Are we screwed? Is AI going to kill software development as we know it?

The advent of AI coding assistance has definitely introduced a sense of dread into the software development tech industry. I mean, listen to this interchange between Tim Ferris and investor Chris Sacca.

What's left?

Here's my grand theory. We are super fucked. I think other than most computer science, the highest level of computer science, almost all of the rest of coding is fucking useless. You and I can go to Chat JB team and be like, I wanna build an app that does this, this, and this, and give me the code.

And it spits out the code. And then I've literally said, hey. By the way, I haven't coded since basic. What do I do with this? And it's like, oh, no problem.

Go here, download this, open this Python thing, and then shove it in here, and then do this. And it just talks you through it. And now it'll be agentic, and agent's gonna do all that for you. You just don't need to fucking do it anymore.

Pretty dire, but I'm not convinced. I'm just not convinced that AI is going to replace the thoughtful expertise of a great web developer, product person, or builder. And I think part of the argument is that people are going to be building their own apps with AI. So instead of buying something off the shelf, they're just going to build their own apps that they will then, use deploy and use. Or companies will be doing the same thing instead of, you know, using software as a service, instead of paying monthly a monthly subscription for something.

With AI, they're going to build their own internal apps that are perfectly customized to, you know, their use case. And we got into this in the panel podcast that I co host with Brian Kasel. Sure. Like, my my kids could probably build a SaaS app using ChatGPT, but could they really?

I don't disagree with you today. I just think that the the trajectory that we're on is it's just gonna get easier and easier for people to not not necessarily build more competitors, but just build their own personal solution.

Mhmm.

Their own tools. Like, it doesn't have to be a SaaS that that's out in the market. It just it just has to be my personal metrics dashboard that I just use for myself instead of buying a metrics dashboard tool. You know?

One of the points that Tyler King brought up in that conversation is that this all this fear about, you know, companies and people building their own personal apps kinda feels like a lot of the fear we had before with no code and no code builders like Bubble.

No code is really more in danger than SaaS. Like, something you might have done with no code to prove out a concept and prototype it. Buy that. Just code it now.

You know, that was the promise of no code is that anybody was going to be able to build their own tools, build their own web apps, build their own mobile apps. And, you know, these builders were going to replace the software developer. And that just hasn't happened. And there's a reason for it. Building a really good product requires a thoughtful, integrated, whole product approach to development.

You can't just take off the shelf pieces and tape them all together with duct tape and create a cohesive, reliable service. And we've used products like this, haven't we? We all have. Somebody rushed out a product and maybe they weren't very good at development. It's just full of bugs.

Maybe the functionality's there. It always doesn't feel maybe very smooth. You know, I think no code has been good in certain pieces. You know, if you're a consultant and you're looking to create a a client portal for your customers to log into, no code seems great at that. And maybe, you've replaced what you would typically hire a web developer for, with, a tool like Bubble.

But it hasn't replaced, these actual SaaS web applications or mobile applications that require some finesse. You know, these need to be performant. They need to be reliable. They can't have bugs. They can't have weird, user experience, quiggles, you know.

They have to be cohesive and they have to work end to end. This is a thought that Colleen Schnettler brought up in our conversation as well.

I think as engineers, we're going to be putting the pieces together more because the AI is very good at these individual things, but it's not very good at the system level thinking is what I've seen so far. And I'm a heavy user of all the things.

So the threat of AI when it comes to software development does kind of remind me of, you know, the the whole no code wave that we saw. And, you know, there were articles, I'll find some, that were saying, you know, no code is going to be the end of software development and product development as we know it, and it didn't turn out that way. I'm old enough that I also remember the waves before that. And there was this whole open source wave. If you like search Hacker News for open source project management tool, you'll find dozens, maybe even hundreds going back fifteen, twenty years where you could download a almost perfect replica of Basecamp or other popular tools.

You could use this magic deploy to Heroku button. I don't know if you remember when those were popular, and it would magically just deploy it, provision the server, do everything, and all you had to do was log in and use your own private version of this app that you hosted yourself. A more recent version of this experiment was when 37 released once dot com, where the idea is you would buy software once and, you know, deploy it to your own server and never have to pay a recurring subscription again. And they released a Slack competitor, Campfire, which you could buy once for $299, deploy to a server, deploy to Heroku quite easily. And instead of paying Slack hundreds or thousands of dollars every month for your company's chat, you could just, you know, buy it once and then reduce your bill to almost nothing.

But in practice, that experiment didn't cause a wholesale change in our industry. Most companies still use Slack or Microsoft Teams. And the reason is companies don't want to be responsible for self hosting their own software. They would prefer to have the peace of mind of a professional service that they pay to do all the security updates to keep the servers up. Companies have enough problems to think about, and self hosting dozens and dozens of their own apps is not making things simpler.

And if you've experimented with any of these, you know, the buy once products or deploy to Heroku open source alternatives, and I have, I've used them for internal projects and to build MVPs, What you end up with is just this proliferation of apps that you're hosting in different places, all which costs you money, by the way. It's not like Heroku and DigitalOcean and AWS is free. You're still paying for the hosting, but you'll inevitably start getting emails like this. Dependabot security warnings, bugs that are cropping up, the need to upgrade all the dependencies, the need for customer support, the need to tweak things. Software requires lots of maintenance, lots of updates, and AI is not doing that work right now either.

In fact, it seems to be introducing more bugs than it's fixing. I know organizations now that have, you know, massive AI generated functions and they nobody at the organization understands how they work because they didn't write them. So when something happens, when something breaks, when there's some sort of deadlock that nobody can figure out, it's going to be difficult to fix those problems, a, because you might not have the expertise on your team anymore. But, b, even if you do have, capable web developers on your team, they have might have no idea what AI slop has been generated in your code base, how to find it, and how to remedy it. The truth is great software is the result of thousands of decisions made by humans.

Thinking through each user flow, each interaction, understanding the customer's job to be done and all of the things that are at stake for the user. What makes software more enjoyable to use? What makes software more efficient? What makes software more effective? And these are creative decisions that are hard to make even for product people that have years and years of experience.

And if you look at what AI is currently generating for text and images and video, sure, it can maybe do an okay job in a very specific slice of the creative process. But as soon as you ask it to generate a completed work end to end, it quickly becomes apparent that AI is not there, and I don't think it's ever going to be. Sure. You can ask AI to build you an app, and you can even direct it and prompt it in, an authoritative expert way. But what's gonna be produced at the end of all that is likely going to be fragile, need updates, have security vulnerabilities, and you're still going to have to host it somewhere.

And that Heroku bill might be just as much as just paying a software as a service company to do the same thing. And then you get all the security updates and all the dependency updates and the customer support. I'm curious to hear what you think about all this. If you're watching this on YouTube, you can leave a comment. If you're listening to this as a podcast, you can actually reply to my post on Blue Sky, and it will add a comment to the podcast episode page.

I'd love for you to try that out. Thanks for watching or listening and I'll see you next time.