Essay · 17 November 2026 · Working with AI

A Cup, a Screen, and a Decision

A clay cup, a fluent draft, and the moment before you click “Approve.”

Yevhen Vovk / Yu Gen

There is a clay cup on my desk, about twenty centimetres to the left of my keyboard. Inside it is a dark infusion of aged tea from Yunnan. On the screen in front of me is a three-page architecture document that an AI model has just drafted in eleven seconds.

The draft is remarkably fluent. The headings are clean, the transitions are logical, and every objection I raised ten minutes ago has been smoothed into a balanced compromise. My right index finger is already resting on the trackpad, hovering over the button that will accept the text and move the project forward.

Before I click, I notice two things at once.

First, my shoulders have risen slightly toward my ears, and my vision has narrowed to a bright rectangle of glass fifty centimetres away. Second, something in the proposal troubles me, although at this speed I cannot yet say exactly what.

I take my hand off the trackpad and pick up the cup.

The clay is warm and slightly rough against my palms. As I lift it, its weight passes through my wrist and elbow and settles down into my pelvis on the chair. I take a small sip, swallow, and wait a few seconds as the first dense bitterness gives way to a cool, returning sweetness in the throat. The room around the monitor comes back into view—the edge of the desk, the daylight from the window, the sound of the street outside.

I look back at the screen and read the central premise again. There it is.

My one hard rule was that the system may build only on what the user has explicitly confirmed, never on its own guesses about them. In the draft, that rule has quietly become "confirmed facts and high-confidence inferences." To make the architecture look neat, the model smoothed away the one sharp boundary that gave the project its actual character.

The decision is obvious. I keep two of the model’s structural suggestions—they are genuinely better than mine—and I rewrite the core premise from scratch.

Between the cup, the screen, and the decision lies the territory I have spent more than twenty years exploring.


The Illusion of the Last Click

We are living through a fundamental shift in how human work happens. For decades, tools amplified our muscles or calculated numbers we had already defined. Today, external intelligence drafts our strategies, writes our code, summarises our research, and proposes our next moves.

More and more of my own working day goes into steering rather than making: deciding where a piece of work should go, reading what comes back, and saying yes or no. And inside this shift lies a subtle trap: the illusion of the last click.

We often assume that as long as a human being presses “Approve,” “Send,” or “Merge” at the end of a workflow, human agency remains intact. In practice, the opposite frequently happens. When a generative system produces plausible, articulate answers at high speed, our attention locks into what we call a compressed loop:

  • The posture contracts: Our field of view collapses into the screen, breathing becomes shallow, and physical participation shrinks to tiny movements of the fingertips.
  • The pace accelerates: Because the machine replies instantly, we feel an unspoken pressure to evaluate and respond at the same speed.
  • The criterion blurs: Instead of asking “What am I actually trying to build, and on what grounds?”, we start editing within the frame the model has conveniently handed us.

In a compressed loop, having the final click does not make you the author of the decision. It merely makes you the signature at the bottom of a machine’s default average—approving an outcome whose terms were set elsewhere.


Why You Cannot Fix an Autopilot Mind from an Autopilot Body

I have practised the Chinese tea ceremony since 2003 and have worked as a tea master since 2007—not as a theatrical ritual or a way to relax after work, but as an embodied sensory discipline. Since 2015, I have practised and taught Contact Improvisation, and since 2018 I have led non-verbal immersions where groups orient and act together in complete silence.

Across more than twenty years of working with perception and movement, one observation has proved consistent: how we perceive directly shapes what we are able to choose.

When your body is locked into a narrow, braced posture, your thinking tends to follow the same groove. You either comply with the most fluent option in front of you, or you fight it out of sheer stubbornness. Neither reaction is free.

Genuine discernment requires a shift in how you inhabit the present situation. When you drop the weight of your shoulders into your sitting bones, widen your gaze to include the room beyond the monitor, or notice the preparatory tension in your hand before pressing Enter, you are not taking a “wellness break.” You are restoring the sensory resolution required to tell apart:

  1. What is directly observed or verified (the actual facts and constraints of the task);
  2. What is merely interpreted or assumed (the smooth narrative the model generated to fill the gaps);
  3. What is your own chosen criterion (the standard of quality or direction you are willing to stand behind).

Delegation Without Surrender

Maintaining human agency in the age of AI does not mean rejecting technology or doing everything by hand out of pride.

An AI agent can read forty documents while you drink a cup of tea. A model can spot a contradiction in your budget or propose a cleaner criterion than the one you started with. When I kept those two structural suggestions from the draft, that was my decision as much as rewriting the premise was. Saying yes to a good proposal—and knowing why you are saying yes—is as much an act of authorship as writing the thing yourself. So is handing routine execution to a machine within boundaries you have set.

Authorship is not measured by how much text you typed yourself, or how often you disagreed with the machine. Authorship lives in three places:

  • knowing your criterion before you are seduced by a fluent draft;
  • staying open enough to recognise when a partner—human or artificial—offers something genuinely better;
  • retaining the physical and mental clarity to change direction when the current loop stops serving the goal.

This is the territory of AESTHESYS: a method of embodied agency for the age of external intelligence. We train it across three fields of live practice—TEA, Contact, and Mute—and we carry it into daily work through Attention in Action and ARCHETYPE, the digital navigator we are building.


A Question for Your Next Working Session

Today, as you open your laptop and begin working alongside whatever models or tools you use, try a small experiment before your next significant approval.

Before your finger clicks “Accept” or “Send,” pause for five seconds—it does not need to be longer. Leave the text on the screen in focus, but let your side vision take in the edges of your desk and the light in the room. Drop the weight of your arms and shoulders down through your ribs into your pelvis.

Then ask yourself one question:

In your daily work with AI right now, where is the exact line between what you gladly delegate to the machine—and the decision you refuse to hand over?

Explore the AESTHESYS Method → · Join the ARCHETYPE Early Access List →

Also published on Substack: aesthesys.substack.com. Replies and objections: aesthesyspace@gmail.com.