2026 is the year AI coding changes everything
I’ve been building web apps since 2002, and it’s always been painful. About a month ago, it became joyful.
When I first started, my only choice was to go to the shop and buy a physical ‘teach yourself how to code’ book, read it and start practising in Macromedia Dreamweaver. And that’s what I did. Learning to code was super fun, I learned ASP, PHP, JavaScript and HTML, only the basics, but enough to build simple data driven web apps.
When I started doing it as a business, I discovered quickly that I wasn’t great at it, and I needed to work with expert developers. So I let those beginner skills dwindle and over the years worked on a range of projects with developers. It was always a slow and frustrating process, so I was regularly looking for alternatives. During COVID I used a drag and drop coding app to build a web app for beer delivery. But it was still a slog. The app was very limited, and a bit of a hack job.
About half way through 2025 I tried again after hearing about all the advances in AI app builders. I started playing with Replit, Cursor,VS Studio and Expo, and I was blown away by what they were capable of. I could single handedly build an entirely functional web app from scratch, using fully native code. But the process was a little frustrating, with the AI constantly telling me I was absolutely correct while writing bug filled code.
My last post: You’re absolutely right: Lessons learned vibe coding our asset management system
Around the time, the industry was very reluctant to really embrace the technology, or at least admit they were. Tweets like this were the norm for most of 2025.
Either developers were running scared (very likely), or people just couldn’t see how quickly these tools were improving.
I thought he was crazy, without knowing anywhere near as much as he does about coding, it seemed pretty obvious to me that while there were small frustrations currently, the pace at which the models were evolving, made it super clear that these AIs would be able to code perfectly very soon.
So I persisted playing around with the few apps I was building. The first was a web app for us to manage all of our assets in my coffee business East Coast Roast. The second was a rostering and training app I also built from the business to enable staff to sign in for their shifts, and complete basic training and SOPs.
Towards the end of 2025, the main AI tool I was using, Replit started working in a way that blew my mind. To give you some idea, back in July I was using Replit, but putting prompts into different AI’s by manually copying and pasting the code constantly to get different code back. The models in Replit couldn’t really nail it each time without creating multiple new bugs. I would look at what model replit was using, change it, copy code out to a few other models then back in, and test again. Better than working with a developer for sure because at least it was fast. I was using mainly Grok which seemed to give better results than the others, I’d have to explain to Grok that I was building in a different tool and work around the fact that Grok didn’t have access to multiple files etc. Replit at the time had an assistant (ask it to do one thing) and an Agent (makes its own mind up and does heaps of stuff at once). I was using the Assistant and didn’t touch the agent, after trying it a few times and it going on a wild goose chase doing all sorts of crazy shit. It was awesome, but it was a hot mess.
By October / November 2025 Replit was doing the entire job for me basically flawlessly. I stopped looking at the files or the code. I stopped even looking at what model it was using. And I stopped using the Assistant and let the Agent run free and build everything for me. And it worked. Both apps are live and we are using them in our business, and both written completely by the Agent inside Replit. I’ve never even spoken to a developer about these apps and never looked at the code. They are even hosted on their own domains with their own branding.
In December I noticed public perception was changing. My Twitter feed became full of people talking about what they were building with these tools.
Twitter can be full of a fair amount of BS but some of the stories were pretty staggering. At the very least it was obvious that people were building some very cool stuff with these tools. And even though my experience was fairly basic in comparison, they seemed very consistent with what I had seen – mind blowing improvement in the performance of these apps in only a few short months.
DHH even changed his mind! He tweeted ‘This is the most exciting thing we’ve made computers do since we connected them to the internet’. ‘it turns out half the resistance was simply that the models weren’t good enough yet. I spent more time rewriting what it wrote than if I’d done it from scratch. That has now flipped’.
We’re entering into 2026 in a situation where anyone can realistically build their own apps, for the first time in history. And that’s only the beginning. These apps don’t just write code, they change how apps are designed, managed, built and hosted.
Why pay a developer when AI can write all the code, is the obvious question. But there are more.
Why get a designer to create a flat design that you have to somehow manually build into a web app, when you can just get the AI to do the design and ask it immediately to tweak anything you need? And they do it basically free and within seconds better than most designers could?
Why have a project manager when projects are going from months or years to minutes and you can build apps basically instantly and change them instantly.
Why outsource to India to save money when you can avoid all the pain of that and have AI write the code for you immediately.
Why have a host when these AI tools can publish immediately to the world’s best hosting platforms (Replit uses Google Cloud) as part of your plan?
Why use WordPress to build a website when you can just ask AI to make you one, and you never have to worry about updates or plugins or WordPress lagging behind other technologies.
Why pay for a SAAS when you can build it for yourself. Why pay per user for SAAS if someone else can build the app far cheaper and quicker and sell it for a tenth of the price?
Why even have SAAS apps if the AI’s can do all of this stuff for you anyway?
And this is just the stuff that is possible right now.
With how far this stuff has come in 2025, who knows what will happen in 2026.
I certainly don’t have any answers. If DHH can’t predict 3 months ahead, I’m not going to try to predict what is going to happen in 2026. Other than to say, what’s for sure is everything is on the table, and everything is going to change. It’s going to be the most disruptive year in web app building since web app building was a thing. And I for one, am pretty excited about it.
I’m going to be putting out lots of content as I continue building in 2026. Jump on my weekly emails to get the latest and follow me on Instagram or Twitter / X or LinkedIn for more content.
None of my content is written by AI – the feature image was made by ChatGPT, I didn’t even modify it in an editor afterwards I just asked ChatGPT to compress it for the web.
