A training space, not a feed

Keep the parts of you
that AI makes optional.

Small daily practices for the capacities that fade when the friction is removed. Judgment. Words. Curiosity. The nerve to be unrehearsed. Do the rep. The reading is one click away if you want it.

The dailies

Six small games. New every day.

Same games tomorrow, different content. Each one exercises something this site is trying to keep alive.

The studios

Pick a mental workout studio

Every studio is open. Each is a set of practices, not a lecture.

Not sure where to start?

The two-minute daily check

Eight quick taps. No score, no diagnosis. Just the two skills where you have the most to gain, and today's first rep for each.

The question that started the book

Would it be worse than cheating?

The author put one strange scenario to more than sixty couples. Build your own betrayal scale, rate it yourself, then see where they landed. The book thinks nearly everyone got the number wrong.

← Back
The daily check

Where has your threshold moved?

← Back
Studio · Skill 04

Think First

Keep the threshold where it belongs: attempt before assistance. Five minutes of your own thinking before the tool gets a say.

Why this matters →

The Attempt Timer

About to hand a real problem to AI? Run one honest attempt first. Your claim, two tries, a prediction. Then the tool is all yours, and you finally have something to judge its answer against.

Threshold check-in

One question, once a week or so. When did you reach for AI this week?

Protect the beginner

Skill is built by doing the slow work yourself. If a machine does it from day one, you never build the judgment to supervise the machine later. Norway now keeps generative AI out of the early school years on purpose, and a New York law firm has its junior lawyers work their first three years unaided. Pick one thing you would normally hand straight to AI, and commit to doing it yourself this week.

Your decision log

Unaided view, model's contribution, and who you sided with. Over time this shows you your own calibration. Where you beat the tool, and where it beats you.

← Back
Studio · Skill 05

Find Your Words

The reaching is the point. Ten minutes with a page that will not finish your sentences, name your feelings, or hurry you.

Why this matters →

The Blank Page

10:00
starts when you start typing
0 words

Nothing you write leaves this device. There is no cloud. There is no draft saved to anyone's server.

Test the word

At some point, someone gave you a word for what you were feeling. It might have been a friend, a therapist, or a loved one. Before you take it on as the truth, check whether it actually fits. A word that sounds right is not always the right one. Answer the three questions, then name the feeling for yourself.

← Back
The Why

The reading, kept in its place.

This site exists because of a book, The Quiet Cost: What We Stand to Lose in the Age of AI. Its argument fits in one line. When a technology removes a difficulty, we rarely ask whether it removed an obstacle or the practice through which a human capacity was formed. The site is the practice. These pages are the why.

EXPLAINER

Why Speak Unscripted

Why the dining room started to frighten the people who used to find it easy, and how the capacity comes back.

Read · 2 min
EXPLAINER

Why Ask a Person

The favor that starts a friendship, the research on being needed, and what the machine answer can't leave behind.

Read · 2 min
EXPLAINER

Why Stay Raw

What rehearsing a conversation removes from it, including the four things you cannot get back.

Read · 2 min
EXPLAINER

Why Think First

The threshold that moves in one direction, and what a 17% worse exam score has to do with you.

Read · 2 min
EXPLAINER

Why Find Your Words

Bottle caps, alexithymia, and the difference between receiving a word and reaching for one.

Read · 2 min
EXPLAINER

Why Stay Curious

I Am the Cheese, the interval between question and answer, and why the staying matters more than the asking.

Read · 2 min

Five hundred years of getting it wrong

We always predict the wrong loss.

Every technology gets a loud, mostly wrong panic, and a quiet cost nobody names until it is gone. Tap any era, or walk the whole timeline.

← Back to the studio
Why this matters · 2 min

The threshold moves in one direction.

"He used to go to AI after hitting multiple roadblocks… Now, he told me, he often goes at the first sign of friction. Before the struggle has really begun."
On William. Chapter 6 of The Quiet Cost.

Struggle is the mechanism, not the obstacle

Decades of research on "desirable difficulties" (Robert Bjork, UCLA) shows that the conditions that feel like efficient learning, the smooth and assisted and fluent ones, often produce the worst retention. Effortful retrieval and failed attempts are what build durable ability. Performance during practice is an actively misleading indicator of learning.

Evidence: experimental, well-replicated

The 17% warning

Nearly a thousand high-school students were randomly assigned to study math with no AI, a normal chatbot, or a "guarded" tutor that made them work first. With the tools, both AI groups did better. When the tools were taken away for the exam, the chatbot group scored 17% worse than students who never had AI at all. The guarded tutor, which helped only after the attempt, avoided the penalty. Same AI. Different sequence. Opposite result.

Evidence: randomized field experiment (Bastani et al., PNAS 2025)

The jagged frontier

In a field experiment with BCG consultants, AI made them dramatically better inside its competence, and 19 percentage points more likely to be wrong on a task just outside it, because fluency hides the boundary. The only defense is your own calibration. Knowing what you think before the model tells you what it thinks.

Evidence: field experiment (Dell'Acqua, Mollick et al.)

So: the attempt

Five minutes. A claim, two approaches, a prediction. Not because struggle is noble. Because the attempt makes the model's answer judgeable, and the judging keeps you the senior partner in the collaboration.

Full argument: Chapters 6, 9 & 12 of The Quiet Cost.

← Back to the studio
Why this matters · 2 min

The reaching was the thing you were doing all along.

"How am I going to pay for it? I don't know. Bottle caps. What does it matter, just come and fix it."
The author's mother, under pressure, finding words no tool would have found. Thirty years later, "bottle caps" still means something in that family. Chapter 7.

Language is how we have an inner life, not just how we report one

Psychiatrist Peter Sifneos described patients who functioned fine but genuinely could not elaborate what they felt. He called it alexithymia. No words for feeling. Roughly one in ten adults scores high on it. Without words, a feeling can only be endured, not examined, communicated, or answered.

Evidence: clinical construct, population studies

Precision helps. Probably.

People who distinguish their feelings finely ("embarrassed, resentful, lonely" rather than "bad") tend to report better well-being and regulation across experience-sampling studies. The catch, honestly stated: this evidence is mostly correlational. Precision is promising, not a proven prescription.

Evidence: correlational (Erbas et al. 2019 meta-analysis)

The difference between receiving a word and reaching for one

All language is borrowed before it becomes ours. Friends and therapists hand us words too. The risk with AI is not the source. It is the speed, a fluent label arriving so fast and sounding so exact that the checking disappears. A word you tested against your own experience does work in you that a word you merely accepted cannot. Recognition is not completion.

Evidence: the book's argument, an open question, honestly flagged

So: the blank page

Ten minutes where nothing names things for you. Then a word treated as a hypothesis, not a verdict. Then the real destination. A person.

Full argument: Chapters 7 & 12 of The Quiet Cost.

← The Why
Five hundred years of getting it wrong

We always predict the wrong loss.

Ten stops. At each one: the loud panic, what actually came true, and the quiet cost nobody named until it was gone.

← Back
The question that started the book

Would it be worse than cheating?

In the fall of 2024, at the edge of a kitchen, the author asked a couple a question that hadn't fully cleared quality control. Over the next two years he put versions of it to more than sixty couples. He learned to do it the way you're about to: ordinary betrayals first, the strange one last.

← Back
Studio · Skill 01

Speak Unscripted

Talking to people in real time is a skill. It fades when you avoid it and comes back when you practice. Here are small, low-risk ways to practice.

Why this matters →

Pick one to try

Start with one that makes you a little nervous but won't cost much if it goes badly. Move up as each one gets easier.

Your track record

How bad you expected each one to be, next to how it actually went. Watching that gap shrink is the point.

← Back
Studio · Skill 02

Ask a Person

A need taken to a person leaves something behind that a need resolved privately cannot. Keep the transaction alive, in both directions.

Why this matters →

The Redirect

What were you about to ask a machine? And who in your life would light up if you asked them instead?

Make yourself askable

Said once, this sentence changes the default. Fill it in, copy it, and actually send it to someone.

"I know a lot about ___, ask me before you Google it."

Someone asked you?

Being needed counts too. The research says it may count most. Log it.

Reach for a person

When the pull is toward a screen for comfort, a vent to a bot, a scroll, name a person instead and go to them. In one trial the supportive bot out-empathized the humans and still moved loneliness not at all. Only the person did.

← Back
Studio · Skill 03

Stay Raw

Prepare yourself, not the other person. The moments that build real trust are the ones you didn't script in advance.

Why this matters →

The Unrehearsed Hour

For two people who know each other well. Draw a topic neither of you has worked out in advance. Phones away, no prep. Let the conversation go wherever it goes. The point is not what you cover. It is the proof that the two of you can talk this way.


The Stop-Line

Got a hard conversation coming up? The urge is to script the whole thing out. Don't. Write down just two things: what you are feeling, and what you actually need from the talk. That is the prep that helps. When you finish, this tool stops you on purpose, so you go in ready but not over-rehearsed.

← Back
Studio · Skill 06

Stay Curious

Curiosity is less about asking and more about not rushing to the answer. Let a question sit for a while before you look it up.

Why this matters →

The Question Shelf

Caught yourself about to look something up? Shelve it instead. When you come back, you'll have to guess before you're allowed to look.


Sit in Boredom

The screen goes dark and nothing happens. That's the whole feature. Your mind will wander. That wandering is the oldest workshop you own.

Parent mode

Today's real question to ask your child. Only ask it if you honestly don't know how they'll answer. Kids can tell a real question from a quiz, and that difference is the whole point.

And when your child asks you something today, try leaving at least one of those questions open instead of answering it right away. That gap is where their curiosity grows, and where they get to surprise you.

← Back to the studio
Why this matters · 2 min

The dining room didn't change. The practice did.

"They describe the dining room the way their predecessors described the kitchen… The way the job is, as one twenty-two-year-old told me recently, 'up for interpretation.'"
A decade of hospitality hiring. Chapter 2 of The Quiet Cost.

Capacity follows exposure

The research on social anxiety is consistent. The problem is not the anxiety, which is normal. It is the avoidance the anxiety produces. Every avoided interaction buys short-term relief and raises the cost of the next one. Every survived interaction lowers it. That is not a moral judgment. It is how nervous systems work.

Evidence: well-established clinical research

The hours really did change

Across 8.2 million U.S. adolescents surveyed from 1976 to 2017, face-to-face social time fell. College-bound seniors in 2016 reported about an hour less per day than their late-1980s counterparts. It does not prove screens caused it. It makes the kitchen's appeal less mysterious. The next demand is explicit, and nothing has to be inferred from a stranger's face in real time.

Evidence: cohort-level descriptive (Twenge et al. 2019)

So: the ledger

Every rep here asks you to predict the dread and then report the reality. The gap between those numbers, watched over weeks, is the exposure lesson made visible: your forecast is almost always worse than the event.

Full argument: Chapters 2, 5 & 12 of The Quiet Cost.

← Back to the studio
Why this matters · 2 min

The ask is a gift in both directions.

"A small inconvenience for a friend put me in a place and posture I would never have engineered on my own." Nineteen years later, they're still married.
The designated-driver night. Chapter 3 of The Quiet Cost.

Doing a favor changes the doer

The Ben Franklin effect: in the classic 1969 experiment, people who were asked to do a researcher a favor liked him more afterward, not less. The mind that helps a stranger updates its assessment: I must have helped them because they're worth helping. Attitude follows behavior.

Evidence: classic experiment (Jecker & Landy 1969)

Being needed is not a burden

Sociologists call it mattering, the sense that others depend on you. A 2024 meta-analysis of thirty studies found it consistently predicts well-being. In a five-year study of 423 older couples, giving support predicted lower mortality; receiving it didn't show the same relationship. When you route your need to a machine, the task gets solved, and no person gets to carry the exchange into the rest of their day.

Evidence: meta-analytic + observational (Paradisi 2024; Brown 2003)

The quiet mechanism

AI doesn't remove your friends. It removes the reason to ask. The need resolved privately at midnight, the small opportunity to be useful diverted to a machine. Each choice is individually rational. The sum is a slow reduction in the raw material relationships are made of.

The book's argument, Chapters 3, 4 & 12
← Back to the studio
Why this matters · 2 min

The conversation went well. And yet.

"Not a failure of the conversation. Not a wound. Something more like a temperature, a slight drop in the room's warmth that she could feel but couldn't locate."
The rehearsed conversation. Chapter 3 of The Quiet Cost.

What rehearsal removes

Four things, by the book's accounting. The information in the unrehearsed reaction, which is only available raw. Co-regulation, two nervous systems meeting difficulty together. Symmetry, because one partner arriving practiced while the other arrives raw is invisible to both and corrosive to trust. And repair, the clumsy reach-back that becomes the evidence a relationship can hold its own weight.

The book's argument, built on couples research

Repair is where the trust gets made

Decades of couples research says conflict is common and largely unavoidable. What distinguishes healthier couples is what happens once strain appears. The de-escalation, the acknowledgment, the reach back toward contact. A pre-smoothed conversation can reduce needless harm and remove the occasion on which two people find out they can repair.

Evidence: longitudinal observational (Gottman tradition)

The line: support vs. substitution

Preparing yourself, knowing what you feel and what you need, makes honesty more possible. Simulating the other person and mapping the route decides the conversation before they've entered it. The Stop-Line tool holds that boundary for you, because in the moment it's genuinely hard to hold alone.

Full argument: Chapters 3, 4 & 12 of The Quiet Cost.

← Back to the studio
Why this matters · 2 min

The reaching is the destination.

"It's because he's alone at the end. You know. The cheese stands alone." Her five-year-old had connected two things she had not, and she let him see exactly what that meant to her.
I Am the Cheese. Chapter 8 of The Quiet Cost.

Curiosity organizes the mind

In a 2014 experiment, states of high curiosity enhanced memory, and not just for the answer. People also remembered unrelated things they encountered while waiting for it. A 2026 meta-analysis across 47 studies confirmed a moderate memory advantage for information people were curious about. The gap between question and answer isn't empty time. It's an activated state.

Evidence: experimental + meta-analytic (Gruber 2014; 2026 meta-analysis)

The interval is where curiosity becomes part of you

Curiosity is not the possession of unanswered questions. It is the tolerance for remaining with them. A child, or an adult, who learns that a question can occupy them for a while, that the wrong guess is information, develops a particular relationship with their own mind. It can move, on its own, toward something it does not yet have. The instant answer closes the gap before the reaching can begin.

The book's argument. The interval, Chapter 8

Boredom does real work

In experiments, people given an undemanding interval improved on creative problems they had already encountered. Incubation, not mysticism. The phone can preempt every such interval because it offers a compelling task at any level of attention. The dark-screen timer just gives one interval back.

Evidence: experimental, narrow tasks (Baird 2012; Mann & Cadman 2014)
← Back
Daily game · new word tomorrow

The Reach

One definition. The word is one you know. Reaching for it, instead of being handed it, is the practice. Retrieval is how words stay yours.

Why retrieval matters →
← Back
Daily game · the answer unlocks tomorrow

Hold This Question

One question a day, and one rule. No looking it up. Commit your answer, and the reveal arrives sealed. Leaving it sealed overnight is the practice. The interval is the game.

Why the interval matters →
← Back
Daily game · new panic tomorrow

The Panic

Five hundred years of people being loudly wrong about new machines. A real, documented fear each day. Name the technology behind it.

Walk the full timeline →
← Back
Daily game · new route tomorrow

The Route

Directions the way a neighbor gives them. Each leg shows once, then it is gone. Hold the whole road in your head, because the questions at the end arrive out of order. The turn-by-turn voice never asked this much of you. That was the problem.

What the turn-by-turn voice took →
← Back
Daily game · new one tomorrow

Support or Substitute

The whole book in one daily question. Here is a real way someone used AI. Did the tool return them to their own capacity, or quietly stand in for it? Support builds you. Substitution replaces you.

Why the line matters →
← Back
Daily game · new pair tomorrow

The Tell

Two versions of the same message. One a person wrote, one a machine polished. Pick the human one. This is not about catching the machine. It is about noticing what the fumble carries, and why the effort was the point.

Practice writing your own →
← Back
Studio · the writing room

The Writing Room

A prompt, and a page that will not write it for you. No autocomplete, no suggestions, no cleanup. The words are yours, or they are nobody's.

Why reaching for words matters →
← Back
Studio · lend an ear

Lend an Ear

Feeling for someone and making them feel heard are two different skills. Most of us are better at the first than the second. This is a place to practice the second, on a hard moment, before you are in one.

The study behind this →
← Back to the studio
Why this matters · 2 min

People feel empathy. They just do not say it.

In 2026, researchers built a platform called Lend an Ear and had 968 people practice comforting someone in trouble, over 2,904 conversations. The finding underneath all of it: people report feeling plenty of empathy, and then systematically fail to express it. How much you feel turns out to be nearly unrelated to how well you communicate it. They called it the silent empathy effect.
Kumar, Poungpeth, Yang, Lambert & Groh, "Practicing with language models cultivates human empathic communication," 2026.

The part that fits this whole site

Half the participants got brief coaching between conversations. It worked, and it worked on a specific thing. It did not make them feel more. It changed what they said. The measurable shift was toward the moves that make a person feel heard, and away from the ones that quietly shut them down. In plain terms: quality, not quantity. Conversations did not get longer. They got better.

And the responses did not become identical or robotic. People still sounded like themselves. The rehearsal did not replace them. It returned them to people, more able to say the thing they already felt. That is the only version of a machine worth keeping around, and it is the whole argument of the book.

Evidence: preregistered randomized experiment, 968 participants (Kumar et al. 2026)

The six moves this studio drills

The three that help. Ask a question that invites them to say more. Name and affirm what they are feeling. Say back what you heard, so they know it landed.

The three that hurt. Jumping to advice they did not ask for. Turning the focus onto yourself. Minimizing the feeling. In the study, coaching pushed people up on the first three and down on these three, and independent raters agreed the results felt more empathic.

← Back
Studio · the unpolished message

The Unpolished Message

A real message to a real person, written by you and sent rough. This studio will not smooth it, shorten it, or make it sound like everyone. The fumbles are the proof a person was there.

Why the polish is the problem →
← Back to the studio
Why this matters · 2 min

The polish is the tell.

People have started running personal messages through AI before sending them. Condolences, apologies, catch-ups, smoothed out. The trouble is that the smoothing is exactly what gives it away. A message with no rough edges reads as a message no one really wrote.
Terry Nguyen, "Why are people using AI to text their friends?", 2026.

Why the fumble carries the person

The effort was the point. A message costs you something: you had to stop, think of this one person, and find your own clumsy words for it. That cost is the signal. When a machine removes it, the words arrive frictionless and land the same way, as though addressed to anyone. The technology writer Kevin Kelly, updating a famous essay for the AI era, put authenticity at the top of the short list of things that stay valuable when copies are free. It stays valuable because it takes a person's time and presence to make, and those cannot be copied.

Evidence: cultural argument, not data (Nguyen 2026; Kelly 2026)
← Back
Studio · companion check-in

Companion Check-in

If you talk to an AI companion or a chatbot for comfort, this is a quiet place to look at what it gives you, and what it might be standing in for. No judgment. Just an honest look, once in a while.

The two things a companion can be →
← Back to the studio
Why this matters · 2 min

Practice space, or the place you stop.

Developmental researchers who study what AI companions do to people frame it as a contest between two outcomes. Stimulation: the companion is a safe place to practice, and the skills carry back to real relationships. Displacement: it crowds out the people, builds dependence, and quietly rewrites what you expect a relationship to feel like.
Sun, Wang & McDaniel, Child Development Perspectives, 2026.

The finding that settles it

In a controlled trial, a supportive chatbot delivered more measured empathy than a human stranger did. And only the humans reduced anyone's loneliness. The bot moved it not at all. Being handed empathy is not the same as being connected to a person. A companion is worth keeping only if it sends you back toward people. The moment it becomes the place you stop, the thing you needed is the thing it is quietly replacing.

Evidence: randomized trial, n=296 (Li, Folk, Singh, Ungar & Dunn 2026) framework review (Sun et al. 2026)
← Back
Resources

Where to go from here.

This site points off itself on purpose. Here is the book, the research the practices are built on, the wider conversation, and real help if you need a person. Everything opens in a new tab. We track none of it.


The book

The Quiet Cost ↗

The full argument. The companion you are standing in was built from its Chapter 12.

The evidence

The studies behind the practices, tagged for what they can honestly carry.

AI-Generated Empathy: Opportunities, Limits, and Future Directions ↗ peer-reviewed review

Ong, Goldenberg, Inzlicht & Perry. A careful map of where machine empathy helps and where it hits its limit.

Practicing with language models cultivates human empathic communication ↗ preprint

Kumar et al. The study behind Lend an Ear. Practicing on a machine made people better at empathy with humans, without making them sound like machines.

A chatbot out-empathized humans and still didn't touch loneliness ↗ randomized trial

Li, Folk, Singh, Ungar & Dunn. Only being connected to a person moved the needle.

Help that boosts you today can tax what remains later ↗ small trial + review

The cognitive-offloading thread: specific skills erode when a tool does the thinking, and design decides which way it goes.

The wider conversation

Context, not endorsement. Voices thinking hard about the same terrain.

Sherry Turkle, Artificial Intimacy

Four decades studying what our machines do to our relationships, aimed straight at the companion moment.

Inventing ELIZA

The 1966 chatbot that convinced people it understood them, and alarmed the man who built it. Everything now was visible then, in miniature.

Kevin Kelly, "Better Than Free" ↗

What stays valuable when copies cost nothing: the human, relational, embodied things a machine cannot make.

For parents and teams

Digital Wellness Lab, Boston Children's Hospital ↗

Evidence-based guidance for families, educators, and clinicians on kids and screens.

Center for Humane Technology ↗

On designing technology that respects human attention and development, with a youth toolkit.

The Anxious Generation ↗

Jonathan Haidt on protecting the growing-up years. Pairs with the case for keeping AI away from beginners on purpose.

If you need a person

If something here stirs up more than a practice can hold, please reach a real person. These are free, and they are people.

988 Suicide & Crisis Lifeline (US). Call or text 988. 988lifeline.org ↗

Crisis Text Line. Text HOME to 741741. crisistextline.org ↗

SAMHSA National Helpline. Call 1-800-662-4357, free and confidential, 24/7. samhsa.gov ↗

Outside the US: Find a Helpline ↗ lists free lines by country.

To find a therapist: Psychology Today directory ↗.

← Back
My account · optional, always

Carry your practice with you.

The site works fully without an account. An account adds one thing. Your practice record syncs, so it survives a cleared browser and follows you to your phone.

What an account stores, and what it never will

Stored: your practice record. Reps done, attempts logged, questions held, words reached, panics called, routes recalled, check-in results, and the notes you chose to keep in your decision log.

Never synced: your writing. The Blank Page, the Stop-Line, the Writing Room, Lend an Ear, the Unpolished Message, and the Companion Check-in all stay on your device and never reach our server. And never your behavior on this site. No analytics, no tracking, no engagement metrics. That part of the deal does not change.

Sign-in: a password you choose, or a one-time email link if you would rather not keep one. Passwords are stored hashed by the sign-in service, never in plain text and never by us.


Your record

Everything you have practiced.

A plain record of the work. What you did, when, and how the daily games went. It lives on your device, and syncs when you are signed in. We count the practice. We never count your time on this site.

← Back
About

The site practices what the book preaches.

This is the companion to The Quiet Cost: What We Stand to Lose in the Age of AI by Michael McNamara, a book about the capacities that quietly fade when technology removes the friction that built them.

Our design rules

Success is you leaving. Every practice points off-screen. Toward a call, a person, a page, a question left open. We count reps, asks, and calls. We will never count your time on this site.

No chatbot. A site about the cost of frictionless answers will not greet you with an answer box.

Nothing leaves your browser unless you say so. By default your logs live on your device and nowhere else. If you choose to make an account, your practice record syncs so it can follow you between devices. Your writing never syncs, and there are still no analytics. That is the deal, and we think it is the right one.

Honest evidence. Every claim is tagged with what the research can actually carry: experimental correlational hypothesis. Certainty is not the product.

The book

The full argument. Five hundred years of mispredicted losses, the evidence on what AI performs versus what you practice, and the six things you can actually do.

Visit the book’s site ↗