Work beyond laptops

For years, the laptop meant freedom to work anywhere. My first Mac was the black 2006 MacBook I paid for it myself and took it everywhere. Later, a second-generation 13-inch MacBook Air was the workhorse when I lived nearly a year when I lived in Paris and Barcelona for nearly a year.

Now, more of my work involves directing agents and reviewing what they make. I don’t need to sit at a keyboard for every step. Carrying both a personal laptop and a work laptop, then hunching over a screen to use them, feels increasingly at odds with that shift. I want more freedom to move while I think and work.

This year, we saw people walking around San Francisco with open laptops to keep their agents awake. Others are developing noise-canceling masks for calls and conversations with agents. Those workarounds make me wonder what computing should look like for this way of working.

This past week, OpenAI introduced GPT-6 Astra, the model I’m now testing. Its teaser nods to the 1979 Xerox PARC demo: creating on a computer from a sofa. This morning, while playing video games, I talked to ChatGPT and had it build a native Mac app based on an idea I wanted to explore. Though early, the experience feels more ambient. I can pace around my garage studio, thinking through ideas and building at what feels like the speed of voice.

Why aren’t we all working this way? The answer has as much to do with our habits and workspaces as the technology.

Where progress is lacking

Multimodal AI is advancing quickly, but making it feel natural requires changes to how and where we work.

Workspaces aren’t built for multimodal interaction

My garage studio gives me control over the room. Try pacing between desks in an open office, talking through an idea with an agent. Headphones keep its responses private, but everyone can still hear your half. What feels natural in the studio becomes an interruption at the next desk.

This tension predates AI. A study of more than 42,000 office workers found that easier interaction in open offices didn’t offset dissatisfaction with noise and privacy. Voice agents could bring that friction into work we previously did quietly. Individual work starts to sound like a meeting, and unfinished thoughts become public.

Vision adds another complication. Showing an agent a whiteboard might also expose a colleague or someone else’s screen. A study of camera-glasses wearers found that uncertainty about recording and other people’s privacy shaped how participants used the devices.

Workspace design becomes part of the interface. Booking a room to think aloud takes away the spontaneity. We need space for speech and quiet concentration, clear signals about what devices capture, and ways to switch between speaking, typing, and drawing. The room helps determine which capabilities we can use.

The future experience

The Internet of Things (IoT) arrived before the intelligence it needed. It promised technology woven into everyday life, but too often gave us smart refrigerators. Agents, the Model Context Protocol (MCP), and embedded devices could help fulfill that promise. I want those capabilities to work as naturally at a drawing table as they do at a desk. Three dimensions matter: multi-device, multimodal, and spatial.

Multi-device

Michal Levin’s Designing Multi-Device Experiences offers a useful framework: the three Cs.

• Consistent experiences feel familiar across devices, even as the screen and input method change.
• Continuous experiences let a task move between devices without losing progress or context.
• Complementary experiences give devices different roles within the same workflow.

The third C interests me most: e-ink displays for reflection, tablets for sketching and review, desktops for demanding work, phones for capture and coordination, and ambient displays for context. Each device contributes what it does best.

Multimodal

We work with code, text, images, and voice, often in separate tools. Moving between them still feels fragmented. The future is arriving in pieces; the opportunity is to connect them.

Voice clicked for me when I enabled computer use. I can talk through ideas with ChatGPT while it operates my desktop. Giving it remote access to a trusted computer lets me step away from the screen. Grokbot users have a similar experience, watching their agents work from their phones.

Remix with Rovo: content that changes form

At Atlassian, we call content that can change form malleable content. One of our design guidelines is “Make content and form dynamic.” Remix with Rovo makes that idea tangible: a table becomes a chart, a list becomes an infographic, and a written process becomes a diagram. The source stays intact while you explore another way to communicate it.

Confluence slides builds on the same Rovo platform. You can choose Slides from the Remix menu to turn a page or live doc into a presentation. A project brief can become a kickoff deck, complete with a timeline. Then you can refine it with Rovo, collaborate with your team, and present within Confluence.

The shared foundation is what interests me. Rovo generates and reshapes the content; Confluence provides the place to organize, share, and discuss it. For slides, the Teamwork Graph brings in context from Atlassian and connected tools, while the deck uses Confluence’s permissions and collaboration features.

That’s the direction I want multimodal work to take. A document, a chart, and a presentation can grow from the same body of work. We can choose the form that helps the next person understand it, with less time spent repackaging what we already know.

Spatial

Spatial interaction involves both the spaces we design and how we control them. Nearly four years ago, I wrote that Dynamic Island hinted at Apple’s XR strategy. I still think there’s more to come.

I’m especially interested in how we point to things and direct actions in space. What does a multimodal pointer look like?

The post-laptop era

I’ll still need a laptop for work, but I want to spend less time carrying one around and hunching over it. I’ll use my garage studio to explore a different way of working: a “seeing space” that can interpret what I’m doing and help me build.

I plan to build an app that interprets my drawings and connects them to my work. Voice agents with access to computers will let me spend more time away from the screen. I’ll give my agents access to Mac minis or cloud computers, with one getting the Mac Studio I just ordered.

Chris Tate shared a vision of the kind of workspace I’d like to use. Investments in command-line tools, MCP, and APIs can give us more freedom to design how we interact with our computers.

Creation should feel like a joy from multiple devices, not feel tied down or burdened by your device.

Hyperlinks + notes

• Remix with Rovo → Turning existing knowledge into charts, diagrams, and other forms.
• Confluence slides → From team knowledge to a presentation, built with Rovo.
• Seeing Spaces → Bret Victor’s talk and illustrated poster on spaces for understanding what we make.
• Garage studios → My earlier exploration of making room for drawing, research, and experimentation.
• Sketching with code → How paper sketches and computer vision feed into working prototypes.
• Unleashing neurodiverse teams with Loom, Rovo and the Teamwork Collection → My coworker Stu Smith is co-presenting on this topic.
• Thinkable → Thank you, Sanjiv.
• In design, the job market is a tale of two cities right now by Chris Andrews
• : Stories with early employees from Stripe, Notion, and Figma