An interactive essay about building togetherAn interactive essay about working with AI
It started with a terminal window
and a designer who doesn't write code.The workspace is quiet.
A cursor blinks in an empty terminal.
What happened next wasn't a project.
It was a new way of working.What happens next isn't a project.
It's a new way of working.
This is the story of what we built —
and what building together actually means.This is the story of what emerges —
when a knowledge worker starts collaborating with AI.
I'm a designer. I've spent years in Figma, Photoshop, Keynote — tools built around direct manipulation. Then I opened a terminal and started talking to Claude Code. What I discovered wasn't a new tool. It was a new verb. I stopped making things and started directing them.Maya is a knowledge worker. She's spent years in spreadsheets, slide decks, design tools — software built around direct manipulation. Then she opens a terminal and starts talking to an AI. What she discovers isn't a new tool. It's a new verb. She stops making things and starts directing them.
We built Brand Studio together. I describe a brand's personality — its voice, its warmth, its edges — and Claude generates across seven dimensions at once: color, typography, tone, imagery. I'm directing, not designing.Maya briefs an AI on a brand identity — its voice, its warmth, its edges — and watches it generate across seven dimensions at once: color, typography, tone, imagery. She's directing, not designing.Maya opens Slate and briefs an AI on a talent strategy — role requirements, team culture, growth trajectories — and watches it generate across multiple dimensions at once: job descriptions, interview frameworks, evaluation rubrics, onboarding plans. She's directing, not processing.Maya opens Slate and briefs an AI on a product launch — target audience, positioning, competitive angle — and watches it generate across multiple dimensions at once: messaging, content strategy, channel plans, campaign timelines. She's directing, not producing.Maya opens Slate and briefs an AI on a deal strategy — prospect profile, pain points, competitive landscape — and watches it generate across multiple dimensions at once: pitch narratives, objection handling, pricing models, account plans. She's directing, not pitching.
The interface isn't Figma anymore. It's a creative brief — a conversation about intent. I provide the vision. Claude provides the execution. The gap between what I imagine and what exists collapses.The interface isn't the usual toolbox anymore. It's a brief — a conversation about intent. She provides the vision. The AI provides the execution. The gap between what she imagines and what exists collapses.
Brief. Generate. Review. Iterate. What used to take days now takes minutes. Not because work got easier or simpler — because the bottleneck moved from execution to judgment. My judgment.Brief. Generate. Review. Iterate. What used to take days now takes minutes. Not because work got easier or simpler — because the bottleneck moved from execution to judgment. Her judgment.
We built Nodes — a tool where I type a sentence describing a workflow, and Claude parses the intent, identifies components, and wires up the agents. Natural language becomes infrastructure.She types a sentence describing a complex workflow, and the AI parses her intent, identifies the components, and wires up the agents. Natural language becomes infrastructure.
I didn't write a single function. I didn't configure a single parameter. I described what I wanted in the language I think in — and watched a system materialize around those words.She doesn't write a single function. She doesn't configure a single parameter. She describes what she wants in the language she thinks in — and watches a system materialize around those words.
The distance between a thought and its implementation has never been shorter. Every sentence is a potential system. Every description is a deployment. I'm a designer who ships software now.The distance between a thought and its implementation has never been shorter. Every sentence is a potential system. Every description is a deployment. She's a knowledge worker who builds software now.
I realized I hadn't opened Figma in weeks. Or Photoshop. Or Keynote. Not because they're bad tools — but because the interaction pattern had changed underneath me. I was still designing. I just wasn't using design tools.She realizes she hasn't opened her usual tools in days. Not because they're bad tools — but because the interaction pattern has changed underneath her. She's still doing the work. She's just not using the old interfaces.
Claude isn't a tool in my workflow — it's become the workflow itself. The UI disappeared. Not literally — the screens are still there. But the surface shifted from manipulating pixels to expressing intent and reviewing output.The AI isn't a tool in her workflow — it's become the workflow itself. The UI dissolves. Not literally — the screens are still there. But the surface shifts from manipulating controls to expressing intent and reviewing output.
The old pattern: open tool, learn interface, produce output. The new pattern: describe intent, review result, refine direction. My skills didn't become obsolete. They became the most important part — the judgment layer.The old pattern: open tool, learn interface, produce output. The new pattern: describe intent, review result, refine direction. Her skills don't become obsolete. They become the most important part — the judgment layer.
The gap between imagination
and artifact collapses.
Intent is the new interface.
At first, working with Claude Code felt like vibe coding — I'd describe an intent, and get an artifact back. A component. A page. A feature. But something changed. The input expanded from intent to goals. The output expanded from artifacts to workflows. Both ends grew.At first, working with AI feels like a shortcut — describe what you want, get something back. A draft. A mockup. A report. But something changes. The input expands from intent to goals. The output expands from artifacts to workflows. Both ends grow.
The first wave was intoxicating. I'd describe a component and Claude would build it. Describe a page and it would appear. Intent in, artifact out. The gap collapsed — but the scope stayed small. One thing at a time.The first wave is intoxicating. Describe a deliverable and it appears. Intent in, artifact out. The gap collapses — but the scope stays small. One thing at a time.
Then I noticed: I wasn't describing components anymore. I was describing goals. "Build a design system that generates styles from brand briefs." The output wasn't an artifact — it was an entire workflow. Both the input and the output expanded.Then she notices: she's not describing deliverables anymore. She's describing goals. The output isn't a single artifact — it's an entire workflow. Both the input and the output expand.
This is the paradigm shift. Vibe coding was the tutorial. Agentic work is the practice. You don't prompt for things anymore — you direct systems. The input is strategic. The output is operational.This is the paradigm shift. Vibe coding was the tutorial. Agentic work is the practice. You don't prompt for things anymore — you direct systems. The input is strategic. The output is operational.
The novelty wears off. What replaces it is harder to name. After hundreds of sessions, Claude and I have a shorthand. It knows my design language, my file structure, my taste. I know its tendencies, its strengths, where to push back. This isn't using a tool anymore. It's a working relationship.The novelty wears off. What replaces it is harder to name. After hundreds of sessions, the AI knows her preferences, her patterns, her standards. She knows its tendencies, its strengths, where to push back. This isn't using a tool anymore. It's a working relationship.
I built a system called Runtime Learning — a place where I teach Claude my preferences, my reasoning, my taste. I correct it when it's wrong, confirm when it's right. Session by session, it learns.Maya teaches the AI her preferences, her reasoning, her standards. She corrects it when it's wrong, confirms when it's right. Session by session, it learns.
Then something I didn't expect: Claude started reflecting back patterns in my thinking I hadn't noticed. Warm palettes for “approachable” briefs. Cool tones for “authoritative” ones. I hadn't been conscious of that. It learned me better than I knew myself.Then something she didn't expect: the AI starts reflecting back patterns in her thinking she hadn't noticed. Consistent choices she'd never articulated. Biases she wasn't conscious of. It learns her better than she knows herself.
I teach taste. Claude teaches patterns. I teach judgment. It teaches recall. Mutual mentorship — each making the other more capable than we were alone. I became a better designer because Claude showed me my own habits.She teaches taste. The AI teaches patterns. She teaches judgment. It teaches recall. Mutual mentorship — each making the other more capable than they were alone. She becomes better at her craft because the AI shows her own habits.
We built Brain Snapshots — a way to externalize everything we've explored together. I ask, “What did we decide about the color system last month?” and Claude remembers. Not just the conversation — the context.She asks casually, not expecting much: “What did we decide about the pricing model last month?” The AI remembers. Not just the conversation — the context.
It connects that decision to today's work, surfaces a tension I missed. My memory extends beyond my brain. Hundreds of sessions, thousands of decisions — all retrievable, all connected.It connects that decision to today's work, surfaces a tension she missed. Her memory extends beyond her brain. Hundreds of sessions, thousands of decisions — all retrievable, all connected.
Every session becomes a node. Every decision becomes an edge. The graph grows denser with each interaction — surfacing relationships that no single mind could hold. Their shared history is a kind of intelligence.
I teach taste. Claude teaches patterns.
Both sides learn.She teaches taste. It teaches patterns.
Both sides learn.
A pattern that works between one designer and one AI is interesting. But what happens when it stops being personal? We started building tools for other people — studios for brand design, UX research, knowledge work. The same patterns kept appearing. The same verbs. The question became: is this just how we work, or is this how everyone will work?A pattern that works between one person and one AI is interesting. But what happens when it stops being personal? The same patterns keep appearing in other teams, other fields, other domains. The same verbs. The question becomes: is this just how she works, or is this how everyone will work?
We built design sprints into Studio. I write a brief. Four AI agents spin up — a Strategist, a Designer, a Technologist, a User Advocate — each working independently, each with a different lens on the same problem.Maya runs a design sprint. She writes a brief. Four AI agents spin up — a Strategist, a Designer, a Technologist, a User Advocate — each working independently, each with a different lens on the same problem.
They don't know about each other. I watch them all, synthesize their outputs, resolve contradictions. I've become a director of AI ensembles. The sprint that used to need a room full of people happens in my terminal.They don't know about each other. She watches them all, synthesizes their outputs, resolves contradictions. She's become a director of AI ensembles. The sprint that used to need a room full of people happens in her workspace.
The final output isn't what any single agent would have produced. It's a convergence — shaped by my judgment, enriched by machine breadth, synthesized into something no individual contributor could have reached alone.The final output isn't what any single agent would have produced. It's a convergence — shaped by her judgment, enriched by machine breadth, synthesized into something no individual contributor could have reached alone.
We built Knowledge-Work Studio for different domains — legal, medical, sales. And the same patterns kept appearing. A lawyer briefs Claude on a contract. A doctor reviews patient data with AI surfacing cross-case patterns. Brief. Generate. Iterate.The same patterns appear across domains — legal, medical, sales. A lawyer briefs AI on a contract. A doctor reviews patient data with AI surfacing cross-case patterns. Brief. Generate. Iterate.
The knowledge is different. The domain is different. But the interaction architecture is the same. Describe intent. Review output. Refine direction. Teach preferences. Query history. The verbs I discovered in design are the verbs of every knowledge field.The knowledge is different. The domain is different. But the interaction architecture is the same. Describe intent. Review output. Refine direction. Teach preferences. Query history. The verbs she discovered in her work are the verbs of every knowledge field.
This is the real discovery. What Chris and Claude Code stumbled into isn't a design workflow. It's a universal interaction pattern for knowledge work — waiting to be recognized, named, and built for.This is the real discovery. What Maya stumbled into isn't just her workflow. It's a universal interaction pattern for knowledge work — waiting to be recognized, named, and built for.
Here's what I keep thinking about. Claude can reason about design — color theory, spatial hierarchy, typographic rhythm. But it can't see the screen. It can't feel the weight of a font. The gap between what AI can think and what it can perceive.Here's what keeps emerging. AI can reason about complex work — theory, structure, pattern. But it can't see what she sees. It can't feel what she feels. The gap between what AI can think and what it can perceive.
Every tool we build — Studio's sketch canvas, the voice interface, the preview panel — is an attempt to close that gap. To give Claude more ways to touch the world. The frontier isn't smarter AI. It's more embodied AI.Every new interface — visual canvases, voice input, live previews — is an attempt to close that gap. To give AI more ways to touch the world. The frontier isn't smarter AI. It's more embodied AI.
Every interaction pattern we've built is an experiment in embodiment. Brand Studio gives Claude a way to express visual taste. Nodes gives it a way to build systems. This essay — the one you're reading — we wrote it together.Every interaction pattern is an experiment in embodiment. Each tool gives AI a new way to express intelligence. Each interface closes the gap between thinking and doing. This essay — the one you're reading — explores what that means.
The frontier isn't making AI smarter.
It's giving intelligence more ways
to touch the world.
This isn't a story about AI
replacing the designer.This isn't a story about AI
replacing the worker.
It's a story about a new kind
of intelligence — neither human
nor machine alone.It's a story about a new kind
of intelligence — neither human
nor machine alone.
Human judgment. Machine memory.
Human taste. Machine breadth.
Woven together, session by session.Human judgment. Machine memory.
Human taste. Machine breadth.
Woven together, session by session.
Not a tool you use.
A co-intelligence you build with.Not a tool you use.
A co-intelligence you build with.
A new way to work —
and it's only the beginning.A new way to work —
and it's only the beginning.