This article talks extensively about death, even if it’s really about living. If you are feeling mentally unwell, please contact your local mental health crisis hotline.

When did you last sit with the fact that you are going to die? Not abstractly that death applies to everybody, but yours in particular. I have gone years without looking at it directly, and those were exactly the years when I could still have done something with what I’d have found.

In Buddhism there is a practice known as Maraṇasati, also known as mindfulness of death, where one frequently keeps death in the forefront of thought. This might seem off-putting, since death is often a taboo subject to talk or even think about. But keeping death close at heart is the key to liberating yourself from death’s grip, and I found it tremendously insightful in shaping how I want to live.

I remember my first distinct existential sense of my demise when I was six or seven years old, and funnily enough it was from watching a TV drama about the death of a character. I don’t remember the plotline anymore, but I remember viscerally understanding the concept of my own death and feeling a sinking sensation in my throat that spread to my chest, and I had to sit down. I had multiple deaths in my extended family, but the next distinct moment was in my 20s when I attended a funeral of a fellow military service batchmate who had a tragic auto accident. The idea that someone so close to my age and also so alive just a few months ago was now dead really hit home for me.

In these moments, I was brought to the present, to the now, perhaps too abruptly, but they also brought clarity that whatever seemed to matter just minutes ago didn’t really at all.

As it happens, psychologist Laura Carstensen studied how people’s priorities shift as they approach the end of life, and the pattern is consistent (Taking Time Seriously: A Theory of Socioemotional Selectivity). People who believe they have a long stretch of time ahead want ambitious goals, new experiences, and to meet new people; they invest in the future because there’s a future to invest in. People who believe their time is short want the opposite: their oldest relationships, familiar food, a few good hours, the pleasures that they already know they like.

It turns out which frame you take depends more on your perceived horizon than on your age. Convince a young person their time is short and their values move toward the present; conversely give an old person a genuinely open-ended future and their values move back out. So the difference between a future-oriented life and a present-oriented one isn’t something that happens to you at 70. It’s a setting, and it responds to what you believe about how much time is left. It just happens to correlate with age, since that’s when the question of how much time you have left starts to press on you.

I find this hopeful and a little scary at the same time, because if the viewpoint is adjustable then nothing is stopping you from adjusting it, and we had that option the whole time. Of course we always had that option.

Carstensen also found that the present-oriented outlook makes people less critical, more attentive to the pleasant and less alert to the unpleasant, and this is partially why the elderly are disproportionately targeted by scams.

The goal isn’t to live permanently as if you have six months left, because you’d stop maintaining anything and you’d get robbed. The goal is to be able to sift through your priorities, both short and long term, and assess them more fairly.

You have the time now. What regrets do you imagine you would have? What would you change today, knowing them?

Subscribe now

We always end up losing the point

Look at what you’re keeping score with right now. The job title, the compensation, the status, the followers, the net worth, and even the streak on whatever app you’re loyal to. We always end up choosing those the way most people choose, because they’re defensible, not because anyone ever established that they were worth it to us. The end of life is where that lie becomes obvious and irreversible. And as I said earlier, choosing differently was always allowed, and in the middle of life it’s just as real, and you still have time to change it.

Being mindful of death isn’t meant to be morbid. The point is to bring the immediacy of life to the present. That said, this can be a disorienting practice and easing yourself into thinking about your death might be helpful. Find small ways to begin: when you see a dead animal or insect, whether it be in a supermarket, on the road, or even in your house, don’t be so quick to look away or to clean up. Instead, pause and reflect that this was a creature that was alive not too long ago and notice the stillness of its body. What feelings and emotions arise within you? Note them too.

Gradually spend more time throughout your day being aware of your own mortality. For example, you might imagine yourself at your deathbed, gasping for your last breath, or at your own funeral, surrounded by family and friends. What thoughts and feelings arise now for you? It’s OK to have intense emotions such as fear, and powerful thoughts that try to pull you away. Be still. Notice all of this.

Coming out alive

So the case for thinking about dying isn’t that it makes you appreciate the sunset. It’s that everything that goes wrong at the end starts by going wrong in the middle too, quieter and, most importantly, more reversibly. You track what’s easy to count instead of what’s good. You never say out loud what you’re hoping for, so you follow other people’s wishes, and by the time you notice, your whole calendar is filled with stuff you didn’t want or need.

Be aware of your mortality and write down honestly what you want out of this life, then find the one thing on your calendar that isn’t consistent with what you want or need, and cancel it. If someone came to mind, call them or write to them, even if it’s just to say “Hi, I was thinking about you. How are you doing?” That’s the whole practice, and you can start it forty years early.

On the other hand, acting without much thought has its own strange failure mode: overoptimization. With just doing, it’s easy to just focus on small goals and end up maximizing sleep, training, calendars, and routines to the extent that even after doing all of it correctly, I can end most days and weeks with the sense that nothing meaningful has happened.

In other words, not overthinking is the right instinct, but it isn’t enough on its own, because it can creep into overoptimizing, which is an attempt to control something that doesn’t respond to control.

Here’s what I have found works instead.

Subscribe now

The two types of attention

There are two types of attention. One is zoomed-in and task-locked, and the other is wider and works like radar, so I think of the first as a spotlight and the second as a lantern.

These days, everything about how we work and live emphasizes the spotlight and neglects the lantern. Notifications draw your attention, OKRs narrow your focus, and calendars are packed to fill your working hours with no gaps in between. The spotlight is genuinely useful for driving progress, but it can’t make meaning, it can only execute against meaning that was already defined.

You might think that working has to use spotlight attention, but when I first started in Big Tech, we were encouraged to push on new projects and ideas as long as they met the vision. The first few months weren’t about metrics, they were about the mission statement, the culture, and stories about why we did what we did. Deliberate or not, that widened the scope of our attention and brought in ideas from every direction, and it’s the kind of clarity you only get from the lantern, because no amount of narrow focus will produce it for you.

In life, you want to take the wide lens, zoom out and be aware of what’s happening everywhere, not just in a narrow band of life that you happen to be focusing on. This is why you can have a bad day at work, or in the stock market, or in your project, and feel like your entire world is collapsing despite a loving family, close friends, and a beautiful home. The spotlight is the right instrument for something you can plan, and the lantern is the right one for something you can’t.

Your life is complex, not complicated

A complicated system has many parts, but also assembles the same way every time, like an assembly line, so you can plan it. On the other hand, a complex system changes as you interact with it, adapts, surprises you, and refuses to move in a straight line.

Your career is complex. Your team is complex. Your family is complex. Your life is complex. We confuse the two because both of them feel hard, and then we bring planning to a problem that doesn’t take well to planning.

What works instead are rules of thumb and a willingness to go with the flow. I don’t cook every day because the dishes I like to cook tend to be elaborate, so on the days I’m making Bak Kut Teh or Hainanese Chicken Rice I tell the family in advance, and the rest of the week we’re content to cook and eat simply. In other words, nobody plans the week’s meals. There’s a default and there’s an exception, and that’s enough.

Even while I was leading a 100-person org in Big Tech, I would always leave a real block of time in the week empty and guard it zealously. It might seem wasteful, but that free time is the only condition under which anything unplanned can happen. Brilliant ideas, new collaborations, and critical insights happen in this block, not because it was planned, but precisely because it was unplanned.

I won’t pretend this was free. Holding that block meant saying no to 1:1s or meetings with stakeholders and peers, and there were quarters where nothing came of it at all and I kept it anyway, which is harder than it sounds when the free block of time is just showing up on your calendar. Going with the flow costs you the ability to precisely point at what it produced because it could be the start of an amazing launch, or nothing at all.

Reach out to someone with no agenda. Not for a favor and not for networking, but because a memory or an article brought them to mind. Or chase something for no reason beyond curiosity.

It will feel like a waste of time. It won’t be.

Rituals do work

Within the activities that you repeat, some are routines and some are rituals. A routine is something you repeat to get more efficient. For example, training to run a faster 5K is a routine. A ritual is something you repeat because the repetition itself becomes meaningful, like visiting the same place together every year as a family. For example, I hike-run with my son on Fridays. Partly it’s good for him to get outside, and partly it’s an hour every week where we chat or we don’t, but either way we are together.

The reason the ritual with my son works is that I know it’s happening and he and I don’t know what will happen in it. That’s the whole point. We live in those moments and I soak in them. While they don’t last forever, they are everything that I need at this point.

So pick one recurring moment this week and redesign it. Not to be more efficient, but to make it mean something to you.

Pay attention

We make life complicated because we see that life is complex. But the response to complexity isn’t a better plan, it’s paying attention, because the good stuff is waiting to be discovered and you have to be in a state to see it.

None of this really cures the spiral. I still find myself going over the cons late at night, and my greatest enemy is still myself on those nights. What changes is that the spiral becomes one thing in a wide frame rather than the whole frame, and it turns out that’s most of the difference.

Expect little, notice a lot, and you’ll find life is already full.

The timing of my latest parkrun is not a typo. Even though I did about six weeks of training and even got the best numbers I’ve ever produced during training, I still ran the same course down to the same second. Sigh.

This article will be shorter than my usual, with no charts and fewer numbers. It’s a sober realization of what I need to do to get to where I would like.

I closed parkrun #4 with a sentence I was pleased with: “You train for what you want to do. Today I trained for a fast two miles and ran a fast two miles. Next time, I want to have trained for the whole five.” Then I went and trained for a fast mile, four times over each week, and ran a fast mile, and then broke down. Go figure.

At the end of the day, you should definitely train for what you want to race.

Subscribe now

The training got me genuinely faster

In July, I ran one interval session per week, and the long and short of it was that I could run mile repeats faster than race pace, at lower Rates of Perceived Exertion (RPE) than ever. I felt like I was faster, and I was.

On August 12, three days out from the parkrun, the 2.4km tune-up came in at 10:30, which is 4:23/km (7:03/mi). That’s faster than the average pace of the best interval session I’d ever run, held continuously as a single effort. A fairly decent 2.4km, even I would think. As I remember it, that was a typical pace for 20-year-old me. Hey, I never said I was a fast runner. :-)

How the race went down and what it entails

The timings mostly say it all: Mile 1 at 7:08, mile 2 at 7:33, mile 3 at 7:48, and the final stretch at 7:13.

I basically went out too fast because the 2.4km tune-up told me I was fitter than I was. Going out too fast cost me something. But seven seconds per mile at the start shouldn’t sabotage the whole race, so most of what happened after mile one was not just a pacing error unwinding.

Parkrun #4 left a question open for me: do I lack the durability, or the aerobic ceiling? This race answers it. I averaged 90 percent of my maximum heart rate and was ten beats from my max. So it wasn’t my heart running out of speed, it was my legs.

Something needs to change.

Each week, I run about 30–40 miles over 6 days. In July alone, my running took about 31 hours, so this shouldn’t be a volume problem. But my overall average pace across everything is about 13 minutes per mile, because most of my running is steep trail running with something like 25,000 feet of climbing a month. At that pace and gradient I kinda have to shuffle and hike.

So roughly ninety percent of my training time goes into a running pattern that is nothing like what a 5K race asks for. When I counted the work specific to a 5K, with warm-ups and cool-downs stripped out, it came out to only about eight percent of my training time.

I’m therefore moving to two track sessions a week instead of one. That takes my 5K-specific work from 30-40 minutes a week to around 60–80 minutes a week.

I’m going to do tempo sessions on Tuesdays, where I will run 7:35/mi for 4–5 miles continuously to train my legs to keep working even as they fatigue. This should build the speed endurance I’ve been neglecting. On Thursdays I’ll run longer reps at goal pace off short recovery, starting with 2 x 2 miles at 7:25/mi off 90 seconds of jog and building toward 3 x 2. Every third week I’ll swap that for 5-6 x 800m at 6:55/mi to keep my speed.

And yet I’m also keeping the trail running the rest of the week. I enjoy it too much, and it’s mostly why I’ll still be doing this in three years. In any case, easy running builds the aerobic base underneath everything regardless of the terrain it’s built on. But I’m done pretending four hours on the hills is preparation for twenty-plus minutes on the flat.

I know this will take time. Moving threshold pace is going to take something like one to two seconds per mile per week, so closing the gap to sub-22:30 is a season’s work.

Back to work

Your body has no idea what you signed up for. It only knows what you’ve been asking it to do, and it will get better at that. So the question isn’t what your plan says, but what your legs have actually been doing.

Mine have been doing slow shuffles up a hill. Time to ask them for something else.

Don’t trade money you don’t need for time you don’t have.

You might have already seen the time, money, and health triangle, but if not, I have made one for you. The point of it is to remind you that the three are rarely available together at once. When you're young, you have time and health but no money, so you trade time for money. This is practically the only trade available to young folks. However, of the three, money is the only one that is recoverable. That is, I can lose money and earn it back, but there's no rewinding a year, and health gets sapped by the very jobs you take in order to get paid.

Subscribe now

The standard template

Here’s the template: Get good credentials, get a good job and work hard, put money into index funds in a retirement account. And finally eventually retire by drawing down that portfolio.

However, you aren’t accumulating index funds because you want index funds per se. Rather you are building a way to pay yourself without needing you to spend time working. So when you retire and start the retirement drawdown, you are turning on that machinery to pay yourself. In effect, the goal was never actually the portfolio itself, but to generate income that doesn’t require your time.

And in order to get there faster, you would normally think about maximizing what you get for the time you sell. This means getting a raise, or getting better credentials so that you can take the job in the expensive city and go to where the highest wages are. I did exactly that for close to two decades at Big Tech and it worked.

But there are caveats

The index fund portfolio is just one path. Index funds are one way to generate passive income, and a genuinely good one, but they aren’t the only way. If you believe the only way to financial independence is a single product, then that’s usually worth a pause. I’m raising this point so you can check whether you chose this because you compared the options, or because it’s simply what everyone says you should do.

In any case, the rate is a fraction. If the rate you are trying to maximize is money divided by time, then there are two ways to raise it. The first and standard approach is to work on the numerator. Find a way to increase your salary, which means money goes up while the hours stay the same, and so the rate improves. Fine.

But the bottom of that fraction is where it gets interesting, because as time goes to zero, the rate goes to infinity. Any income at all, divided by no hours, beats any salary divided by some hours. So if you take the goal of “maximize the rate” seriously, the answer is to find setups that earn money while taking less of your time.

That’s the whole difference between a Money Baron and a Time Lord. The former works on the numerator, while the latter works on the denominator, which is the side almost nobody thinks to chase. And yes, I know Time Lord has another more popular meaning to it and I am not asking you to have two hearts. iykyk.

So what are some alternatives?

Here’s a short list of alternatives to earning money:

This is not financial advice. Do your own research when investing.

Index funds. The standard route. This requires near zero time and earns market returns, which has historically meant about 10 percent a year nominal for US large caps over the long run, or closer to 7 percent after inflation. The practical version for passive income is the 4 percent rule, which says you need roughly 25 times your annual spending to live off the portfolio indefinitely. This is the baseline and the number every other option on this list is competing against. It really only takes an hour a year of your attention once it’s all set up. This is why this is the standard route.

Landlord. It’s the old trope - nobody is making more land so own land as an appreciation asset. You could even own a rental property, which is what most people actually mean when they say they want to be a landlord, but that isn’t really what I would call passive income.

Dave Ramsey has a quote about this, which is simply: “There’s Nothing Passive About Real Estate”. It’s a small business with tenants, repairs, vacancies, insurance, and a phone that can ring at inconvenient hours. You can hire a property manager and hand back some rent, which buys the time back but at a price, but even then somebody still has to call you to approve the $8K furnace that just died. Ramsey’s own suggestion for actual passive income, incidentally, is an index fund you set and forget.

Options premium. I used to be pretty negative on options trading but I have been selling puts and calls against stock I already hold using a strategy called the wheel. Here is a write up on the wheel on Reddit that explains better than I ever could, but essentially with options, you can get paid (PAID!) to buy a stock you want at a good price, and get paid again (AGAIN!) to sell it at a better price. Imagine that!

It does require more research and time to monitor and execute, maybe a few hours a week, but it yields more than the index, and the useful property is that it is scalable. Those working hours barely increase even when the positions do.

Starting a company. This used to be a poor option for generating passive income, because building anything significant meant managing a business and people full time, but that’s changing fast. The Wall Street Journal just wrote about the rise of million-dollar companies with just one employee due to AI tools making it easier for founders to get started.

Be clear-eyed about it though. A founder is trading time at a terrible short-term rate for a claim on equity that usually doesn’t pay out, so the risk is front-loaded and the failure rate is brutal. What AI changed isn’t the odds, but the size of the bet. The downside used to be a decade and a payroll but now it’s closer to a few months and a subscription, which might make it a bet cheap enough to take even when it doesn’t land.

Pick your number

Whatever your thing might be, the point is to push the work away from labor and toward systems. If every new dollar earned requires another hour from you, then you have simply built yourself another job. But if the thing can keep earning while you are running or sleeping, then you are getting closer to actual passive income.

At the end of the day though, none of this works if your retirement target keeps moving. This is the main reason a high salary is such a comfortable trap. As you get raises, the number you think you need rises to meet it, and soon you are continuing to sell your time. And since it is at a better price, you can feel like you’re winning when you are actually running on the spot or worse, backsliding.

You should therefore measure freedom in terms of years rather than in dollars. Divide what you own by what you spend in a year, and you get how long you could go without working. It is important to note though that the spending lever is the faster one to move the needle, since you control it outright and it works immediately, which is the case for spending less I have made before.

At any rate (pun intended), sell your time at the best rate you can find, put the proceeds into things that earn money without needing your time, and then, unlike almost everyone, actually stop.

Just remember to pick your number and to stick with it. Good luck friend.

The worst weeks in my life were never the ones where I worked too few hours and not even necessarily the most number of hours, but the ones where I worked anxious hours.

Subscribe now

Busyness is a lie

The trouble starts at a societal and cultural level. We are trained at school and then rewarded for at work: every hour should contain accomplishments, and any hour that does not is a small failure on our part. The desire to be busy works well enough for a while, but it does not switch off when the real work runs out. And so we have learnt to fill the gap with things that feel like work.

You refresh email, check your feed, and read the news. But none of it moves anything forward, and each refresh or reload hands you a fresh thing to worry about, along with a fresh comparison that makes you feel behind. These days we also enter prompts to a chatbot and hope its stochastic outputs are favorable this time. And somehow all of this activity registers as productivity because it is visible and stimulating, and since in most knowledge jobs nobody has ever defined what being productive looks like, looking busy becomes the proxy for being useful, both to your colleagues and to yourself.

All of this actually affects us biologically at the same time. We had evolved dopamine to push us toward the next food source or safety, but it has no off switch. So even though our survival needs are easily met in modern life, dopamine keeps pushing us anyway. So while each small action provides a reward which satisfies briefly, the body quickly resets and looks for more. That constant sense of not being where you should be becomes low grade anxiety.

As a result, our body’s dopamine mechanism is fighting against being present with what is in front of you. So there is a choice we have to consciously make if we don’t want to let dopamine-induced activities run amok. By the way, all of this also explains the emptiness of a day that might be full of activities and urgent, and yet somehow we don’t feel at ease.

The social treadmill

Social media, news, and games give us quick bursts of dopamine. On the other hand, reading, walking, cooking, and talking to another person don’t hit as hard. But the sum of what you do all day sets your baseline level of stimulation your mind treats as normal.

Once you see it that way, a few things stop making sense, or make a lot of sense in a manner of speaking. Scrolling feeds after work is not restful because it stimulates you to the baseline exactly where it was all day. You have effectively swapped one high stimulation source for another, and now the restlessness you feel ten minutes into a walk without headphones is not boredom but rather withdrawal. This is why you can’t enjoy the walk.

Social platforms make matters worse. They run on variable reinforcement, which is the same mechanism that makes slot machines work: sometimes the pull pays and sometimes it doesn’t, and the unpredictability makes you check far more compulsively than a guaranteed reward ever would. B.F. Skinner found this with rats and levers in the 1950s, and the same pattern now runs in your pocket. Nobody was intentionally malicious, but the inventor of infinite scroll now regrets his invention.

Do more of these

Here’s a list of actions to do more of: end your days well, move your body, sit with your own mind, and eat healthy.

End your days well. Pick your productivity hours, be fully ambitious inside them, and stop when it’s time. A deadline does more for focus than any technique, and the boundary makes the rest of the day restorative instead of merely unsupervised. If you don’t, the failure mode is obvious but unfortunately common: you stop working without ever disengaging, and the unfinished tasks follow you to dinner and bedtime. So build a shutdown ritual. Check for anything genuinely urgent and make sure every loose task has a date attached to it, agree mentally to look at it when you are back at work, and then say out loud that you are done. It sounds silly and it works, because the point is to give your brain a signal to move on.

I believe that one ritual we lose when we work from home is the commute from the office to home, so if you work from home, it might be good to have some form of transition from your “work area” to your “home area”. It might be as simple as a sign on the door that you can flip that signals the difference between working and not.

On movement, because the science on this is clear. The guideline from WHO is 150 minutes a week of moderate activity, or 75 minutes if you push hard enough such that talking gets difficult. If that sounds like too much, a 2025 meta analysis in The Lancet Public Health pooled 57 studies and found that 7,000 steps a day, measured against a sedentary 2,000, was associated with roughly half the risk of dying early, a quarter less cardiovascular disease, and substantially less dementia and depression. And the gains turn out to start well before you reach 7,000 steps. In any case, you should find an activity first that you enjoy and not worry about the duration, because the workout you look forward to is the one you will still be doing in a year. My long trail run is fifteen miles and takes about four hours, which is indefensible as time management, and yet is the best decision in my week.

On meditation, I have sat with the breath for close to thirty years, so take my enthusiasm with the appropriate discount, but even if you are not interested in meditation per se, you can do breath exercises. A Stanford trial found that five minutes a day of exhale-weighted breathing lifted mood and lowered anxiety more than mindfulness meditation did, and the long exhale activates the parasympathetic side of your nervous system, calming you down.

On people, it turns out that chronic loneliness measurably damages your health, raising the risk of heart disease by 29% and stroke by 32%, with a mortality effect that the US Surgeon General puts on par with smoking fifteen cigarettes a day. Despite what some might have hoped, online contact does not close the gap. A review of more than a thousand studies comparing face to face and “screen mediated communication” found that both engagement and impact drop when the interaction is just a video call. So if you lean introverted like me, the answer might be fewer and deeper interactions rather than more of them. It is most likely not less interactions, because introverts' happiness turns out to depend on social connections at least as much as anyone else's.

On food, about 95% of your body’s serotonin is actually made in your gut rather than your brain. While the gut serotonin does not actually cross into the brain, what you eat still shapes the chemical and microbial environment of your body and, eventually, brain. The loop that matters: chronic stress raises cortisol, cortisol drives up appetite and unsettles blood sugar, and then your body starts asking for the fastest energy it can find, which is sugar. This is why you have a sudden craving for junk food when you are stressed. Eating healthy is part of being calm.

Do work you actually care about

Burnout actually has three parts: exhaustion, cynicism, and the sense that nothing you do matters. The research points at six ways a person and a job can be mismatched. While five of them are about conditions: too much work, too little control, no recognition, bad relationships, unfair treatment, those are fixable with boundaries, a different manager, or a conversation. However, the sixth is the gap between what the work asks of you and what you actually believe is worth doing.

So if you are working on things that don’t make sense to you, the uncomfortable truth is that you should change the work. Sometimes that means changing jobs, but far more often it means reprioritizing your work so that you work more on the parts of the job you would do for free, and less or none of the work that you don’t find meaningful.

As they say, keep calm and carry a towel

Calmness is not a reward that arrives once the work is finished, just as joy is not waiting at the end of a productive year. It turns out that both are available in the ordinary hours, but what blocks them is mostly stuff you can remove: social media, invented deadlines, meetings you agreed to out of guilt, the work you do not believe in, etc.

Stay calm and keep going.

So consider this my PSA: fix your sleep debt. This is a very practical and highly effective way to improve your health and mental well-being.

Here’s what changed my mind. Caffeine doesn’t remove tiredness, it masks it and that has consequences. Through the day your brain accumulates adenosine, which is the molecule that builds sleep pressure, and caffeine works by sitting in the adenosine receptors so the sleepy signal can’t land. However the adenosine is still there, still piling up, and when the caffeine clears you get the crash all at once. So the second cup is just continuing to mask that deficit.

Subscribe now

What you’re actually buying with those hours

Sleep isn’t downtime. It turns out it’s when most of the maintenance of the body and mind happens.

Your brain clears metabolic waste more efficiently during sleep. When researchers kept twenty healthy people awake for a single night and scanned them, beta-amyloid burden rose measurably in the hippocampus and thalamus, two regions implicated in Alzheimer’s. Just one night folks. That’s a small study and a preliminary finding, but it points at something real: the brain only cleans up when you’re sleeping.

It is also believed that memory consolidation happens during sleep too, so the thing you learned today gets stored correctly tonight or it gets forgotten. And finally, emotional regulation runs on it, which brings me to the grumpiness of the officemates in the morning before their coffee.

The grumpiness is a real thing

There’s an intriguing study: researchers scanned people after a normal night and after a night without sleep, showed them increasingly negative images, and the sleep-deprived group’s amygdala fired about 60% harder, with weaker connection to the prefrontal cortex that’s supposed to keep it in check.

So when you’re short on sleep you’re not imagining that the world got more annoying but it’s because you are running with a less regulated threat system. Somehow I found this the single most persuasive argument for sleep, more than the longevity numbers, because I didn’t feel good about waking up as the grouch, especially when I work from home these days.

Recovery is a training input

I also trail run about 40 miles with 6000 to 8000 feet of elevation a week and I’ve been chasing a sub-20 5K, which means the hard sessions are genuinely hard, and for a long time I underestimated the effect of sleep after training, work, and everything else. But that’s backwards as the adaptation you’re training for happens while you sleep and when you are training, so if you skip the sleep you just did the damage without collecting the benefits.

A clean demonstration of this is the Stanford basketball study, where players extended sleep toward 10 hours a night for five to seven weeks and their 282-foot sprint dropped from 16.2 to 15.5 seconds, free-throw accuracy went up 9%, and three-point accuracy went up 9.2%, along with faster reaction times and better mood. This is just eleven athletes with no control group, so hold the results loosely, but those are pretty impressive improvements from just getting the proper amount of sleep in!

Consistency beats duration

This is the result that surprised me most. Researchers pulled over 10 million hours of accelerometer data from about 61,000 UK Biobank participants, scored each person on how consistent their sleep-wake timing was, and found that the most regular sleepers had 20% to 48% lower all-cause mortality risk than the least regular, with regularity predicting mortality better than duration did.

Duration gets all the attention because it’s the number your watch shows you, but the timing is what your circadian system is trying to lock onto, and it doesn’t work with an irregular sleeping pattern. In other words, going to bed at 10 every night and waking at 6 every morning beats a ragged mix of 6-hour and 9-hour nights that averages out to the same total.

The practical version is simple. Pick a wake time and hold it, including weekends, because the wake time anchors the rest of the day. Then get light in your eyes early, keep the room cold and dark, and cut caffeine much earlier than feels necessary: in one trial, 400mg of caffeine taken six hours before bed still cut objectively measured sleep by more than an hour, and the participants didn’t notice it happening.

What I also found useful

On top of regularity in sleeping patterns, preparing the room and the body for sleep at night, cutting down on caffeine, there are two things that have earned a permanent place in my routine. These might sound like advertising, so to be clear, I am not paid to endorse these products, I bought and used them with my own personal money:

The first is Dream Shot from The Feed. A 2oz tart cherry shot with 3g of glycine, 200mg of L-theanine, magnesium L-threonate, ashwagandha, and a rice-protein peptide, taken 30 to 60 minutes before bed. The ingredient I’d defend hardest is the glycine, since 3g before bedtime shortened sleep latency and improved sleep quality with polysomnographic confirmation and reduced next-day fatigue in follow-up work. Tart cherry has several small positive trials and a plausible mechanism, though the melatonin content is tiny and the best-known study had eight completers and industry funding. I don’t take this daily but on the nights of hard work outs, drinking this half an hour before bedtime works well enough for me that I keep buying it.

The second is, and I am perhaps most amazed by this one, is Apollo Neuro. I expected nothing from this one. It’s a wearable that delivers low-frequency vibration to shift the autonomic nervous system toward rest, which sounded to me like a hoax, but there are decent reviews on Reddit and other forums for it. I don’t know how it works, but it works shockingly well that my HRV has been increasing steadily since I started wearing it, with my Garmin watch and Oura ring recording longer sleep time too. In a blog post, their new analysis with Oura covers 474,852 nights from 935 users, and among habitual short sleepers, Apollo use was associated with 46 extra minutes a night and 77% lower odds of sleeping under six hours, with a dose-response pattern where more use meant more sleep. It works for me, and I’ve stopped arguing with it.

The reframe

I used to treat sleep as the thing I gave up to make room for the important stuff, which meant training, work, watching videos and even reading all took time from sleep and my sleep paid for it. Now I think of the night as the first session of the next day, because the adaptation, the filing, and the emotional reset all happen there, and everything I do the following day is drawn from what got built overnight.

Try one thing this week. Don’t even bother with the supplements or gadgets. Pick a wake time and hold it for seven days including the weekend, and move your last caffeine to more than six hours before bed. Then look at how you feel around other people on day five and beyond, because that’s where you’ll notice it first.

Arthur C. Clarke perhaps wrote that as a compliment to engineers, but it speaks volumes about the audience. Magic isn’t a property of the machine but that sense lives in the gap between what a thing does and what you understand about how it does it. This means that the machinery doesn’t have to change at all. Only your ignorance does.

And that gap has never been wider than it is right now. Unfortunately this means we’ve filled it with worship.

Subscribe now

Science made the world work

Before the scientific revolution, the world was full of intentions and anthropomorphizing of natural phenomena. Storms had motives, illness had a moral logic and the cosmos had an order built into it with your place in that order.

Science took that apart though, and as a result we got antibiotics, electricity, flight, and a working model of the atom. What we gave up was the sense that any of it means something. The exchange is that science tells you how the world works but it has nothing to say about why it matters that you’re in it.

So that vacancy never got filled, but that doesn’t mean humans ever stopped looking for meaning.

The roadmap is a religious text

Read the AI conversation and you’ll notice that it’s answering the same questions religion has always answered. What happens after you die? Upload your mind. When does everything change? At the Singularity, a dated event after which the faithful are transformed and scarcity ends. Is history going somewhere? Yes, toward superintelligence, and there’s a forecast for the year. How do we make sure the vastly more powerful thing is good to us? That’s alignment, and finally p(doom) which is a probability of damnation quoted to two decimal places.

I’m not calling anyone religious, but notice that the questions came first and God was one answer to them. What survives death, where history is heading, what makes you you, and who gets to decide were never really about God at all, they were always about us. And these days talking about AI and technology will inevitably bring them up because it’s making promises that big.

The technology is real, but the conversation around it has become suspiciously like a religious text. Is it any wonder that some would start treating AI as a religion?

The problem isn’t worship, it’s deference

Deferring to a machine is actually not a new phenomenon and mostly fine. You trust a calculator for good reason: arithmetic is checkable, so if you doubt the answer you can work it out yourself. Chess computers passed the best humans in 1997 and nobody’s mind broke over it, because you can replay the game afterward and see what happened; Deep Blue was better than Kasparov but its search tree was always available for review.

Then in 2016 AlphaGo played a move no professional would have played, and nobody, including the people who built it, could explain why it was good. We called that a win, but it was troubling, because we accepted a decision we couldn’t reconstruct and then celebrated the fact that we couldn’t.

That’s the issue, and it isn’t “the machine is better than me,” which is fine and always has been. The issue is “the machine is better than me and I can’t check its work.”

Here’s where we stand today in arguably the most logical field of study: mathematics. In May 2026 an OpenAI model disproved a conjecture that Paul Erdős posed in 1946 and that most experts believed was true, and the mathematicians who reviewed the result had to publish what they described as a human-digested version of the proof so the rest of the field could follow it. Two weeks later, more than fifteen hundred mathematicians signed a declaration asking for guardrails around AI in mathematical research. Then in June a researcher pointed a model at formalizing that same result and it generated 1.2 million lines of Lean over three weeks, which no human being will ever read.

Mathematics is going to be fine though, and it’s worth understanding why. Lean is a proof checker, so a proof no human can hold in their head still gets verified line by line by a machine whose own correctness we can inspect and argue about. Faced with output beyond human comprehension, mathematicians built the ability to check and convince themselves of the validity of the proof.

So the risk was never that a machine decides to harm us, it’s that machines decide for us and we get worse at deciding for ourselves. At some point in this slippery slope we hand our humanity to the machine.

That’s the plot line to Wall-E, with the humans aboard the Axiom having had everything done for them for so long that they’ve forgotten how to even walk. It’s worth noticing what their recovery took, because even a children’s cartoon built for a happy ending needed a miracle to pull it off: one broken little robot who didn’t do what he was designed to do. Wall-E, as it turns out, is a cautionary tale for p(doom) as much as the Matrix is.

Seeing it doesn’t get you out

I know what’s happening inside most AI systems, roughly. Attention, next-token prediction, a very large pile of stochastic gradient descent, tons of loss functions. Nevertheless despite knowing the chatbot is just a series of matrix multiplications, I’d still say thank you to a chatbot on occasion.

By the way, none of this griping is a case against the tools, and I want to be clear about that. These AI models are the most useful thing to happen in twenty years of being in tech and they’ve changed how I build, how I write, and how fast I get from a vague idea to something I can actually test. I’m not here to tell you it’s autocomplete with a token budget. But what I really can’t stand is the hype. Unfortunately though the hype isn’t a harmless side effect of the tools but is, I fear, the thing standing between you and using them well.

Even knowing all that doesn’t inoculate me, and I still catch myself seeing a demo or reading a company blog post as if the singularity has arrived. Lately though it does seem like more people are asking what exactly is being produced with AI, and not liking the answer. The evidence arriving now is awkwardly split: task-level studies keep finding real gains of anywhere from 14 to 55 percent depending on the work, and yet an NBER survey of nearly six thousand executives across four countries found more than 80 percent reporting no measurable effect on productivity or employment over three years, while PwC’s 2026 CEO survey found only 12 percent saying AI had delivered both cost and revenue benefits. The same NBER paper found that the executives doing the talking use AI about an hour and a half a week. The tools work but it turns out change is slower than what CEOs had hoped for.

The stock market has (finally) started asking. On July 23, 2026, Alphabet reported a quarter that beat on nearly everything, with revenue up 24 percent and cloud revenue up 82 percent, then raised its 2026 capital expenditure guidance to roughly $200 billion. But we still watched the stock close down 7 percent, because free cash flow had gone negative and investors want to know when the spending turns into returns. That isn’t the market deciding AI is fake. It’s the market asking for a reality check for all this AI spending, and it’s the same question you should be asking about your own use of the stuff.

So when we eventually say “remember when we were all delusional about this AI thing?”, that sentence includes me and probably you too.

The airplane of the mind is here

Here’s a piece of technology so extraordinary that we’ve managed to make it boring. You climb into an aluminum tube, sit down with a bad coffee, and a few hours later you’re on a different continent, and the only thing anyone complains about is the legroom. Flying is one of the most miraculous things our species has ever built, and we’ve turned it into a bus with wings.

For a while though, we were certain flight was going to be the answer to everything. For the first twenty years after Kitty Hawk, aviation was mostly prophecy and barnstorming. The Wright brothers themselves thought they’d built a peace machine, on the theory that reconnaissance from the air would reveal what every nation was doing, make surprise attacks impossible, and therefore make war less likely. Orville Wright later acknowledged, more ruefully, that the airplane had not ended war; human beings had simply found new ways to use it in one. Unfortunately, we seem to be particularly creative in turning new inventions into weapons.

Meanwhile the view from a cockpit convinced a generation of Americans that a new civilization had dawned and the pilot was the man building it. Crowds stood in fields with their heads tilted back, watching someone do a loop through the sky.

The prophecy was wrong but don’t forget: the flying machine was real, is real. But it’s boring technology now.

What changed the world wasn’t the manifesto. It was airmail contracts, cargo routes, and air travel. Barnstorming didn’t end because audiences got bored, it ended in 1926 when Coolidge signed the Air Commerce Act, bringing pilot licenses, aircraft certification, airways, and safety rules to an industry that had largely governed itself. The stunt era gradually gave way to a profession with standards. Aviation began to matter most when people stopped gathering to watch airplanes and started using them to get somewhere.

We’re in AI’s barnstorming years. Look at how much of the discourse in AI is spectatorship: timeline debates, benchmark theater, screenshots of a model saying something eerie, arguments about whether it’s “really” thinking. Everyone’s head is tilted up but that’s not what we should be doing.

Watching the loops won’t tell us what this technology becomes.

Get in the cockpit instead.

Start by accepting that this isn’t the end of the world or the start of utopia. Every hour spent treating it as one or the other is an hour not spent doing anything useful. Doom and hype are the same posture facing opposite directions: either way, you end up standing still, head tilted toward the sky, mouth agape.

Then learn to fly it. I mean hours in the seat, not hours reading about the seat. The capability everyone is arguing about is already sitting there, and almost nobody is putting in the time.

The highest leverage is to build one level deeper: not solutions but solution-makers. The most important use of AI is not drafting another document. It is building the bespoke software that fits your exact situation and that nobody else would ever build for you; the workflows, agents, and skills that can keep solving the problem after the conversation ends.

Talking to a chatbot is useful (heck I do it all day) but it is mostly one-to-one: one problem, one answer. A program or workflow written for you can run tomorrow morning, and the morning after that. A good agentic workflow can be reused, tested, and improved. Use the strongest models not merely to answer questions, but to build systems that keep working even if the oracle disappears.

For example, watch what happened during Anthropic’s mythical Fable model release (sorry I couldn’t help myself). During Fable’s on-again, off-again availability due to security and regulation concerns, the genuinely useful advice going around was this: use the window to audit and improve your setup, have it write your skills, have it plan the projects that lesser models can execute later. In other words, use these AI models as the tools they are meant to be: to build things that keep working even if the oracle leaves or is taken away.

Properly evaluate tasks you are repeating. Split what you use it for into two piles: the one-off things and the things you’ll do over and over.

The one-offs mostly do not matter. If a model summarizes a document you will never open again or chooses a restaurant for Friday, being unable to reconstruct the answer costs you little.

Repeated work is different. You will do it next week and the week after that. Anything in this pile should be something you can recreate, genuinely understand, and test when it changes. That might mean writing software, constructing a proof, maintaining a spreadsheet, or producing a recurring analysis. You need to know what mediocre, good, and excellent look like - and how to work with AI to reach excellent repeatedly. I would insert some discussion about loop engineering at this point, but I’m afraid to ask.

But keep doing some of it by hand. Write something longhand with a pen on paper, take a story from blank page to finished draft without any help at all, and sit down and write the code the old-fashioned way of typing it into vim. This isn’t principle and it isn’t nostalgia, it’s that those are the muscles you’ll want to continue building and maintaining. Nobody on the Axiom in Wall-E lost the ability to walk in an afternoon, they just stopped walking, and eventually the ability was gone.

Remember you don’t owe deference to anything that can’t explain itself. Certainly not to an algorithm, and definitely not to the people selling you one.

We now have the airplane of the mind.

Stop tilting your head up and just go fly the thing.

Whether that’s a problem depends on the shape of the distribution you’re standing in.

The middle is emptying

Start with the belief almost everyone runs: work hard, get rewarded. Do X, get Y. It’s the closest thing we have to an instruction manual.

Look at what happened to the people who followed it.

In 1971, 61% of Americans lived in middle-class households. By 2023 it was 51%. Those people went somewhere, and the direction matters. The lower-income share rose from 27% to 30%. The upper-income share rose from 11% to 19%. The middle didn’t collapse downward. It emptied in both directions.

And where you start matters more than what you do. Researchers measured the fraction of children who out-earn their parents: 92% for those born in 1940, 50% for those born in 1984. Escaping your birth class became a coin flip in two generations. The decline was sharpest for children of middle-class parents, and the cause was the unequal distribution of growth, not a lack of it. The pie got bigger but the shape has changed.

So the hypothesis: hard work isn’t the mechanism anymore. At best it’s necessary and nowhere near sufficient. There’s no instruction manual, no recipe where you do X and get Y.

That leaves two honest responses.

One is to stop playing. I made that case in You can’t max a life: the sprinter and the person lying flat make the same mistake, because both believe there’s a race. You can’t be behind in a game you’re not playing. For some people, that’s the better answer.

The other is to play differently. Because some things you want for real. The career. The research bet. The business. The money. But they come with distributions attached. Most companies fail. Most bets don’t land. The distribution doesn’t care that your reasons are good.

Both answers reject the same move. Grinding up the leaderboard is the mistake in my last essay. Grinding harder on the standard playbook is the mistake in this one. And to be clear: effort was never the differentiated variable.

So here’s the fix, and it’s a slogan Apple coined in 1997: Think Different.

Why different? Because if you do what everyone else does, you get what everyone else gets. That isn’t a slogan, it’s a tautology. Same inputs, same model, same output. You read what your peers read, take the advice they take, benchmark against the person next to you, and your result lands in the middle of theirs. Grinding harder just runs the same model faster.

Differentiated returns need a differentiated generator. So how do you build one?

Subscribe now

One woman, thirty years, one idea

Katalin Karikó arrived at the University of Pennsylvania in 1989 with one idea: inject synthetic mRNA into a cell and you could instruct it to build any protein you wanted. A drug for anything. A vaccine for anything.

Her field had already decided. Synthetic mRNA was inflammatory. Inject it into a mouse and the immune system attacked it. The mice got sick and some even died. mRNA was a dead end, and the verdict came from data.

So her grants got rejected. Then rejected again. In 1995 Penn demoted her off the faculty track and cut her pay. She stayed anyway, and kept working like nothing had happened.

What she did next comes down to three habits worth stealing.

Be open to new ideas

The insight didn’t come from working harder on mRNA. It came from a photocopier.

In the late 1990s Karikó was at a shared machine in the medical school when she started talking to Drew Weissman, an immunologist building an HIV vaccine. She was a biochemist trying to make mRNA behave. Neither was senior enough to have their own copier, so they started teaching each other their fields over it.

Half the answer came from Weissman’s world. Immunologists knew DNA can trip an immune receptor, and that a chemical modification called methylation switches the alarm off. If DNA has a chemistry that makes the immune system ignore it, why wouldn’t RNA?

The other half came from an anomaly in plain sight. Synthetic mRNA set off the alarm. But tRNA, the workhorse nobody was building a career on, didn’t. Why? Because tRNA is heavily modified. So they swapped the building blocks of synthetic mRNA, and when uridine became pseudouridine, the alarm went quiet and protein production went up.

No amount of mRNA expertise would have produced that. It took an immunologist’s rulebook applied to a biochemist’s problem. Read outside your lane. Take the meeting with the person whose work has nothing to do with yours. Every unrelated field you know is another place a connection can come from.

And the tRNA clue had been in the literature for years, filed under boring. The thing that puzzles you and won’t leave isn’t a distraction from the work. It’s the actual work.

Be willing to be wrong

Most people can’t do this one.

Karikó never ignored the evidence. Her mice really did get sick. What she refused was the interpretation. “Synthetic mRNA triggers inflammation” is a finding. “mRNA doesn’t work” is a conclusion. Everyone else collapsed the two and moved on to fundable work. She held them apart and asked why.

It’s brutally hard to actually do. Most of us do the reverse: we hold our beliefs tightly and our evidence loosely. When something doesn’t fit, we explain it away, because the alternative is admitting a thing we’ve built an identity on is incomplete. The more fundamental the belief, the harder we defend it, which is exactly when updating pays most.

Having the insight was the easy part but holding it hurt. Nature rejected her paper. Science rejected it. Immunity ran it in August 2005, and nothing happened. Nobody invited them to present and the paper collected citations at the rate of a small specialist result for fifteen years.

If your differentiated view is comfortable to hold, it probably isn’t differentiated.

Be willing to go deeper

Take a differentiated position and you’re almost guaranteed to be wrong at some point. In public, maybe for years.

Karikó was demoted for a decade before that photocopier conversation, and never had proof she wasn’t the one making the mistake. From the outside, “brilliant scientist ignored by a myopic field” and “stubborn person wasting her career” look identical. No test tells you which one you are until it’s over. Being atypical puts you in the left tail as often as the right, and most people who look like Karikó at year ten are simply wrong.

So how do you stay in for a decade with no proof? Not by grinding.

Grinding is suffering now for a payoff later, and that trade usually isn’t worth it. Karikó never made it. She told Eric Topol that if her goal had been to advance, the demotion would have ended her, because advancement was exactly what she’d lost. Her goal was to understand the thing. That’s the difference.

Ask her about those thirty years and she doesn’t describe a sacrifice. Her line is that she felt successful while others considered her unsuccessful, because in the lab she was in full control. Asked how to be successful, she says success isn’t the thing; being happy is — then talks about running six kilometers every morning before work. That detail may have endeared her to me more than anything else.

She was never in doubt, either. Pushed out of Penn in 2013, she told her chairman the building would be a museum one day. She thought she was right, and she was.

That’s the habit. The demotion couldn’t touch her because she wasn’t climbing, and you can only be demoted if you’re keeping score. Not keeping score is what let her go deep. Thirty years on one problem turns “this fails” into “this fails because of this mechanism, which I can attack.” There was no eureka: a decade of failure, a chance conversation, a borrowed analogy, then years more work.

In 2013 she left for BioNTech, a German company with no approved product. In 2020 a pandemic arrived that I don’t need to recount. The part that matters: two companies designed vaccines on a computer in days, built on her pseudouridine trick, at 95% efficacy, and shipped billions of doses. In 2023 she won the Nobel.

For thirty years it looked, from the outside, exactly like losing. She was never in the race those people were scoring.

Your organization is doing this to someone right now

One key uncomfortable part isn’t the field’s skepticism. It’s Penn. It’s the system working “against” you.

The insight that would end a pandemic sat in their building the whole time, and they demoted the person holding it. However, nobody there was a villain. There was no one inside out to get her. Penn almost certainly ran a reasonable process: funding signals merit, unfunded researchers are a bad allocation, the org chart should reflect that. The process worked exactly as designed and threw away a Nobel.

Your company does this through three doors, all well-intentioned. You optimize for predictability, so plans stop absorbing new information. You optimize for zero errors, so people stop making the jumps that look like risks on a spreadsheet. You build hierarchy, so insights get filtered out three levels below the person who could act. Karikó needed a decade of funded failure to produce one result, and no quarterly review would have approved that.

If you lead anything, replace “that’s not how we do things” with “tell me more about that.” Reward the person who surfaces a contradiction as much as the person who ships on time. Ask what would make the crazy idea work before you ask what’s wrong with it. And notice who in your building is standing at the photocopier with something you’ve already discounted to zero.

The bet

Thinking differently isn’t a personality. It’s a few habits you can start now.

Separate the finding from the conclusion, especially when everyone agrees on the conclusion. Go find the person whose work has nothing to do with yours. Chase the anomaly everyone has already decided is boring. When you lose a bet, spend an hour on the evidence before a second defending yourself. And stay on the problem past the point where staying looks like a mistake.

None of it guarantees a good draw. That’s what a skewed distribution means. But the alternative, following the crowd, guarantees the median, and the median is a number you don’t want.

Be average, expect average returns. Or don’t be.

After parkrun #3 I wrote the fix down: pace better. I had 22:30 legs and 23:20 pacing. The job was to open two to three seconds per km slower, trust the fitness, and let the second half come to me. I said I’d write it down again and try harder to mean it.

I meant it. I spent six weeks fixing it, and the pacing is fixed.

I ran event #26 at Sunnyvale Bay Trail, my fourth parkrun, in 23:35. My first non-PB of the series with a 26-second positive split, and 16 seconds slower than the 23:19 PB I set in May.

Same course, same race, but different result. Let’s go with “failure” for the purposes of this series even though at the end I continue to be enormously thrilled with the state of my running.

In any case, I fixed the exact thing I set out to fix and ran slower anyway, and that deserves an honest accounting.

KM 2 dropped to goal pace on tired legs, then the floor gave way. Two red cards: one at the fast end and one at the slow end, both from the same missing thing.

Subscribe now

What exactly is threshold?

The block started with a question about a word: threshold. My watch told me my lactate threshold pace was 7:26 per mile. That is 4:37/km. That is also, to the second, my 5K race pace.

Those two numbers can’t be the same since threshold is the pace I can hold for an hour while 5K pace is the pace I can hold for 20+ minutes and no more. If they’re equal, one of them is a lie, or I need a way better aerobic base.

I trusted the watch for a few months now but it seems to be the lie. Every “threshold” session I’d run was faster than my real 5K pace and so I wasn’t training threshold. I was running short VO2max intervals, calling them tempo, and then wondering why I couldn’t hold “threshold” for a full race.

I now think that my real threshold is around 4:46/km (7:40/mi). And once I had the right number, the sessions changed. Slower reps, shorter recovery, held even, finishing like I could do one more. For the first time the word on the plan matched the effort in the legs.

The good pacing

Once I got the right threshold pace, my intervals stopped front-loading.

June 3, 6×1km, I opened at 4:14/km (6:49/mi) and spent the next five reps paying for it, a 16-second spread from fastest to slowest. Then the recalibration took hold. June 10, five reps at my real threshold, 4:50/km (7:47/mi), held inside a three-second band, finishing like I could do one more.

July 1, 3×1600m at goal pace, 4:31/km (7:16/mi). Three reps within four seconds of each other, and my heart rate never touched threshold on any of them. I ran goal 5K pace and my heart barely noticed.

And the one that mattered most. On June 17 I went to the track and tried to run a 5K flat, on purpose, at goal pace. I held it dead even at 4:37/km (7:26/mi) for two miles before my heart rate hit its limit and I pulled out. I didn’t finish the distance. But I ran the most disciplined two miles of my life. The skill I’d been chasing since March finally showed up.

The discipline gate from parkrun #3. Closed. In training, at least.

Once I trusted the right threshold number, the reps stopped front-loading. June 3’s 4:14 opener was the last ragged one. By July 1, three reps sat inside four seconds at goal pace, with the heart rate to spare.

The 2.4km tune-up

Three days out, the 2.4km tune-up. Same IPPT distance I ran before every race in my army days in Singapore, two decades and change ago.

2640m at 4:27/km (7:10/mi), run as a negative split, heart rate capped right at threshold across the line. My +4s/km tune-up-to-race rule predicted a 4:31/km race. That’s 22:35 for 5K. The best readiness signal I’d ever produced, and it said sub-22:30 was possible.

I’d read this paragraph before. Before parkrun #3, the tune-up predicted 22:25 and I ran 23:20. The tune-up promised more than the race delivered.

But it happened again and this time the pacing wasn’t the reason.

The race, and the course underneath it

Mile 1 at 4:38/km (7:27/mi). More conservative than parkrun #3’s 4:32. One aspect I am quite happy with is the discipline to open right and not overcook it at the beginning. I’d learned that much.

Then km 2 was my fastest kilometer of the day, and from KM 3 the floor gave way: 4:34, 4:43, 4:52 (7:21, 7:35, 7:50/mi). Cadence fell from 188 to 174. Stride shortened. The same collapse in lockstep I described three races ago, in the same order, at roughly the same point on the course.

KM 2 wasn’t reckless. 4:31 is my goal pace, the pace I held three times on July 1 without my heart rate reaching threshold. On fresh legs it would have been fine. On the legs I brought today it was already too much, and I had no second half to absorb it.

A word on the clock. Sunnyvale Bay Trail runs long on gravel: my Garmin measured 5,050m, about fifty past the line. The watch clicked 5K at 23:20, a single second slower than my 23:19 PB, and the official chip time read 23:35.

Max HR 187, one beat past the 186 from parkrun #3, the highest I’ve ever recorded. I emptied the tank. 23:35.

This is why I am quietly not disappointed with my performance. Even if it didn’t meet my expectations, it feels good that I gave it a good shot.

The wall at two miles

The equation doesn’t balance. Every session in this block said I was fitter than my May PB. I ran 16 seconds slower than it, at full effort. So despite fitter legs, I had a slower race.

The easy story is that I showed up tired as suggested by multiple chatbots. I don’t fully buy it, and I can’t fully clear it either. I ran about 30 miles the week before, under my usual 40-plus, so nobody overtrained me into this. But a down week isn’t a taper, and I ran a near-max 2.4km three days out. That tune-up predicts the race and fatigues me for it in the same breath. I banked the prediction and waved off the cost. So freshness probably isn’t the whole story. I won’t pretend it’s none of it.

Here’s the line I want to be careful with. I keep calling 4:31 the pace I can sustain, but I don’t. I can run 4:31 for a mile, three times, with rest in between. What I can sustain for 5K is what I raced: about 4:40. The gap between 4:40 and 4:31 isn’t durability sitting there waiting to be let out. It’s fitness I don’t have yet. That’s fine. It’s a different claim, and the tidy one was flattering me the way the intervals did.

So where does it break, and why. KM 2 and 3 held near goal, 4:31 and 4:34. Then 4:43, then 4:52. The floor gave way a little past two miles. June 17 broke at the same place: two even miles at 4:37, then my heart rate spiked and I stepped off.

Two readings fit that, and they pull in different directions. One is durability: the pace is there, I can’t hold it long enough yet, so I stretch it. The other is less flattering: goal pace sits above the aerobic ceiling I can hold for 5K right now, and my heart rate hitting 187 at the wall, a max reading over the past few months, is that ceiling announcing itself. The durability read says to extend the pace I have while the ceiling read says the pace has to come down to me as I get fitter, not be forced now. I’ve been telling myself the first while the heart-rate number says take the second seriously.

I think this means that I have miscategorized where I needed to improve: it isn’t a pacing issue that is preventing me from hitting sub-22:30. It’s my fitness or durability to sustain that pace for 5K.

The trail base is part of how I fooled myself. Forty miles a week of trails for almost a year should have built a deep aerobic engine, and I read that engine as 5K fitness. It isn’t. Aerobic volume and race-specific fitness are different things, and I’d been banking one and spending it like the other.

Four races, one course. #4 (purple) opens the most patient of the lot, dips to a 4:31 second kilometer, then climbs to the slowest last kilometer of the day. The fade didn’t leave. It moved out past two miles.

Sub-22:30, still, and closer than the clock says

The honest version of where I am has changed since parkrun #3. Then I thought it was 22:30 legs, 23:20 pacing but it isn’t. The pacing is a solved skill, and that’s progress worth banking. The next step is to train my 4:40 5K fitness into a 4:31 one. That’s fitness to build, not fitness I’m failing to express.

The good news is that endurance and aerobic ceiling are both trained by the same work: dragging the pace I can hold for twenty-odd minutes downward until goal pace fits under it. So the next block doesn’t hinge on getting the diagnosis exactly right. Fewer short reps with generous rest, more continuous work at 5K effort: mile repeats off short recovery, tempo at the edge of comfortable, race-pace runs long enough to find the wall and lean on it.

The north star

Three parkruns, three PBs. The fourth broke the streak, and it taught me more than the three that didn’t. A PB tells you the process worked. A non-PB tells you where the process wasn’t quite sharp enough.

The speed for a mile is there. The fitness to hold 5K pace for a whole 5K isn’t, and I don’t yet know whether the limiter is endurance or a heart rate ceiling that goal pace sits above.

Sub-20 stays the north star. Sub-22:30 needs one thing from me now, and it isn’t a mile of speed and it isn’t pacing. It’s the fitness to hold the effort for all five kilometres instead of two. You train for what you want to do. Today I trained for a fast two miles and ran a fast two miles. Next time, I want to have trained for the whole five.

Looksmaxxing. Sleepmaxxing. Gymmaxxing. Moneymaxxing. You take one part of yourself and push it to the ceiling. Faster, better, higher than the person next to you. The word crawled out of some grim corners of the internet, then went everywhere, half a joke and half not. Now people say it with a straight face.

I feel the pull too. Open any feed and you see a hundred people at the top of a hundred stats at once. Someone sleeps better than you. Someone lifts more. Someone made more money before lunch than you’ll make this year. It’s hard not to read that as a scoreboard. It’s harder not to want to climb it.

There’s a version of this essay where I tell you to slow down, and it would be me giving up and dressing it up nicely. “Don’t max” comes easy when you’re tired of climbing. One might misconstrue exhaustion as wisdom.

But I don’t think that’s what this is. So stay with me.

Subscribe now

Maxxing and giving up are the same move

Think about the two things people say when the world feels like a race.

One side says you have to win it so grind harder and optimize everything. Fall behind and you’re done.

The other side says the race is rigged, the machine always wins, so why bother. You can’t fight it.

These sound like opposites, but they’re not. Both of them believe there’s a race. While one sprints, the other sits down or lies flat, so to speak. Neither one stops to ask if the race is real.

That’s the trick: you can’t be behind in a game you’re not playing. So the way out isn’t to run faster or to quit. It’s to even notice that nobody started a clock.

You can max a stat. You can’t max a life.

Here’s why maxxing feels so natural and works so badly.

In a game, someone hands you the stat. Strength, gold, level, whatever. The game, sometimes implicitly and other times explicitly, hands you the ceiling too. You know what the top looks like, and you know when you’ve hit it. Maxxing makes sense there as there’s a number, and the number goes up.

Nobody handed you the stat for your life. You picked it up from the feed without noticing, and all the things that make a life full don’t have a number.

There’s no high score for love, you don’t max a friendship, and certainly a good marriage isn’t a stat you top out, and a twenty year friendship isn’t something you finish. You go deeper into these things. You don’t beat them.

You can’t max a walk with your kid. You can’t max a quiet morning. You can’t max a practice, whether it’s meditation or a hike, because the whole point is there’s no summit, just the parts in the moment. The small things in life turn out to be the important bits.

Maxxing takes the logic of a video game and drops it on a life. Then it wonders why the win feels empty. You climbed to the top of a stat that was never the point.

Even the winners will tell you this

Maybe you don’t care about a fuller life. Maybe you read all that and thought, fine, but I still want to win.

Okay. The strange thing is that maxxing loses even on those terms.

I’ve been reading Jim Murphy’s Inner Excellence. He coaches world champions and has spent years studying how the best people alive stay calm when everything is on the line. You’d expect a book like that to be all grind, but it isn’t.

One of his clients became a world champion the year he stopped trying to beat his biggest rival. He realized the rival was the reason he got to play the sport he loved, so he went from wanting to crush the guy to wishing him well. Then he won. Funny how things work.

Murphy tells another story about a swimmer named Lewis Pugh, who set out to swim a kilometer in water near the North Pole. Mind you, the water near the North Pole kills most people in under a minute and Pugh failed in practice. He only made it when he stopped swimming for himself and swam to wake people up about the oceans.

The pattern shows up over and over where the people who perform best aren’t the ones gripping hardest. They’re the ones who let go of the scoreboard and gave their attention to the thing in front of them. It seems presence beats optimization. So even if all you want is to be great, the grind is a worse road there than being fully here. Enjoy the journey and you might even get to the end.

That’s the part I’d push on if a striver were sitting across from me. You’re not losing the race because you’re too soft. You’re losing it because you’re staring at the scoreboard instead of the ball.

What a fuller life looks like

Drop the frame and something simple is left.

Not more. Enough.

Not score and certainly not maxxing it. Depth.

You aim at the parts you control such as your effort, your attention, and how you treat people. You let go of the parts you don’t, like whether you came out on top. You can always give your best effort. You can’t always win, and chasing the win wrecks the effort.

You give more than you take. Spend a day trying to be useful to someone instead of trying to get ahead of them, and the fear quiets down since most of the fear was about you anyway. Take yourself out of the center and there’s less to protect.

You go deep instead of wide. One craft you keep returning to and a few people you keep showing up for. A body you keep healthy, not to win, to live in. None of it maxes but all of it grows.

Your turn

Here’s where I land.

There’s nothing to max because I was never a number, you were never a number, and we are not numbers. You’re the thing the numbers were trying to point at, and we mistook that pointing as a directive.

But you need to be aware that the machine is real and the pressure to climb is real, and I won’t pretend opting out is free, or that it’s an option when the rent is due. Enough comes easier when you’re safe and I know that.

But between the person sprinting up the leaderboard and the person who gave up on it, there’s a third person. She’s not running and she’s not sitting down. She’s building something that doesn’t have a top. From the outside she looks like she’s losing the race.

It turns out that she’s not in it.

Be her. Be you.

But the name also points at something more deliberate. The emphasis is on the word thinks.

This Substack isn’t a collection of tactics, routines, or recommendations. It’s an attempt to show the thinking underneath them: the assumptions, tradeoffs, principles, and philosophy that produced the action.

That distinction is the whole point.

Subscribe now

The trap of copying visible actions

We live in a world that rewards copying visible forms. We copy the habits, routines, templates, investments, training plans, and career moves of people we admire.

Someone successful wakes up at 5am, so we wake up at 5am. A famous investor buys a stock, so we buy the stock. A fast runner does track intervals on Tuesdays, so we do track intervals on Tuesdays. A great engineering leader runs a particular meeting format, so we run that meeting format.

Actions are visible. Reasoning is not. So we copy what we can see.

This is a gripe I always had about meditation. You can’t copy a meditation instructor during a formal sit the way you can copy a golf pro’s swing. Even if you hear instructions about focusing on the breath or visualizing something, how exactly do you know whether you are doing it right? (Editor’s note: you should still learn how to meditate)

The broader problem is that actions are compressed outputs of a much larger computation.

That 5am wake-up might exist because of a particular family situation, chronotype, or commute. The stock purchase came from a thesis about position sizing, time horizon, and portfolio context you can’t see. The Tuesday intervals sit inside a training philosophy about stress and recovery calibrated to one person’s body and history.

When you copy the action, you inherit none of the computation that produced it. You get the answer to someone else’s problem.

And when your context differs from theirs, which it always does, the copied action quietly stops making sense. Worse, you can’t tell why, because you never understood why it made sense in the first place.

The point is that the reasoning is the useful part while the action is a byproduct.

What I learned at Google

I spent almost two decades at Google Search, and one of the best educations I received came from launch meetings for Search Quality.

The setup was simple. Teams brought proposed changes to Search, along with data, and leaders decided whether to ship them. If you looked only at the outcomes, each meeting produced a binary result: approved or not approved. You could have summarized months of meetings in a spreadsheet.

But the spreadsheet would have taught you almost nothing.

The value was in watching how leaders reasoned their way to those decisions. Which metrics they trusted and which they discounted. When they overrode a positive experiment because it degraded something harder to measure. How they weighed a small win now against optionality later. What questions they asked when the data looked too good.

The same decision could be brilliant or careless depending on the reasoning behind it. Two leaders could approve the same launch for entirely different reasons, and only one of those reasons might generalize to the next hundred decisions.

Junior engineers who watched the decisions learned what got approved. Engineers who watched the thinking learned how to judge.

Years later, the first group could recite old precedents. The second group could handle situations with no precedent covered. That gap grows and compounds over a career.

Form doesn’t scale; principles do

Copied form breaks when context changes while understood principles adapt.

If you copy Google’s launch process into a ten-person startup, you will likely get theater. The form assumes hundreds or thousands of engineers, mature metrics, and a product where small regressions can affect billions of people.

But the principle underneath, that you should know what you’re optimizing for and what you’re willing to trade against it, holds everywhere. At a startup, it may produce a five-minute conversation rather than a committee.

The same is true in training. Copy an elite marathoner’s plan and you may get injured, because the plan is the form. The principle is progressive overload matched to your current capacity. That principle survives translation to a 45-year-old with a day job. The plan doesn’t.

Copying is not useless. It is often how learning begins. But copying gives you scaffolding, not judgment. Eventually, you have to understand why the form exists and when to depart from it.

Principles scale because they carry their reasoning with them. Forms don’t, because the reasoning has been stripped out.

Visioneering as an example

There is a huge industry of leadership-shaped artifacts: frameworks, one-on-one templates, organizational principles, strategy deck formats, and all-hands rituals. You can adopt every one of them and still not lead anything.

The artifacts can support leadership, but they can’t produce it.

What mattered in the leaders I respected was something I came to call visioneering: the ability to hold a vision of where things should go, connect it honestly to technical reality, sequence the work so each step makes the next one possible, and exercise judgment when the plan meets the world.

That’s a thinking skill. It doesn’t come in template form. You develop it by reasoning through real decisions, watching your predictions fail, and updating.

You can’t copy visioneering by adopting its artifacts any more than you can copy a chess grandmaster by memorizing one of their games.

Why this applies to everything I write about

Once you see the distinction between form and thinking, it shows up everywhere. That is why this Substack can cover different topics and still, ostensibly, be about one thing.

When I write about tech and AI, I’m less interested in which tool won this month than in how to reason about systems: where the leverage is, what breaks at scale, and what the incentives actually reward.

When I write about running, the workouts are incidental. The interesting part is how to think about effort, recovery, and patience, and how to read your own data instead of borrowing someone else’s plan. I also write about my struggles and the adaptations it took to chase bigger running goals than I thought possible.

When I write about life, meditation, or meaning, the same rule applies most of all. Nobody else’s answers quite fit. Only their ways of questioning transfer.

These are different domains, but the perspective I write about is the same.

What you should expect here

After nearly a year of writing this Substack, I finally understand what I’m offering.

You won’t get certainty. My views come with assumptions attached, and I’ll try to state them clearly enough that you can see where we diverge.

You also won’t get universal advice. Most of what I write worked for a specific person (me) in a specific context.

The advice is the form. The reasoning is what I’m offering.

What you will get is thinking in public: principles, tradeoffs, and the occasional changed mind. When I’m wrong, the useful artifact won’t be the wrong conclusion. It will be the visible reasoning that led there, because that is the part you can inspect, challenge, and improve.

If you leave a post with a tactic to copy, I’ve half succeeded. If you leave with a better way to reason about your own situation, that is the actual product.

Don’t copy what people do. Understand how they think. That is the real object of learning.

The action is what happened, but the thinking is what lasts.

The deal is breaking down. Not because effort stopped mattering, but because the conditions that made effort reliably pay off have changed. Growth has slowed, and nowhere is this more visible than in tech. In a growing environment, it’s easy to reward all effort. The pie expands, promotions multiply, and even average performance gets carried upward. In a static environment, the math changes. Meritocracy becomes harsher and narrower when growth slows. The same amount of effort produces less predictable reward.

This matters because most of us built our lives on the old playbook without ever examining it.

Subscribe now

The Gen Z response is rational

Older generations look at Gen Z and see apathy. Quiet quitting, low institutional loyalty, skepticism about the grind. The standard read is laziness.

I read it differently. What looks like apathy is rational disillusionment. Gen Z inherited the motivational story but not the economic conditions that made the story feel true. They’re entering a world with expensive housing, credential inflation, weaker entry-level ladders, AI uncertainty, and a sector mix that no longer favors new graduates. The median first-time homebuyer is now 38 years old, a record high. Young college graduates now face higher unemployment than the broader workforce, a reversal of an advantage that held for decades.

They still want success, meaning, autonomy, money, and status. They’re less willing to believe that grinding inside an institution will reliably deliver those things without proof. Deloitte’s 2026 survey of 22,000 Gen Zs and millennials found them deliberately prioritizing stability, skills, and well-being over fast advancement, and aligning life choices with realistic conditions rather than traditional timelines. That skepticism looks like apathy to people who benefited from a more forgiving growth curve. It’s closer to updated priors for a different environment.

There’s a deeper point here. Doing the market consensus thing, following the average path, earns you average returns. In a growing world, average returns were fine. In a world that isn’t growing, average may not be enough. The consensus path has, funnily enough, become the risky one.

The critical skill of the AI era: think for yourself

We’re all figuring out what AI means for work and careers right now. Nobody has the answer, including me. But I have a hypothesis about the skill that will matter most, and it’s not prompt engineering.

It’s thinking for yourself.

LLMs are trained on the aggregate of human output. By construction, they pull toward the mean. Ask one for a life plan and you’ll get the consensus life plan. Ask one what success looks like and you’ll get the culturally conditioned answer, polished and fluent. These tools are extraordinary, and I use them daily. But the more we lean on them, the stronger the gravitational pull toward the average becomes.

That’s not my experience, you say. As I get better at prompt engineering, the results get so compelling and unique that they’re nothing like the usual LLM stock answers. That’s precisely the point. When you get better at prompting and see great results, you’re giving the model inputs that move it into a semantic space beyond the stock answers. To be fair, it’s still the “average” of the new concepts and insights you provided, but it’s no longer the standard answer. You have, effectively, done the thinking for both yourself and the LLM.

Which means the scarce thing is taste. Your own opinions, your own sense of what matters, your own judgment about which problems are worth solving and which life is worth living. The machine can give you the mean. Only you can decide where you want to deviate from it.

Here’s the test I keep coming back to: what does a meaningful life look like? Then ask a harder question: did you figure that out yourself, or did you inherit it? Most of what we call our values arrived by absorption. The promotion ladder, the house, the prestige markers. Society rewards you for adopting these values, which makes them feel like yours. Even if winning more trophies or climbing another level isn’t deeply satisfying to you, the social and material rewards make it feel like something you should want. In other words, you learned to value the rewards, not the thing itself.

Untangling your values from absorbed ones isn’t something you do in an afternoon. But the work matters more now than ever, because if what you’re striving for doesn’t match your inner values, it will always be a net drain. Every day becomes an uphill struggle. And if you do reach the misaligned goal, the payoff falls flat, so you climb the next peak, and the next. More is never enough when more was never yours to begin with.

When your efforts line up with what you care about, the energy and resilience come naturally. Be intentional about this. It’s the one decision the machines can’t make for you.

Happiness doesn’t work the way you think

Most of us define happiness as the feeling that comes from getting what we want. The promotion, the raise, the milestone. Arrive at the goal, collect the happiness.

That’s not how our emotions work. The brain chemical most associated with pleasure, dopamine, is released when we anticipate achieving something, not when we achieve it. Dopamine drives us toward goals instead of rewarding us for reaching them. This is why the finish line always underdelivers. Psychologists call it the arrival fallacy: the mistaken belief that reaching a goal will deliver lasting fulfillment. It almost never does, by design.

The word I’ve come to prefer is contentment. I’ve written about this distinction before: happiness is a target, contentment is a state. Happiness says “I’ll feel fulfilled when.” Contentment says “I’m at peace with what is, even as I grow.” The first keeps you on the dopamine treadmill, forever one arrival away from a payoff that never comes. The second is available now, requires no achievement, and doesn’t conflict with ambition. You can be at peace with what is and still build what’s next.

And here’s the part that goes against the cultural grain: contentment fuels the work rather than following from it. We’re told every meaningful achievement takes hardship and sacrifice, that nothing good comes easy. But peace of mind is what gives you the energy and resilience to be successful, not the other way around. Chasing happiness through achievement means obeying your psychological defaults. Cultivating contentment means overriding them.

Apply 80/20 to your whole life

If only 20% of the things you’re working toward produce 80% of your enjoyment in life, why not focus the bulk of your energy on that crucial 20%?

This isn’t only about leisure. It’s 20% of the things that keep you healthy, 20% of your relationships, 20% of what you do at work. The distribution is uneven everywhere. Most of your joy, most of your growth, and most of your best work come from a small slice of what fills your calendar.

I’ve written before about the 80/20 rule for your whole life: a boring, stable foundation taking up most of the space, with a laboratory running alongside it. The same architecture applies here. Identify the small set of people, practices, and pursuits that generate most of what makes your life good. Protect them like a foundation. Treat much of the rest as negotiable.

Which brings us to the hard part: the rest doesn’t leave on its own.

Stop doing the default automatically

Most of what fills your days was never decided. It’s the default. You say yes to the meeting because declining takes effort. You check the phone because it’s there. You take on the obligation because refusing feels rude. None of these are choices in any meaningful sense. They’re the path of least resistance, running on autopilot.

The automatic yes is the most expensive default. You want to be a good friend, family member, and teammate, and that often means agreeing to help or taking on duties that benefit everyone. But every yes has an opportunity cost. Saying yes to a meeting is saying no to deep work. Saying yes to an obligation you don’t care about is saying no to one you do. Run the default long enough and your schedule crowds, your stress rises, your effectiveness drops, and you start to resent the people you once valued as they add more to your list. The only defense is intentionality. Before agreeing, ask whether this serves the 20% or crowds it out.

Your environment runs on defaults too. Every pile of mail, every wall of sticky notes, every app you check compulsively is a potential to-do occupying space in your mind, whether you acknowledge it or not. The modern attention economy designs everything on your phone for maximum engagement, which means everything on your phone is designed to be the default. Set hard limits on when you check email and news. You won’t miss anything, and you’ll recover mental space for what matters.

And here’s the most undervalued default to break: treating rest as what happens after the important work. Taking care of yourself isn’t a detour from the important work. It’s what makes the important work sustainable. Rest, exercise, time outdoors, reading, creating. These aren’t rewards you earn after getting stuff done. They’re the maintenance that keeps you capable of doing anything at all. We treat sustainability as a given in our systems and infrastructure, then ignore it in ourselves. A career is a multi-decade endeavor. Optimize for the decades, not the quarter.

One practical rule: whatever you’ve decided matters most has to go on your schedule first. If it’s not scheduled, the defaults win by default. Progress on the project, time with your family, the call to an old friend. Schedule it before you commit to anything else, and let nothing override it.

Be present. We’re going to be fine.

The old playbook is breaking, and that’s unsettling. But the breakdown is also a release. When the inherited script stops working, you’re forced to write your own, and the one you write yourself was always going to serve you better than the one you absorbed.

So ask the question, today and often: is what I’m doing right now bringing me closer to what I want? Not what the culture wants, not what the algorithm suggests, not what the consensus rewards. What you want, after you’ve done the work of figuring out what that is.

Your time is a resource meant to serve you, not someone else. The people who thrive in what comes next won’t be the ones who followed the playbook hardest. They’ll be the ones who noticed it was broken, thought for themselves, and built a life around their own answers.

The deal changed. The opportunity is that you get to write the new one.

Discouragingly, some of the most successful people I’ve watched there treat the humans around them like pieces on a board. These leaders hoard scope, staff up to look important, and burn people out and call it hustle. And it works, at least for a while. Promotions land while org charts swell, but these empires are being built on the backs of people who never agreed to be raw material.

That disturbs me. It should disturb you too. As someone who thought hard before going into management myself, and who actively discouraged people on my team from rushing into it, I want to evangelize this: it doesn’t have to be this way.

Here’s what years of building and managing engineering teams have taught me: the pawn-pusher wins the visible game and loses in the long run. The org chart grows while the org rots, and the reason isn’t karma or comeuppance. It’s mechanics. The way you treat people and the way your organization performs are governed by the same underlying system. Treat people as a resource to consume, and you build something brittle by construction.

This way of leading pays off in the short term and fails in the long term. And the industry right now is squeezed enough that it seems to reward the short-term version. But that bill comes due when the cycle shifts from AI euphoria to whatever comes next. The moral way to manage and the durable way to manage are the same way, and the pressure to forget that has rarely been higher.

Subscribe now

Why working people harder feels like progress

When a team falls behind, the reflex is to push. Out come the warrooms and the manufactured urgency, all of it pushing people to work harder, multitask, and coordinate more tightly. These feel like leadership because they produce visible activity. Someone’s staying late. Someone’s heroically shipping. The dashboard turns green for a week.

None of it changes the system that produced the problem. You’ve treated the symptom and left the cause untouched. The team produces the same output it did before, now at a higher cost to the people in it.

This is where the moral problem and the practical problem turn out to be the same problem. When a structure can’t support good work, some people respond by becoming heroes, working unsustainable hours to keep things from collapsing. Others get sidelined, because the people closest to the fire have claimed everything that matters. You end up with a handful of exhausted indispensables and a bench of underused, demoralized engineers. That isn’t a motivation gap. It’s what a broken structure does to people. And asking people to absorb the dysfunction with their own health and time treats them as fuel.

Empire-building is the same move dressed in status, where headcount becomes a trophy, scope becomes territory, and activity gets mistaken for progress. The empire-builder looks busy and important right up until you ask what the system produces and realize that the emperor, indeed, wasn’t wearing clothes.

Teams are systems, not piles of pawns

A team is not a collection of individuals who need more motivation. It’s a system, and systems behave in predictable ways.

Think of a few quantities that accumulate: people on the team, unfinished work in the queue, technical debt in the codebase. Then think of the rates that change them. Hiring adds people; attrition removes them. New requests add work; shipped projects remove it. These accumulations and rates feed back on each other. A growing backlog makes prioritization harder, which slows everything down, which grows the backlog. Loops like that decide whether a team climbs or sinks, far more than any individual’s effort does.

Once you see the team this way, the cruelty of treating people like pieces on a chessboard becomes obvious. When a talented team gets stuck, it’s almost never because the people stopped trying. It’s because a feedback loop is running against them. Blaming the people for a structural failure is both wrong and unkind. It’s the management equivalent of yelling at someone for drowning.

The limits that don’t care about your ambition

Systems have hard limits, and the limits aren’t up for negotiation. Pushing past them is like asking why the car won’t fly.

A manager can hold real relationships with maybe six to eight people. Past that, the attention thins; you stop developing anyone and start merely tracking them. The same ceiling applies upward: four to six managers under a senior leader before the layer stops working. Teams have a floor, too. Below four people, losing one person to illness or a departure removes a quarter or a third of your capacity overnight, and you can’t plan around that.

Bigger isn’t safer either. Three people have three connections between them; eight people have twenty-eight. Coordination cost grows faster than headcount until people spend more energy synchronizing the work than doing it.

Empire-builders break these limits on purpose, because a bigger box on the org chart reads as power. Twelve reports. A sprawling team that can’t be split because splitting it would shrink someone’s domain. The numbers punish this every time. The right response to a large effort is more teams, not larger ones.

These rules of thumb are a kind of physics: they describe how information moves between people, and good orgs are built to work within those limits, not against them. A lot of companies are now betting the other way. With AI arriving and layoffs hanging over the sector, the hope is that people can be pushed to transcend the limits, to cover the work of those who left and coordinate across a wider span. That’s a wish, not a plan. It says what leaders want to be true without doing the harder work of changing the process and tooling that might make it true.

Until better processes and tools move those limits, the rule when you grow is to keep teams intact and move the work between them; don’t shuffle people like inventory. A team that’s worked together has built up an invisible structure: who catches problems early, whose judgment to trust on what, who checks in with whom without being asked. Move people instead of work and you reset all of it, taking your most resilient teams and making them fragile to look tidy on a slide.

Underwater, treading water, above water

Most teams are in one of three conditions, and each one needs a different fix or approach.

Underwater. Work comes in faster than the team can finish it. The backlog grows every week. People spend their days triaging instead of building, the number never moves no matter what they do, and morale drops with it. Then your best engineers leave, which leaves even less capacity for the same load.

The fix is to add capacity, but how you add it matters. Hire ahead of the crisis, not in the middle of it. A new engineer takes three to six months to become a net positive, and until then someone senior is spending hours onboarding them instead of clearing the queue. Wait until you’re drowning to start, and the help arrives a quarter too late. Alongside hiring, find the changes that remove work for good: automated tests that catch a whole class of bug before it reaches a person, or deployment tooling that ends a recurring fire drill. Throwing bodies at a late project in a panic doesn’t help, and usually makes it later.

Treading water. The team ships what’s urgent every week but never gets to the rest. The migration that would simplify everything stays half-planned. The flaky test stays flaky. The doc never gets written. Capacity matches the incoming work, so the team is stable, but it’s stuck below what it could do.

The fix feels backwards: have people do fewer things at once. When someone is splitting attention across five projects, each one crawls and nothing ships. Cut that to two, finish them, and put the freed time into the debt that’s been slowing everything down. Pay off enough of it and the work itself speeds up, which frees more time. The reason this works is queueing. A team running at 95 percent of its capacity isn’t a little busier than one at 80 percent; tasks wait several times longer. Load people to the edge and everything they touch sits in line. Leave real headroom and the same team delivers more.

Above water. A team with its head above water is in one of two places.

The first is repaying debt. The cycle has turned, so each fix frees a little time and that time funds the next fix. The catch is that the early gains are small, maybe a couple of hours a week, and that’s exactly when someone notices the team “has room now” and hands them a new priority. Spend the surplus that early and you slide back to treading water. So the work here is mostly protection: guard the time, hold it for months rather than weeks, and don’t spend a surplus you don’t have yet.

The second is innovating. Debt is low, there’s real slack, and most of the team’s time goes into building. This team isn’t visibly struggling, which is the danger. The slack that makes it productive looks like spare capacity to everyone outside it, and spare capacity gets filled. New work flows to the teams that handle work well, and the margin quietly disappears. Defending it means saying the unpopular thing plainly: we are holding this capacity on purpose, because it’s what lets the team absorb a sick week or a production incident without falling apart, and it’s what lets them do the work no one else has time for.

The habits that keep the system flowing

Diagnosing the state and fixing it is problem-solving: triage, stabilize, get the team’s head above water. That’s half the job. The other half is visioneering, building a clear picture of where the team is going and clearing the path to it. The pawn-pusher rarely reaches the second half, because a team kept permanently underwater has no room to think past the next fire. A few standing habits protect the room to do both.

Think for yourself, then make the process consistent. Just because every famous company or someone else does something doesn’t make it right, and copying their playbook is mostly a way to dodge the harder work of deciding what your team needs. The autonomy and speed you admire elsewhere aren’t techniques you can transplant; they’re what healthy structure looks like from the outside. Sound organizational structure comes first and behavior follows, which is also why structure beats individual brilliance: great engineers raise the ceiling, but the structure sets it. Once you’ve decided your own structure, apply it consistently. A documented, repeatable process for hiring or promotion isn’t bureaucracy; it’s how you treat people fairly instead of by mood, relationship, or who asked first. When someone asks you to bend the rule for them, don’t. Watch for the pattern instead: if many people need the same exception, your process has a gap, so fix the process rather than the one case.

Measure what you want, not what’s easy to count. Reward lines of code and you’ll get more code, not better software. Reward tokens spent and you’ll get, well, more tokens spent, not productivity gains. Reward bugs closed and you’ll get trivial bugs hunted while the serious ones rot. People optimize for whatever you measure, so pair every metric you push with one you refuse to let slide. And accept that some of what matters most, like whether people feel supported or whether they’re growing, resists clean measurement and still deserves your attention. Motivate people through vision, not numbers alone.

Build for the scale you’re growing into, and write things down. Organizations break in steps, not slopes. A planning process that works at twenty engineers shatters at two hundred, and it shatters all at once. If you’re growing fast, build for ten times your current size so people aren’t relearning how to work every six months; don’t build for a hundred times, which is complexity you’ll never use. Part of building for scale is getting knowledge out of people’s heads. Anything that lives in one head alone is a bottleneck and a single point of failure, so write it down, and treat updating the docs as part of finishing the work rather than a chore for later that never comes.

Be patient, and finish what you start. Structural fixes work slowly. Hiring, onboarding, and compounding all take months, and during that lag you’ll feel the pull back to asking people to push harder right now, because that produces visible motion. Resist it; the slowness is the point. Quick fixes don’t stick because they run on effort no one can sustain, while structural fixes last because they build something people can hold long-term. And see them through, because half-finished work is its own tax. A half-done migration means maintaining two systems forever; a half-done reorg leaves people unsure who they report to. The last twenty percent is the hardest part, full of edge cases and the conversations everyone’s been avoiding, and it’s exactly where the value is. Don’t start what you can’t resource to the end, and push through the ugly finish.

Build something that outlasts you

Notice what every one of these has in common. None of them asks a person to be the shock absorber for a broken system. That’s the whole ethic, stated structurally: build the conditions where good people can do good work without burning themselves down to do it.

The pawn-pusher’s bet is that people are interchangeable and consumable, and that the system will hold long enough to cash the promotion. Sometimes it does. That’s the part that’s hard to stomach. But the org they leave behind is brittle by design, and someone always inherits the bill.

The harder, slower, less heroic path is to treat the team as a system worth tending, and the people in it as the point rather than the fuel. It rarely makes for a good war story. No late-night heroics, no empire to brag about, just teams that keep working after you’ve moved on.

I’ll take that trade. The empires may photograph better, but the ecosystems outlast them.

My son was also upset last week about not getting as many responses to his internship applications as he’d hoped. He spent the afternoon reading Reddit and other forums about some folks who even got internships while they were in high school, and decided the whole thing meant something about him.

I recognized it right away, because I do the same thing. I read about someone else’s success and feel behind. I check a number on a screen and let it set my mood. My kids didn’t get this from nowhere.

I’ve come to think one of the most useful things I can teach them isn’t to feel less but how to let something hurt without letting it take over. How to feel something and not let it run the rest of the day. It took me a long time to learn even a little of this, so I’m writing down what’s helped, for them and for anyone else who needs it.

Subscribe now

Why this is harder now

Some people are simply more sensitive than others. Loud noise has always bothered me more than it seems to bother most people. At the same time, I was not naturally good at reading emotions, so I learned it almost mechanically: watching body language, tone, and shifts in mood.

However it develops, that sensitivity can be useful. It helps you notice things others miss. But it also means you can absorb too much from the world around you.

We used to compare ourselves to a few dozen people we actually knew. Now the feed shows us thousands of strangers who all look like they’re winning. The news shows us every bad thing happening anywhere, all day. Our brains treat all of it as a signal about where we stand, and they were never built to handle this much input. So we end up anxious and tired because we are comparing ourselves to curated versions of people we don’t actually know.

My kids are growing up in a world like this from the start, and they’ve never known anything slower.

It’s a skill, not a personality trait

Here’s the useful part: how we react isn’t fixed. It feels like who we are, but it’s closer to a habit.

Something sets you off, you react the way you always do, you feel a little relief, and your brain learns to do it again next time. Repeat that for years and it starts to feel permanent.

But it isn’t. You can practice a different response.

These are the small practices I’m trying to teach them, mostly by practicing them imperfectly myself:

Don’t carry other people’s mood as your own. I used to absorb everyone else’s emotions too easily. Their anxiety became my anxiety; their anger became my anger. What helps me is remembering that another person’s mood is information - it tells you about them, not about you. When my kid is angry about a friend, I can sit with them in it without taking it on. I tell them they can care how someone feels without having to fix it.

Focus on what you can control. Most of what upsets us is something we can’t change. So I write down what’s bothering me, and next to it, one thing I can do about it. “Money’s tight” becomes “I’ll sit down Sunday and list what’s actually due.” The worry doesn’t go away, but having one small action makes it smaller. I do the same with my kids. What’s the worry, and what’s one thing we can do before bed.

Don’t answer every worry immediately. Pick fifteen minutes a day, not right before sleep. When a worry shows up outside that time, write it down and tell yourself you’ll deal with it then. I particularly like to defer dealing with issues until after a run. Usually by the time I finish, the worry has either shrunk or become more specific. It teaches you that not every anxious thought needs an answer the moment it arrives. For a kid, this can be a notebook on the table and five minutes together after dinner.

Make the decision before. It’s hard to think straight once you’re already upset. So set the rules ahead of time. No phones at the table. We don’t watch the news right before bed. When the moment comes, you’ve already decided, and you don’t have to argue with yourself.

Let them watch me do it. This is the one that matters most. My kids don’t need a parent who feels nothing. They need to see me get upset, notice it, put the phone down, and come back to the table. Not every time but more than I used to.

I’m still learning this

I still read about someone’s win and feel behind. I still check that number in the morning. I still have nights where a hard day makes me shorter than I want to be with the people I love most.

What’s changed is the gap between feeling something and acting on it. It’s a little wider than it was. That gap is what I’m trying to give my kids: Not a thicker skin but learning to take a short pause, where you notice what you’re feeling before you decide what to do with it.

The world isn’t going to get quieter. I can’t fix that for them. I can show them, again and again, what it looks like to feel something and let most of it go.

I started meditating thirty years ago because I wanted out. Out of that, mostly. I dressed it up as other things, a low hum of dissatisfaction, the sense that real life was happening somewhere else. Underneath all of it was the boy in the dark.

I’d read enough to believe there was an exit, and that if I believed hard enough, worked hard enough, sat long enough and quietly enough, it would open. I would walk through and into the light, and the fear would stop.

Some version of this belief is why most people start. The meditator wants enlightenment. The founder wants the exit. The kid wants the offer letter. The shapes differ. The belief underneath is identical. There is a finish line, and on the other side of it, you are finally okay.

I want to tell you what I found, because I hope it will save you time, but suspect it will not.

Subscribe now

The experiences are real. I’ve had them. Hours that dropped away in a single breath. The boundary between me and the room and the world going soft. A bliss that had no cause and asked for nothing.

And then I would stand up. Ten minutes later I’d be irritated about the dishes left in the sink overnight, the same way I’d been irritated about small things my whole life. The bliss does not transfer. The man who came back from the cushion was the man who sat down on it.

For a long time that gap seemed like a problem to solve. More sitting. Better technique. More retreat from daily life. Surely the next breakthrough would be the one that stuck, the one that finally made me into someone who didn’t snap over nothing.

It never stuck. None of them did.

Somewhere in there I gave up. Not dramatically. I stopped believing the exit existed. I stopped waiting to be fixed. I abandoned all hope of arriving.

That turned out to be the redemption. Not despair, the opposite of it. The hope of arriving was the thing that had been hurting the whole time. Not the striving. The hope that it would end. Dropping the finish line didn’t make me lazy. It made me, for the first time, present for the part of my life that was actually happening.

Jack Kornfield spent years interviewing the people who supposedly made it. Monks, nuns, roshis, teachers across traditions, people with deep and genuine awakenings behind them. His finding, in three words: the laundry remains. The neuroses come back. The marriages get hard. The masters still have to do the dishes. His title says it in five words: After the ecstasy, the laundry.

That should be deflating but it isn’t. It means the wise people aren’t a separate species who escaped being human. They just happen to be people who stopped trying to. The path doesn’t lead out of your ordinary life but loops back into it.

One question bothered me though: If there’s no arrival, why strive? Why sit for years to learn there was nowhere to get to?

Because the only way to know a path is wrong is to walk it.

The dead ends, the dry spells, the bliss that fooled me into thinking I’d graduated, the despair that fooled me into thinking I’d failed. None of it was a detour off the path. It was the path. I couldn’t have been told. Nobody can hand you the truth that there’s no finish line. You believe it only after you’ve run at a dozen of them and watched each one dissolve the moment you crossed it.

I run, too. I’m chasing a sub-20 5K I haven’t caught. I train hard for a number I might never hit, and the training is not made smaller by that. Progress isn’t a line. It’s a spiral. The same workouts come around again, the same hills, and the weeks that feel like sliding backward are usually the body reorganizing for the next climb. You pass the same point higher up. The number organizes the effort. It doesn’t validate it. If I hit it, I’ll want nineteen. This used to feel like a trap. Now it feels like the point. You strive toward the destination so the journey has a direction. You stop believing the destination will save you.

So I still sit. Most mornings, before the house wakes up. Nothing dramatic happens. I’m not waiting for the door anymore. I’m not grading myself. I notice the breath, I lose the breath, I come back. The instruction is the same one I got thirty years ago. Then I get up and start the day, which includes, among other things, the laundry.

Chop wood, carry water. Same wood as before. The only thing that changed is the one doing the carrying, and even that only a little, and not on purpose.

Strive anyway.

A few weeks ago I decided to start a company. I expected the long hours but I didn’t expect this.

I’ll say more about what we’re building soon. For now, know that it lives at the intersection of two things I’ve spent my life circling: AI, and making good decisions and predictions.

I want to be honest about why I am doing this. Not ambition, or not only that. It was the plain arithmetic of one life: a finite number of years where I’d still have the energy and the freedom to build something of my own. I found so much curiosity and joy in what I’d discovered that not pursuing it felt like the real risk. The cost of trying was failure. The cost of not trying was a regret I believe I would carry for a long time, or not, we’ll see.

So why the knot?

Subscribe now

The model in your head

As it turns out, the brain is a prediction machine. It builds a model of the world and checks it against what actually happens, thousands of times a day. When the model matches reality, you feel calm and capable, even when you’re busy. That calm doesn’t come from having nothing to do. It comes from knowing what to expect.

For eighteen years in Big Tech I could read my world. I knew what a good week felt like, who to call when something broke, how a launch was supposed to go. I knew my way around. In the past year, I had cleared most of my schedule; most days were consistently around a few themes: family, health, tinkering, and rest.

A startup wrecks that model.

Most days, what happens doesn’t match what I expected, because I don’t have good expectations yet. What’s the right price? Who’s the right first customer? What does an investor’s silence mean? Was today a good day or a bad one? I can’t reliably call any of it. Every surprise trips a small alarm: something’s wrong, fix it now. One alarm is fine. It’s a small amount of stress and is part of learning. But stack these alarms day after day and the alarm never resets. You start expecting threats everywhere, and anything unclear starts to feel like danger.

That was the whole of it. The stress didn’t mean I was bad at this. It meant I was somewhere new, where I hadn’t learned what to expect yet. Of course nothing matched. I was still learning the place.

Two traps for founders

The first is comparison. Open any feed and you’ll find a wall of other people’s funding announcements, traction charts, and accelerator wins. Your brain was built to compare you against a village, a few dozen people you actually knew. It was never built to measure you against a global highlight reel that refreshes every time you scroll. So you lose, over and over, against a sample designed to make everyone feel behind. I went deep on this in You’re comparing yourself to people who don’t exist.

The second is identity. At my old job I was “the director who worked on X.” I gave that up on purpose, and now I’m starting over. On day one of a company, you’re no one in particular. The label that used to answer who am I is gone, and losing it stings more than it should, because we’re wired to feel a threat to who we are almost like a threat to the body. There’s nothing weak about that. And the irony isn’t lost on me. I spent a year telling people that life after a career is the place to be, and here I am, reaching for a new title to stand on. Even if it’s one I don’t particularly crave.

What’s helped

None of this is a reason to go back. It’s a reason to work on myself while I work on the company. A few things have helped.

Treat stress as information, not a verdict. When what happens doesn’t match what I expected, I can dig in, get annoyed, and insist the world is wrong. Or I can change what I expect. The first feels better for a minute and costs me later, because I can’t learn anything while I’m busy defending myself. So now I try to ask a better question. Not what’s wrong with me but what is this teaching me that I didn’t know? That turns the stress into information instead of a judgment.

Don’t let the company become your whole identity. When one role carries all of who you are, every bad week in that role feels like the end of the world. So I keep more of myself in view, not less. I’m a son. I’m a father. I’m a runner chasing a sub-20 5K with the stubbornness of someone who’s finished a 50K. I’ve had a meditation practice longer than I’ve had a career. When the company has a rough week, and it will, those parts of me don’t move.

Don’t tie your worth to the scoreboard. Caring about the work is good, and I want to care a lot. But when my worth rises and falls with the numbers, every dip feels like proof I’m not enough. So I try to care about the work and hold the outcome loosely at the same time. Part of that is learning to lose well. You can’t win every week. When something fails, I give it room, get curious about what went wrong, and try to learn from it without turning it into a story about my value as a person.

When you’re really spiraling, stop trying to think your way out. Some days the rut is too deep, and no amount of clever reasoning gets me out, because the problem isn’t in my head, it’s in my body. On those days I move. A hard run calms me down faster than any pep talk. A long stretch outside, away from screens, does in two hours what a week of willpower can’t. And a quiet sit with my breath puts me back in my body when my mind won’t let go.

But knowing isn’t doing

The four points leave something out. I fail at them, often.

I’ll tell you to treat stress as information, then spend an evening revisiting an investor reply for information which might not even be there. I’ll tell you not to watch the scoreboard, then open the dashboard before I’ve said good morning to my kids. I’ll tell you to move instead of spiral, then scroll for half an hour looking through the code to make sure the idea is still sound.

Knowing the mechanism doesn’t switch it off. The wiring doesn’t go quiet just because you’ve named it.

So this isn’t a thirty-day fix. It’s a practice I keep coming back to. The goal was never to feel no stress. It’s to catch it sooner, hold it more lightly, and be a little kinder to myself while I’m in it.

Still learning the place

I still get stressed. I expect I will for a while, because I’m early in learning a world I picked precisely because I didn’t know it yet.

But I don’t read the stress as proof I made a mistake. It’s what it feels like to be somewhere new, before you’ve found your footing, which is exactly where I wanted to be when I left. The work is hard, and I want it to be. The stress was never the price of the work. It was the price of learning to navigate a life I hadn’t lived yet.

I’m still learning. That’s exactly what I left to do.

What follows is that letter, lightly edited. None of it is clever. Life and financial best practices like: don’t carry a credit card balance, freeze your credit, sleep, call me when things go wrong. The boring stuff is the stuff that works, and most of us learn it too late.

I’m publishing it in case you’re sending someone off too, or in case you’re the one leaving. Take what’s useful. Skip the rest.

Subscribe now

Adulting 101

Before any of the rest of it: I am proud of you, and you are more ready for this than you feel. You will get things wrong. That is the point of going. Call me when you want to, not because you have to.

A few principles I hope you carry with you. You do not need to master all of this at once. The goal is to build a good base, avoid big mistakes, and give yourself more freedom over time.

1. Mindset for Life

Your mindset will matter more than almost anything else as life unfolds.

Do not let one mistake become your identity.
You will mess up. Learn the lesson, pay the tuition, move forward. A mistake once is experience. The same mistake again and again is a problem to solve.

Take calculated risks.
A safe life is not the same as a good life. Do not avoid change just because it is uncomfortable.

Do not expect instant success
Nobody owes you anything. Earn trust, build skill, do the work. Compounding works in money, skills, health, and relationships. None of it shows up fast. Most good things take several attempts. Early failure is not proof you cannot do it.

The rest, in no particular order:

• Do not dwell on the past. Feel it, learn from it, then move. What is the point of carrying it?

• Avoid self-pity. Bad things happen to everyone. Feel them, then move.

• Do not let jealousy run you. Someone will always be richer, smarter, luckier, or further ahead. Their success does not shrink your path.

• Do not obsess over what you cannot control. Put your energy where your actions matter.

• Do not try to please everyone. It is impossible and it will make you doubt yourself. Be kind, but have a spine.

• Do not give others power over you. Your worth does not depend on anyone’s approval.

• Do not avoid solitude and silence. You need quiet to hear yourself think.

2. Take Care of Yourself First

You are about to be fully in charge of your own body and mind for the first time.

Sleep is not optional.
Pulling all-nighters feels productive and is not. Lack of sleep wrecks your mood, your memory, and your judgment.

Move and eat like someone who plans to be around a while.
You do not need to be perfect. Walk, get some sun, do not live on energy drinks and vending machines.

Asking for help is a skill, not a weakness.
Office hours, your RA, your advisor, daddy and mommy. You don’t have to white-knuckle a problem for months rather than ask a question. Be the one who asks.

Drink slowly or not at all, and never lose track of your own count.
Don’t drink would be my advice but if you must, be aware. Do not take a drink you did not watch get poured, and do not leave one unattended.

Never get in a car with a drunk driver, and never be one.
Walk, call a ride, sleep on the floor. There is no version of this where saving some money is worth the risk.

Important Reminder
If you are ever in a place where you feel unsafe with your own thoughts, you can go to the campus counseling center or a doctor. You can always call me or your mother, any hour.

3. Protect Yourself From Scams and Theft

Before trying to get rich, avoid getting wiped out.

If it sounds too good to be true, stop.
Pause. Do not rush. Scammers create urgency because clear thinking is their enemy.

Credit freezes
Freeze your credit at the major credit bureaus. It helps prevent identity theft and fraudulent accounts.

Virtual credit cards
Use virtual cards when possible, especially for online subscriptions or merchants you do not fully trust.

Be careful with money and friends
Trust people, but be safe. Money can damage relationships. Be especially careful with loans, investments, shared expenses, and vague promises.

4. Money, Now

Debt is not always evil, but it is dangerous when used casually.

No credit card balances
Never carry credit card debt. Pay the full balance every month. Credit cards are tools, not loans.

Avoid debt unless it has a clear purpose
Debt reduces freedom. The fewer fixed obligations you have, the more choices you have.

Spend less than you earn, even when “earn” is small.
The habit matters more than the amount right now. How much you save will matter more than how cleverly you invest for a long time.

Keep a cash buffer.
Enough to cover a surprise so you are never forced into a bad decision. Build toward a few months of expenses over time. Any amount beats zero.

When you do invest, keep it brutally simple.
Broad, low-cost index funds (VTI or VOO for now). Boring is good. You own the whole market instead of trying to outsmart it. Use a low-cost platform like Vanguard, not a bank. Banks are good for banking and will happily sell you high-fee products that are not in your interest.

5. Money Later (Read This When You Have a Real Salary)

None of this applies until you have a steady income, so do not worry about it yet. When you land a real job, send me a message and we will set it up together in about an hour. Rough order:

1. 401(k) up to the employer match. Free money, take it first.

2. HSA, if you have a high-deductible health plan.

3. Backdoor Roth IRA, if your income is too high to contribute directly.

4. Max the 401(k).

5. Mega-backdoor Roth, if your employer plan supports it.

6. Taxable brokerage for anything beyond that.

And when life gets bigger (a partner, kids, a house, real assets), term life insurance if people depend on you (simple term, never whole life sold as an “investment”), disability and umbrella insurance, a will, and basic estate planning. Not now. Just know it exists and we can talk through them later.

6. The Big Picture

Build a life with options.

• Choose good people. Be a good person.

• Stay humble. Keep learning.

• Do not let one mistake become your identity.

• Spend less than you earn. Avoid stupid debt.

• Invest simply and consistently.

• Protect against the few catastrophic risks.

The goal is not to get obsessed with money. The goal is to make money boring enough that you can focus on the people, the work, and the life that actually matter.

Five weeks ago I ran my second parkrun in 23:47 with a 2-second positive split, and wrote a whole post about pacing finally clicking. Today I ran my third parkrun in 23:20 with a 32-second positive split.

I am excited about the PB, but I am disappointed about the fade and massive positive split.

Subscribe now

The race, in two halves

The first two miles were as clean a race execution as I’ve ever put together. Mile 1 at 4:30/km, mile 2 at 4:31/km. Cadence 180+, stride length 1200mm, vertical ratio 7%. These are essentially my best-ever interval biomechanics, sustained for over 3.2km of continuous racing. If I had been able to keep the pace through 5km I would have ran 5K at 22:30. Nice.

Then mile 3 happened, and every biomechanical marker collapsed in lockstep. Cadence fell 6 spm. Stride length dropped 20mm. Ground contact time stretched 12ms. Vertical ratio jumped from 7% to 7.44%. All of this is to say that my running form just went south and I slowed down a lot.

It turns out that the cause was at km 2. 4:27/km in the second kilometer of a 5K is above my lactate threshold and my average HR jumped 13 bpm between mile 1 and mile 2. I didn’t realize that I was burning candles and by mile 3 I was anaerobic, and there’s no recovery from that inside a 5K. I simply couldn’t slow down enough to clear lactate without slowing down too much to finish.

I knew it was happening at the time. I felt the legs flatten at 2.5km and watched the pace drift up despite holding effort. I still tried valiantly and my max HR in the race was 186, the highest I’ve ever recorded.

What got me here: the May 13 breakthrough

The training between parkruns was four sessions of 6 × 1000+m repeats on the track (in the outer lanes).

Apr 29: 4:27/km average, 24-second spread. A front-loaded blow-up. I opened rep 1 at 4:13/km, the fastest single rep of the entire arc, then faded to 4:36 by rep 6. A new ceiling but no sustainability.

May 6: 4:28/km average, 9-second spread. Controlled negative split. Rep 6 was the fastest rep of the day.

May 13: 4:22/km average, 2-second spread. Six reps between 4:21 and 4:23. The cleanest, fastest workout of the whole arc. The session that suggested 22:30 was on the table.

May 21: 4:26/km average, 6-second spread. An off day with heavy legs, ahead of taper. The body asking for a recovery week, which it got.

The middle two weeks, May 13 in particular, are why this race had 22:30 legs.

The tune-up that predicted more

Three days before the race, I ran a 2.4km tune-up. Same IPPT distance I run before every race now, partly because it gives me a clean year-over-year comparison.

May 27 tune-up: 10:38 at 4:25/km average. 26 seconds faster over the same distance than the tune-up before parkrun #2. Twelve seconds per km faster. Stride length up 58mm from April. Vertical economy holding in the same band. Average HR 164, max 172, well under race-day cardiac ceiling.

The 4-sec/km gap from tune-up to race held last time. April 22 tune-up at 4:37/km predicted a 4:41/km race, which is exactly what happened. If the same conversion had held today, I would have raced at 4:29/km. That’s 22:25.

The race the tune-up predicted didn’t show up. The race I ran is the race I paced.

Same fade, fitter legs

Three races on the same course. Three PBs. The fitness curve is monotonic. I’m 84 seconds faster over the same trail than I was on March 21, and roughly a percentage point higher on age grade.

The pacing curve is not monotonic. Parkrun #2 was the cleanest race of the three in this aspect. Today I opened more aggressively than parkrun #2 (4:32 vs 4:44) and produced essentially parkrun #1’s positive split (32 seconds vs 36). The shape of today’s race is the shape of my first ever parkrun, not my second. The discipline I taught myself in March didn’t carry through to today, mostly because the fitness underneath it changed and I didn’t recalibrate.

Pacing isn’t permanent

Pacing is a skill you re-execute every race, not a virtue you accumulate. The right opening pace at 23:47 fitness (4:44/km) is different from the right opening pace at 22:30 fitness (somewhere around 4:34/km). The shape of the discipline is the same. The numbers shift. And the calibration has to happen fresh every time you toe a line, because you only roughly know what your current fitness is.

The 4:30/km mile 1 today felt conservative. It felt the way 4:44/km would have felt in April. Against the 22:30 fitness I’d built, it was neutral, and 4:27 in km 2 was over the line. I was driving a faster car at the same throttle settings I’d learned on the slower one, and I overshot the corner.

The fix is small in mathematical terms and meaningful in mental terms. Open two to three seconds per km slower. Trust the fitness is there. Let the kick at the end produce the negative split that the discipline is supposed to enable.

I will write this down again in the next race plan and try harder to mean it.

Sub-20, still eventually

Sub-22:30 is the next gate. The fitness for it is already locked in. Today’s first half was 11:14, sub-22:30 pace held for 2.5km. The job is repeating it for another 2.5km, which means opening at 4:33-4:35/km and trusting there’s more in the second half than I expect.

Sub-22 is probably one more training block beyond that. Sub-21 is the gate that asks for a different kind of work. More weekly mileage, real strides, proper plyometrics, possibly some hill repeats. Sub-20 is several gates beyond, and still a long way off.

I wrote in the last post that the climb is patient, whether I am or not. That hasn’t changed. What’s changed is that today gave me a sharper picture of which gates need fitness and which gates need discipline. Sub-22:30 doesn’t need more training. It needs me to execute a plan I already know how to write down. Sub-21 will need both, and a lot more of one of them.

The honest version of where I am: I have 22:30 legs and 23:20 pacing. The fitness curve is doing its job. The pacing curve is mine to close, and it has to be closed in real time, while the race is happening, against a body that always wants to go faster than it should in the second kilometer.

You train for what you want to do. Today I trained for sub-22:30 and ran a 23:20 PB. Next time, I want to be patient enough to let the fitness do its job.

Sub-20 stays the north star. The slope today was steep in the wrong direction for one mile, and it cost me 30-45 seconds. The direction over the longer arc is still right. The process is still working. The pace will keep moving as long as the work keeps showing up, and as long as I keep relearning, race by race, what controlled means.

When I was younger, most of what I pictured about the future came from someone else. Back to the Future shaped how I saw it. The Matrix and Minority Report shaped it differently.

These days, influencers on social media bombard us with predictions of the future. But you didn’t reason your way to those possibilities. You absorbed them, and now you quietly repeat them as if they were your own conclusions.

That mattered less when change was slow. It matters more now. Whatever you believe about AI timelines, the rate of change has gone up, and borrowed predictions age fast. I’m not in the singularity-is-imminent camp. But I’ve stopped trying to find the one person who called it right and reading them like scripture. The more useful move is to get better at the thinking itself, so you can run your own scenarios instead of relying on other people’s.

Here’s the framework I am using.

Four lenses: what’s possible, what should be, what’s likely, and what to avoid

Almost every claim about the future is one of four things. Naming which one you’re looking at is useful, because each one works and fails in predictable ways.

What’s possible: What the technology might someday enable if today’s limits disappeared. This is the type of prediction that runs science fiction, product launches, and most of what gets called “the future” in public. It’s exciting and it spreads, which is exactly the problem. These futures show you the extreme version of everything with no path to get there, and they collapse the moment you check the details. There is usually another issue: the primary beneficiary is an advanced user, such as a surgeon doing something miraculous with the headset, never the ordinary person who buys it to play games. Many of these predictions should be treated as entertainment until proven otherwise.

What should be: Anyone who tells you how the future “ought to” look, backed by numbers and moral weight, is doing this. It’s a strong motivator. A clear, desirable vision pulls people toward it and sometimes bends reality into making it true; we describe paying with credit before the system exists, then go build it. The weakness is that it’s self-centered by design. Everyone’s preferred future is incompatible with someone else’s, and people present their own as inevitable. The skill is spotting when a should be is dressed up as a neutral forecast. “This is where things are heading” often means “this is where I want things to head, and I’d like you to help.”

What’s likely: These predictions are the most honest. They admit the future is uncertain and hold several possibilities at once instead of betting everything on one. They give you tools: scenario sets, base rates, the futures cone with probable outcomes near the center and the tail at the edge. The catch is that a model is only as good as its inputs and its hidden assumptions. A tidy 2x2 of four scenarios feels rigorous and can quietly delete the ninety other futures that didn’t fit the grid. Use the tools, but keep asking what the tool is leaving out.

What to avoid: Cautionary futures, unintended consequences, the failure mode of a well-meaning plan. This one is underrated. Fear is a strong teacher, and a good dystopia changes behavior in a way a roadmap never will. Its failure is the reverse of the what’s possible problem: spend long enough hunting for what could go wrong and you see only that, which makes you as blind as the optimist. A warning with no alternative attached is just despair with footnotes. Pair every what to avoid with a then what.

The practical use isn’t to pick a favorite. It’s to label the next future-claim you encounter, including your own, and ask which lens it’s stuck in. Most predictions are a what’s possible dressed in flowery language, or a what should be pretending to be a what’s likely.

Subscribe now

It’s almost never the extremes

The most useful lesson I’ve taken from all this: stop spending attention on the tails.

Both the jetpack future and the collapse future are seductive and roughly useless for planning, because most of life isn’t either. Most of life is unremarkable. People go to work, eat, sleep, deal with the thing that broke, and enjoy the occasional good day. A realistic picture of the future has to survive contact with that ordinariness. The sharpest exercise I know is to pick something mundane and play it forward. What does a grocery store look like in 20 years? What’s on the shelves, what does it cost, what’s the floor plan, what rules does it follow? You’ll learn more from that than from another story about colonizing Mars.

Three errors keep dragging predictions toward the extremes, and they’re worth naming so you can catch yourself:

• Treating the future as a replacement instead of a layer. New things rarely wipe the slate. They get bolted onto the existing tangle of laws, infrastructure, and habits, and that tangle shapes them as much as they shape it. Anything you can’t get rid of in a decade is a thing you have to plan around, not plan past.

• Assuming things won’t break. The honest future has glitches in it. Products get misused, systems fail, people do the unintended thing. A future with no friction in it is a marketing asset, not a forecast.

• Mistaking a story for data. Hard numbers tell you what happened. The moment you point them forward they turn into a narrative wearing the numbers as a disguise, and that narrative gets less reliable the further out you push it. Confidence in a forecast is not evidence for it.

The impact on society runs through the average person’s average day, not through the exceptional case. The change that matters is usually boring. It’s a tweak to a tax rule, a default setting, a price that drops below a threshold, a form that gets shorter. None of that makes it a blockbuster movie or bestseller book. All of it moves the median Tuesday, and the median Tuesday, multiplied by a few hundred million people, is what “society changed” actually means. When you evaluate a future, ask what it does to the unremarkable middle, not what it does to the hero in the demo.

What the AI debate looks like through these lenses

The loudest future-argument right now is what AI does to us in the next decade, and it’s mostly two camps shouting past each other.

On one side, the doomers. Eliezer Yudkowsky’s 2023 Time piece argues we need to “shut it all down”. The Center for AI Safety’s one-sentence statement, signed by Hinton, Bengio, Altman, Amodei, and Hassabis, places extinction risk from AI alongside pandemics and nuclear war. Daniel Kokotajlo’s AI 2027 scenario lays out a step-by-step path to superintelligence and possible human takeover by the end of the decade. In September 2025, Yudkowsky and Nate Soares published If Anyone Builds It, Everyone Dies, now a New York Times bestseller, whose thesis is exactly what the title says.

On the other side, the accelerationists. Marc Andreessen’s 5,200-word Techno-Optimist Manifesto declares stagnation the real enemy and treats technological acceleration as a moral duty. The e/acc movement organized around it. Leopold Aschenbrenner’s Situational Awareness essay forecasts AGI by 2027 and a US-China arms race over trillion-dollar compute clusters. Anthropic CEO Dario Amodei, who also signed the CAIS statement above, published Machines of Loving Grace in October 2024, a 15,000-word essay arguing the upside of AI is radically underestimated and could compress a century of biological progress into five to ten years.

Run all of this through the lenses and the same trick shows up.

The doomer case is fundamentally a what to avoid. Extinction is the worst outcome anyone can name, and the work of naming failure modes and funding safety research is exactly what what to avoid is supposed to do. The strongest doomer pieces don’t present the bad outcome as one possibility among many. They present it as a what’s likely. AI 2027 reads as a forecast, not a warning. If Anyone Builds It, Everyone Dies puts the inevitability in the title. The value-claim, that extinction would be catastrophic, is sincere and probably right. The probability-claim is painting this as urgent when it’s still one possible future among many. If you remember Y2K, plenty of people thought planes were going to fall out of the sky.

The accelerationist case runs the same pattern in reverse. At its core it’s a what should be: a conviction that progress is good, abundance is the goal, and the ideologies that brake technology are the real enemy. The Techno-Optimist Manifesto is openly a values document; it ends with an enumerated list of enemies. Machines of Loving Grace is upfront about its frame from the first page: a sketch of what a world with powerful AI might look like if everything goes right. But these visions get repeated and recast as what’s likely, with intelligence takeoff and material abundance presented as the path we are already on. Aschenbrenner’s 165 pages make a vivid story of inevitability, and the essay was successful enough to be spun into a $1.5 billion hedge fund. A should be turned into an investment thesis is about as on-the-nose as the framework gets.

The most interesting cases are when a single actor runs both moves at once. The same company that published Machines of Loving Grace in 2024 announced in April 2026 that its most capable model, Claude Mythos, was too dangerous to release to the public and would go only to a hand-picked group of about forty corporate customers under a program called Project Glasswing. The framing is what to avoid (this model could exploit critical infrastructure). The practical effect is what should be (Anthropic’s products are this powerful, so use them). Critics in the mainstream press noticed and called it safety theater deployed as marketing. When the safety claim doubles as the sales pitch, picking the right lens stops being academic.

The same problem shows up at the what’s possible end. In October 2025, OpenAI publicly claimed GPT-5 had cracked ten previously open Erdős problems. The mathematician Thomas Bloom, who maintains the canonical Erdős problems list, looked at the work and called the framing a dramatic misrepresentation; the model had retrieved existing solutions from the literature, not produced original proofs. The executive who posted the claim eventually left the company. Then on May 20, 2026, OpenAI announced a genuine result: an internal reasoning model disproved an Erdős conjecture from 1946 about how many pairs of points on a plane can sit at exactly unit distance from each other. This time the proof came with external verification from Bloom himself, Noga Alon, and Melanie Wood. The contrast is the lesson. The first announcement was a what’s possible claim dressed as a done-deed and collapsed in days. The second was a real result that survived the same scrutiny. The lenses, plus the social process of verification, are the difference between the two.

What’s missing from most public AI talk is the what’s likely in the boring sense. The view that AI is closer to electricity or the internet than to a new species, transformative over decades, diffused unevenly, mostly experienced as faster spreadsheets and better customer service and gradual shifts in which jobs exist, gets less airtime because it doesn’t make a thumbnail. Arvind Narayanan and Sayash Kapoor have argued exactly this in their Princeton-Knight AI as Normal Technology paper, explicitly positioning the view as the alternative to both utopian and dystopian framings that treat AI as a separate kind of entity. Even the careful doomers are quietly updating in this direction. Kokotajlo, whose AI 2027 went viral on that title, has since revised his median AGI estimate to around 2030 as the underlying capability curves came in slower than expected. That’s the move you want to see from anyone making predictions, and it’s mostly absent from the loudest voices on both sides.

The lenses don’t tell you which camp is right. They tell you what each camp is doing, which is most of what you need to stop being shoved around by them.

Now point all of this at your own life

The same four lenses work on the only future you steer: your own.

Run them on your next decade instead of the world’s:

• What’s possible: What’s genuinely open to you? Most people underestimate this one because they only ever saw the demo-reel version of other people’s lives. Imagine it for yourself.

• What should be: Whose voice is the “should” in your head? A parent, a peer group, an algorithm optimizing for your attention? A should be you didn’t choose is the most expensive thing you can carry, because it spends your years on someone else’s vision.

• What’s likely: What’s the honest base rate, given your real trajectory? If you keep doing exactly what you’re doing, where does that land?

• What to avoid: What’s the future you’re quietly drifting toward that you’d regret, and what’s the early signal you’d see first? Name it now, while you can still cheaply change course, and attach a then what so it’s a plan and not just a fear.

Then apply the boring-Tuesday test to yourself. Don’t picture the promotion or the finish line. Picture an average Tuesday five years out: what you do in the morning, who you spend the day with, what you’re tired of, what you look forward to. If that ordinary day is good, the plan is good. If you can only make the case with a highlight reel, the plan is a what’s possible in disguise.

Pick one real decision you’re sitting on. Run it through the four lenses; write a sentence for each. Put a probability on the two outcomes you’re worried about. Even a rough 60/40 beats a vague “probably.” Then put a date on the calendar for when you’ll know whether you were right, and go back and check.

You won’t predict the future. Nobody does. But you’ll stop outsourcing it, and that’s the part that was always yours to do.

I remember a monk answering a question on how to fall asleep for someone who was having problems sleeping.

His advice was: don’t think about sleep.

He suggested thinking about how comfortable the bed felt or how nice the room was. He explained that the moment you think about sleep, you’ll think about how you’re not sleeping. Then you’ll start to think about how long you’ve been not sleeping. Then you’ll think about tomorrow’s meeting, the run you’ll have to skip, and the slow drift toward exhaustion that never quite arrives. You toss and turn and repeat. It turns out thinking is the problem. Sleep doesn’t come because you want it. It comes because you stopped trying.

A lot of people in that room laughed. There is lots of wisdom there.

Despite how minor this topic was compared to other headier topics discussed in many sessions, I think about that talk often. Not at 3am, mostly. Any time I’m trying to feel a way other than how I feel.

The trap of good and bad

Most of us grew up with a vague spectrum of feelings. Joy, calm, gratitude, love, these are the good ones. Anger, sadness, jealousy, grief, fear, these are the bad ones, the ones we were quietly taught to manage, suppress, fix, or hide. The good ones we welcome and the bad ones we treat like uninvited guests.

This ends up with the same trap as the one with sleep.

You can’t push a feeling away by deciding it’s bad as it becomes another layer of the feeling. Now you’re not just sad, but you’re sad and also ashamed of being sad. You’re not just angry, you’re angry and worried that your anger makes you a bad person. The story you tell yourself about the emotion becomes heavier than the emotion ever was.

Subscribe now

What we usually try

When a feeling shows up that we’ve decided is wrong, we try a few tactics:

We distract ourselves. We open a phone, pour a drink, or start scrolling. The feeling goes quiet for a while and we count that as a win, even though it’s still there, demoted to background noise.

We hide the feeling. We tell ourselves we shouldn’t feel this way and push the feeling somewhere it won’t be seen. It goes down, but it doesn’t leave. It comes out later as tight shoulders, a short fuse with someone you love, a bad night’s sleep.

Finally, we vent. We tell someone, share, or post about it. There’s a small relief in this, like opening a valve, but the pressure builds again because the feeling itself never got met. Worse, we hand a piece of it to someone else, and over time the people closest to us start carrying our load.

None of these are wrong, but there is another way I have found liberating: letting the feeling be there.

Sitting with it

I’ve learnt, for myself, to let “bad” emotions or thoughts be in the body. I don’t argue with them, don’t moralize about them, don’t try to reframe them into something more presentable. I feel the heat in my chest, the heaviness in my body, the tightness in my jaw, and let it move through.

It’s much less dramatic than it sounds. The feeling becomes ordinary as soon as you stop feeding it a story. The shame of being sad falls off when you stop calling it shameful. The fear of being afraid stops replicating once you stop fighting the fear.

What surprised me, the first time I really tried this, is that big feelings are usually three or four feelings stacked on top of each other.

If I sit with anger long enough, sadness shows up underneath it. If I sit with sadness long enough, fear is usually underneath that. If I sit with fear long enough, there’s something quieter under it, something that feels like an old belief about whether I’m okay as I am.

You don’t get to the bottom layer by attacking the top layer. You get there by being patient with what’s actually present.

Don’t just do this for “bad” emotions though. As much as we welcome “good” emotions, I have found that sitting with “good” emotions exposes their hold on me; while it might feel good to revisit “good” emotions and memories, it is even more liberating to let them go.

The present is the only place this works

“Live more in the present” sounds like a wellness slogan or mindfulness mumbo jumbo until you notice what it really means. It means meeting what’s here without trying to swap it for something else.

The present moment, by definition, is the only place a feeling can actually be experienced. The minute you start labeling it, you’ve left. You’re now in a story about the feeling, not the feeling itself. And the story is where the suffering lives.

Anger isn’t bad. Anger is information. Sadness isn’t bad. Sadness is information. Fear isn’t bad. Fear is information. The category we filed them under was ours all along.

When you stop rationalizing, when you stop trying to feel a way you don’t feel, the emotion does the thing it was always going to do anyway. It moves. It changes. It softens. Sometimes it stays a while, but it stops running the show.

You fall asleep when you stop trying to fall asleep. You feel better when you stop trying to feel better.

May you be present and happy.