How-to AI much?
By the time a useful how-to feature on AI makes its way through the culture, the practice it describes is already being replaced.
Only about five months ago, it felt mind-blowing that I could prompt a single agent into building a working prototype of a complex product design – a command line interface that accepted typed command and returned results, delivered in a React app. Then “vibe coding” became a mainstream term for describing software into existence from concept to production. Just weeks later, the frontier moved again: not one agent building one thing, but many agents working in parallel, opening pull requests, reviewing code, delegating subtasks, and beginning to resemble a compressed software team doing continuous deployment at machine speed.
Our minds are wired to imitate, but the lesson is not that everyone should copy the newest agent hack. It’s to observe that the ceiling keeps moving higher in shorter intervals.
We should assume that whatever felt ambitious last month is probably too small this month. So, don’t bother memorizing new workflows. Instead, learn how to expand the scale, complexity, and ambition of what you ask an agent to do. Because the models are changing overnight, so should your expectations, prompts and process. Wait. Process? Is there even such a thing as “process” anymore? That term assumes something got hardened. But these days, it never does.
Zoom out again. For decades, consumer technology organized itself around the app. There was an app for banking, an app for shopping, an app for travel, an app for messaging, an app for work. Each app had its own interface, navigation model, settings, workflows, and cognitive burden.
Agentic systems point toward a different pattern.
The old phrase was: “There’s an app for that.” The new phrase is: “There’s an AI for that.”
That does not simply mean an AI chatbot replaces a software screen – or hundreds of them. Rather, user experience shifts from doing the task yourself to directing, constraining, supervising, and trusting agents that do the work on your behalf.
DIY becomes DIA: do it agents.
This shift changes what designers, product managers, founders, and ordinary users need to understand. The interface is no longer only the arrangement of buttons, forms, flows, and screens. The interface is the relationship between human intent and machine actions.
What does the system think I want?
What is it allowed to do?
When should it ask?
When should it act?
How do I inspect its work?
How do I recover from a bad decision?
How do I know whether I am still in control?
These are cultural questions, design questions and trust questions. This space is where I will explore them.
JSN is about the human experience of emerging technology, especially where AI, crypto, commerce, art and product design collide. Sometimes that will mean studying interfaces. Sometimes it will mean dissecting agentic workflows. Sometimes it will mean looking at the cultural impact of a world where software stops waiting for us to click repetitively and starts acting on our behalf in continuous loop.
The app era taught us to navigate software. The agent era will teach us to delegate to it without sacrificing being human.