Completeness Week Continued: Always More Tabs to Open
The power of endings to give drudge work purpose and release your mind.
“It was like I had a dozen browser tabs open in my head, all fighting for attention.” — Developer complaining of AI Brain Fry.
If the memory of a dozen open Chrome tabs follow you uncomfortably after you close the laptop, you’re like the waiter in Weimar Berlin who inspired the Zeigarnik effect.
The waiter stunned the Gestalt psychologist Kurt Lewin and his colleagues because he could remember every drink each of them had when they asked for the bill. He had written nothing down. You realize that a waiter like that is remembering multiple people’s different drinks for hours: an SAT of memorization every night. When they asked later to quickly remind them of their orders, however, he gave them a blank stare.
“You paid the bill,” he explained when they asked how he could have forgotten so much so quickly. All the details had vanished.
Bluma Zeigarnik wondered if this was a common trick of the mind. Perhaps an open “tab” makes memories stick and a “Check, please” releases them. Until the task is done, you remember it. If you want to forget something, complete it. Experiments in which she asked people to complete 5-minute tasks and interrupted a few of them seemed to confirm it. People had way better recall of the interrupted tasks. We now call this the Zeigarnik Effect: the idea that your mind clings anything incomplete. And that’s as true for bar tabs as for Chrome tabs.
The effect hit a research dead end in the 1960s, but the last studies then sugggest that overachievers are particularly cursed with it. Which makes sense, given that it’s overachievers who are probably at the forefront of building AI agents and are complaining of intense, productivity destroying Brain Fry.
Agents and multiple chat windows are multiplying Zeigarnik’s constant "nag" of interruption, leaving multiple tasks incomplete at once and burning holes in our memory. Worse, AI’s optimized to encourage you to start a dozen "amazing" projects, without a clear sense of ending them.
This wave of AI tech began with chats, an open-ended conversation similar to social media. And judging from the suggestions at the bottom of every Claude, Gemini and ChatGPT reply, engineers are being rewarded for engagement rather than satiety.
We are setting ourselves up for endlessness, endlessly. The tools we’ve built to protect ourselves from the endelessness of social media won’t be enough to manage AI. We need new digital wellness concepts. Might I humbly suggest the power of an ending?
Closing an Epic
Tech engineers are so bad at endings that they pay people who know nothing about code to provide them.
They’re called project managers.
They constantly tell engineers when they’re done. This is not easy.
Engineers can revise their work like Marcel Proust turning his one-page novel into a ten-volume epic through constant revision. The thing, while always getting better, is never quite existing.
To counter this, project managers fill engineers’ days and weeks with satisfying endings. The term of art for a large coding project is literally an “epic.” The joy of a final deliverable is such that it barely matters what it is. Whether you’re Agamemnon watching Troy burn or a back-end engineer seeing nine nations going live on your billing system, the deliverable delivered is everything. The key to satsifying work is knowing when it’s over.
An AI bot or agent, however, will give you exactly what you wanted, and then suggest improvements or new ideas or create weird unexpected problems. It’s a project anarchist.
AI is designed to be a wilderness of beginnings.
From the Infinite Scroll to the Unfinishable Task
And, yet, productivity is up. So what’s the big deal?
To get why these superpowers can feel so unsatisfying, I suggest you try ordering a bottomless soup bowl at a psychology lab near you.
Brian Wansink, the architect of this famous study, retired in academic disgrace in 2019 after his data was found to be a mess of falsification and statistical one-off quirks. Yet, a 2023 replication by Lopez et al. saved one of his TED-friendly nightmares from the flames of falsification: people eating from bowls that secretly refilled ate 33% more soup than people who emptied their bowl. Crucially, the bottomless bowlers felt less full. Their internal cues of satiety were no match for the lack of a visual "stopping cue." Eat more, feel less full.
This is also the infinite scroll.
Aza Raskin says he invented the infinite scroll while working for his medical startup called Humanized. The idea was to remove friction from technology: no more clicking on to the next page, no need to make decisions. He has since, rather grandiosely, apologized for all subsequent social-media design, taking responsibility for the “behavioral cocaine” that leads our thumb twitch to the next video every few seconds.
We scan TikTok for a minute and look up an hour later. We consumed more than we needed and feel nowhere near full.
The people who create this content mirror that feeling. A recent survey of wellness influencers found that 10% of them have suicidal thoughts, double the general population. These are wellness influencers.
Like the burned-out parents of toddlers, they are overseeing lots of overlapping tasks that are never done. They show up emotionally for their audience long after they feel connected to the work. When is a wellness influencer done, after all? They’ve doomed themselves and us all to constant improvement.
Brain fry is the first warning sign that AI is bringing this infinite scroll of consuming and creating social-media content to all our tasks. How can we manage it?
From endless friction to actual ending
For about a decade, the solution to the infinite scroll in our personal lives has been mindful friction: turning your phone face down, screen timers, page breaks, pauses. It works. I LOVE LOVE LOVE a simple device called Brick. I touch my phone to it and it stops any social media from working on my phone.
It solves everything except the core problem of social media: its drowning endlessness.
Wouldn’t it be better if you could, instead, one day say, “I did everything I set out to do on Instagram today. It was a fantastic success. And off I go to read my Goethe…” ?
Yeah, right. Because we don’t know how to end our sessions on social media satisfyingly, we will never end our AI sessions either.
I’d suggest that what we all get better at is designing endings for ourselves, in more places in our lives.
Yes, it’s artificial. But all endings are artificial. My fellow Manxman Frank Kermode’s amazing book “The Sense of an Ending” has an amazing passage about the ticking of a clock which captures how we instinctively turn mechanical time into something meaningful.
Let us take a very simple example, the ticking of a clock. We ask what it says: and we agree that it says tick-tock. By this fiction we humanize it, make it talk our language.
Of course, it is we who provide the fictional difference between the two sounds; tick is our word for a physical beginning, tock our word for an end. We say they differ. What enables them to be different is a special kind of middle.
We can perceive a duration only when it is organized. It can be shown by experiment that subjects who listen to the rhythmic structures such as tick-tock, repeated identically, 'can reproduce the intervals within the structure accurately, but they cannot grasp spontaneously the interval between the rhythmic groups,' that is, between tock and tick, even when this remains constant...
The clock's tick-tock I take to be a model of what we call a plot, an organism that humanizes time by giving it form; and the interval between tock and tick represents purely successive time, disorganized time of the sort we need to humanize.
What Kermode finds in the satisfaction of tick-tock is exactly what is amputated from TikTok. It’s all tick. And no tock. Kermode explores how humans create endings out of nothing, to transform unbearable chronos (time passing) into satisfying kairos (progress). Religions do it. Books do it. Plays do it. Beckett doesn’t. Nor does Instagram. Nor does ChatGPT or Gemini.
Daniel Kahneman found a similar two-beat, tick-tock structure in our memories: The Peak-End Rule. We don't judge an experience by its length; we judge it by its most intense moment and its conclusion. Almost everything else will get dumped from our memory and assessment from it. A psychologist will note that a subject might have noted hours of happiness in a day in a diary but then sum up the day as “Bill yelled at me at lunch and when I left the tension was still there.”
If an experience has no end, the brain cannot assign it a value. It stays "live," burning background CPU, a blur of mental static, like those unpaid bills in a Berlin cafe.
The idea of closure is obviously powerful for trauma, and as a coach I’ve found the minor power of an ending also can release people from all sorts of anxieties, like the waiter dumping the memory of 15 drinks. Whenever I’m lost as a speech writer, I find I’ve usually spent too much time fussing about a spectacular opening. When I refocus on the ending, things suddenly fall back into place and the openings rewrite themselves (Howard Hawks: “Third-act problems are almost always first-act problems”).
We can forget that those AI agents and open tabs were all begun by us. Rather than demanding time from us, they’re just sitting there. It’s up to us to design small, but satisfying endings. Whether it’s full psychological closure, or little milestones. A succession of endings will always feel better than a succession of beginnings.
Like an ending in a book, each ending will be made up and imposed on what is in reality a continuous and endless project called life. But it’s much more satisfying than an essay that never
Coming up: Designing satisfying endings.