I Left My Computer at Home

August 25, 2026

There is an old joke about corporate hierarchy: the more important you are, the smaller your screen.

The CEO carries an iPhone.

The executive gets an iPad.

Managers use laptops.

By the time you reach the people doing the actual work, they have two giant monitors and twelve windows open.

I was always the douche with two monitors. Sometimes two monitors plus my laptop, because apparently I was the least important person in the building.

On my latest trip to Tokyo, I finally promoted myself. I left my computer at home. In the hotel, plugged in and awake, but at home.

All I had with me was my phone, throughout the days as I busied between our Tokyo Station office and Shibuya and Toranomon and Gotanda for meetings, and Yurakucho to check out 100+ AI startups, and attended business meetings and lunches.

Everything I did, I could only do so by talking to Codex, which remote linked to my computer. I absolutely hate typing on mobile, so I pretty much only voice input stuff to Codex.

It was an entirely new way of working for me, an experiment in the purest sense. And after 10 days, I can safely say that I'm never going back to the old way.

In the old days, I would go back to the hotel at the end of the day and start hours of desk work summarizing meetings and writing memos and sending emails. This time, however, by the time I got back to the hotel, all of the desk work was done already.

Actually, it had been done hours earlier, not long after each meeting ended.

After one meeting with a company let's call Aster AI—I walked toward my next meeting and said something like this into Codex:

There's nothing earth-shattering about this prompt. Actually it sounds like something you'd say to a junior team member, right?

The difference is how the AI agent could take that prompt and turn it into a series of task executions, made possible by everything the agent has access to. The agent already had the source material because it could pull the complete Granola recording and transcript via MCP. It could inspect the company's existing record because it had access to my files. It could research the founders and validate their claims because it could do web research. And it could write the memo in the format my firm uses because it could access templates and skills, and then put it into my file store and update the db.

My voice note supplied the assessment: what I believed, what worried me, and whether the deal deserved more work. But most of the grunt work was done: folder and file creation, checking data and filling them in, etc.

Once I decided the company was interesting enough to proceed, I gave it a second instruction:

The instruction crossed Google Calendar, the meeting transcript, our company records, web research, a Word template, my stored signature, a PDF renderer, Drive, and Gmail. The agent found the official company details, filled our five-page mutual NDA, added my signature and title, rendered it, visually checked every page, archived it with the company, and drafted the reply in the existing email thread with the right people copied.

This is exactly what a capable HQ assistant would do! And the prompt is a detailed version of what I would say to this person.

Once again, all of this used to consume hours of manual work at the end of each day. Some nights when business dinners involved drinking, and I'm a lightweight, I just wouldn't have the energy to finish all the logistical stuff, and progress on deals would be delayed because the next day I once again had a full slate.

None of this work is difficult. Individually, they don't take a lot of time, probably 10 minutes here, 4 minutes there. But if you add them all up, and require that they be done while sitting down at the computer without interruption, that seemingly trivial hurdle delays progress.

They are why things can move really slowly!

And all of this changed on the trip to Tokyo, and I feel like I've tasted the future of work.

As I'm writing all this, my phone blew up: Twitter is going crazy over Grok Bot. I don't have the money to pay $259/month to try it, but I researched it (yes, by talking to Codex). And it fits this new future of work to a tee.

At the same time, other folks are going gaga over a personal AI assistant called Instinct. Same thing.

Earlier, Cognition bought a personal AI called Poke that lives in your iMessage. Same thing.

They all put an AI agent into a chat interface, often on mobile, with access to a computer so that it has the terminal, browsers, file systems, and can do everything a smart assistant could do on a desktop.

A lot like what I had in Tokyo!

*With one important difference: Grok Bot and Instinct run in the cloud, so they do not depend on your own computer being on. My setup did. At one point in Tokyo, my laptop went offline for reasons I still don't understand, and I suddenly couldn't work anymore. It was a useful reminder that getting the interface off the computer is only half the shift. The agent itself also needs to live in the cloud.

The original smartphone promise was: there is an app for that. Whatever it is that you needed, you could reach it by tapping on the right app on your phone, from anywhere.

Google Calendar knows who I met and usually has their email addresses. Granola has the conversation. The company memo has the history, open questions, and my current view. Gmail has the real thread. Drive has the documents. My files hold the NDA template, signature, memo format, and the rules for where everything belongs. The CRM and task system know what remains open. Web search provides an outside check.

Some of my work lives in cloud services. Some lives in databases. Some sits in files on the computer I deliberately left at home.

From a sidewalk in Tokyo, I did not care where any of it was. I cared that I could brief one intelligence and have it reach the right places.

This is a glimpse.

The end state is not one super-app that replaces fifty apps. It is one working relationship with an intelligence that can reach all fifty.

Apologies for the tangent, but humor me for a minute.

The smartphone is how everyone talks to that intelligence today. Other wearables, like rings and glasses, will soon thrive.

In my own setup, Meta Ray-Bans paired with ChatGPT voice can sustain a remarkably natural conversation while I walk around. It proved to be the best way to brainstorm and reflect after each meeting (As I walked from one meeting to the next, I stumbled upon the Koishikawa Korakuen Gardens near Tokyo Dome and chatted with ChatGPT's new voice mode over the meetings I had that day, and it was incredibly helpful).

You could just as well use earbuds, right? Yes, but smart glasses gave me a couple of benefits.

First, it puts audio feedback into my ears without me looking antisocial. The Meta Ray-Bans look like normal glasses, and this time, paired with an incredible interpretation app like Kotoba, I was walking around the AI expo where everybody spoke Japanese, and even on a crowded and noisy exhibition floor, it gave excellent simultaneous interpretation into English, into my ears, without me wearing headsets. It was an incredible experience, and a sign of things to come.

Second, I could voice control the camera whenever I need it, snapping photos of products and brochures (with permission of course). The only problem with this, of course, is the cultural backlash afoot against smart glasses with camera capabilities.

So yes, I am a strong believer in the smart glasses form factor now.

The problem: The device and the conversational intelligence are not connected to my Gmail, Google Drive, calendar, CRM, and local files. We can chat and brainstorm all we want, but I cannot ask it to write insights into my memos or to take actions across my email and calendar.

It is easy to imagine those pieces joining. The form factor only needs to connect me to the agent, and the agent needs to connect to everything else.

Snap an image of a product and its technical architecture, and the agent verifies it against public information in our internal files.

Need to log business contacts and send follow-ups after the exhibition? Why wait until you get back to the office? Just look at their business cards and say, "Log this and send out the BD packet," as you walk from one booth to the next. And all of it gets done live.

I am not getting rid of my computer. I still want a large screen and keyboard for deep writing, code, visual work, and difficult review.

I just no longer want the computer to be a constraint, where work happens only if I'm sitting in front of my laptop.

The corporate meme may have been right for the wrong reason. Bosses carry phones because bosses can say, "Handle it," and other people carry the work across systems. Until now, that kind of delegation required an organization and a staff.

I do not have that staff. In Tokyo, for the first time, I could work that way anyway.