Six months ago, I'd never opened a terminal. The blinking cursor was something I associated with IT support, system administrators, people who genuinely enjoyed configuring things. It felt like the opposite of creative work.
Now it's where I do most of my building. I open a terminal, describe what I want, and an AI agent reads my files, writes code, deploys sites, iterates on designs, and ships products. The command line didn't change. What changed is that it learned to speak human.
The permission gap
The barrier to the command line was never intelligence. Plenty of smart people avoided it for their entire careers. The barrier was permission. The terminal felt like it belonged to a specific tribe, and if you hadn't earned your way in through years of Computer Science or self-taught coding, you weren't welcome.
AI collapsed that permission gap overnight. When you can type "build me a responsive homepage with this colour palette and this layout" and watch it happen in real time, the command line stops being intimidating and starts being exhilarating. It's the most direct path from thought to thing that's ever existed.
The orchestrator model
What's happening in the terminal now isn't coding. It's orchestration. You're directing agents that read, write, test, deploy, and iterate. The workflow looks more like a creative director briefing a team than a developer writing functions. You set the vision. You make the taste calls. You decide what stays and what goes. The agent handles the execution.
Ian Rogers, the former Chief Experience Officer at Ledger, put it clearly at SXSW: "Future humans are orchestrators of agents." That's not a prediction about some distant future. It's a description of what's happening right now, today, in terminals and IDEs and chat interfaces around the world.
I've shipped over twenty products this way. Not prototypes. Live, deployed, public products with real users. Each one built through conversation with AI in a terminal window. The quality of the output depends entirely on the quality of the input: your taste, your specificity, your ability to articulate what "right" looks like.
Agents are not hype
There's a strain of scepticism that treats AI agents as vaporware. "It's just a chatbot with extra steps." "It's like saying Netscape was hype because we already had AOL and Gopher." The people saying this are standing in the way of the thing they claim to understand.
Agents are qualitatively different from prompts. A prompt generates a response. An agent takes an objective, breaks it into steps, uses tools, makes decisions, and produces outcomes. The gap between "write me a function" and "build me a product" is the gap between a calculator and a colleague.
The new creative toolkit
Here's what my creative toolkit looks like now. A terminal. A text editor. An AI model. A deployment platform. That's it. No Figma (though I appreciate what it does). No project management software. No ticketing system. No standup meetings. Just a clear idea, a set of taste references, and the ability to articulate both in plain language.
The people shaping what comes next aren't the best coders. They're the best orchestrators. The ones who can hold a product vision in their head, direct multiple agents toward it, make the editorial calls that separate good from great, and ship before the moment passes.
That's not a technical skill. It's a creative one. And the command line is where it lives now.
I explored this shift in depth in The Command Line is Sexy, a 21-slide presentation on why the blinking cursor became the most powerful creative tool of 2026.
