Vibe Coding with AI and the Future of Software

The software industry is changing and with the inclusion of many new coding agents it’s just moving faster and faster.

While I would love to be able to code everything by hand because of the enjoyment that it brings me, I also realize that this is not the case moving forward. We have now entered an era which is our version of the Industrial Revolution. However, this time it’s an AI revolution.

Times are changing, and I truly believe that you can either sit idly on your hands and get left in the dust or you can adapt with the new AI revolution.

I’m choosing the latter.

Listen, I’m not an AI fanboy. I’m just seeing the writing on the wall and adapting. This is the same thing that happened for me when mobile phones came out. I saw the writing on the wall and I jumped at it. I quickly realized that mobile phones were revolutionary and a new technological advancement for humankind. Now look at us, we wake up and within 30 seconds are looking at our phones, most can’t go to the bathroom without it and losing your phone immediately invokes panic in most people.

Back in the Industrial Revolution era there were many who didn’t believe that it would improve life or change things. While there are still people that make things by hand, the vast majority of things are made by the help from what spawned from the Industrial Revolution. Farming tools, manufacturing machinery, and so forth. Moving forward we’re going to see software change in the same way. We will still be software developers, however, we will just be implementing things through the use of an AI agent and then cleaning it up and redirecting it when it goes off tracks.

Our titles might still say “software developer”, but what we might actually become is a “software orchestrator”.

Just as a composer orchestrates a symphony, we are going to do the same thing, but with software. A composer knows that a clarinet needs to play at a certain time, and that a snare hit needs to hit at a certain time, and that the flutes need to be at a certain level, etc. The composer may not know how to play all of those instruments that well. But he knows how they connect, and what sounds good, and how it should be orchestrated.

The same thing is happening with software with AI agents.

As a software developer/orchestrator you’ll understand the fundamentals, how everything connects, how the technology should work, but you may not know the exact syntax. Previously, you would have to go look up the syntax and type it yourself. Now, you can just have AI do it, and if it doesn’t work, then you can dive into the details and figure out why for that particular situation. Sometimes AI might do something really crazy and completely go off the rails and you have to tell it to stop and undo some things. This is where the software engineering experience kicks in. But for most scenarios, it’s probably going to do something close to what you need. However, it won’t always be perfect, which is why human intervention is necessary.

How I’m Adapting to the AI Revolution with Vibe Coding (AI Assisted Development)

I learn by reading and watching, then doing, then tweaking, and iterating, and then doing something brand new on my own after I have a steady foundation. It’s not a complete understanding, but enough for me to get going, then I learn as I go. It’s a process for me.

Its “just in time learning”.

In order to really wrap my head around the capabilities of what is possible with AI agents, I gave myself the following task (and it’s one I think that is useful for all engineers) –

You’re going to create an application using an AI agent such as cursor and you’re not allowed to write a line of code.

The last line is important. You’re not allowed to write a line of code.

As developers, we want to “dive into the code” and figure it out and feel productive and like we’re creating something. To truly harness the power of AI agents and see what they were capable of, I had to remove this as an option for me.

I wasn’t allowed to write a line of code. Ok, got it. Ugh.

What this ultimately means is if the agent does something incorrect I can tell it to fix it or I can use the tool (Cursor, etc) to undo the changes but I am not allowed to write code. This will in fact slow me down and does slow me down significantly. However, this has forced me to improve my prompting technique and how I’m iterating with the AI agents to develop software.

Using this protocol, I recently developed multiple applications and here’s two of them:

• Gramifier.com – Transform your ideas into Instagram-ready quote posts in seconds.
• CoparentKit.com – Co-parenting made simple for separated families.

Gramifier is a simple utility app that’s 100% client side with JavaScript (Stimulus JS) and hosted via a static site. This was written in under 2 hours.

Coparent Kit is a SaaS application that allows separated/divorced parents to coparent effectively and contains multiple features. This was written over the course of 2 weeks 2-3 hours a day here and there. I’d say I’m about 20 hours into this project, max. This would have taken me 80-100 hours to implement previously. (PS: I’m looking for a founder to take this over as I don’t have time to run it, but it has some legs – if you’re interested please let me know.)

I decided to use Ruby Static Pro for Gramifier and Ruby on Rails (with Jumpstart Pro) for Coparent Kit because I feel these are some of the best tools out there to do what I needed to do.

I’ve also been experimenting with AI coding agents with compiled languages such as Kotlin and Java. What I have found is that the AI coding agents are much more capable with dynamic languages such as Ruby, Python, and JavaScript. While they still work with compiled languages, their effectiveness is not as great as it is with dynamic languages.

The Future

Ultimately, I have a lot to say about the process of how I use the coding agents to develop software that I’ll be sharing soon.

That said, I don’t feel that we’re at a situation where non software engineers can create a fully functional, maintainable app yet.

Can they create a simple utility or connect some things together?

Yes, they can, absolutely.

But this all starts to fall down once you have to start modifying and changing the application if it’s not built correctly, or if they hit a random runtime error. Things just fall down then.

I can’t tell you how many people have reached out for help because they’ve gotten stuck and the agent can’t help anymore.

The quality of the code that agents output is also subpar in a lot of cases (though sometimes it can be insanely good too).

Some of the code that the AI agents write is pure garbage and many of the times I have to interrupt the AI agent and tell it to stop doing what it’s doing because it’s doing it incorrectly. At this point the AI agent will agree with me and then start doing things correctly once I prompt it with some sample data or how I want it to implement said feature or technology.

All this to say AI agents are changing our industry and I doubt we’ll be going back to the way that we used to do things by hand.

So at this point you have to make the decision if you’re going to adapt or if you’re going to get left behind.

AI agents are definitely here to stay because it makes us far more effective and efficient at our jobs.

I’ll be writing more about my process of how I work with AI agents tips and tricks And the various tools Models and Prompts that I use I’ll also be posting videos on YouTube of how I’m doing this So please sign up to my email list because I will let you know when these things are released Follow me on X and Instagram for day to day updates and more.