The WordPress template previews looked exactly like the website I wanted. Clean layouts, good typography, photos that filled the screen. The marketing promised I could build the site of my dreams.
A software demo is a dinner service the kitchen has rehearsed for weeks. Every plate goes out on time. The cooks know where every pan hangs, the walk-in is stocked for exactly this menu, and nobody orders anything off it.
Then you rent that kitchen for your own Friday night. The walk-in is stocked for someone else's dishes. The burners are arranged around food you don't serve, and the dish you're known for needs a pan nobody bought. You spend the night improvising around equipment you didn't choose. The plates go out, but they aren't quite your food.
That was my website. I imported a template, added my content, and tried to make it mine. Getting from that starting point to the design in my head meant working through editors and settings I found hard to control. The site I ended up with looked basic, and I was paying every month for it.
Buying software is renting that kitchen. The vendor maintains the equipment, and you still have to work out how to serve your menu in it.
For a long time, I considered that a reasonable trade, for my website and for the businesses I worked with. Agents have changed what I can build with the resources I have. The site you're reading is the result: I rebuilt it myself, hosting it costs a fraction of what it used to, and it finally looks the way I wanted.
Buying the head start
A business needs to serve lunch. Building its own kitchen can turn into a project about plumbing, refrigeration and electrical work before anyone starts cooking.
Custom software has its own version of that setup: infrastructure, a database, the application itself, and connections to everything else. Each piece takes its own skills, and assembling them used to take a team. I wrote about what that assembly involves separately. A company could spend so long building the capability that the process it wanted to improve barely changed.
That was the reason behind my preference for buying. A platform could be expensive and still earn its price by getting people to useful work sooner. Even a contract costing hundreds of thousands of dollars could make sense if it avoided years of development.
My website was the same decision at a much smaller scale. I wanted a site, and I didn't want a project about servers. WordPress on a hosting plan gave me somewhere to start, and the host kept it running. That head start was worth paying for.
An industrial kitchen for a small menu
The difficulty starts when the kitchen's size and layout bear little resemblance to the work you need to do.
A broad platform has to accommodate many customers. Its configuration options let them adapt a shared product to different needs. For a customer that needs those capabilities, that flexibility is useful. For someone with a narrow need, getting a few things right can mean navigating a much larger system.
WordPress has to serve everyone from a personal blog to a news site to an online store. Its themes, page builders and plugins are how it does that. I needed a handful of pages and a blog, and I was working through tools built for all of those other sites to get them.
Businesses run into the same mismatch with more at stake. The subscription is one expense. Making the product fit adds others: configuring fields, connecting data, training users and maintaining those decisions when something changes. A larger customer may have the budget to fund custom work or negotiate for a feature. A smaller customer still needs its particular problem solved, with little influence over what the vendor builds next. Whatever the product can't handle usually ends up in a spreadsheet beside it.
A food truck becomes an option
A focused application used to be a small request with a large staffing problem. The business might need only a few screens, but delivering them meant hiring for most of the skills a large application needs.
That's the part agents changed for me. I can work through what the software needs to do, have an agent build it, inspect the result and revise it. I still need to understand the decisions and check the behavior, but applications that used to exceed my technical reach are now work I can take on.
The food truck is the useful comparison. It has a flattop, a fryer and a serving window, and everything on board is there for the six things on the menu. The cook can reach all of it without taking a step. It doesn't need every station in an industrial kitchen. It does need enough equipment to serve that menu reliably, every day. A focused application can be built the same way, sized by what it actually has to do.
The kitchen I was already paying for
My website is the smallest version of this change, and the easiest to see.
I was paying roughly $10–$20 a month to host a WordPress site that got very little traffic. That bought me a site I didn't like and a set of editors I'd stopped wanting to open.
With Claude, I rebuilt it from scratch and deployed it on AWS Amplify. Hosting now runs around fifty cents a month at my current traffic. My AI subscription costs about what the old hosting did, and it does far more than run one website. The hosting figures are approximate, the domain and my time are separate, and what I paid was for my particular setup, not a price everyone pays to use WordPress.
The bigger change is control. The posts are plain text files. When I want a different layout, I describe it, look at the result and change it. I'm no longer searching for a setting a template may never have offered, because there's no template between me and the site.
A low-traffic personal website has far fewer demands than a business application. What it shows is the same change in feasibility: building has become a practical way to get the fit I wanted, at a cost that compares well with renting.
Owning the smaller kitchen
The food truck still needs someone to fix the refrigerator.
Owning software means deciding who handles a failed import, a dependency update or a change in how another system shares data. Someone needs to know how to restore records and judge whether an agent's proposed fix actually works. My website asks very little of me here. A business application asks much more, and those responsibilities belong in the cost of building.
Buying is still the better use of resources when the product fits the work and the vendor takes on jobs you'd struggle to maintain yourself. Extending a platform can also be enough. A fair comparison counts the configuration and workarounds on the buying side and the ongoing care on the building side.
What I'd reconsider is ruling out custom software before looking at the need. The resources available to build it have changed. Start with the menu: what you actually need the software to do, every day. Buying may get you there sooner. But when shaping a large platform around that small menu becomes the project, building the right-sized kitchen deserves a serious look.