AI made building for one person cheap enough to bother.
If you’ve spent any amount of time using AI beyond search and dialogue - meaning you’ve spent time building things with Agents - one thing becomes clear. AI is lowering the cost of building software for an audience of 1. The opportunity for a bespoke software revolution emerging that will make all types of custom software reliable, easy to create, and in many cases, accessible through shared infrastructure so that you don’t have to build or maintain software completely on your own. This is and will continue to cause a significant shift in how individuals and businesses think and interact with many types of technology.
The how and the why of this is simple. Cheap access to capable coding LLMs changes the economics of creating, adapting, and maintaining custom solutions. Long-tail features that previously couldn’t justify the time or expense become practical to serve and that’s true even when the only customer is yourself.
Learning By Building
I recently created my own personal recipe website using an LLM subscription where I store custom gluten-free and dairy-free recipes. If you’ve ever had to worry about these types of dietary restrictions, you know that customizing food can be a massive pain. With AI, I give it the ingredients I have, the thing I want to make, my dietary restrictions, and then - voila - a recipe spits out.
I took it a step further on my custom recipe website and added a button to the recipe pages so that anyone can copy paste the recipe into their own LLM for further customizations.
Thanks to AI, I can skip the awful GF/DF recipe sites bloated with ads that crash my browser. That’s really the point I’m driving at. With AI, I’m able to build something I want ( a simple recipe site) without the extra BS. All for the price of a $20 per month AI subscription and the cost of the domain name.
There’s something about using AI in this way that woke me up to the bespoke audience of 1 tech trend that’s forming. Once you see it in one place, you start to see it everywhere.
I saw it when I built my recipe site and again in my job search where I built a bespoke workflow to search for and adapt my application materials to jobs that are good fits for me.
In this flow, the AI organizes and packages my constraints on city, role, company type, and salary along with several versions of my resume and a work portfolio. The agent takes that info, finds jobs based on my requirements and packages the application materials. It’s an autonomous workflow that once built, I can tweak and further customize along the way.
First We Play On Blank Canvas And Then Platforms For Bespoke
We’ll start by using AI to build our own websites, reports, newsletters, automations, and widgets. But I don’t anticipate everyone wanting to build everything from the ground up just because they have access to AI. Moving forward, SaaS will work to enable an audience of 1 experience by creating foundational software, apps, and tools that let you and your AI start on 2nd base as opposed to building from scratch.
The actual business opportunity with real financial incentives is making it easier for people to create bespoke software with tools to adapt features from existing apps, embed top of the line security practices, and simplify backend design and maintenance. This is a big piece of the AI mass adoption puzzle.
My bet is that because of AI, many technology companies will now shift towards platform-like models where they as experts design tech explicitly for their users to plug and play with their own AI to create bespoke experiences. Platforms that make it easy for users to bring their own tools, data, and UI/UX formats.
Here are a few expressions of that platform model.
Shopify Sidekick makes it so that merchants don't have to build storefronts from zero. They inherit checkout, inventory, and themes, then customize the store in plain, natural language.
Roblox Studio Assistant / Build lets creators inherit the engine, distribution, and economy which includes prompt-to-game tools to start them with prefabbed components and not in a blank Unity project.
Cursor rules and Claude Projects are platforms that provide agent harnesses. You bring repo rules, files, and taste in a way that lets you skip re-explaining yourself every session.
Medicine is also a good expression of the bespoke platform movement. The FDA just shared draft guidance for plug-and-play gene therapy. It’s one platform application that covers a family of mutation-specific edits instead of a separate regulatory visit for every typo in the genome. Because the platform is shared, the customization layer gets cheap, and medicine can stop pretending there's a median patient and start getting adapted to individual needs.
Hardware works under this format too.
Formlabs' Form Now service prints an uploaded model and ships it in a day or two, without owning a factory. Pair that with LLMs that can already spit out 3D designs, and the path from idea to object in your hands skips tooling, inventory, and a median customer by providing a shared production layer with cheap customization enabled on top.
This “platform” as a method of unlocking bespoke and personalized buildouts across all types of technologies is where we’re heading. We’re all going to become worldbuilders.