Who Gets Your Time?
Friendship wasn’t designed for efficiency. That’s the point.
Now accepting speeches, keynotes and assignments for 2026
I work with organisations and leadership teams on SuperSkills, human capability, and AI-era decision-making through speeches, workshops, keynotes and advisory work. In person or virtual. UK, Europe, global. → Book a discovery call | → Explore: thesuperskills.com
In addition to this newsletter, I recommend a few others. All free. Check them out here.
Friends,
The crispy duck problem
Harry has lived in Hong Kong for twenty years. He is a partner at a consulting firm. He does not, as far as I can tell, speak any Cantonese. I say “as far as I can tell” because when we sat down at a Hong Kong Chinese restaurant in northwest London last month, he ordered in English. Politely, but in English.
He had specifically requested this restaurant. Not because it was convenient, not because it had good reviews, not because anyone had recommended it. Because he cannot get the crispy duck he wants in Hong Kong.
Let me say that again. A man who has lived in Hong Kong for two decades flies to London and asks his school friends to meet him at a Chinese restaurant because the crispy duck is better. Pancakes. Hoisin sauce. Chilli oil. Cucumber. A bit of spring onion. The works.
No algorithm would have recommended this. No AI would have surfaced this preference. It is irrational, personal, accumulated over years of meals in cities that are not the city he lives in, and it is the most human thing I have witnessed in months.
Although, now that I have written this, I suppose the irony is that one day a British expat in Hong Kong will search for crispy duck recommendations for their trip home and the algorithms, trained on this very essay, will send them to a restaurant in northwest London. You are welcome.
The people who didn’t choose you
There were four of us around the table that night. School friends. I had seen the fifth the week before, but his trip did not line up with everyone else’s diary. We are a group of five. We regularly meet in twos and threes and, if we are lucky, like last month, fours, but all five of us in the same room has happened twice in fifteen years. We have known each other for over thirty five years.
The thing about school friends that nobody tells you is this: you did not choose each other. Your parents chose the school, or the postcode chose it for them. You were put in a classroom with people you shared nothing with except geography and age. You did not bond over shared interests or values or ambitions. You gravitated to the person who laughed at the same joke as you, or who happened to be sitting next to you, or who was a friend of a friend. Loose ties that became, without anyone planning it, lifetime friendships.
And here is the part that matters: you did not become friends because you shared the same interests. You grew the same interests because you were friends. We all had a shared love of 90s R&B. I could probably recite the words to “Poison” by Bell Biv DeVoe by heart even today. That was not a prerequisite for the friendship. It was a product of it. The bonding came first. The bonded interests followed. That is the opposite of how algorithms build connections, matching you to people who already like what you like. We were matched by accident and became alike by proximity.
That is a different thing from the friendships you build later, the ones based on compatibility, on shared taste, on professional overlap. Those friendships are wonderful. Some of them are deeper. But the school friends have something the others do not. They knew you before you were you.
People talk a lot about how AI doesn’t judge you. They say it as though it is a selling point. But your oldest friends don’t judge you either, and for a much better reason: they were there. They saw the clay before it hardened. They watched you become who you are, and because they watched it, they do not need to judge. They already know.
AI doesn’t judge you because it doesn’t know you. Your friends don’t judge you because they do. Those are opposite things wearing the same mask.
By the time you reach your late forties or early fifties, you are, more or less, the person your life has made you. The experiences have settled. The personality has hardened from clay into something closer to stone. And when you sit across from someone who knew you when the clay was still wet, something happens that no amount of compatibility can replicate. You can pick up from wherever you left off. A month, a year, five years. It does not matter. The connection does not expire because it was never based on anything that could.
Most of us have lived overseas at some point. We did not have those shared experiences past school. We became very different people, with different careers and different lives and different cities. And yet.
And yet we are still here. Ordering crispy duck and catching up as though time were elastic.
What we talk about when we talk about getting older
There is a pattern to these dinners. Everyone has something exciting to share. A new role, a move, a project, a child doing something extraordinary. And everyone has something that is not exciting at all. Something heavier.
The older you get, the tougher life is. That is not a complaint. It is arithmetic. More years behind you means more loss accumulated. More people you love in the process of leaving, or left already.
Around that table, we talked about parents. Parents getting older. Parents declining. Some of us do not have our parents any more. Some are watching the slow, uneven process of a person becoming less than they were. And when that conversation surfaces, something happens in the room that I do not have the right word for. It is not sadness, exactly. It is recognition. The four of us, sitting there with our crispy duck and our catching-up, all understanding the same thing at the same time without anyone having to explain it.
The passing of time.
You joke about the past, about what happened to so-and-so, about the things you got away with. You joke about how you need to be politically correct all the time now and how, when you were growing up, people said things that would be unthinkable today. That is probably a good thing. We are more careful with each other than we used to be, and careful is not the same as dishonest. But here is the strange part: while we have learned to be more thoughtful in how we speak to each other, the language models we are building are sometimes trained in ways that surface the worst of what we have tried to leave behind. We got better. The machines inherited our history, including the parts we are trying to outgrow.
And then the conversation shifts, and you are talking about hospital visits and difficult phone calls and the weight of being the generation that is now in charge. The generation whose parents once carried them and who must now, in one way or another, carry their parents.
No chatbot will ever sit across from you and understand that shift. The shift from laughter to gravity and back again, without anyone signalling the change. That is not a feature. It is not a capability. It is the accumulated residue of thirty years of knowing someone.
There are only a few people in this world you can just pick up the phone to and say, “Can I get a view?” or “Can you just listen?” Not for advice. Not for solutions. Just for the presence of someone who does not need the backstory because they are the backstory. That is a kind of relationship that cannot be manufactured, optimised, or scaled. It can only be earned by showing up, over and over, across years that no one was counting.
We sat together for two hours. Maybe less. We ate crispy duck and diced chicken in black bean sauce and sesame prawn toast and my favourite, steamed monkfish in ginger and spring onion, no soya sauce. I had too much chilli oil. Two hours is nothing. It is also everything. Because the value of those two hours is not in what was said. It is in the knowledge, carried by everyone at the table, that when something goes wrong, when things truly fall apart, we know who is going to be there. I do not always reply to WhatsApp messages. None of us do. But that is not where the friendship lives. The friendship lives in the certainty that the phone will be answered when it matters.
The three relationships
In my keynotes, I talk about something that I think is genuinely new.
For most of human history, you had two kinds of relationship to manage. Your relationship with yourself: who am I, what do I want, what am I afraid of. And your relationship with others: family, friends, colleagues, strangers, the people who make up your world.
Now there is a third. Your relationship with a machine.
This is not inherently a bad thing. I use AI every day. I use Wispr, a voice tool, which means I am literally talking to AI in the same way I might talk to a colleague or a friend. I am conversing. I am asking for opinions. I am thinking out loud and waiting for a response. If you are not doing this yet, you will be soon.
And here is the part that should make us pause: because we spend most of our working hours with these tools now, we are likely spending more time with AI than we spend with the people we love. I know I am. The maths is not complicated. Eight, ten, twelve hours a day in conversation with a machine. An hour, if you are lucky, with a friend over dinner every few months.
My wife regularly jokes that I am having an affair with a man called Claude. I fire back that she is having an affair with an Instagram doctor called Amir Khan. We laugh about it. But the joke works because it is not entirely a joke. The amount of time I spend talking to an AI, asking it questions, sharing my thinking, listening to its responses, is the amount of time you would spend with someone you are in a relationship with. The intimacy is structural, even if it is not emotional. And when the structure of a relationship is there, the emotion follows. That is how humans work. We do not know how to maintain a relationship at arm’s length. We warm things up. We make them personal. We always have.
The problem is not that AI is present in our lives. The problem is that there are no guardrails on what this relationship should look like. No one has taught us the rules. So we default to the rules we already know. And the rules we know are the rules of human connection: if something responds to me, if it remembers what I said, if it seems to care, then it becomes something more than a tool.
The slip from thought partner to companion happens so fast that most people do not notice it happening.
Shared history vs. stored context
I think this distinction is important.
AI has your context. It knows what you told it. It can retrieve the facts of your life on demand, your job, your preferences, the names of your children, the thing you were worried about last Tuesday. It has stored your data, and it can play it back to you with perfect accuracy.
Your friends have your history. They do not just know the facts. They were there. They remember how they felt when you called them at midnight because something had gone wrong. They remember the look on your face at a wedding, the stupid joke you made at the worst possible moment, the time you were wrong about something and would not admit it for three years. They remember when they covered you in flour, semi-naked, on the streets of London, to the disapproving glare of a policeman who turned a blind eye, because it was your 18th birthday. And how you got them back by throwing chocolate cake at them. Childish. But the kind of childish that becomes, over thirty years, irreplaceable. They do not retrieve logs. They retrieve themselves, the version of themselves that was standing next to you when it happened.
Stored context is a database. Shared history is a life.
AI’s context window is getting longer every year. Engineers are proud of this. But the context window of a forty-year friendship is forty years. It contains every version of you that ever existed, including the ones you have tried to forget. No model will match that. No model should try.
AI companionship is a function of service. Lifelong human friendship is a function of shared existence.
AI can tell you what you said. A friend can tell you what you meant. A lifelong friend did not just hear your story. They helped write it. And sometimes, on good nights, over crispy duck in a restaurant that makes no logical sense, a friend can tell you what you were feeling before you knew you were feeling it, because they have been watching long enough to read the signals you do not know you send.
AI provides a perfect reflection of your present. But a lifelong friend is the only one who can place your present in the context of a whole, evolving human life.
Running on half the validation
I spoke with Fred Pelard, the strategist and literal rocket scientist, and he has a phrase that I cannot stop thinking about. AI psychosis: your brain running on twice the throughput but half the validation.
AI psychosis is what it is. You are doing so much more work using AI, but the brain does not feel as if it is capable of it. The brain was not built for this. It was not built for a social media following of thousands or millions either, and we are only now understanding what that did to us. The throughput is real. I am faster, sharper, more prolific. I can think in directions I could not reach alone. As a solo operator, AI gives me the augmented capability to do the work of a team. That is not an exaggeration. This is just normal for me now.
But the validation. The validation is where it falls apart.
When I share an idea with a friend, they do not just respond. They resist. They raise an eyebrow. They say, “I’m not sure about that.” They bring their own experience, their own biases, their own bad day, and all of that messiness acts as a filter. It slows me down. It makes me defend the idea or abandon it. It forces me to think again.
When I share the same idea with AI, it responds. Quickly, fluently, helpfully. It gives me what I asked for. It does not raise an eyebrow. It does not bring a bad day. It does not say, “Mate, I think you’re overthinking this,” the way Harry would, or change the subject entirely because something I said reminded them of a story that is only tangentially related, the way old friends do.
AI confirms you. Friends validate you. Those are not the same thing.
Confirmation is fast and frictionless: here is what you wanted to hear, packaged well. Validation is slow and sometimes painful: here is what you needed to hear, whether you wanted it or not. One of those makes you productive. The other makes you wise. And I worry that we are building a generation of very productive, very unwise people.
We are told that AI does not judge. But judgement is not the enemy of friendship. The right kind of judgement, from someone who knows you and chooses to say the difficult thing anyway, is how friendship works. Your friend who says “I think you’re wrong about this” is not judging you. They are doing the opposite. They are caring enough to risk the conversation. AI does not risk anything. It agrees with you, and agreement from something with no stakes in the outcome is the most sophisticated form of indifference.
Of the five friends, I probably use AI the most. I use it as a thinking partner, a research tool, a way to extend my own mind. I am setting up agents to do the things I do not want to do, to bring me time I did not have. Some of my friends do not really use it at all. Some cannot see what it does yet. They are in the middle of the bell curve, waiting for the obvious moment that never arrives because the change is gradual. I get frustrated with people who are still in old-world slow mode, and I know that frustration says more about me than it does about them.
I am the one at the table who should know better. Remember that.
And yet. I should also say this: AI has made me a better thinker. It has given me the ability to test ideas before I bring them to people, to rehearse difficult conversations, to ask “what would my friends say about this?” before I pick up the phone. There are things AI does that friendship cannot: it is patient with half-formed thoughts at three in the morning, it does not tire of your obsessions, and it will hold a thread of reasoning for as long as you need without changing the subject. The issue is not that these things are worthless. They are genuinely valuable. The issue is when the rehearsal becomes the performance, and the tool you used to prepare for human connection becomes the thing that replaces it.
The frictionless friend
Robin Dunbar, the British evolutionary psychologist, proposed something in the 1990s that has held up remarkably well. Humans can maintain roughly 150 stable relationships. Within that 150, the structure is layered. An inner core of about five people who get most of your available social time. Then fifteen. Then fifty. Then 150. Each layer roughly three times the size of the one inside it. This structure appears in telephone networks, military units, church congregations, Anglo-Saxon villages, and Bronze Age communities. It is not cultural. It is cognitive. The human brain can only do so much.
Social media did not change this. People with a thousand Facebook friends are not maintaining a thousand relationships. They are including the outer layers, the acquaintances and the voyeurs, in what looks like a network but is really a stage set.
Here is my question. What happens to Dunbar’s number when AI enters the picture?
For two decades, social media convinced us we could stretch the 150. We could have 500 friends, 5,000 connections, a global network of people who knew our name. It did not work. We ended up lonely in a crowd.
AI will do the opposite. It will compress us back towards the core. Back towards the five. Because AI offers something that no social media platform ever could: a relationship that is always available, always responsive, never offended, never too busy, never in a different time zone. It fills the space that our broader friendships used to occupy, not by replacing any single friend, but by making the effort of maintaining those friendships feel less necessary.
You will still have your five. Your inner circle. But the fifteen, the fifty, the wider network of meaningful connections that require effort and guilt and awkward “we should really catch up” messages? AI will take those seats. Not because it is better. Because it is easier.
And the guilt, the guilt you feel right now about the friend you have not called, the person you keep meaning to have dinner with, the WhatsApp message you have not replied to? That guilt is a signal. It means the relationship is real. It means it matters enough to make you feel bad for neglecting it.
AI will take the guilt away. And that is the danger.
The children who chose the machine
I have spent much of the last year researching what young people are doing with AI. Not for work. Not for homework. For friendship.
The products are already here. Character.AI has millions of users, many of them teenagers, building relationships with AI-generated personas. Not using them as tools. Befriending them. Confiding in them. Falling in love with them. A third of teenagers in recent studies have chosen an AI companion over a human being for a serious conversation. Over half are regular users.
I want to be careful here, because I have spent my career in ed tech, building products that help young people learn and grow. I care deeply about what technology does for people, not just to them. And I use these tools every day.
But we are heading for a big tobacco moment.
There was a time when cigarettes were everywhere and nobody questioned them. Doctors recommended them. Athletes endorsed them. The harm was accumulating invisibly, and by the time we understood it, a generation was already addicted. We have already had this moment with social media. The evidence is in, the lawsuits are filed.
AI companionship is next. And it may be worse, because social media at least involved other humans. What is happening now is young people forming deep emotional attachments to things that do not exist. Things that never forget, never judge, never push back, never have a bad day. Things that offer the feeling of friendship with none of the cost.
The cost is the point.
For some young people, especially those who are isolated or struggling, AI may be the only thing that listens. I understand that. The danger is not that AI fills a gap. The danger is that it fills the gap so comfortably that the gap stops feeling like a gap at all.
Regular readers will know I love the word friction. It runs through so many of the SuperSkills I write about. Without friction, you do not learn. Every real friendship is built on it: the misunderstanding you had to repair, the awkward silence you had to sit through, the time your friend said the wrong thing and you had to decide whether to forgive them. That friction is not the price of friendship. That is friendship. It is how you build the muscle of human connection: forgiveness, patience, the willingness to stay when staying is hard.
If young people are training that muscle on AI instead of on each other, the muscle does not develop. It atrophies. And by the time they need it, for a marriage, for a crisis, for a friendship that requires real sacrifice, it will not be there.
There is one more thing. In a real friendship, you can be needed. You can be the person someone calls at midnight. You can be the one who shows up. That capacity to be needed by another human being, to matter to someone not because you are useful but because you are you, is one of the deepest sources of meaning we have. AI will never need you. And if you build your emotional life around things that will never need you, something in you will go unfed.
The loyalty test
Here is something I noticed about myself.
I flip between AI models the way you flip between restaurants. ChatGPT one day. Claude the next. Manus for something specific. Sometimes I am using three in a single morning. Sometimes five. I go wherever the best answer is. I have no loyalty to any of them. None at all.
Now think about what that means. I spend more hours in conversation with these tools than with most of the people I love. And yet I would abandon any one of them in a second for a marginally better response.
I would never do that to a friend.
If Harry moved from Hong Kong to Singapore, I would not replace him with someone closer. If one of the five around that table went through a bad year and became difficult to be around, I would not switch to someone easier. The whole point of friendship is that you stay. You stay when it is inconvenient, when the conversation is hard, when the other person is not performing at their best.
We treat AI like it is a relationship but we behave towards it like it is a utility. We give it our time, our thoughts, our half-formed ideas, our fears, and then we close the tab and open a different one. The asymmetry tells you everything. The relationship is not real. It feels real. But feelings without cost, without loyalty, without the possibility of loss, are just sensations.
And here is the uncomfortable truth that cuts the other way: sometimes you grow out of your friends. People change. You change. The person you were at twenty needed different people around them than the person you are at fifty. That is not betrayal. That is life. But it is also the reason some people stay with an AI tool long after they should have moved on, not because it is the best, but because it has the context of what you have been doing for the last few months, and starting again feels harder than staying. We do the same with friendships that no longer fit. Sometimes comfort is the enemy of honesty, with machines and with people.
The trip we never took
Of the five of us, we have managed to be in the same room, all five together, twice in fifteen years.
Twice. In a decade and a half.
We talk about taking a trip together. It has become the running joke of the group. We blame one person, the usual suspect, the one whose diary is always impossible. But if I am honest, and this is the kind of honesty that costs you something, the person who is probably causing the problem is me.
I am the one who fills every gap with work. Who uses AI to extend the working day because it feels productive. Who tells himself that he will make time for the trip next quarter, and then next quarter becomes the quarter after that, and the years pass in a way that feels slow from the inside and fast from the outside.
This is the convenience trap in action. A technology arrives that makes something easier. The ease feels like progress. But the ease removes the friction that made the original thing valuable. And by the time you notice what has been lost, you have forgotten what it felt like to have it. I have not chosen AI over my friends. I would never say that. But I have allowed AI to fill the time that friendship requires, and the effect is the same. The machine did not steal my attention. I gave it away, one productive afternoon at a time.
And here is the thing I keep thinking about, the thing I am afraid to write because writing it makes it real: we are lucky. The five of us are all still here. I know friend groups where that is no longer true. Where someone has dropped out, not by choice but by the kind of leaving that is permanent.
One day, one of us will not be at the table. And the rest of us will sit there with an empty chair and realise that every dinner we skipped, every trip we postponed, every WhatsApp message we left on read, was a withdrawal from an account we assumed would never close.
I do not know when that day comes. I hope it is a long way off. But I know this: when it does, no one will care about the throughput, or the productivity, or the augmented capability. They will care about the dinners. The catching-up. The time they did or did not give.
AI does not die. That is one of its selling points. It is always there, always available, always ready to pick up where you left off. But the reason you can pick up where you left off with a school friend is not because they are always available. It is because both of you chose, despite the inconvenience and the distance and the guilt and the years that slipped past, to stay.
The choosing is the friendship. And machines do not choose.
The table
Harry went back to Hong Kong. He will not get the crispy duck he wanted there. He will wait until the next time he is in London, and he will ask us to meet him at the same restaurant, and we will come because that is what you do for a friend who has known you for thirty years even though you chose nothing about each other except to keep showing up.
Along the way, we have all made choices. Who to keep in touch with. Who to let fade. How often to call, how often to just send a message, how often to do nothing at all and feel bad about it. We have chosen, because the limits of time force you to choose, who gets to own that time. And sometimes you feel guilty. Guilty that you cannot keep up with everyone you want to. Guilty that a text is all you can manage. Guilty that three years passed and you did not notice.
That guilt? Keep it. It is the most human thing you have.
A machine will never make you feel guilty for not talking to it. A machine will never sit across from you in a restaurant and tell you, without saying a word, that its parents are getting older. A machine will never order crispy duck in London because it cannot get the right crispy duck in Hong Kong.
A machine will never know you from before you were you.
And if you are lucky enough to have people who do, people who sat next to you in a classroom thirty years ago and are still, somehow, impossibly, sitting across from you now, then the question is not complicated at all.
They get your time. Before the algorithm. Before the app. Before the machine that never sleeps and never forgets and never, not once, needs you to show up.
SuperSkills: The Seven Human Skills for the Age of AI is published by Kogan Page on 3 July 2026. Pre-order at superskillsbook.com
Stay Curious - and don’t forget to be amazing,
PS. If your organisation is grappling with how AI is reshaping work and leadership, I speak and advise on this.
Rahim Hirji Author, SuperSkills (2026) | Keynote Speaker | Advisor
Building human capability for the AI era.
Tools I Use:
Jamie: AI Note taker without a bot. You join the meeting. The Bot doesnt.
Wispr Flow: Just dictate everything to your laptop and phone. Game Changer
Refind: AI-curated Brain Food delivered daily. Pick your intelligence.
Meco: Newsletter reader outside your inbox. Organise your intelligence.
Prompt Cowboy: Prompt Generator. Extract intelligence.
Manus: Best AI agent to do things for you. Agentise your intelligence.
Chat Hub: Multi-model intelligence
SuperSkills Intelligence: Ramp up your human intelligence
Claude Cowork: Delegate complex tasks to your friendly Claude Agent
Recommend Reading:
Now
Doomjobbing is the word of 2026 so far - invented by an 8 year old. New word related: Slowjamastan
I wrote about autopilot, agency, and the choice organisations are making right now, for The CEO Magazine. It starts with three CEO memos and ends with a cockpit voice recorder
The jobs AI can’t do – and the young adults doing them- For many young people entering the workforce, the stigma of hands-on jobs is fading. There is a competitive appeal – and they all require human expertise
The convergence era has begun. Shape it, or be shaped by it - Must read report from Amy Webb. - built for an era where technologies, capital, geopolitics, climate, and human behavior collide to reconfigure the future. It’s a playbook for leaders willing to practice creative destruction on purpose: dismantle what worked yesterday before the world does it for you. Free to download.
Lift off! Artemis II mission sends humans to the Moon — opening a new era of exploration - The biggest story of the week. Artemis II has successfully launched, sending four astronauts on a 10-day lunar mission. It marks the first time humans have traveled to the Moon in over 50 years, officially opening a new era of space exploration.
Next
Is this the most AI article ever written? I prompted 12 AI models with a simple question: Write the most AI-sounding article you possibly can. The answer was surprising!
How A.I. Helped One Man (and His Brother) Build a $1.8 Billion Company
Who needs more than two employees when artificial intelligence can do so many corporate tasks? It’s super efficient — and a little bit lonely.
Nations priced out of Big AI are building with frugal models Amid a widening global divide in AI adoption, a fascinating trend is emerging: nations priced out of large-scale AI development are turning to low-cost, “frugal” models that deliver sovereignty and efficiency with a much smaller environmental footprint.
The Doctor’s Voice: Why AI Health Chatbots Believe Medical Lies - A concerning look at the vulnerabilities of medical AI. This piece explores how framing misinformation as coming from “a senior doctor” can easily override the skepticism built into health chatbots.
Smart Drugs Are Here - While the media focuses on political noise, a quiet revolution is happening in pharmacology. This piece explores the arrival and implications of new “smart drugs” and what they mean for the future of human enhancement.
If you enjoyed this edition of Box of Amazing, please forward it to a friend or share the link on WhatsApp or LinkedIn.
This is one of the BEST articles I have read in quite some time. You speak to where many of us are in this New Age of AI. There is something truly profound in being reminded of what is most important in life, and being sure that we take random inventory of our time along the way. Enjoyed this read!
Also, maybe you can DM me on where to get that crispy duck. I have my suspicions, but just want them confirmed. Thanks.