. It showed me that far more complex systems can be solved using this approach with only a few minutes of thought. And that opens up the world of solutions that would have been unthinkable only a few months ago.

These tools are collapsing the cost of writing software virtually down to zero. The large-scale adoption of these approaches will change the world. In fact, I will argue that it already has.

As you think about that, have an awesome rest of your day. Go make some great things happen. We’ll talk to you again tomorrow.

Stay connected and discover more about my work in real estate and by visiting and following me on various platforms: