Last week I built [a small app](https://apptape.alltuner.com/) in a couple of hours. Not a prototype, a finished thing that does one job for a handful of people. It cost me an afternoon I would never have spent on it five years ago.

I've been calling this shift **I2S**, Ideas to Software. SaaS was the economics of building one generalized product and renting it to thousands of customers. I2S is the inverse: software generated for one specific need, cheap enough that it never has to pay for itself. We love naming things. A name makes a shift feel like it belongs to us, to right now.

Except this one doesn't.

In March 2004, Clay Shirky published an essay called [Situated Software](http://shirky.com/essays/situated-software/). He watched his students at NYU build applications for a single classroom, a single dorm, a single group of friends. Software that ignored every rule of what he called the Web School: no scalability, no generality, no plan to survive contact with strangers. It didn't need any of that. It was built for a situation, and when the situation ended, so did the software. Years later Robin Sloan gave the domestic version of the idea a warmer name, [home-cooked software](https://www.robinsloan.com/notes/home-cooked-app/): an app made like a meal, for the people at your table, not for a market.

So the idea is old. What was missing was the economics.

Situated software stayed a curiosity because even a small throwaway app cost weeks, and nobody with real work to do could justify weeks. I know this intimately. Most of the hard calls I've made as an engineering leader were never about what to build. They were about what not to build, and I got genuinely good at it. Twenty-five years of sharpening one skill: killing perfectly buildable ideas because the distraction wasn't worth it. Now I watch that hard-won expertise depreciate in real time, and I'm not sure whether to laugh or update my CV.

That's what changed. AI didn't invent situated software, it made it affordable. Simon Willison keeps making the economic version of the point: [code is now cheap enough](https://simonwillison.net/guides/agentic-engineering-patterns/code-is-cheap/) that you should try things your gut says aren't worth building, and his own [collection of tiny single-purpose tools](https://tools.simonwillison.net/) is situated software being practiced in public, dozens of little apps that will never meet a stranger. Others are pushing the idea further, arguing that [software is becoming paper plates](https://auren.substack.com/p/disposable-software-software-is-now), built to last the project rather than forever.

The wall between having an idea and holding the artifact dropped from weeks to hours, and suddenly the long tail Shirky described twenty-two years ago is actually being built. One-user tools, one-weekend apps, software as personal as a home-cooked meal.

I2S is just situated software with the price tag removed. Which is fine. Naming things is how we notice them. But the map existed long before the road was open, and the interesting question was never whether we could build it. It was whether it was worth trying.

For the first time, the answer defaults to yes. My inner gatekeeper is going to need a new job.