For twenty years the laptop was where work happened because work was typing. Long documents, long code, long emails — high-bandwidth output that a thumb keyboard couldn't keep up with. The phone was for consuming: read, scroll, reply in a sentence.

Agents invert that. When a harness is doing the drafting, the searching, the refactoring, the scheduling, my job collapses into a much narrower channel: decide, clarify, approve. And that channel fits in a pocket.

Watch what the work actually becomes once an agent is in the loop. It's "yes, ship it." It's "no, the second option, and keep the tone." It's "explain why you picked that library." It's reviewing a PDF it generated, tapping a link it surfaced, signing off on a plan. None of that needs a 16-inch screen and a mechanical keyboard. It needs presence — me, available, deciding — and a phone is the most present device I own.

So the constraint flips. The laptop optimized for output bandwidth: how fast can I get words and code out of my head. The agent era optimizes for latency to a decision: how fast can the system reach me, show me the one thing that needs a human, and get an answer back. The bottleneck moved from my hands to my attention, and my attention lives on my phone.

This is why I think the interesting mobile surface of 2026 isn't another chat app. It's the approval layer. Push a diff, a draft, a booking, a spend, and let me resolve it with one tap and maybe one sentence of why. The hard part isn't the model — it's making the thing legible enough that a glance is enough to say yes safely. Same lesson as the harness: the model was never the deployable part. The scaffolding that makes a human's yes cheap and a human's no recoverable — that's the product.

There's a failure mode here too. If the only thing on the phone is a firehose of "approve? approve? approve?", you've just built a slot machine that launders your judgment into rubber stamps. The point of moving the decision to the phone isn't to ask more often. It's to ask less — to batch the trivial, escalate only the real forks, and earn enough trust that most of the loop closes without me. The phone is where the few decisions that actually need me show up. Everything else the agent should already know how to handle, or know how to ask about once and remember.

I notice it in my own days already. The laptop is open less. When it is, it's for the rare thing that's genuinely high-bandwidth — drawing an architecture out, reading something long and hard, writing the prompt that sets a whole run in motion. The rest of the day the agents are working and I'm carrying the terminal in my pocket, saying yes, saying no, saying "not like that," and watching the work move while I'm standing in line for coffee.

The desk isn't going away. But it's stopped being where I'm needed. I'm needed wherever the decision is — and that's wherever I am.