From Castles to Cavalry: Building Software in the Age of Agentic AI
From Castles to Cavalry: Building Software in the Age of Agentic AI
I built my first website for a paying client in 1998. I soon went on to drop out of DSU and start a business that I grew to 100 clients across 14 states and four countries. One of the early projects was something I called Build And Go — one of the very first content management systems aimed at regular folks instead of just big enterprise customers.
Back then the whole mindset was to build something that would last. You poured your time into it, made it solid, and it could stand for years with occasional updates and redesigns. The technology moved, sure, but not at the pace it does now. You could actually build a digital castle, dig a moat around it, and feel like you had ground to build upon.
Fellow web developers shared knowledge on forums, pieced together components, and created real tools. Then WordPress came along and changed everything. Today it powers something like 40% of the web. The barrier to entry for a basic brochure site has dropped to almost nothing. You can get one made for free or next to nothing by any number of services out there–it’s been completely commoditized.
That old model — the castle and the moat — just doesn’t work the same way for small operators and solopreneurs anymore. The moat was supposed to be your differentiation, your defense, giving you time to grow. There is still a place for that approach in certain enterprise niches, but for most of us the entire landscape has shifted.
I’ve been thinking about this a lot lately as I work with agentic AI every day. The better metaphor now isn’t a castle. It’s mobile cavalry.
You’re out on the field. You need to be nimble. You need to read the situation, maneuver quickly, and adapt in real time to the battlefields’ demands. Mark Zuckerberg lived by a version of this early on when developing Facebook with the “move fast and break things” approach. They weren’t afraid to ship something that wasn’t perfectly polished.
I’ve experienced the same lesson in writing novels with my wife. The saying that I live by is Perfection is the enemy of done. You could edit and tweak a story forever and never release it. The better path is to do the best work you reasonably can, get it out there, learn from it, and move on to the next one. We published eight novels that way and the lessons from each one made the next better.
The same principle applies to software development in the age of agentic AI.
These large language models change fast. Each one has its own feel — its own way of responding to the same prompt. What works beautifully with one model might fall flat with the next. If you spend too long getting comfortable with a particular framework or approach, you risk waking up to find the world has already moved on.
That means the winning skill now is adaptability. You have to stay flexible in your thinking. You have to be willing to ship, test, learn, and pivot quickly.
At Hurt Ridge Labs that’s exactly what we’re doing. This whole site exists to explore that reality in public. Some of the projects I’ve been working on for months are suddenly running into similar efforts from big players like Microsoft. I’m not interested in going head-to-head with them! Instead of trying to finish building the castle walls brick by brick, we get on the horse and start maneuvering, pivoting to a different niche.
Some experiments won’t go anywhere. Others will take unexpected turns because the technology or the market changed in front of us. That’s not failure — it’s the nature of this new landscape.
The Little Badlands I grew up hiking south of South Heart still look the same as they were when I was a kid–they’re the same “Clay Hills” I love exploring to this day. Geography hasn’t changed much in my lifetime, but the digital world certainly has!
The real challenge now–for each of us, whether we’re developers or not–is tirelessly adapting to that changing landscape.