Hello!
Today I'm thinking about a line from Elon Musk that people are buying Full Self-Driving with a car attached, not the other way around. People buying the outcome (it drives me places) instead of the tool (a car) is going to impact a lot more products, especially software. At a high level, this is how AI upends existing companies. Why does this matter so much for software? Because software teams spent decades making tools people use to get the outcomes they want. Photoshop is a good example. It has thirty-plus years of interface built around manipulating pixels, but when you want to edit a photo you're thinking "make her hair flow to the left," not "select these pixels and apply a transform." The tool sits between you and what you actually want. I've been chasing this idea for a while. Back in 2024, I wrote about how the continuous march of "more features, more UI" leaves people with more functions they don't know about and don't use, and how AI lets you control software with hundreds of features by just telling it what you want to do. Last month I pushed it further: start from the outcome ("here's the news you asked for") and let people engage with the tools later if they want to. Outcome first, tool second. The car example just makes it really concrete. Nobody wants a steering wheel, they want to get somewhere. As we've said for years, we don't call carpenters "hammer users." So if you're building software today, ask yourself: what's the "it drives me places" version of your product? Because if you don't ship that, someone else will, with your tool attached as an afterthought.
How have your views changed on AI agents?
My views haven't so much changed as they've had to keep sprinting to keep up. Back in April 2025, I charted what I called The Evolution of AI Products: from machine learning behind the scenes to chat interfaces to agents that pair models with tools. At the time, I explicitly said it was just a timestamp of where we were, and that I'd probably have to revisit it soon. Turns out "soon" was even sooner than I thought.
The biggest shift in my thinking is around orchestration. A year ago, we were wrestling with how to spin up sub-agents and run them in the background. By early 2026 I was writing that we're fully in the agent orchestration phase, where agents orchestrate other agents. So the frontier moved from "how do people interact with one agent" to "how do people coordinate a whole team of them."
That reframed the design question for me. It used to be about how people interact with AI. Now it's about where and how people fit in. Watching software developers has been the clearest signal here. Their role has abstracted upward fast: first humans wrote code, then machines suggested, then people talked to AI directly, then agents got tools, then came orchestration. Each step pulls the human further from the individual actions and closer to steering the whole thing.
Which raises a question I genuinely don't have an answer to yet. If you use the metaphor of being a manager or a CEO, how large a company can you actually run this way? Can you be a CEO of CEOs? I don't think we know yet. There's a real open question about where the line is, how much people can actually process and manage before it all becomes noise.
And that ties to a tension I keep running into: whether to show the work of agents or hide it. When AI plans which tools to use and fires off more based on results, the complexity of what to show in the UI explodes. I initially assumed the split between people who want to see all that work and people who just want results was temporary, that scrutiny would fade as trust built. But it still seems like there's two camps, at least for now. So I've had to let go of the tidy assumption that this resolves itself.
If there's a throughline in how my views have evolved, it's this: I keep expecting things to settle, and they keep not settling. We take the previous forms and mental models we have and try to shove them inside the new technology, and it takes time for people to find the more natural forms. We're not at that stage yet with AI agents. So my honest position is less a fixed set of views and more a willingness to keep redrawing the map every few months.