I once lived in Berlin. I was 19, and alchemically thin. I earned essentially no money, getting fired from dead-end jobs and doing nude photoshoots for local 'artists’, subsisting on partying and 10 cent Brötchen. I shared a single bed top and tail, read Russian literature, and played Shogun. It was amazing.

But, after a while, I noticed something. I would often meet people twice my age, still thin, still partying. This made me question things. No matter how amazing it is, you can't live in Berlin forever. That's why I had a child.

Not immediately. In fact, my wife and I delayed having a child for as long as biologically reasonable. By the standards of television, we are not the sort of people who have kids. We are homeless. Not in the literal sense, but in the metaphorical bourgeois sense of finding our own way through the world. Since leaving home we have switched countries every couple of years. Neither of us have ever had a stable job in a stable company. Our family live on the other side of the phone, not the other side of the street. Friends 'back home' marry their high school sweethearts, have careers you can explain, and live near their parents. They have kids.

By our own standards, too, there were many good reasons not to have kids. Some positive: we had lots of things left to do and wanted to make the most of our time as a couple. Some neutral: neither of us were especially fond of or familiar with other people's children. And some negative: we feared that our lives would end when our children's began. Perhaps we would willingly give up our personalities, passions, and preferences. Perhaps they would be taken from us. Both were terrifying. We liked who we were.

But, 14 years after leaving Berlin, biology said now or never, and we chose now. You can't live in Berlin forever.

Leaving Berlin

This state of mind was not ideal preparation for fatherhood. You can't live in Berlin forever, but having a child was like leaving to live on an iceberg. I felt all the wrong feelings. When my wife held our tiny son for the first time, covered in blood and guts after he had near murdered her, she declared her instant love. I sensed what I was supposed to feel, but I wasn't feeling it.

Every newborn is different, but the first month or four are generally entirely oriented around sleep deprivation. This is true especially when you live 5,000 miles away from friends and family who would otherwise help. Our son would sleep, scream, poop, scream, eat, scream, vomit, scream, on a 2 to 3 hour loop. At no point could he be horizontal without vomiting, nor could he sleep while still. Newborns sound like they are dying when they sleep. My wife and lived in shifts, opposites to each other. This is all normal. But for every family, including ours, at some point something 'not normal' happens - whether in pregnancy, birth, or infancy. It might be as dangerous as it is absurd: my son didn’t poop from January to March! This is a new kind of stress, mortal fear, and that too is normal.

The first month turned me into an asshole. It is the only time my wife and my's relationship has truly frayed. I wrote a note to myself because I knew at some point I wouldn't remember. The title was "don't forget - don't have another baby!". 19 bullet points, 1,000 words. Reading back, it is a soup of parochial lunacy and guilt over missing feelings, combined with the realisation that I had left Berlin forever. This poisonous soup literally made my heart stop. A few weeks in, I woke up sweating, heart thumping crazily, and then nothing. Thump, thump, thump. Nothing. Thump. Emergency room. Turns out as well as a new baby, I had a new arrhythmia.

Double the love

My son is now almost one. As he grew and my sleep returned, I lost the feeling of my arrhythmia, and found the feelings I had missed. In fact, I found the feelings I had always missed. I love being homeless and finding my own path. But, after a while I came to realise that there are some things I will never have.

Friends, family, and familiar faces, sustained across time. Real relationships. These things matter most, and most people don't even notice that they have them. There is no substitute. There is no shortcut to a long time, and there is no route to community and belonging without commitment. Once you are homeless, you can never un-sever your roots. While you were away, life went on.

Every day love is scarce for me. I love my friends and family, but not every day, not every year. Those I've known longest I see a few times a year. Those I see more often I haven't known for long. Without a home you get, if you are lucky, one roadside companion, one every day love. Mine is my wife.

And now, I have a son! And I love him! I love him every day, forever until we travel apart. I love him when he can only sleep in my arms, when all he wants to do is dance to a single Peppa Pig song over and over, and when each moment he is new again. Every day now I feel double the love.

Having a child gave me the roots I thought I could no longer have. It seems ironic now that having children felt like something for friends 'back home'. For people who have always had stability, community, and every day love. But what's another million to a billionaire? The reasons a child seemed so incompatible and so impossible for me are the reasons why this has been so important. Before having my son I saw only what I had left to lose, and had given up on the kind of love and relationships I had already lost. I now see this was the exact gift that only he could force me to accept.

Most things aren't like startups, and it'd be terrible if they were. If the lessons from civil engineering didn't hold from one building to the next, we wouldn't have anywhere to live.

And yet, people still seem to read startup advice or, at least, want it enough that there are endless books and companies and universities dedicated to giving it. I've even given a lot of it myself over the years, whether people asked me to or not. The conclusion, therefore, has to be that startup advice just doesn't work.

Easy to say, hard to do

One explanation of why startup advice doesn't work is that it's mostly just wrong. But this doesn't feel right to me. It would be surprising if no one ever learned anything about how to start a startup. And even if people learned hardly anything, as long as some of them wrote it down, it'd still be adding up over time.

Another explanation is that most advice is right but still bad. This happens because most advice isn't really a strategy, but an outcome. If you're a man, being tall is often better than being short, but 'be tall' isn't good advice. Which means 'make something people want' isn't good advice either, even though it’s the most famous startup advice ever. Making something people want is what all good startups do, and hearing the phrase might make you more aware of it, but it doesn't help you figure out how. Whenever I find myself giving advice about metrics, like "get X users by Y date", it's mostly because I don’t know how to do it.

Giving advice about how to do things is especially hard because advice that was right before might be wrong now. Startups aren’t like civil engineering - you can build identical buildings one after the other, but you can only build each company once. Successful startups have to succeed slightly differently to every startup that came before. If engineering was like startups, the laws of physics would change slightly every time someone laid a brick. So it’s hard to come up with startup advice that lasts forever: sometimes good advice for one company becomes bad for the next.

This means there’s a high bar for advice to work: it needs to be right; tell you how to do something; and apply to you specifically. And yet, even if you get advice like this, I bet it still wont be effective. That’s because the main reason startup advice doesn't work is that startup advice is easy to give other people but hard to follow yourself.

Even the best advice is hard to follow because you are living in first person. This happens all the time, and it's most common in the situations that are most important. If your friend is dating an asshole, it's easy to see. Unfortunately, usually they don't see it themselves because they are them and not you. And often, even when they do see it, they stay with the asshole much longer than they should. I don't blame them. If you’ve ever been in love, you’ve been delusional. You might have even been the friend dating the asshole. And if you've ever broken up with someone, you'll know that there's no good way to do it. Don’t date assholes - easy to say, hard to do.

Startup advice is like this. For example, in the early days of Entrepreneurs First we discussed moving to San Francisco, and never did it, because if it didn't work out it would have probably killed the company. One day, EF might have to come to SF, we said. But not today. Today, we go to Singapore! Meanwhile, a common piece of advice we'd give on the cohort would be 'run into the spike'. We stole this from Genius, reframing their original 'do the thing you least want to do next' as 'do the thing that might kill your idea/cofounding team'. I now see the irony.

EF delayed coming to SF until it was around 10 years old, long after I'd left. This was, it seems so far, the best decision they ever made. It could have killed the company. Run into the spike - easy to say, hard to do.

This is one reason why most founders who succeed once still usually fail when they try again. Even if you know what you're supposed to do, even when you give great advice to people in your exact situation, everything important is harder in first person.

I'm no good at following my own advice, and often I don't know what founders I'm advising should do. But I'm currently working on some new ideas, as well as advising some founders, so I was thinking of ways around this. I want a way to make startup advice work, and I found one.

How to take your own advice

If startup advice doesn’t work because it’s is easier to give to other people than to follow yourself, to figure out what to do you should flip your situation around. Every startup CEO has a competitor. Even if you don't know who they are, you can be sure they exist. In some San Francisco townhouse is some other founder hacking away on the exact same thing you are. Any idea worth working on will have people working on it, especially once it starts to seem promising.

Usually, it's not a good idea to focus too much on this. Startups kill themselves before they kill each other. But, when you're wondering what to do as a founder, you can use your competition to tell you what to do. Just imagine your scariest hypothetical competitor, and then ask yourself what they're doing right now that makes you so afraid of them. If you're the CEO of a company trying to work out what to do today, ask yourself what you would be most afraid of if your CEO nemesis did. Are they launching something? Are they closing that big customer? Are they raising a round from someone awesome? What makes them scary to you will be what makes you scary to them.

This works because it is a reflection of reality. Your nemesis is your own advice on exactly how to win, you just need to copy them. At some point, winning means beating everyone else, so you'll need to become the most terrifying founder on the planet one way or another. And, conveniently, you don't have to know anything about startups to know what's terrifying.

May you break up with that asshole and move to San Francisco.

Some people have a knack for languages. They say annoying things like "Portuguese is just Spanish with an accent", or "we speak French to the kids". Fuck those people.

I was not allowed to take my French exams in secondary school for fear I would bring down the school test scores as well as my own (hi there hothouse British school). Instead I was made to study Latin (did I mention my ridiculous school?) because it had no speaking component to fail. Nevertheless, I barely passed, scoring the worst mark in my class.

Languages? I just don't got it.

When the love of my life informed me, therefore, that in order for us to marry and start a family I would need to learn Dutch, I realised I would die alone. I had already brushed off her family's requests - the Dutch have higher English literacy rates than the British! Isn't Dutch a dead language? I'm pretty sure even Dutch people just revert back to English when no one is looking. If a tree falls in a forest and no one is around to hear it, does it make a sound? No matter how I protested, my girlfriend remained dead serious.

So, I had to learn Dutch. But how? I had some major constraints:

• I don't like teachers,

• Other ways of learning a language don't work,

• And I need to learn a language, sharpish.

I'm sure if you find your Mr Miyagi, teachers are great. But I never found mine. I was always getting in trouble at school. My primary school teacher said I needed 'squashing', but was unfortunately unsuccessful. Due to her failed squashing I wound up getting kicked out of the aforementioned hothouse secondary school - the silver lining being that my terrible Latin didn’t bring down their grades. Teachers were off the menu.

And everything else kind of sucks for actually learning. We all have a friend with a 500 day Duolingo streak who can't have a basic conversation. We all used those textbooks that tell you the items in your pencil case before they tell you how to order a beer. We all know that weeb whose identity became teaching abroad. I ain't doing it.

And why should I? Do the old ways even work? Estimates say it takes about 1000 hours to learn a language. That's three years doing an hour a day! Even if I had three years to spare I wouldn't spend it on Dutch.

I therefore had no choice but to examine, from first principles, how do you learn a language in the least amount of time possible, when you don’t want to?

An example of my then girlfriend’s ideal family arrangement.

What is a language, anyway?

Lots of time in language classrooms and textbooks and apps is spent on grammar and pronunciation and spelling and cultural context. I am going to claim that these are, in the scheme of things, complete bullshit.

The more I thought about it, the more I realised a language is just words. While this sounds trivial, it is a startling and underutilised fact. If you don't know the words, you can never speak the language. If you do know the words, you might get sentences upside down or say them funny, but you have a chance of saying something. There's no escaping it: to learn a language you have to learn the words.

So what words do you have to learn?

The good news is, you don't need to learn a lot of words to get started. Something like 1000 words make up about 75% of all the words used in regular sentences, and 2-3000 words will cover 95% of everyday conversation. The bad news is, you need to know a lot of words to get near 100%. Not a single native speaker knows 100% of the words in their own language, and they generally know 10-20,000.

Language schools have a system, CEFR, aimed at measuring fluency. And research suggests 3000 words are necessary for B1 aka 'Bad'. The final boss of CEFR, C2 aka 'Completed', needs about 5000 words. Knowing a word in this context means recognising and comprehending at least one word in a word family (run, ran, running = 1 word family). These aren't words you can actively use off the top of your head and conjugate perfectly. Research suggests your active active vocabulary might be less than half your passive.

Want to speak a language? You've got to learn at least a few thousands words.

So, I thought, may as well start there.

Memorising 10,000 sentences

I would be going to my in laws for Christmas, and I wanted to surprise them by speaking Dutch.

After thinking long and hard about ways of learning Dutch that didn't require me to know any Dutch words, I realised it could not be avoided. So I committed like a ten year old microwaves a fork. Over the next few months, I went on a vocabulary spree and memorised roughly 5,000 Dutch sentences, each testing a different fill-in-the-blank word. There was no fancy gamification, no teacher, no cultural exchange. Just raw dogging rote memory.

Christmas came, and I was ready to test my theory. I felt quietly confident. I gave my usual apologies for only speaking English at the start of the evening, and waited for conversation to switch to Dutch as is both reasonable and predictable at a Dutch family Christmas. I had my interjection rehearsed - I would crack a timely joke in Dutch for the big reveal. Unfortunately, I understood nothing. What my girlfriend failed to mention is that her family speaks Twents, the native tongue of the Dutch region Twente. Though technically a Dutch dialect, Twents is gobbledygook even to native Dutch speakers. It is perhaps gobbledygook even to Twents speakers - interactions in Twents seem more like a mutual dare than a genuine form of communication. Fail to understand random mutter of guttural vowels? You must be a poser from the city.

I remained silent for the meal and, in the care home, explained to my girlfriend that it was a lost cause - I had memorised thousands of words but didn't recognise a single one. She explained that her family spoke Twents, and said I should try again, with her parents, in Dutch. Annoyed at her omission but not wanting to have wasted my time memorising 5,000 useless sentences, the next day I tried again. This time it worked. I switched into Dutch mid sentence, like a happy exorcism, and her father cried in joy. I spoke Dutch!

Now, I don't want to overstate things - I spoke Dutch like an idiot. But I spoke Dutch. I had, for some definition, learned a language. This had never happened to me before. And my excitement about this carried me to memorise >10,000 sentences, which got me fluent enough to stop caring. I still make mistakes all the time (de or het anyone?), but I speak Dutch enough to go on holiday with friends and family, and for us to never have to switch to English even if we could. I am now successfully married to my Dutch wife and have an American kid who speaks Dutch at home, even if I'm a little rusty. I don't know what CEFR that is but, for me, it is enough.

You're probably thinking, "this Alex guy is full of shit. Who would learn a language by rote memorising sentences?". Me. I would do that. It took something like 120 hours over a year to learn 10,500 fill in the blank sentences, and about 60 more to get them into long term memory, less than a minute per sentence overall. I reviewed more than 115,000 flashcards, studying 80% of the days and averaging >200 reviews at a time. This is a lot of time, but 150 hours normally gets you to A1 aka Amoeba in CEFR. Here are the receipts from the flashcard software I used:

The first principles of language learning

You might think that memorising sentences is about the dumbest possible way to learn a language. Why else do we have all these teachers and apps and textbooks? My theory is that learning a language is very simple but very boring. Teachers and apps and textbooks are much less boring, and make you feel like you're learning without you having to do the painful actually learning part. Duolingo doesn't care whether you learn French, they care whether you use Duolingo, and so over time Duolingo has gravitated towards whatever makes you keep your subscription. People don't subscribe to very boring apps.

To understand why my method worked, and how to do it yourself, you need to understand some fundamental truths of language learning. I'll explain them below, from most important to least important.

Vocabulary is King

You can't speak a language fluently without somehow memorising thousands of words. So your initial focus should be on memorising those words. Here's a rough guide:

# of Word Families / Words % Coverage of Text Description / Text Type 3,000 ~95% Informal spoken / scripted dialog (movies, TV, soap operas) 4,000 ~96% Academic spoken English (lectures/seminars) 5,000 ~98% Informal spoken / scripted dialog 8,000 ~98% Academic spoken English

Source and source.

Frequency is Queen

You can get more coverage from more common words. If you learn the 1,000 rarest words in a language, your spoken language coverage will still be close to 0%. It makes sense, then, to learn the words according to a Frequency Lists. Frequency Lists are lists of words from most to least common in a language, and they're easy to find for most languages.

There is some dilemma with Frequency Lists about whether to count lemmas (run, running, ran = 3) or morphemes (run, running = 1). From my own experience, you can just learn the most common lemmas, roughly in order, until around 5k lemmas. At this point you'll start to be able to converse and read things. 10k lemmas was enough for me to feel that my vocabulary was not an issue.

You do not need to start with words that 'feel' basic but aren't common, or avoid words that feel harder but are common. Children learn words in order of conceptual complexity, because they are limited by their conceptual understanding. As an adult, you have all the concepts. So you should learn words in the most logical way. If 'enterprise' is more common than 'rhino', learn enterprise.

NB. This is one of the reasons why learning Chinese/Japanese characters is tough. Kids have to learn them in order of their conceptual complexity, which means they learn the Kanji in a somewhat random order. Systems like Heisig's Remembering the Kanji allow you to learn radical (component) by radical.

Learn vocab in context

You never want to learn vocabulary in isolation, only in sentences. "Miss" might mean "not have" or "fail to catch" or "young lady" or "long for". You need to intuit that you can't say "I missed my keys" and you can say "I missed the bus". These intuitions only come from seeing words in their proper context, in sentences.

Memorise actively

The most important lever you have in memorisation is the Testing Effect. Anything passive, like reading, will be much less useful for memorisation than anything active, like being repeatedly asked "what's dog in Spanish?".

Ultimately, speaking is a test of your active vocabulary, vocabulary you can spontaneously produce. Passive vocabulary, vocabulary you can comprehend, is useful but should be a byproduct of trying to increase your active vocabulary.

I recommend that you use Cloze Test type cards: that way you can learn vocabulary in context while being tested on it. Think "Today, I went to the ___________ and bought some milk and eggs. I knew it was going to rain, but I forgot to take my ________, and ended up getting wet on the way.". I like to have an English translation as the blank, as it removes ambiguity and research is mixed on if staying in your target language entirely is good or bad.

Memorise with Spaced Repetition

Memorising thousands of words is very time consuming. We can make it take 10-30% of the time of conventional methods by using with Spaced Repetition Software (SRS). Basically, SRS will test you over and over, and learn both how good you are at memorising things, and how tough the individual things you're trying to memorise are.

For discrete, large scale memorisation tasks like thousands of vocabulary in context, SRS is kind of miracle. It's hard to overstate how useful SRS is if you want to memorise things. It is the most important innovation in memorisation and learning, and you should be using it.

Amazingly, the best software for this is open source - Anki running the FSRS algorithm. Sadly, Anki has quite a steep learning curve. I had previously muddled through Anki memorising other things. Frustratingly no one has kept what makes Anki great functionally while making SRS more accessible, though most of the big apps started out with this premise.

Learn incrementally

In order to memorise items quickly, we want to learn the least amount of new information possible per sentence. We don't want sprawling sentences with lots of new words. Ideally, we have reasonable length sentences (you can look up the average for your language) with only one word we don't know. This is sometimes called i+1 learning.

When you don't have any vocabulary at all, this is hard to achieve. But it gets easier as you go along. Early sentences might look something like this: The cat sat on the mat. > The dog sat on the mat. > The dog sat on the floor. > The dogs sat on the floor.

At the start this is near impossible. When your i is close to 0, learning i+1 is incredibly tough. Many sentences will contain multiple words you don't understand. Do not be disheartened. Brute force the first ±100-500 words and it will get easier.

Go fast

We should be trying to go fast when testing if we know vocab, even if that means we fail more often. We’re trying to make recalling words automatic, not deliberate. Studies show automaticity is a key measure of fluency and we should be measuring retrieval speed. As per my stats, my average review time for a piece of Dutch vocabulary is <5s. I either know it or I don't.

If something you’re trying to memorise seems confusing, or a sentence isn’t well structured so it takes too long, just change it or delete it. Just go fast.

Never listen to non natives

If you want to learn a native accent, you should only ever hear native speakers. Every sentence you encounter should have audio to go with it and you should try to emulate that when you read it. You want to bake the lilt of the language into your brain.

One reason foreigners sound so foreign is because they practice together. If you are copying mispronunciations, you're destined to mispronounce.

What about grammar?

Don't worry about it. Whether or not you can spontaneously produce correct grammar doesn't depend on whether you can slowly get grammar exercises right in a book. The best way to get grammar is 'feel' grammar through massive amounts of exposure. If you can't figure out how a sentence works by reading through it, look up how it works. But don't practice grammar - your time is going to be much better spent expanding your vocabulary in context, and thus also your exposure to grammar, than by completing grammatical exercises.

What's the catch?

Following these principles is, unavoidably, absurdly boring. And adherence - sticking to the plan - is necessary to make progress. Learning a language is like bodybuilding: you might know how to pick things up and put things down, but the only way to get strong is to do it. That's the catch.

Making this practical

You might have noticed that teachers, textbooks, and apps break the first principles of language learning. There is no perfect solution to this (yet). Learning is hard work, and people don't pay to work hard. The good news is, you are not people. You are here and therefore you are different. I believe in you.

The easy way

Clozemaster is the closest approximation to my method that's available. Sadly it tests multiple choice (passive recall) or typing (slow) rather than Cloze deletion, is automatically generated so can be janky, and doesn't use FSRS. But it's otherwise not bad.

The nerdy way

This is pretty manual, but free and you can customise it to your heart's content. For an example or a starting point, you can download my English to Dutch deck and a Dutch/English to Spanish Anki deck here.

Creating a flashcard deck

We want a flashcard deck which is a frequency list with example sentences in the Cloze Test format, sorted by sentence difficulty.

• Download a sentence list from Tatoeba.org or Open Subs in your target language.

• Create a frequency list based on this, or use one available online.

• In a spreadsheet (example here) or with code:

  • Sort the sentences by the hardest word in them, with some additional variables for sentence length and average word difficulty. Call this the ‘difficulty score’.

  • For each word in the frequency list, find the lowest difficulty score sentence for that word.

  • Make a cloze sentence by blanking out the word.

  • Sort the frequency list by the sentence difficulty.

  • You now have a frequency list with example cloze sentences for each word, sorted roughly by i+1. Occasionally, early on it’ll be i+>1 unless you have a vast number of sentences.

  • If you are missing translations, add them in with something like DeepL or even Sheets translate. It doesn't matter much because you won't rely much on them.

• Set up your Anki cards in the below format.

• Import as a CSV to Anki.

• Generate AI audio with this Anki Add On. AI audio is pretty good now. Much closer to the pronunciation you want than non natives and Twents speakers.

Reviewing

• Review diligently for some amount of time you can sustain almost every day (eg 10 minutes a day). Don't do more, the SRS backlog will punish you for your ambition.

• Ask Chat GPT for explanations and don't be afraid to change / delete sentences you don't like.

• Do reviews quickly, maximum 10 seconds per review.

• Only use "Again" and "Good".

• Again is for when you didn't understand the sentence or couldn't produce the right word quickly.

• Good is for when you understood the sentence and could produce the right word quickly.

• Read the sentence out loud as you review and listen to the native speaker.

• If a card seems confusing or bad, change the sentence or delete it.

• Don't give up.

Card formats

You want the front of your card to look like this, using the Dutch word "regenlaarzen" (a sadly very common word in the Netherlands):

_____ FRONT _____

Ze moeten hun wellies niet vergeten.

_______________

And the back to look like this:

_____ BACK _____

Ze moeten hun regenlaarzen niet vergeten.

regenlaarzen

wellies

A clickable sentence translation

*The audio of the sentence plays.*

_______________

The flip from english to target language for the target vocabulary can be done automatically. You can download my English to Dutch deck and a Dutch/English to Spanish Anki deck here. You can copy the card formatting and set up to get the automatic language switching.

FAQ

Doesn't [Insert App / Teacher / Thing] do it better?

Potentially. I'd love to hear about it if they do! If you can find a real flesh and blood adult who actually learned to speak a language in a reasonable time through using an app / teacher / thing, tell me.

Isn't memorising lots of words boring?

Reeeeallly boring, yep. It's hard to keep up.

Isn't it a problem that it's boring?

Definitely. In the end If there were a way to make it not boring, that would be a very successful tool.

Won't AI make learning languages irrelevant?

Quite probably.

Won’t AI make learning languages easier than rote memorisation?

I hope so! I did start learning Dutch in 2020, when AI didn’t speak Dutch either. I may fork Anki at some point to try to make it more AI and accessible, but I’m sure and hope other people are working on this.

Fitness professionals are lying to you

In the 20 years since I first tippy toed to impress girls, I became the world’s least fit OCD fitness professional, all in pursuit of big and strong (but not fat). I’ve competed in Thai Boxing fights, kicking people in the head until I broke my knee, and Brazilian Jiu Jitsu tournaments, securing a podium finish by way of lack of competition. I’ve paid a personal trainer to help me in powerlifting. To be fair, they dragged me through a 175kg / 385lb deadlift and a 150kg / 330lb squat. Unfortunately this required engorging my naturally ghostly 65kg / 145lb frame until I became the embodied stench of dairy, somehow weighing almost 90kg / 200lb. I’ve tried every plausible diet - from calorie tracking to keto to vegan, and some implausible ones too - the GOMAD (Gallon Of Milk A Day) being my favourite and my wife’s least favourite. I’ve worn five finger shoes on purpose, frolicked through GMB animal movements, and learned the underhand sneak in rope flow. I’ve bulked and cut and bulked and cut. I’ve tried all the supplements, optimising blood tests for over a decade, hoping to find the secret to big and strong (but not fat).

My most dedicated effort was in 2018. Ahead of a grappling tournament, under the supervision of paid fitness and dietary professionals and with meticulous adherence - every calorie counted and every workout supervised - I did a bulk and a cut. This means you get as big and strong as you can, putting on muscle and fat, then lose weight and keep the muscle but not the fat. The end result is big and strong (but not fat). I did DEXA scans, the most accurate measure of body composition, near the start, middle, and end of this process. Big and strong (but not fat) incoming.

Here’s what happened:

A year of work to gain some visceral fat?

Spot the difference before and after.

What the fuck? I was identical practically to the gram at the beginning and the end of the process. To boot, I moved some fat from my limbs to my organs. That’s not a good thing!

It turns out, nothing works. My body knows what it wants to be, and that’s not big and strong (but not fat). I have the physique of computer code, looking like a 1 or a 0. All those jacked guys are on steroids and or won some genetic lottery I didn’t.

Shook by this realisation, during COVID I gave up, isolating in Amsterdam with nothing but the Netherlands finest cheese, bread, wife, and peanut butter to comfort me. I reached my heaviest ever, my only exercise being ‘mobility work’ in the park whenever I dared to go outside and respire.

Being fat is not without downsides. In the short term, I developed a conspicuous bald spot where my belly shaved itself using my waistband. In the long term, it turns out that getting as fat as I did makes you fatter forever - if you gain enough fat, your fat cells run out of room and multiply more or less permanently. Something needed to change.

Bryan Johnson is (probably) also lying to you

In 2021, COVID free, I decided I’d get to a healthy weight by eating less and then just stay there. No more losing weight, no more gaining it. I gave up on big and strong (but not fat), and just wanted to sustain a healthy weight.

My weight from 2015 to 2025, stable since 2021.

As I was losing the unhealthy weight, I discovered Bryan Johnson's Blueprint and, by 2022, I was experimenting on myself. Bryan claimed extraordinary results, turning back his biological age - whatever that means, and the hope of living forever. Perhaps Blueprint would be the panacea I had been searching for.

It was not. Blueprint had lots of ideas, but just as many problems. At first, Blueprint’s issues felt benign: the recipes were disgusting and a pain to make. Following even half Bryan's ‘protocols’ was a full time job. Worse still there were no explanations, let alone citations, so who knows what anything was supposed to do and why. When I looked into things, I found that potential health benefits were often speculative, proven only in mice, or there was no long term safety data in humans.

Bryan tried to simplify and commercialise, releasing his own Blueprint brand of products loosely based on a subset of his protocols but… should you trust Bryan? On this point there is a un bouquet de fleurs. Some of his actions were more marketing and aesthetics than health, like getting fat injections in his face. There was some sort of paid opt in ‘trial’ which allegedly got swept under the rug after results didn’t go well. Bryan was on testosterone (aka steroids) at various points. The biomarkers and blood tests were from unspecified labs and potentially mixed and matched from different dates. Rejuvenation Olympics leaderboards appeared to shuffle at convenient times. The meaning of these ‘biological age’ tests is disputed. When Bryan released his own supplements and foods, there were quality control issues. Blueprint employees allegedly signed weird legal agreements. Bryan has some claims against his character.

I am not criticising Bryan. He has done many amazing things and I hope - thought doubt - he will turn out to be completely legit. But, a Blueprint customer I am not.

So what is good about Blueprint? For me, Bryan and his Blueprint was new inspiration, new hope, and a new starting point.

From 2022-2024, my approach was to try everything indiscriminately. I tried to copy Bryan in various ways. I’d take blood tests and dozens of new supplements. At one point I was gulping down something like fifty pills a day, trying to optimise biomarkers. I’d eat some variation of the Blueprint ‘Super Veggie’ and ‘Nutty Pudding’ most days. I’d go to the gym and do upside down shin curls, hold some stomach vacuums while brushing my teeth, use laser beams on my hair, hop in an ice bath I made out of a freezer, and so on. I tried everything and more. It was very hard work.

Below are the measurements from my impedance scales over these three years. The absolute kilogram numbers from these scales are generally not very precise, but the relative numbers give us some idea what happened.

I had tried every possible supplement, every workout routine, every quack therapy I could think of and afford. Once more, nothing changed, nothing worked.

It's not a me thing. I looked at my now dad school friends and realised over 20 years the fitness industry has failed us all. The skinny guys are still skinny, the fat guys still fat.

Just give up and procreate

In December, I had a child. My wonderful son detonated every aspect of my life. I no longer had time to be indiscriminate. If I wanted to be healthy, I needed to get myself together. I needed a minimalist plan. If I didn’t, I was going to lose every aspect of my wellbeing.

So I reflected on my fitness journey, cut out the bullshit, and formed a concrete plan I could trust and stick to. I called it 'Dadprint'.

I wanted Dadprint to be:

• explainable. I want to know why I’m doing what I’m doing.

• flexible. I want to be able to do things I love, not feel constantly restricted.

• safe. I don’t want to consume or do things that might hurt me long term.

• sustainable. It has to be easy and affordable enough to maintain as a lifestyle.

I planned out Dadprint, stuck to it since my son was born, and… it worked. Not worked like ‘I didn’t lose every aspect of my wellbeing.’ Worked like ‘I transformed my body like never before.’

Here’s what happened according to a DEXA, with my professionally monitored bulk and cut - my previous physical peak - as a reference:

TLDR: I dropped to almost 10% body fat, about as low as is sustainable year round.

My body fat plunged to 10% - about as low as one can sustain year round without bordering on unhealthy. I lost a bunch of visceral fat. And, comparing to other DEXA users (much fitter than the general population), I became exceptionally lean and disproportionately muscular:

That reads: less fat per kg / m2 (mass controlling for height) than 95% of DEXA using males my age; less visceral fat 86% of DEXA using males my age; while having more muscle than 56% of DEXA using males my age; and more muscle in my arms and legs than 65% of DEXA using males my age. I would guess fat and muscle mass correlate. But, basically, I am very lean and somewhat muscular.

I changed aesthetically, too. Here’s a side by side, first thing in the morning in every photo, from 2017 around my current weight (before I put a lot of weight on), 2018 (after my professional bulk and cut), and then now (after six months of fatherhood):

While the X kilo numbers aren’t precise, directionally they are useful.

I wish I’d collected more data, photos, and blood work. I just didn’t expect this to happen… What are the odds I would be in my best ever shape six months after I became a father? Not high. And yet.

My impedance scales agree on the timeline. Though these aren't precise in absolute kg terms, they are typically directionally correct and they suggest the birth of my son in December was indeed the turning point.

A gym bro and a scientist might, for different reasons, scoff at these results. Being leaner doesn’t mean I am healthier, at least not by definition. You might say I just lost body weight and body composition is related to that - though I think the before and after photos disagree. You might say that, compared to a ‘good’ cut, I lost much too much muscle compared to how much fat I lost. But, my baseline is nothing changes my physique at all. I have never had a body like this in my life even though I spent vast amounts of money, effort, and time on it.

It’s hard to say what caused what. Dadprint is pretty much a watered down Blueprint. I didn’t add anything new, just took things away. Why would that work better than the hardcore version? Less, tougher exercise? Fewer, better meals? No speculative supplements or laser beams? Hard to say. I am N of 1 and, as with anything biological, I can’t be sure what worked for me will work for you. But something did, finally, work for me. I share Dadprint with you hoping you can take what suits you and make N of 2.

My Minimum Viable Dadprint

The Most Important Thing

The dominant variable in any health or fitness plan is time: the longer you do something for in years, the more impressive your results will be. This means whatever you choose to do should be something that’s sustainable for you. I have tried to cut out as much effort as possible in a way that suits me. This leaves some gains on the table, but is what is sustainable for me. You should take Dadprint as a set of ideas rather than a plan to follow precisely.

The basics

For a precise, step by step guide with explanations, skip this section. Here I share the absolute fundamentals for those of you who don’t want the details. It’s also what I aim for when I’m on holiday.

• Sleep:

  • Sleep 8 hours in a 24 hour period, any way you can.

• Eat:

  • Pick a healthy target weight range. This should be between your Devine formula weight and your weight assuming a normalised FFMI of 22 and a body fat of 10%, ± 1.5kg / 3.5lb.

  • Weigh yourself after using the loo in the morning. If you're above your target weight by 1.5kg, eat 2 normal sized meals each day. If you're below by 1.5kg, eat a third or fourth meal. Eat a third meal on tough exercise days.

  • Your two base meals should primarily combine red and blue berries, nuts and seeds, cruciferous vegetables (think broccoli and cauliflower), mushrooms, something fermented, and olive oil, plus a big source of protein. For an every day protein source, prioritise vegetarian protein sources (ideally tofu), over fish (ideally salmon), over lean meat, over fatty meat.

  • Your third meal can be whatever, within reason. Reason is self regulating - if you need to lose weight and you aren’t, your third meal should reduce in either frequency or quantity.

• Exercise:

  • Do strength training with compound movements twice a week. Do four exercises total, two exercises per session. For example: Session A is some form of squat and deadlift; Session B is some form of push and pull. For each exercise do 3 sets of 5-10 reps, with a 2 minute break in-between, and add a tiny bit of weight every time you go. This is ~1h of strength testing so you could even squash it into one session a week if you are a savage.

  • Do Tabata style HIIT cardio after strength training. That’s 20 seconds max effort / 10 seconds low effort for 4 minutes on something like an air bike or sprints.

  • Do lower intensity cardio, eg your favourite cardio oriented sport like tennis, once or twice a week.

• Supplement:

  • Vitamin D 2000iu daily.

  • Omega 3s combined 1500-3000mg daily, depending on whether you regularly eat fish.

My Full Dadprint

Here we are. The nitty gritty. This is, down to the tupperware, as much detail as I can write about what I’ve been doing and why. If something needs an explanation, let me know and I’ll add it.

It should go without saying that it’s impossible to stick to every bit of the plan all the time. That’s fine. The goal is to build good habits, and then over time it’ll get easier.

Sleep

Sleep quality and quantity come before anything else. Getting these right will make you physically and mentally better off on practically every dimension. When you don’t get good sleep, your mental health will suffer, it will be harder to adhere to a diet, and your exercise performance and recovery will suffer.

Our main goal is to get at least 8 hours sleep in a 24 hour period. A tiny fraction of the population can get away with less, and you’re probably not one of them.

Some other things that might make you sleep better, in rough order of importance to me.

• Avoid alcohol.

• Your bedroom should be quiet. Create a quiet environment and / or use white noise to make disturbances minimal.

• Your bedroom should be dark. Use black out blinds or curtains, or an eye mask.

• Your bedroom should be cold enough that you need a proper duvet, probably under 19c / 66f.

• Go to bed and wake up at roughly the same time.

• Avoid high caffeine drinks like coffee in the 8 hours before bed.

• Avoid drinking lots after your last meal, if you’re not, otherwise you’ll be peeing a lot.

• Try to have a wind down period before sleep of, say, 30 minutes without major stimulation. I like to do some stretches while watching Nomad Push.

Of course, it won’t always be possible to sleep well. With a newborn it wasn’t for me a lot of the time. If you have the ability to change your situation (my wife and I slept in shifts for months), or somehow squeeze a nap in by trading off something else, it’s probably worth it.

Food and weight principles

We want a baseline diet which is:

• easy to prepare and transport,

• tastes good,

• is in aggregate nutritionally complete,

• supports long term health and fitness,

• and has flexibility baked (!) in.

But before we get to my recipes, I want to give some principles.

Picking a target weight

Our diet is the key way we can manage our weight. Relative to our diet, exercise will have little effect. Massive fluctuations in weight can be harmful long term, and so we want to pick a healthy and desirable weight range and stay there.

Unfortunately, people have gotten much heavier in recent decades, even controlling for height. Widespread obesity in the general population and steroids in the influencer population have contributed to unhealthy expectations regarding bodyweight and muscularity. Picking a healthy range by intuition is therefore hard. You should probably be lighter than you think.

To anchor our bodyweight expectations, try the Devine bodyweight formula, which roughly predicts your weight based on your height, assuming you are 20-29 and entering college in the US in 1950. It is therefore a reasonable anchor of an average pre obesity epidemic weight. It’s likely much lower than you would guess your weight should be by modern intuition. You may think that if you’re older than 29, as I am, you get a free pass on being heavier. Unfortunately, there’s good evidence that metabolic rates don’t change much at least until around 60, and weight inflation throughout mid life is largely lifestyle choices.

The second anchor we have is normalised FFMI of 22 at 10% body fat. This is going to be heavier than the Devine formula, and represents your target weight if you were in peak shape. You can calculate your target weight at 22 normalised FFMI and 10% body fat here. Why these numbers? An FFMI of 22 is about as muscular as a non professional athlete with average genetics can expect to build in reasonable time, and 10% body fat is about as lean as is healthy to be year round. You may find you can reach a normalised FFMI of 23 or higher, in which case congratulations. But the point here is to anchor you to something realistic, not what you see on socials.

With those high and low anchors in place, pick a target weight. This weight ±1.5kg / 3.5lb is your weight range for the foreseeable future. If you’re lucky enough to be genetically gifted, push up the range when you find you can maintain the high anchor while being close to 10% body fat.

For body fat measurements, if you want to be precise, once a year or so you will want to do a DEXA. Impedance scales just aren’t that accurate, especially as your body fat percentages get low. Otherwise visual guides are fine, and over time - especially if benchmarked against a DEXA - you’ll find what it looks like for your fat distribution. For me, when I am close to 10% veins start to appear all over my body.

Regulating our weight

The system I will now describe evolved from just trying to maintain a constant bodyweight. It is quite trivial once you've set it up.

Every morning, after using the loo, weigh yourself. This is important information in deciding how you will regulate your weight. We will not emphasise regularly counting calories or restricting ourselves to specific foods.

I suggest that you craft two base meals which are your baseline diet every day. You can use mine, recipes below. Together they should be nutritionally complete - hitting your RDAs on most things. They should put you in a (500ish calorie) caloric deficit if you are overweight. If you are in your target weight range for long enough this will become caloric restriction. Caloric restriction implies you're not losing weight, even though you're eating less than you would expect to maintain your weight, because your body is adjusting various processes to maintain the weight it ‘wants’ to be. This is only possible if the caloric restriction is small enough that the body can compensate. Caloric restriction seems to increase lifespan in animals, and it doesn't seem to be bad for humans. You can estimate your calorie needs here.

The algorithm is this: we eat our two base meals when we are above our target weight range; and we eat more than our two base meals when we are below it or doing intense exercise that day. In aggregate this may offset some of our baseline caloric restriction, though hopefully not entirely. If you have massive weight fluctuations due to some quirky routine (eg doing long distance running and refeeding after), it can be helpful to use your average weight from last week - smart scales will make this information easily to find. If we are not moving towards our target weight range then we reduce or increase the size or frequency of our extra meals accordingly.

We don’t care too much what our extra meals are made of - it’s important to be able to eat food we like and not stress out about it. Over time I’ve found that I more often want to eat healthier extra meals because my body is used to it. I also want to exercise more to justify cravings. If you must snack beyond your meals, try fruit or homemade popcorn.

Of course, some days things don’t go as planned, or you will want to eat an extra meal even if you shouldn’t according to the algorithm. No worries! Tomorrow the algorithm will compensate for today.

Over time our body learns what weight and body composition it ‘wants’ to be. It reduces or increases our energy use to maintain this weight. Big weight fluctuations throw off this natural regulation. This is why picking a weight we can and want to sustain, getting there, and then slowly tweaking is a better long term strategy. Once you have stablised for a year or so, you could try fluctuating between the bottom and the top of your target weight range as a form of bulking and cutting, but you’ll find it’s remarkably hard to shift your weight around once your body has gotten used to where it ‘wants’ to be.

The point of this system is threefold: we get great baseline nutrition; we regulate our weight; and we get to eat other foods we love without guilt.

Protein

There is a lot of nonsense about how much protein you need to eat. This is a good analysis. The gist is, 1.6g / kg or 0.725g / lb of bodyweight per day seems to be roughly the point at which eating more protein makes no difference. To be clear, that’s not the ‘minimum you should eat’, that’s the maximum you _can_ eat before eating more has no effect.

There are good reasons not to eat more protein than you need to. Protein sources tend to have their own issues: meat has epidemiologically associated with all sorts of health issues and dirty and unethical supply chains; fish have contaminants and heavy metals risks and dirty and unethical supply chains; and isolated vegetable sources can be comparably high in heavy metals. A diverse source of ‘enough’ clean protein is where you would like to be. I think tofu and single ingredient whey are your cleanest daily sources.

Food and supplement sources and quality

Generally we want to compose our base meals with foods related to those eaten by our distant ancestors. This is a kind of historical safety heuristic which prevents us from eating too much food that our bodies aren’t evolved to handle. It also implies we should eat organic and avoid unusually processed foods and supplements when there is a reasonable dietary alternative.

If we’re going to consume something every day, we need to source the best ingredients possible, and be paranoid about heavy metals and contaminants. This can be tricky, remember Bryan Johnson messed it up. In the US, there is little regulation on metals in supplements and food - it's not unusual for products like chocolate or protein to test high for lead or arsenic, we just don't often test. Ideally high risk products we consume regularly are third party tested by a certified lab for ingredient purity, heavy metals, and other contaminants. Where possible I have checked this. It is painful to check CoAs for every batch of every item, so I encourage you to check yourself and wish it were easier to do. If you find verifiable cleaner sources than me, let me know!

I recognise that a bunch of the ingredients in the shot / smoothie are powders. This isn’t ideal, but I think freeze drying fruit, which is how powders are often made, seems fairly innocuous. In the spreadsheet of ingredients I highlight which ones are ‘high risk’ for metals or contaminants or toxicity, based on my research. I also give a simpler recipe for the smoothie based on what I deem to be only the important ingredients.

Pill Form Supplements

In addition to the quality risks outlined earlier, pill form supplements generally have very weak evidence. There is usually no evidence that supplements have an effect in humans, admittedly because these things are hard to study, and good reason to think that things taken in pill form may in fact act differently in the body than when they are ‘packaged’ in food. For most supplements we have no good evidence about the right dose in humans, or the right time frame for which to take them. Bryan Johnson takes, from what I can tell, everything that _might_ work whether it is demonstrated to be safe or effective in humans. Bryan took Rapmyacin for example, a popular ‘off label’ longevity drug, and now thinks that this made him older rather than younger. Spending lots of money to take probably useless, maybe beneficial, and potentially harmful supplements is not the right risk reward trade off for me.

Because of this risk reward, quality control issues, and the general depressing experience of taking lots of pills, I want to minimise the amount of pill form supplements I take.

From my calculations, the baseline diet I have is good enough to get my RDA of most essential vitamins and minerals. So we don’t need to supplement much.

I do take, however, some supplements at doses shown to be effective and safe. These are:

• Vitamin D3 - 2000iu. Almost everyone is deficient, but it is possible to take too much. This is a good low maintenance dose.

• EPA / DHA - 1500mg to 3000mg combined, dependent on whether you eat oily fish a lot.

There are some well researched and safe extras you can try if you want to get exotic:

• Kyolic Garlic - 1200mg. Good for you but hard to find in your diet without being stinky.

• Ubiquinol (CoQ10) - 100mg. Good data that this is safe long term and helps with cardiovascular risk.

• Blood help. If you give blood regularly or are vegan, something like this might be a good idea.

Recipes

As above, I eat these 2.5 meals a day as a baseline. The start to finish cooking time per day is <15 minutes with no microwave.

The 2.5 meals are:

• Morning shot

• Smoothie

• Veg Bowl

Here is a spreadsheet with all the ingredients listed for all recipes. It includes links to the brand I bought last, basic nutritional info, the ‘health logic’ behind each item in the recipe, and you can scale the recipe by number of portions (the yellow F1 cell) to see how much to buy in total.

Bonus: if you follow these recipes, you will poop like a champ.

How to make preparation easy

Easy preparation is really important to making a good diet sustainable. If it takes you 1 minute to make a smoothie, that’s easier than getting take out, and you will consume a lot of smoothies. Therefore, while it sounds ridiculous, it’s worth getting your set up right so you can ‘set and forget’ your diet.

In all my recipes, therefore, you should adjust the number of portions you buy and make at once, and the gram amounts of each item in the recipe, so that it’s easy to buy and dump into storage containers without extra work.

For example, let’s say the recipe says 6g of husk per serving, and 10 grams of cacao, and your particular brand of husk comes in a 300g pack and your source of cacao in a 600g pack. You could make 50 servings at 6g of husk and 12g of cacao, or 60 servings at 5g of husk and 10g of cacao. This kind of variation is not going to make or break your health or fitness. The only time it would make a difference is if you’re significantly increasing the amounts of a high risk item, eg Brazil nuts (selenium risk) or cacao (metals risk).

Get some relevantly sized containers for each recipe (details below). Then, when you are buying ingredients you can scale the recipe to, say, 45 portions and just dump everything you buy in the relevant container. You also know roughly the gram amount per portion of, say, all the nuts in the smoothie. So you can then just stick the right size measuring spoon for the nuts in the container. This way it takes almost no time to mix the ingredients in advance - you just dump them into the container, and no time to prepare the meal on the day - you just take one scoop out.

While it sounds slightly absurd to go to this level of detail, making your baseline diet easy and time efficient is the best way to make sure you actually eat it.

Morning shot

This is essentially stuff that’s good for you but tastes gross so we don’t want to put it anywhere we have to chew on it. Amazingly, it somehow works as a firebrand ‘ginger shot’ type experience. Best to take it first thing in the morning since it has caffeine in it and some of the ingredients work best on an empty stomach.

If you are conservative, skip or change this without reservation. Chlorella is probably the most speculative ingredient I consume, primarily out of hope that it chelates metals, but that is very speculative.

Here’s a full recipe:

• Chlorella 2g

• Spirulina 2g

• Matcha 1g

• Beetroot 2g

• Moringa 1g

• Apple cider vinegar 15ml

Meal prep

Dump all the powders in a container, find the right spoon for the serving weight. You can prep months worth of servings in one go.

On the day

30 seconds to make: A glug of apple cider vinegar into a glass, a scoop of the powders, and top up with a tiny bit of water to taste. Whizz with one of those milk frothers. Down it.

Smoothie

This is basically an antioxidant, fibre, and protein bomb. Some of the fruits I stick in there are a little exotic, because I’ve having the same thing almost every day for years, but the foundation is very simple:

• Blue and red coloured berries. Antioxidants and polyphenols.

• Nuts. Fiber and good fats.

• Protein. Make sure your protein sources are as high quality and clean as possible!

• Creatine. This is one of the few supplements that you should almost certainly take. There is good evidence will make you better at a long list of things.

Here’s a full recipe:

• Collagen 12.5g

• Pomegranate powder 10g

• Cacao 7.5g

• Ginger powder 5.0g

• Açai powder 5.0g

• Psyllium husk 5.0

• Resistant starch (green banana powder) 5.0g

• Inulin 5.0g

• Creatine 5.0g

• Aronia powder 2.5g

• Maqui powder 2.5g

• Amla powder 2.5g

• Sunflower Lecithin 2.5g

• Ceylon Cinnamon 1.3g

• Macadamia nuts 30g

• Chia seeds 25g

• Hemp seeds 10g

• Walnuts 5g

• Flax seeds 2.5g

• Brazil nuts 1g

• Mixed berries 150g

• Protein powder (whey) 20g

• Vanilla bean powder 0.3g

• Water 150-200ml

Meal prep

Get four containers: protein; other powders; nuts; and frozen berries. Dump all the relevant ingredients in each container, find the right spoon for each serving weight per category. You can prep months worth of servings in one go.

On the day

1 minute to make: Scoop a spoon from each container into a [blender](https://tornadoblender.com/) with approximately 150-200ml or 1/3 pint of water. Blend for 30 seconds. Done!

This smoothie gets pretty gloopy if you leave it for a while so either make sure your container is airtight or make it and eat it.

Veggie Bowl

This is your main meal. You’ve got to find the flavours that work for you, but this is verified tasty if you’re into Asian food. In case you’re not, here’re the important parts:

• Cruciferous vegetables. Antioxidants and fibre again.

• Mushrooms. Good for your gut, immune system, inflammation.

• Protein source. Good for your muscles.

• Something fermented. Good for your gut.

• Olive oil. Good fats, anti inflammatory.

• Bonus ingredient: you can at a pulse or bean like lentils or chickpeas. I find it to be a faff to cook these from scratch, so I often don’t bother, but they make the dish more delicious.

Here’s a full recipe:

• Tofu 284g

• Broccoli 250g

• Cauliflower 150g

• Cooked black lentils / chickpeas (if you can be bothered) 150g

• Kimchi 60g

• Shiitake / Maiitake mushrooms 50g

• Evoo 45g

• Dashi Miso 40g

• Onions and garlic flakes 20g

• Nori (2 sheets) 5g

• Cayenne powder 2g

Meal prep

Buy however many containers will fit in your freezer. Buy however much veg you can fit in your containers. Chop them all up and put them in containers. I keep my protein source separate, but you can keep it together if you buy bigger containers. My freezer only fits about a week’s worth sadly.

On the day

10 minutes to make. Empty a container into a steamer. I go with 9 minutes in the steamer since I have tofu. Make the ‘sauce’ in a bowl / tupperware (olive oil, kimchi, miso cayenne etc). Put everything in the bowl / tupperware.

This tastes good (enough) cold, so often I’ll steam it in the morning or evening and then take it to work.

Third meal

Whatever you like!

Exercise

Exercise is your biggest active lever in your long term after sleep and diet. Strength training improves your muscle mass and bone density. Cardio training improves your cardio vascular endurance and your VO2 max. Exercise makes you happier. All these strongly correlate with a higher quality, longer life.

If you’re reading this, I assume I needn’t convince you to do exercise. But don’t put it off! It’s important to start early, too it’s much harder to get strong and fit later in life, especially if you’ve never been strong or fit before, because of factors like ‘muscle memory’.

So what exercises matter? We want to pick exercises which:

• have low injury risk and are relatively easy to execute, because getting injured is much more costly than missing out on marginal muscle gains.

• train lots of things at once, or are very efficient, and in combination cover our entire body, because we want to get the best returns on our time.

• can be scaled to our desired difficulty via progressive overload, because this is how we get stronger and fitter.

If you’re new to strength training, I recommend getting a good coach for at least one session, and ideally one session every few months, until you’re comfortable with any new movements. Alternatively you can use r/formcheck.

Key Strength exercises

I landed on four core strength exercises. Between these four exercises, we can develop a generally strong body in under an hour per week of strength training.

Romanian Deadlifts (RDL)

Regular Deadlifts are amazing, but hard to execute technically and have scary associated injuries. I spent my entire life doing them, and never really got my form to a place that felt good. If you can, great. But, hear me out, Romanian Deadlifts are simpler with lots of the same benefits. There are no mobility constraints and the cus are easy: soft knees, push your butt backwards until you can’t anymore, keeping your back straight, then come back up. As you reach your grip limits, it’s fine to use straps.

Bulgarian Split Squats (BSS)

Back Squats and Front Squats suffer the same challenges as Regular Deadlifts. They’re fantastic, but very few people will get them right even with a lot of help and no-one likes the thought of failing a rep under a bar. BSS, on the other hand, are hard to get wrong: drop your back knee almost to the ground and come back up. If you fail a rep, no problem. You can load a BSS with dumbbells and weighted vests. And if you doubt their muscle stimulus, you’ve never tried them.

Pull Ups and Dips

I put these two together because they are two sides of the same coin. Between them they work out more or less your entire upper body in just two movements, in wonderful proportions. They reward hardcore form - for maximum stretch induced gains start your pull ups in a dead hang and end at least chest to bar, and take your dips as deep as you can. They have the added benefit of scaling up and down nicely - a simple dip belt can make them infinitely hard. And they can be done on park equipment or even with cheap gymnastic rings at home.

Accessory exercises

I like to do a single set of Seated Good Mornings, for mobility and back robustness, or some form of hanging L sit / leg raise, for nice abs, at the end of my workouts. I have hamstring mobility issues, and I am vain. YMMV.

Cardiovascular training

We have two cardiovascular objectives, VO2 max and cardiovascular endurance. I have always hated cardio. The good news is, for VO2 max we can train it very fast, and for cardiovascular endurance we can substitute something we find fun.

VO2 max - Tabata

Even short sessions of max effort cardio can have dramatic effects on your cardio vascular system. Tabata, where you alternate 20 seconds of max effort with 10 seconds of rest for 4 total minutes, will kick your VO2 max into a new dimension. I like doing it on an air bike, and the 20/10 4 minute program is built in, but any kind of sprint works.

It’s safest to do Tabata at the end of your strength session, but you might experiment with doing it up front to save time if you are well trained. If you want to invest more time in your VO2 max, you can try the Norwegian Protocol, but I (happily) don’t want to.

Cardiovascular endurance - Sport

Cardiovascular endurance is generally trained in heart rate Zone 2, Zone 3 at a push. This is very light effort. If you play some other sport once or twice a week, that should cover your cardio vascular endurance. Best to just pick whatever you enjoy and can stick to, for me it’s been bjj.

Programming

Intensity, sets, reps, and rest

If we’re strength training twice a week, we can either do two long hard sessions, or two short sessions. The marginal gain from training a muscle group twice a week might be around 1.5x vs training once a week. It does also seem like more training volume is basically better, but again, the gains become fairly marginal fairly quickly.

For me, I don’t think the fatigue of training muscles twice per week, or with high volume, is worth the injury risk or time. I have a tendency to get overuse injuries like golfer’s elbow, and become physically run down faster than most. Increasing my time in the gym also makes me less likely to go because it’s less convenient. If I do have extra time going for exercise, I would rather enjoy an extra session of sport.

And in the end, slow or fast you will get to a similar place if you can be consistent over a long period. Whether you are adding a little weight every week, or a little weight twice a week, you’ll push your limits soon enough. I think it’s better to optimise for going consistently without getting injured or sick. Of course, if you recover fast and are well trained, ignore this.

As for intensity within a session, if you are new to training you may want to just stick to a basic program of 3 sets of 5-8 reps per key exercise. Your main mechanism for growth is progressive overload - if you can successfully complete your sets and reps for a given exercise with good form, next session just add the smallest amount of weight possible. Eventually you will be training mainly within 5 ‘reps in reserve’ which are the reps that drive growth.

As we are going for a small number of exercises at high intensity, we rest for a full 2 minutes in between sets.

Maximum intensity

If you are well trained, here is some more controversial advice which I follow. If we are working each muscle group roughly once per week, we can afford to train very intensely. We have time to recover. RDLs and BSS are very intense and regular progressive overload works well. Going to failure or similar encourages poor form which, especially in RDLs, we should avoid. For Pull Ups and Dips, which have a lower intensity and injury risk once you are comfortable with them, I recommend trying a high intensity approach.

High intensity Pull Ups and Dips is three sets: AMRAP; weighted; then negatives. This approach tests your muscular endurance, power generation, and strength at every point of the movement.

Our first set is As Many Reps As Possible (AMRAP). This is what it sounds like. Only after your first session, you now have a Personal Best (PB). In future sessions you will do your AMRAP set and, if you fall short of your PB, take 5 breaths and go again until your total reps matches your PB. These are bastardised Myo Reps. Our second set is 5 weighted reps, as explosive on the upwards (concentric) part of the rep as possible. We want to launch ourselves and put our hands into our pockets. Our third set is 5 negative reps. Here we don’t care how we get to the top of the rep, we just try to go as slowly on the down (eccentric) part of the rep as possible. Do these reps weighted if you can resist for long periods. It can help to count your way down from a high number, or consciously try to get back up every time your upper arms hit a number on the clock.

Performance tracking and progressive overload

As our program is very simple, you only need to track your best (should be your most recent) weight for each strength exercise. Four numbers! For cardio, we just track our Tabata max calories on the bike.

Every time we go we aim to beat our previous numbers slightly. If you can sustain this over a period of months and years, you will approach the limits of your strength and cardiovascular fitness.

My Exact Program

Workout A - vo2 max and power strength

• Tabata 20/10 x 8, ±5 min

• Sciatic glide / Warmup RDL, ± 5 min

• Romanian Deadlifts, 3 x 8, 120s rest. ±10 min

• Bulgarian Split Squats, 3 x 8, 60s rest between sides, ±15min

• Seated good mornings / Back hyperextensions 1 set, ±1 min.

Total ±40m

Workout B - vo2 max and aesthetic strength

• Tabata 20/10 x 8. ±5 minutes.

• Pull Ups AMRAP / myo into PB, 5 reps weighted explosive, 5 weighted negatives, 120s rest, ±10m

• Deep Dips AMRAP / myo into PB, 5 weighted explosive, 5 weighted negatives, 120s rest, ±10m.

• Toes to bar / L sit scapula retraction hang 1 set 5 rep negatives or 1 minute.

Total ±30m.

Other things

Tests

DEXAs

If you get a DEXA scan, try to go to the same place next time. There can be variation caused by different equipment so you want to keep things as consistent as possible.

Blood markers

I regularly - one a year or two - get a blood test to test a variety of my biomarkers. I have had at various points very high cholesterol. I have always had high SHBG and therefore sometimes low free T. If anyone knows root causes of high SHBG in particular, let me know!

Three things on blood tests.

• Try to get your tests done at the same lab, since results and equipment can vary lab to lab.

• Try to understand what the ‘normal’ ranges represent if you get any abnormal results, since being outside normal can be anything from a potential measurement error (what time of day did you get your T tested?), to a ‘the green on the chart just represents one standard deviation of people rather than something that correlates with wellbeing’, to a real problem.

• Try not to optimise within normal ranges. We don't understand enough about biology to be certain about the variation of even major biomarkers (does diet influence cholesterol? Is high normal T better than low normal?), let alone unpack the effects of taking multiple poorly studied supplements and how they might interact.

Interventions

Water fasts

A 72h water fast (with electrolytes) seems to be enough to get most of the speculated health (eg autophagy) benefits from fasting. There are obviously risks so it's a good idea to be careful. I do this once every six months.

NB: blood donations and water fasts take it out of you, so I would spread them apart from each other (eg if you do a blood donation in January, do a water fast in April).

Blood donations

There’s evidence that donating blood is one of the few ways to actively reduce PFAS. I do this once every six months. I would recommend you supplement as explained earlier to recover from this.

NB: blood donations and water fasts take it out of you, so I would spread them apart from each other (eg if you do a blood donation in January, do a water fast in April).

Cold plunge

Maybe the only benefit to a cold plunge is that it feels great. That's still a good thing! You can make one cheaply for your home out of a freezer. Cold showers don't feel great, so don’t bother with those.

Recommended for off days.

Barefoot shoes

Not a recent thing and won't give you abs but, while I have your attention, get barefoot shoes. It's so obvious once you've worn them for a while that regular shoes are hurting you. My favour are the Tyr / Squat University shoes.

Drugs

Finasteride and hair loss

I used to be big on finasteride, and I took it for a long time. Unfortunately, I recommended it to a friend and they had some life altering side effects. If you look online you will find a lot of anecdata about this. I think new evidence is emerging which suggests mechanisms for how DHT inhibitors can affect the body long term, and it is intuitive that nuking a very powerful hormone would have some side effects. This may be true in an extreme way for a subset of the population, but you have no way of knowing if you are in that subset. I suspect there are disincentives to discover what is really going on here and strongly discourage anyone from taking it. This experience was also formative of my reluctance to take supplements in general as often safety data is both limited and biased.

Alcohol and smoking

For a long time I drank and smoked in varying quantities, sometimes too much. I can’t profess to know how to stop for everyone. But, what made me quit both almost entirely was getting hobbies that are way harder if you drink and smoke. Competitive BJJ and chess did it for me, although ironically BJJ has historically had a weed habit and chess an alcohol habit. What’s the point in entering a chess competition if I’m hungover and can’t think straight? Or training BJJ if I can’t breathe?

Sparkling water

If you’re addicted to soft drinks then sparkling water, and to a lesser degree kombucha, is a great hack. I used to have a fridge specifically for red bulls and Coke Zero. Get into Topo Chico, friends.

Helpful links

This sheet contains everything you need to start: calorie estimates and protein guides, recipes, exercise tracking, and some Dr LLM generated risk profiles.

Your knight for my pawn, and you have to take it. Things are about to get worse, sunshine. This is my game now. You have been out-calculated. I don't want the piece, I want your soul. My attack hurts like a school disco and I'm going to sac sac mate until... wait. What? 10 seconds left. But, I'm winning, right? I must be like +5 right now. Just play any move. But how do you have 10 minutes on your clock? Anyway, just play any move. That one will do, fine. For. Fuck's. Sake. Any move but that move! How could I instantly find the one move that loses? Literally, I could have played anything el...

At first, chess seems to be about the moves you play. But, after a while, you'll notice something weird. You can know more, think deeper, and play better moves, and still lose. That's because better chess is not about playing better moves. It's about playing good enough moves, quickly enough.

Playing moves faster means you play worse moves. If you have to play a game with only 3 minutes on your clock, you'll play about as well as someone 700 points lower rated than you would in a 90 minute game. Put another way, the 3 minute version of you beats the 90 minute version of you roughly 2 in 300 games (that’s what 700 points mean). Ben Johnson and Ken Regan go in to detail here.

It's hard to think of anything that can affect your chess strength this much. How many openings have you learned? How many strategies have you formulated? How many tactics have you solved? How many endgames have you studied? And yet, for better chess, the most valuable skill might be how you use your time.

This feels cruel. You can improve your tactics, openings, and everything else, but if you don’t combine them in the right way, you’ll still lose. You’re like a mathematician who never learned BODMAS. You might perfectly add, subtract, multiply, and divide, but if you do it in the wrong order you get the wrong answer.

Unfortunately, after two years of maniacal chess playing, I do not have this skill. How could I? When I play chess, I just play. I have no idea what I should be doing, let alone how long I should be doing it for. And, somehow, among the endless drift of chess content, it seems few have much to say about to actually practice using the right amount of time on the right things. Only that you should.

It’s not clear what playing chess even means. Only I’m pretty sure it’s not what I’m doing: think an arbitrary amount of time until there's a move I reckon is best, or least bad, then play it. This means I run out of time, all the time. I checked Aimchess and I'm behind on the clock 90% of the time. I get sucked in to all the possibilities, especially when any move will do. The more I've learned about chess the more there is to think about and the slower I’ve got. I win the game the hard way, through calculation and planning. I then lose the game the easy way, because I'm watching the clock more than the board. Clearly, this is wrong.

Other people play chess differently, but equally badly. I run an informal chess club, the Tuesday Chess Club (which ironically runs on Monday or Wednesday but never Tuesday). And Shanjit, a quiet guy who speaks only to tell you you're mated in 5, is top dog. When Shanjit beats you, he smashes you. We play with a 10 second increment, and Shanjit finishes every game with more time left than he started with. But he loses that way, too. Win or lose, it takes Shanjit no time at all. Maybe Shanjit doesn't actually like chess much, so when he's playing chess he tries to avoid it. I dread the day he decides to think for 11 seconds a move.

So how long should you spend per move? And what should you be thinking about?

How long to spend per move

The two triggers

A lot of what I write here is a summary or extrapolation of Dan Heisman's A Guide To Chess Improvement. It's really phenomenal, so go buy it and read the whole thing.

You might say it's not possible to say precisely how long you should spend per chess move. Some moves are harder than others. You'd of course be right. But I suspect you're being pedantic, and we can get a practical answer and improve our game.

If you are optimising for time alone, you should move as quickly as possible. But it is possible to spend too little time on a move. If you miss a win or blunder because you moved faster than necessary, that's a mistake. On the other hand, if you aren't optimising for time at all, you should think almost infinitely. But it is possible to spend too long on a move. If you miss a win or blunder later because you now have insufficient time, that's a mistake.

So we need one principle to stop us moving too fast, and one principle to stop us moving too slow. Dan calls these 'move triggers'. You should move when:

Trigger 1 (if you move too fast): You have proven which move is best, so no further time is necessary. You would be willing to stand up in front of a room full of grandmasters and argue that it would not be possible to find a better move; or

Trigger 2 (if you move too slow): You have not proven which move is best, but time constraints (good time management) make it efficient for you to play the best move you have found so far. There is a time 't' which is the reasonable time to think for that move.

The first trigger is straightforward. You move as fast as possible when you're sure it's opening prep, or when you're sure of the best move. If, like Shanjit, you're used to playing too fast, you probably move before Trigger 1. And, you might wonder 'what do I do with the rest of my time?'. We'll get to that.

The second trigger is more complicated. If, like me, you're used to playing too slow, you might wonder about the mistakes you’ll make by playing a worse move because you've hit Trigger 2. That's the trade off you make to avoid losing in time trouble. Smaller mistakes now instead of bigger mistakes later. But if reasonable time t is our trigger, how long is t?

How long is t?

To answer this, we first need to understand three principles: average moves per game; Botvinnik's Rule; and critical moves. At times some of the numbers below might sound fiddly, but don't worry - you don't need to remember them. The guidelines they generate will be straightforward.

Average moves per game

There are roughly 40 moves in the average chess game (source here). If you use all your time in 40 moves, that's an average of ~2.5% of your time every move. If your average move time is much faster than that, you're often going to leave time - and chess skill - on the table. If your average move time is much slower than that, you're often going to run out of time. It's crude, but our first estimate for t is just 2.5% of your total time. Assuming you play some moves faster because of Trigger 1, that should give you a little buffer. But we can do better.

Botvinnik's Rule

Botvinnik was a chess world champion, and his time management rule was this: In slow games, use about 20% of your time for the first 15 moves. In fast games, use less than 20% of your time for the first 15 moves (Dan has more here).

This rule is fairly intuitive. Games are not usually decided in the first 15 moves, 'the opening', and it's the most predictable and familiar part of the game so preparation and principles can speed up your play. I personally find this one tricky to stick to, though, and there are some situations where this can be ignored. For example, where you are in book for a long time or out of it very early, or where the position is highly critical move after move.

But Botvinnik knows what he’s talking about, so let’s take it as a good aim. 20% of your time divided by 15 moves is ~1.3% of your time per move. So now we have a better estimate, t is ~1.3% of your total time for the first 15 moves, and ~2.5% of your time after that. Our buffer is quite large now: by moving faster in the opening we saved time compared with if we'd just moved at our original 2.5% speed the whole time. We save ~1.1% of our time per move for 15 moves, which is ~17% of our total time. Add that to the time saved by moving faster than Trigger 1, and we've got a lot of buffer.

Critical moves

Some moves are more important than others. Some positions have many easy to find moves which are playable. These are not critical. Some positions have one complex move which can win or lose the game. These are critical positions. Obviously we want to spend more time on moves in critical positions than non critical positions.

Identifying critical positions is an important skill. We have some tactical heuristics:

• When the best move may be clearly better than the second best move.

• When you have a tactical (short-lived) advantage which will disappear if not exploited now.

• When you see a move which seems to win.

And some strategic heuristics:

• When the game changes from known theory into unknown territory, from opening to middlegame, or from middlegame to endgame.

• When any pieces are exchanged, especially queens.

• When there is any change, or possible change in the pawn structure – especially in the center.

In some cases, it may be warranted to use almost all your remaining time on a critical move - for example when deciding the almost certainly final critical move in a concrete endgame position. But, most games will have several critical moves. One easy to remember suggestion here is just to double t on critical moves compared with non critical moves. If t for non critical moves is ~2.5%, on critical moves t is 5% of your total time. Given the buffer created by our faster opening play, and Trigger 1, we have time for roughly 7 or 8 critical moves per game. This seems about right based on our heuristics.

Just give me the answer

With these three principles, we have a practical guide to time management. And, surprisingly, the numbers work out quite neatly. Roughly speaking, if the time control is 10 minutes, we use ~10 seconds per move for the first 15 moves, double it for ~20 seconds per move for the remaining non critical moves, and double that for ~40 seconds per move for the critical moves. If the time control is 5 minutes, it’s 5, 10, and 20 seconds per move. And so on.

Obviously these numbers are just guidelines, and should be flexed somewhat either way based on your intuition in the moment. But if you're going far longer or far shorter, you should be questioning if that's the right decision.

Here's a table for popular time controls:

If you want to change the parameters, for example to change the time control, length of or time spent in the opening phase, average move per game, etc you can copy and edit the sheet here.

What to think about and how to practice

So, if t is the reasonable time to spend on a move... what should we actually be thinking about for t? And how should you practice thinking about whatever that is, but for no shorter or longer than t?

I know I play too slow, but I didn’t know how to fix it. I just play, slow. Like Shanjit plays, fast. You’d think me playing more faster games and Shanjit playing more slower games would be the solution. It isn’t - I’m living proof that bullet chess and time trouble are simultaneous addictions. But, I’ve found a way to practice and improve. And it starts with knowing what to think about.

Poor time management is a symptom of not knowing what to think about. If you move too fast, you aren't thinking all the thoughts you need to play your best. If you move too slow, you're thinking thoughts you don't need to play your best. Practicing time management means practicing what to think about.

When I'm making a move, my thoughts often wander. I go down a chess rabbit hole and emerge in time trouble. Sometimes it's worth it, usually it's not. On the other hand, Shanjit doesn't wander anywhere - he plays the move he thinks is best before he's sure that it is. The answer to playing faster, for me, and the answer to playing slower, for Shanjit, is the same: be clear about what exactly is worth thinking about. Whether you're too fast or too slow, you need to practice your thought patterns.

So what should you think about? We don't want to think any unnecessary thoughts, but we also don't want to miss anything. Our thoughts should be both necessary and sufficient.

The necessary and sufficient thought pattern and acronym I use is TASOTISM, which knits together thoughts from Dan’s book. I often count TASO TISM on my fingers when I play (though sadly I am not good enough at chess for people to think I am cheating with mysterious hand signals). Every move, TASOTISM is necessary and sufficient to play your best. Here’s what it means:

Time

When is my Trigger 2? Is the move critical?

You can check this while waiting for your opponent to move. You know how long t is for your time control, so you can deduct that from your current time left on the clock to know your Trigger 2 time. If you think the move is critical at any point during your thought process, you deduct t again to double your time that move.

All

What are all the things my opponent’s move does?

It's easy to miss a hanging pawn, or that a bishop has become active where it wasn't before. Often beginners miss their opponent's plans, and this step forces us to consider them.

Safe

Is my opponent’s move safe? Is there a tactic?

A move is safe it doesn’t lose material (eg a pawn), concede something horrible positionally (a powerful outpost) or lead to being mated. There are a variety of ways to check this: checks, captures, threats; the burger method; safety tables; etc. This is just one step of the process of making a good move. That's why complicated tactics aren't that practically important, and why your intuitive sense of whether a tactic is present is. If, in a 30 + 20 game, t for non critical moves is ¬60 seconds, the kind of tactics you could sense within a useful amount of time might be tactics you could sense within 20 seconds and solve within 60 extra seconds. In 5 + 3 blitz, where t is ~10 seconds, if you can't sense a tactic in 5 seconds and solve within 10 extra seconds, it's not worth the time it takes to look for.

Objective

What is my objective this move?

Maybe you need to complete development, apply pressure to some weak pawns, to start an attack or to play prophylactically. This step forces us to consider what we know strategically, so that we don't prioritise the wrong things, like grabbing a pawn while our king is weak or starting an attack while we’re underdeveloped.

Time

How much time do I have left before Trigger 2? Is it critical?

Halfway done, let's check how much time we have left this move.

Initial

What are all the initial candidate moves that might accomplish one or all of my wants?

This stops me wandering around random moves, that don't align with our wants. And it forces Shanjit to consider more moves than one.

Safe

Which of the initial candidate moves can I reject because they are not safe?

We apply our tactical radar now to our own moves. Importantly, this doesn’t mean trying to guess how the game might go several moves into the future. It means checking our candidate moves are safe.

Move!

Of the remaining candidate moves, which do I think is best?

You have thought all the thoughts you need to think to make a good move in a reasonable time, so now it's time to pick one.

That’s TASOTISM.

If you’re used to letting your mind wander, it is brutally utilitarian. But that’s a good thing - thinking is your enemy, and now you may only think to serve the greater good. If you dont usually think much, the extra thoughts can seem superfluous, especially if you're used to playing fast time controls. If that's the case, ask yourself, which step should you skip? The answer, I think, is none.

This thought pattern is not a prescription - it is a description of the minimum and maximum thoughts you must have to play your best chess. I suspect that's why lots of strong players recommend beginners play slow time controls: strong players can complete the necessary and sufficient thoughts very fast and mostly subconsciously.

For weaker players, you can't play fast without skipping a step. If you must play faster time controls, I recommend just SIS. Is their move Safe? What are my Initial candidate moves? Which is Safe? That said, hoping you’ll develop better thought patterns by playing faster time controls is is like hoping you’ll learn Flight of the Bumblebee by playing along.

Lion's bones

It can seem strange to put such a rigorous structure on our thinking in a game like chess. Are we not caging our inner chess lion? No masters are sitting there counting TASOTISM on their claws! I think this is misguided. A rigorous thought pattern is not a cage, it is an x ray. It shows our lion's bones. Strong players don't notice their thought patterns - you only notice bones when they're broken. The goal is not to think TASOTISM in t, the goal is for good thought patterns to become so internalised that they form the skeleton our muscles hang on. Bones force our chess lion's muscles to work harmoniously - bones make it impossible for muscles to work any other way.

Good thought patterns allow you to figure out which muscles are weak. Do you lack quick tactical sense or calculation? Are you unable to formulate strategies or come up with candidate moves? Are you spending too long in the opening? Are you indecisive? It is impossible to answer these questions without a good thought pattern. If you calculate too long, you might mistakenly feel better in calculation than your peers. If you calculate too little, you might mistakenly feel worse. If you are playing endgames with no time on your clock, you endgames might appear bad because your openings are too slow. When your time and thoughts aren’t assembled properly, it’s impossible to build muscle in the right places.

If you complete a good thought pattern in a reasonable time and still lose, it's time to build some muscle. Until then, it’s not tactics or openings or endgames you need most. It’s time with your thoughts.

2023 update

For blitz games I go with SIS.

For longer games, I moved the A before the S, so TASOTISM instead of TSAWTISP.

I recommend playing long 0 increment games to practice time management. Online this would be 30 + 0. There’s no way to come back after mismanaging your time with no increment. This makes them a good forcing mechanism for practicing time management.

It’s worse to run out of time than it is to play slightly ahead of time, and some games run long, so I recommend extending the move count we plan for to 45, so that we have some buffer if it looks like we’ll go to a long endgame.

Nate Solon recommends thinking in terms of time per periods or time checkpoints, rather than moves. For example, where should I be on the clock at move 15? This is useful. If we’re planning for 45 moves, the most logical periods are 15 (opening), 16-25 (middle game), 25-35 transition or mate), 35+ (endgame). So 25, 25, 35 make good time checkpoints. Including these checkpoints and planning for 45 moves makes the time table look like this:

If you liked this you might like my first article about how to improve at chess here.

I thank Dan Heisman, Ben Johnson, Ken Regan, and Nate Solon for providing the ideas and inspiration for this post.

My addiction is every chess amateur’s favourite topic - openings. The first five to twenty five moves, depending on who you ask. I’ve played many different openings many different ways. I have freestyled, innocently mating my opponent half as often as I mated myself. And I have memorised, splurging through five Chessable ‘Lifetime’ repertoires that remain incomplete (for a Lifetime?).

The truth is, my openings still suck. As, I’m sure, do yours. But it’s not our fault. It is, I concluded conveniently, our repertoires’ fault. They are not built for humans. OpeningBuilder is my attempt to fix this.

Who’s winning here?

Why your opening repertoire sucks:

You are learning moves you will never play.

Most opening repertoires are either exhaustive or insufficient.

There are the Chessable courses or books with thousands of lines averaging 15 moves deep, answering engine responses to engine novelties that no human has ever even played and maybe never will. Fine, if you are a Grandmaster who wants to dedicate their life to learning a Chessable course. Not fine if you want to play chess.

And there are scant Lichess studies or Chessable courses designed for ‘club players’. Somehow they’re still too long, but only cover cheap lines that end in checkmate or queen traps. You mostly end up out of book on move 3. Also no good.

Almost all repertoires are organised according to the author’s whim. Never mind what lines you’re actually likely to get in your games.

The reality is that chess is a game with a lot of possibilities, and real players play weird moves. Even the most common main lines are rare. If you calculate the probabilities (I’ll show you how later), somewhere around move 5-10 your odds of being in a common ‘mainline’ round to about 0%, even if you play perfect mainline theory every time. Memorising 20 moves deep is as useful as memorising the numbers on winning lottery tickets.

A practical repertoire teaches you only what’s worth you remembering. No more, no less.

You are not a chess engine.

Most opening repertoires treat the engine’s evaluation as the ‘truth’, seeking out risk free routes to a small engine evaluated advantage in as many lines as possible.

This makes sense: a 100 centipawn advantage (something like having 1 pawn extra) is enough for a strong engine to win most of the time. 100 abstract centipawns are, however, for everyone but the most elite human players, practically useless.

For humans, some types of positions are easier to play than others, and some types of engine evaluated advantages are impractical to convert, especially for weaker players. Check your last game for how often you or your opponent dropped 100 centipawns in a single move according to the engine. Chances are it happened multiple times.

Even in the recent Candidates tournament, where games take several hours, the world’s strongest players could not reliably win from small engine evaluated advantages. Outside elite level, many positions which are certain wins for an engine are positions humans are statistically highly likely to lose! Check the position in the photo above.

A practical repertoire gives you positions you can actually win, not positions you should win.

What your opening repertoire should be:

Only the moves you actually need to know.

Using data on move play rates, we can build comprehensive repertoires with 10x less memorisation.

We now have access to the Lichess database: hundreds of millions of games by players of various ratings. We know which moves players play, in which time controls, at which level.

And with this data, we don’t need to create repertoires to a naive depth of, say, 10 or 20 moves, or whenever the course creator decides the line stops. We can create repertoires to a depth which reflects the probability that you will encounter a given move in your games.

For example:

If you play 1.e4, in Lichess Blitz across all ratings, your opponent will respond with …e5 about 31.5% of the time. If you then play 2.Nf3, your opponent will play the Petrov …Nf6 11.7% of the time. So, assuming you play e4 and Nf3 all the time, you’ll need a response to …Nf6 in around 3.7% of your games. If you then play 3. Nxe5, your opponent will play …d6 43.7% of the time, so you’ll need a response to that in around 1.6% of your games. This is the cumulative probability of the line.

Repertoires should be built according to the probability of you actually needing to play the move. For example, a line could stop after the move is likely to appear in less than 1 in 1000 games.

This allows repertoires to differentiate between moves you need to memorise, and moves you don’t. And doing this alone makes a massive difference.

A complete White or Black repertoire, with a response to every move that occurs at least once every thousand games, is shockingly small. Typically, it’s around 250-350 lines at an average of 6 moves deep, with only one or two new moves per line.

The ‘Keep it Simple 1.d4’ repertoire on Chessable is 1009 lines, averaging 13.7 moves deep. That’s almost 10x bigger than it needs to be for it to cover the moves you’ll see once or more every 1000 games. What is simple about that repertoire?

A practical repertoire is compact and easy to learn, based on play rates.

Moves that actually help you win.

Using data on win rates, we can create courses that are easy to learn, play, and win with.

The Lichess database doesn’t just tell us play rates. It tells us win rates. We know which moves have what probability of leading to a win, in which time control, at which level.

Even simple win rates are surprisingly powerful.

Win rates tell us how easy to play the position is. If an engine evaluates the position as a draw, but Black wins 75% of the time over 100 games, who has the better position? Easy to play positions are valuable: if we forget our lines, we want positions where it’s easy for us to find the winning moves and hard for our opponents.

Win rates tell us which engine advantages are human advantages. Would you rather play a move which is winning by 600 centipawns according to the engine but has a win rate of 10% because you need to find an unintuitive plan to win, or a move which is winning by 100 centipawns but has a win rate of 90% because your opponent needs to find an unintuitive plan to hold?

Win rates tell us the value of a rare move. We want to be in book while our opponents are out of book, but this is risky: rare moves are often rare for a reason. Win rates help you figure out if a move creates more problems for your opponent than it does for you, especially when it’s not the engine’s top move. Except at the highest level, you don’t want total novelties, you want sound rare moves backed by good win rates.

If we take a simple algorithm like: ‘always pick the move that has the highest win rate in the Lichess database’, the results are surprisingly powerful. The lines are easy to learn, play, and usually end up in a win. Left unsupervised, picking by win rate often finds lines with dubious gambits humans find hard to refute. With some simple engine supervision like ‘never pick moves that leave you more than 100 centipawns down’, the algorithm finds lines you’ll pay good money for on Chessable.

A practical repertoire is easy play and win with, based on win rates.

You can now automatically make a practical opening repertoire for any opening

So, I built OpeningBuilder - www.openingbuilder.com. OpeningBuilder is a free, open source program that will generate you an opening repertoire for any openings you want to play.

You can input any PGN starting point like ‘1.e4’, or ‘1.d4 e5’, or even hundreds of PGNs at the same time. OpeningBuilder will generate an opening repertoire by selecting the winningest moves in every position that you’re likely to get.

There are some powerful customisations:

• You can decide the repertoire depth as a cumulative probability based on play rates. So for example, a depth of 0.01 would give you a repertoire where the lines will end once the next move has less than 1% probability of occurring.

• You can decide what time controls and rating ranges you would like play rates and win rates to come from. You can build a blitz specific or 2000 specific repertoire. Or use all the data available.

• You can control the soundness of the repertoire. If you never want to be more than 50 centipawns down, or sacrifice more than 50 centipawns to play a move with a higher win rate, you can do that.

• You can finish lines with the engine. If there isn’t enough data to reliably pick between moves, you can opt to have lines ‘finished’ by the engine.

• And a bunch more.

The output is a PGN file (or multiple files if you started with multiple PGNs) which you can upload into Chessable, Lichess, Chesstempo, ChessBook, or any other learning software to memorise. The lines are concise and strong, backed by data and engines.

It’s super fun to play with, so give it a try. You can make a short (5% depth) e4 repertoire for White, or an e4 e5 repertoire for Black in about 15 seconds.

FAQ

Can it do any opening? Do you have an example?

Absolutely, as long as it’s been played before. It can generate you a grinding Reti course, or a razor sharp Latvian, depending on your settings.

I created a Blitz repertoire with almost no concern for soundness, entirely using gambits you’d struggle to find a course for anywhere. It’s amazing fun for online blitz, with wild (often winning) games. I created a slightly more sound Blitz repertoire for white which might appease pickier players. I created a sound ‘tournament’ repertoire which is complete for White and Black in about 600 lines total, averaging 6 moves deep. The critical responses are covered and where it matters the depth goes into the double digits. I’m not linking that ;)

Does it come up with interesting stuff? How do the lines compare to paid courses?

You bet. There are a bunch of fascinating discoveries you can make by playing with the settings and starting points. Who knew about the Portsmouth Gambit (1. e4 c5 2. Nf3 Nc6 3. b4), a chaotic version of the Wing gambit which scores well even at high level play?

That said, if you dial in stricter soundness settings on the engine side, you’ll get much more solid approaches similar to those recommended by Masters in Chessable courses. Honestly, some popular courses (which I shall not name) are more or less just following win rate recommendations with strict soundness limits.

Can you make this into a web application?

It now is! OpeningBuilder.com

Is the algorithm just climbing win rates?

More or less. We allow you to tweak things though, so you can bias the algorithm to pick moves with more data (or less), only pick moves with minimum play rates (by percentage or absolute), only pick moves within certain engine criteria, rating ranges, or time controls etc.

There are probably more sophisticated win rate functions that could be incorporated. Right now it just uses a confidence interval to account for variances in win rate sample size for each move and prioritise moves with larger sample sizes.

Why does the code look like it was written by someone who can’t code?

Because I didn’t know how to code until I made it.

Did you write this from scratch?

Nope. I started with the Chess Trap Scorer which was super helpful to get going.

Is this useful for very strong players in classical chess?

No. If you’re trying to prepare for your opponents’ novelties, this won’t help.

How long does it take to make a repertoire?

If you don’t use an engine, it can take seconds or minutes to build a full repertoire. If you want to use the engine features, it depends on how good your computer is and how deep you run the analysis. For deep (eg 35 for) analysis you might have to leave it for a few hours.

Does this miss lines where you have to walk a narrow but sure path to victory that most players don’t know about?

Sometimes. For most players I don’t think these lines are practical, though.

Does this miss lines that are ‘busted’ on move 24 by a novelty?

Sometimes, especially if you don’t use engine assistance.

Does this help learn middlegame plans?

No, but middlegame plans are probably not worth memorising move for move as part of a ‘repertoire’ anyway. Just check model games, watch videos, etc.

Does this annotate the lines?

No. The only annotations are play rates and win rates.

Are there other obvious flaws with the statistical approach?

Yes. The most obvious improvement would be something which looks deeper into the tree so that we calculate the win rate assuming we play only the lines we play (right now the win rate includes lines we don’t intend to play). That said I couldn’t figure out how to do this without hitting API limits, it may result in some hard to play or unintuitive lines or bad data, or just be slow. Would love to see someone try.

Another problem is that sometimes the deeper lines that are the highest win rate are less sound than the initial position, so strict engine settings will make the some win rates less accurate as some later moves are not playable. So it’s possible to pick a move with (say) a 60% win rate but find that the deeper moves that win rate was generated by are not actually within your soundness threshold, and alternative moves are a lower win rate.

How does this handle transpositions?

Quite well. It’ll always pick the same move in a given position with the same settings, unless the stats have changed meaningfully in the time the program takes to run, which is unlikely. On the other hand, it might lean away from transposing to a known line if a winninger alternative is found.

Why did you not do XYZ when you open sourced this?

Because I’ve not done this before.

Can I contact you about a bug/feature/something else?

Yes.

A hard question.

I failed last year. Closed down the startup I started. An adulting, unarguable, and public failure. I let people down - my investors and team members were mostly friends. Thinking about it now I feel ashamed and a bit sick. This isn’t a complaint. I’ve had plenty of good luck in life, some earned though lots not, and I deserved what I got this time.

So, am I happy I did it? That seems easy. No, I’m not happy I did it. Starting a startup was a painful way to lose my time, and (mainly) other people’s money.

But it’s not that simple, even if it wasn’t worth it.

Calculating About Magic

Magical products are curious experiences. When you try to explain why they’re magical, it’s often impossible. Yet once you experience their magic, it’s self evident.

Why is TikTok spellbinding? You can (in retrospect) tell a story about short, vertical, music accompanied videos, displayed in an order you’ll like. But no analysis beats the experience. If it did, TikTok would have been obvious without hindsight.

People have written books on trying to reason your way into magical products. There are Customer Developers with Minimal Viable Products, Hackers Doing Things That Don’t Scale, Zero to Oners Keeping Secrets. But there is something difficult and pointless about this reasoning. If it worked so well, magic wouldn’t be so rare. Following a method is a safety blanket, but the method is not the point, magic is.

And magic is easy to see but hard to prove. You may know 1+1=2, but you almost certainly can’t prove it. Probably, without help, if you spent your life on it you couldn't. This is startup complexity theory - there are answers that can be computed easily (P) and answers that can be checked easily (NP). And, you don’t need to. It’s up to writers and artists and academics to prove things we already know are true. Not the founders of TikTok.

Concentrating on the startup method can, sometimes, mean failure. You can’t prove future magic with tests and interviews. And if you try to prove things that don’t need proving, you’ll go slower than competitors that build the right thing the wrong way. The riskiest strategies win. It’s no use trying to explain TikTok when you can build it and show it. Reasoning ends when magic begins.

For smart people, this is hard to accept. In school, getting the right answer isn’t the point - you need to show your calculations. But in startups, it’s enough to make something magical, and not enough to do what you’re supposed to. No one really cares how you win, as long as you do. And if you fail, no one cares how you did that either. There is no right way, beyond winning and losing. No marks for calculations.

If writing about music is like dancing about architecture, eventually, talking about startups is like calculating about magic.

A Gambler’s Logic

When what matters is the magic, not the method, you must seize your chance.

So, 2020. My friend and I built a site that within a month had hundreds of thousands of people watching social esports content millions of times, with zero marketing. Magic. We'd made the site in a weekend and the esports competition was over soon - there were no calculations to be found, and there was no time to do them. We had to decide, do we go all in? Our answer, of course.

This opportunism is a common startup story. Take Entrepreneur First company Souschef, which made automation robotics for F&B retail, and started about as well as you could hope for. Within the first few weeks of EF, they had a sales deal worth $1million, ready to sign. Magic. Did they sign it? Of course.

Today, both my startup and Souschef are dead. We didn't capitalise on our magic, for reasons good and bad. Without magic you can’t do anything, but magic doesn’t tell you what to do next.

Judging by our results, following the magic and starting a startup was the wrong decision. Should we be happy we did it? No. But results based analysis doesn't quite make sense for startups.

In a competitive world, you win by making correct decisions before it's reasonable to. To stand a chance of winning, you have to take risk. By the time you can accurately assess the risk, it’s too late. Someone else skipped that step. No marks for calculations.

Taking risk means sometimes things don't go your way. That doesn't mean, given the situation, you shouldn't have placed the bet. In poker, the starting hand with the highest win probability is two aces. But, on a six person table, even with two aces, your odds of having the winning hand in the end is still less than 50%. Were you wrong to bet on aces because you lost on the final card? Of course not. It's what you're there to do.

Starting with magic is like starting with two aces. Most startups die young, because they can’t find any magic at all. If you find some magic, you’ll never have a better shot. If you might want to start something someday, you absolutely must take your chance. Even if you don’t succeed, it's not wrong to gamble.

Prison Breaks

So once you’ve failed, should you regret it? I think not. Starting a startup is like attempting a prison break with a life sentence: it probably won't work out, but you gotta take your chances.*

I’m still locked up. And while I’m here, I often think about all the people at EF who didn’t succeed either - by definition almost everyone. I can’t help but feel a little guilty. Have I encouraged thousands of people to do something they regret? Do they feel a little sick about it too? I'm sure some do.

But I hope, for most, they see EF for what it’s meant to be. A chance I’m grateful I had. An opportunity, that might never come again, to start with aces, to prison break. Sometimes failure, sometimes magic.

*Incidentally, I think this is the best argument for people who have no opportunities to make millions to still play the lottery, even though the expected value is negative.

"Costream and share any content: esports; TV; movies; even music."

"Banking, payments, investing, crypto, and currencies, all in one app."

I can already hear the investors groan. And yet, founder after founder makes the same mistake. Me too... that second (bad) idea was mine. So, why are these ideas bad and what should you do that I didn't?

Intuitively, a good startup idea offers users more of something. If your product does more than a product users already love, won't they love your product? Won't users that love your product now love it more if it does more?

Frustratingly, startup ideas aren't intuitive. Offering more turns your idea into an "And Idea". And And Ideas are - almost - always bad.

And Ideas

Good startup ideas can usually be summarised in one sentence. You just say how your product benefits your users. For example, Sonos might be "Play music around your home". Nest might be, "Save energy automatically".

Now, what if you wanted to start a startup to compete with Sonos, or Nest? One way would be to just offer more than either of them: a device that plays music while saving energy. While you're at it, make it a light bulb and take down Hue, too. Your idea: smart music and energy and lights all in one. And more.

This is an And Idea. You can't explain it in one sentence without using 'And'. That isn't a good thing. The words and, all in one, and and more deafen even the toughest investors' ears.

But, why? Wouldn't Sonos users love something that can do everything Sonos and Nest can do, all in one, and more? If only.

The Marginal Feature Fallacy

The best products are the best solution to a problem, not the second best to several.

When you experience a problem or desire deeply enough to make you go and buy something, it is that problem or desire you're addressing. When you want to listen to music, you search for a sound system. And you buy the best sound system for you.

Your product has to be best at its "First Feature" - the reason your users find and use it. When users really want your First Feature, they'll often use your product despite its limitations. That's why Sonos doesn’t need a thermostat. Sonos is just the best at playing music around your home.

If your product isn't the best at its First Feature, people won't buy it no matter how many more "Marginal Features" you add. You don't buy a Swiss Army Knife for your kitchen.

This doesn't mean your product must only do one thing. Great products pull you deeper and deeper. This is how 'Super Apps' like Amazon, Grab, and WeChat win. You don't start using them because everything is all in one place. You use them for one thing first, then stay for everything else.

But, as Andrew Chen notes in his classic post - users never get to the end of your funnel if they don’t finish the start. Marginal Features don’t matter unless the First Feature is compelling. And if you're building Sonos, Nest, and Hue simultaneously, you're at a huge disadvantage. You're going to be worse at music than Sonos, worse at energy than Nest, and worse at lighting than Hue. You'll lose out to each of them on their First Feature.

That's why And Ideas don't work. Offering Marginal Features doesn't differentiate you, because you compete on your First Feature. You won't be the best at something by building something else.

The best products differentiate themselves from their alternatives with an Or, not an And. Sonos is an Or Idea. The second half of "Play music around your home..." is "...Or fiddle around with Bluetooth and give up".

This is why new technology creates so many important new companies. New technology creates new ways of solving old problems. New Ors.

Find the Or in And

So how do you go from an And Idea to an Or Idea? You have two options: subtraction or addition.

Subtraction is hard. Ideas organically become And Ideas because founders aren't exactly sure why their users use their product, or they don't like the answer. So they add more.

I got this wrong both ways. In 2020 I made a social media site that, within a month, had hundreds of thousands of users watching esports matches with streamers. Usage was growing double digits week on week with millions of games watched. The only problem was I had no idea why. Inevitably, since I'd made the site for a specific esports event, once it ended users left. So, I added more features. First music, then more esports, TV, and movies. This seemed like a good idea: I had some users for every feature. But nothing beat the early metrics. I'd started with an Or - an amazing new way to do something specific, and ended with an And - a product that did a lot of stuff reasonably well.

I should have subtracted. Rather than adding features, I should have improved my First Feature and subtracted everything else. Psychologically, this is hard to do, especially once you have users. Adding features is easy, but poisonous.

Addition only works in one rare circumstance: when it creates a benefit greater than the sum of its features. Control4 is a smart home automation company. They basically built Sonos and Nest and Hue simultaneously, hardware and software, and are NASDAQ listed at a valuation of ~$650m. It's not as big as Sonos (~$3.5bn), Nest (~$3bn), or Hue (~$5bn), but Control4 is clearly big enough to have been a good idea.

So what is Control4’s idea? It is not Sonos and Nest and Hue all in one, and more. Their idea, their Or, their benefit to the user, their First Feature, is the user experience that doing everything together unlocks. Control4 doesn't sell music and energy and lights, it sells romance as you walk through the door; Al Green comes on; the fire crackles; and the lights dim. Their idea is that, together, music and energy and lights are greater than the sum of their parts. Their Or is that combining everything manually is impractically hard. Not everyone who wants music wants romance, but Control4 is the best product for those who do.

Apple is the ultimate example. The original pitch: an iPod, a phone, a browser. Through integration of hardware and software design and production - far beyond that of any other company - they made an And Idea into a market defining new user experience. They called it an iPhone, and found their Or.

Like most people, late last year I watched the Queen's Gambit. Like most people, I made a Chess.com account assuming I’d become the new Beth Harmon. Unfortunately, more than literally almost everyone, I sucked at chess.

At first, this was unsurprising: I only vaguely knew what the pieces did, and didn't know more complicated rules like castling. But I'm a smart guy (hi white male privilege). If I just brushed up on the rules and played a few games I'd be beating Bobby Fischer in no time.

So I did. In December I played about 500 games of 'Blitz' (fast) chess. That's around an hour a day - it was lockdown after all. What did I learn? I suck at chess! Fucking. Suck. I was not even bad for a beginner, I was one of the worst people in the world.

After 500 Blitz games my ELO rating (the number that shows how strong you are) was 328. Out of the regulars on Chess.com, 11 million players, I wasn’t even 5th percentile. AKA, 95% of players on Chess.com were better than me. Worse still, I wasn't improving, I was getting worse. I slumped to 328 on January 1st 2021, after my month of Blitz.

The cursed game that took me to 328. In the spirit of Never Resign, I should have challenged my opponent to find Nb4#. Only I couldn’t find Kd3.

Now, my ego is quite large. And quite fragile. And I'd genuinely tried. I'd fallen in love with chess, but chess didn't love me back: it was starting to prove that I wasn't so smart after all. Well, we can't have that. I decided to figure out how to get good.

Given the size of my aforementioned ego, I wouldn't be writing about this if I hadn't made some progress. After a bit of research, I started a 20-30 minute a day practice routine. In early September/October I played two Over The Board (physical) chess tournaments. After the second tournament, my Rapid rating was 1508, about the same as my recent online ratings.

So, what's that mean? It means I went from being in the bottom 5% to pushing the top 5% of players on Chess.com in around 9 months. More importantly for you, dear reader: I now understand how adults can meaningfully improve at chess.

OK, some caveats. First, I'm no Beth Harmon: 1500 is not going to impress strong chess players. It's empirically harder to improve the stronger you are, and I have no evidence my methods will work for other people. Second, I now play almost exclusively Rapid (fast) games, instead of Blitz (faster). These ratings can't be compared precisely and from my just-had-a-few-beers-and-am-sitting-on-the-bus/train/toilet Blitz games my Blitz rating seems to lag by as much as 200 points. Third, the way I did this probably isn't fun for most people.

But, I did improve meaningfully. And, it turns out, in chess, almost no-one improves. Only about 10% of players ever gain more than 100 points, and only about 1% of players gain more than 200 rating points given years. Adult chess improvement is especially controversial. Since almost every chess Grandmaster became strong mostly during childhood, most strong players never had to do it!

So, if you too are an adult who sucks at chess, and want to improve, here's how.

I wish I had a single graph. To my shame I periodically made a new account, illogically hoping for different results. In October I hit 1492 on Chess.com and 1508 ECF OTB. Recently I’ve switched to Lichess where the rating system isn’t easily comparable, and sunk to 1497 OTB..!

What we're training

If you want my exact routine, skip to the next section. But, if you want to design a routine that works for you, you'll need to understand why my routine worked.

Especially at lower levels, chess is a game of short term patterns, not long term strategy. Sure, you can make a plan. But, mostly, it won't matter. It turns out trying to win a chess game on strategy alone is like trying to win a boxing match without throwing punches.

Why? Because your chess games are almost always decided by patterns called 'tactics' - a short sequence of moves that turns your roughly equal game into a completely unequal game. If you've played chess, you'll know the feeling: you’ve blundered a tactic every time your opponents took your Queen for free. Tactics and blunders are two sides of the same coin: blunders are move that create tactical opportunities.

Thanks for the tip Chess.com Coach.

And if you can't spot tactics, you can't win games. You can play openings like Magnus Carlsen, but play endings like Carlos Magnussen and you’re still going to lose. In the Woodpecker Method, GMs Axel Smith & Hans Tikkanen estimated what percentage of games are won or lost by tactical mistakes at different ratings:

GrandMasters 42%

2200-2400s 44%

2000-2200s 63%

1800-2000s 72%

300-500s 100%*

*my estimate. GMs Axel and Hans didn’t review my games.

Sadly this doesn't mean GMs don't do tactics. It means tactical skill is chess’s foundational skill, which they’ve (modestly) transcended. This classic De La Maza article has more.

But what is tactical skill? Let's check the science.

Well, the simpler the tactic, the more likely a strong player is to consider the best move immediately. In a position with a tactical opportunity, this study found that the first move that pops into a 2500 players head is as strong as the move that a 1548 player takes 5 minutes to find. And, fascinatingly, both strong and weak players improve decision making at similar rates given more time.

So what? The stronger the player, the more automatic the recognition of tactical patterns. They don't see individual pieces, they see chunks. Strong players recognise tactics like you recognise faces, or words, or tastes. Capablanca's famous adage "I see only one move ahead, but always the best move" turns out to be true.

If fluency in a language is measured in accuracy, speed, and prosody, then stronger players are more tactically fluent in chess. They find better moves, right away, in more relevant positions. Which begs the question, if tactical fluency is your goal, how should you train?

To get accurate:

• Test yourself. You'll learn more efficiently using Active Recall than Passive Recall.

To get fast:

• Study ‘Comprehensible Input’. You need to be capable of independently deriving the patterns you’re testing. Start from the absolute basics. Krashen explains here.

• Test response time. We process facial expressions and complex sentences in milliseconds - why not chess positions? Remember, we're trying to train your automatic pattern recognition and chunking, not your calculation or knowledge.

To see a pattern in more positions:

• Learn by category. To learn the pattern of 'mate in one with a queen', test yourself on lots of them in series, with instructive commentary. Context and subtle differences build understanding, on top of raw memory.

• Then test randomly. Use a collection of randomly ordered patterns to encourage your brain to spot familiar patterns in new and more complex positions. You're trying to generalise pattern recognition when you don’t know what you’re looking for. Here's more on why categories and randomness work.

To see more patterns:

• Use Spaced Repetition. For any given test, you should repeat the same test with more and more space in between tests. This dramatically reduces study time. Gwern has the best write up on Active Recall and Spaced Repetition here.

Once you're accurate, fast, and can spot the pattern in almost any context, you're fluent in a tactic. The more tactics you're fluent in, the more games you'll win.

My Routine for Tactical Fluency

This is the routine and tooling I used to train tactical fluency. To understand why this works, check the previous section.

The simplest way to start

The simplest recommendation would be to train Chess Tempo ‘Blitz’ puzzles on Easy mode. Chess Tempo Blitz scores you on whether or not you can solve a puzzle, and how fast you solve it relative to the average solver. If you are faster than average, you get more points. Much slower, you might even lose points. Get it wrong, lose even more. The puzzles are generally well curated, with lots of human commentary.

Another alternative would be to solve Lichess puzzles on the ‘very easy’ setting. This will only give you puzzles that are simple to solve and can be solved quickly, and will penalise your puzzle rating severely for making a mistake. It works for pattern recognition, because even simple errors will make the next puzzles you see a lot easier.

Both these options are free or close to, and easy to get started.

My approach

Before you think this is an ad - I'm not affiliated with Chessable in any way, it's just perfect for what we need. Each tweak of the setting is explained so you can replicate this in a different system if you wish.

I’ve made some small improvements to this approach since this article was published. Check the ‘Update’ section at the end of this article for those tweaks.

First, software.

Chessable is a Spaced Repetition platform for chess. You do tactics, and if you get them right it shows you the tactic again in the future (a 'review'). You can even set time limits for answering. Chessable makes it easy to test accuracy and speed with spaced repetition. This is where we are going to spend all our time.

Second, courses.

We need a course which explains and groups tactical patterns, and a course which presents them randomly. This means we learn patterns and how to apply them across contexts.

For categorised patterns and explanations, I have found GM Susan Polgar's Learn Chess The Right Way series to be incredible, but there are probably others. I am on Book 4 right now, but I especially loved Book 3 on defensive tactics.

For random tactics, I found Tactics Time to be good. It is basic stuff, jumbled up, which is what we need.

Readers have recommended Common Chess Patterns for beginner / intermediate, and Improve Your Chess Tactics for intermediate / advanced players. Both contain categorised patterns and random tactics and are very well reviewed.

Third, Chessable settings.

I recommend studying tactics with some specific settings. These settings force us to solve many problems we can understand slowly, fast.

You'll need a Pro account, but it's great value. You'll only need to change these once.

Once you've got a course, set your course settings to:

Study: Key Moves. This means we don't get tested on long unforced ends of variations.

Review: Whole Variation. This means we always get tested on each tactic from start to finish, rather than each individual move separately. Ideally we play through the whole variation in our head before making our first move.

Reps: 1. This means if/when we get shown the answer to a tactic, we only get shown it once. We wan't to find the answer rather than drill it.

Depth: Full Depth. This means we don't stop the tactic if it is more than N number of moves long. We want to study tactics start to finish.

Soft fail: Retry on alternative good move. This means we don't fail a tactic because there's multiple ways to mate and we pick the 'wrong' one.

Time: 8 seconds. This means we can only get a tactic correct if we answer within 8 seconds. This is important - we don't just want to get the answer right, we want to get it right almost instantly. I've messed around with different numbers here and it seems like 8 hits the sweet spot of 'I have time to see the position', 'I have some time to visualise the tactic from my first intuited move to the end', and 'if my first intuited move is wrong I don't have time to think of another'. The right number here is unclear, but 8 seconds isn’t unreasonably quick. Puzzle Rush asks you to solve progressively harder puzzles (going far beyond our level) for 5 minutes and the top scores are sub 5 seconds per puzzle.

Tactics: Solve problem. This means we don't want to be shown the answer to a tactic in advance, we have to solve it (or fail it) first.

Schedule:

Level 1: 1 day 16 hours

Level 2: 4 days 16 hours

Level 3: 10 days 16 hours

Level 4: 24 days 16 hours

Level 5: 56 days 16 hours

Level 6: 128 days 16 hours

Level 7: 296 days 16 hours

Level 8: 681 days 16 hours

This is a much more aggressive schedule than Chessable's default, both beginning with a larger interval and increasing the space between each correct review by ~2.1.x. This is actually conservative based on my experience with Anki, which uses an even higher default interval growth multiplier of 2.5. There are two major benefits to aggressive spacing, based on the idea that your fluency is cumulative:

• In the short term it encourages understanding, rather than (just) rote memorisation. It’s better to have longer intervals even if you fail more. We’re not trying to optimise for the probability of getting any specific puzzle correct, we’re trying to optimise for cumulative absorption of the pattern.

• In the long term it reduces workload per tactic which increases the amount of tactics you can maintain. If you average 12 seconds a tactic (time slack given failures) at Level 6, you can maintain 12800 tactics in just 20 minutes a day! Maintaining these 498 checkmates now takes me 3-4 reviews a day or about 45 seconds, with an average interval of 142 days per tactic.

• NB. 16 hour thing is because Chessable doesn't make everything due by 'day', and so we basically do it manually by assuming you'll review at a similar time each day and scheduling each tactic to be due before then (eg a tactic due in 1 day 16 hours is in practice due in 2 days). Without this managing how many reviews you have per day becomes impossible.

Finally, look for the 'MoveTrainer' Settings (a little gear icon when studying on web and in the app’s Settings on mobile):. Then change these settings:

Time up action: stop timer. This means when the 8 second timer runs out, even if you get the right answer you will still have failed the tactic... but, you will have unlimited time to find this right answer. This is important: if after 8 seconds you are just shown the solution, your brain will not have to understand the tactic and you may not be training Comprehensible Input.

Enable retry: on. This means that if you answer incorrectly, you will get another attempt to find the right answer. This is also important. While we are training our automatic pattern recognition, we want to force our brain to understand the tactic in cases where our intuition fails.

Max retries for a mistake: 1. If we generate a wrong answer, we want to force ourselves to properly solve the tactic and find the correct answer, rather than to allow ourselves to just try multiple incorrect answers until one passes.

Retry action: stop timer. As above, we want unlimited time to find the right answer if previously gave an incorrect answer, to aid understanding.

Time up action for retry: stop timer. As above. understanding.

Highlight legal moves: off. This means when you click or grab a piece, the moves it can legally make will be highlighted on the board. This is a crutch you won’t have OTB so turn it off.

However you do it, get started and keep going.

Set aside some amount of time you can afford every day, and then split it across studying tactics by category and then randomly.

For me, I set a 10 minute timer, and then review the Polgar categorised tactics courses, from easy to hard. If I finish my reviews within the 10 minutes, I learn more categorised tactics. After 10 minutes, I finish the problem I'm on.

Then, I set a 10 minute timer, and then review randomly ordered tactics from Tactics Time. If I finish my reviews within the 10 minutes, I learn more random tactics. After 9 minutes, I finish the problem I'm on.

That's it.

It takes around 20-30 minutes a day, depending on how I scale up or down the time. One last thing: Spaced Repetition punishes inconsistency because reviews clump up. Spend less time on Chessable than you can spare each day, in case you miss a day. If you have extra study time one day, don’t spend it on Chessable unless you can afford more review time in the coming days. Both under and overstudying create review backlogs, which are horrible to clear.

My Chessable review heatmap.

FAQs

What about openings?

I learned a bunch of openings with White for 6 months or so, also via Chessable. Amazingly I won more games with Black, where I had learned nothing, than with White. I got frustrated with this, and switched my openings entirely. It had literally no impact on my rating, and I continued to improve.

I did a quick calculation using the Lichess database and realised, outside GM level, the most common lines still occur less than 0.5% of the time after 8 moves, even assuming you remember your moves 100% of the time! Basically, memorising lines is pretty useless.

I do still do a few minutes a day on low depth openings, because it gives me psychological safety - "I know what I'm doing" - and it’s fun to try different things.

What about pawn structure? Or strategy?

I have no idea!

What next?

Once I've finished the Polgar series and Tactics Time, I plan to apply this method to more advanced courses, like:

Basic Endgames, The Checkmate Patterns Manual, Dazzling Defenses against Checkmate Patterns, The Woodpecker Method, Common Chess Patterns, Improve Your Chess Tactics

I can't vouch for these yet though.

Do you think you can become a titled player with this method?

Probably not. If chess fluency is like language fluency, there’s probably a reason most GMs started very young. Getting a title means beating native speakers.

Should I play games?

Games are obviously fun, but not obviously good for your improvement. I have averaged about one online Rapid game per day this year. OTB tournament games have a different feel and are more memorable, so I would prioritise those from a learning perspective.

What time control for games?

Early on I played almost entirely Blitz, but the prevailing wisdom from legendary coaches like Dan Heisman seems to be that longer games are better for learners. I now mainly play Rapid.

The risk of Blitz and Bullet is that you drill your bad intuitions without corrective feedback or understanding - literally the opposite of what we're trying to do! The issue with really long games is that you can't play many of them, and in most positions there's diminishing marginal returns on quality thinking time.

Playing without increment has a similar effect. We don't want to force ourselves into playing endgame Bullet because we thought about a position for a few minutes.

For me, Rapid with an increment strikes a good balance. Sadly, I am not immune to the charms of toilet Bullet.

Where to play and what to play with?

Chess.com and Lichess are both great sites.

That said, in my first OTB tournament, I found that having learned chess entirely on a 2D plane with 2D pieces meant that my tactical fluency was dramatically worse in real life. I started using Lichess, because it's available on mobile web with a good looking 3D chess board which matches the plastic OTB tournament sets (see pic).

I genuinely think the same 3D pieces would improve Chessable's utility a lot (if anyone from there is reading!).

And turn off Highlight Legal Moves. You don’t get that OTB.

What about other tactics and tools? Puzzle Rush? Tactics Trainer?

I don't know, really. While I think these could be good benchmarking tools, I don't know if they're comparatively good tools to train tactical fluency. They tend to encourage speed or understanding, rather than both together. Puzzle Rush doesn't have a way to promote understanding, and Tactics Trainer ends up showing you puzzles that are beyond your Comprehensible Input.

I do like to play Lichess/Chess.com tactics on my phone in the background while watching TV. I don't focus on it, and just allow myself to take as much time as I want to solve a puzzle - often a whole evening or not at all if the show is good. I am around 2200 on both apps right now.

What about Chess content?

I confess, I also love a Gotham Recap. And who doesn't love watching GM Simon Williams go 0-8 playing the Reverse Double Bird Sicilian Gambit into a 2200 player?

How did you come up with this stuff?

I transplanted a lot of methods from language learning. Some things I took from folks like Eshepard, JA-Dark, (the early days of) AJATT, and Heisig. For Chess specifically, I got a bunch of ideas about how to learn Chess from the Perpetual Chess Podcast by Ben Johnson. The episode with Logozar sharpened my approach further.

Haven’t you just spent a lot of time playing chess lately?

A lot for some people and a little for others. But how much I play doesn’t correlate with my rating, which is true for most people.

What did you do in January?

Tactics on Chess.com, and some Woodpecker Method style stuff which lead me to my current approach.

Can you share your online IDs?

I have never played chess during a meeting.

Is this a fun method?

Not even for me! But if you like improving, or winning, maybe you'll get over that.

The “plastic tournament” set on Lichess.

Updated method 1 year later

I’ve since tweaked some elements of this approach:

- increased the time to solve to 10 seconds, but hidden the solve timer from screen (easiest done by scrolling so the timer is off screen - better done by just removing the timer element with an ad blocker). The goal is to solve faster, not (just) to guess faster and feel time pressure.

- focused on fully visualising the critical line of the puzzle to it's end position. So, rather than playing the first best move before the time runs out, I have to be able to fully visualise the critical line to its end position. If it’s a mate this would include seeing the king has no escape squares, rather than just the mating move, for example.

- limited the books I’ve studied to exclude random tactics collections (eg tactics time), favouring collections focused on specific patterns (eg Learn Chess The Right Way). This ensures I’m initially learning a comprehensive group of patterns presented in a way I can easily absorb, and then the spaced repetition takes care of automating them. For a full list of books I’d recommend, that I am working my way through, go here.

- now only use use ‘review all’, rather than review structured tactics then mixed. Over time this jumbles the tactics together as I get throngs wrong, and as the due times overlap. You get the best of both: structured patterns as you’re learning and then interleaved mixed testing.

- increased the initial interval to as much as 7 or 30 days, scaling at 2 - 2.5 rate. You can find the full intervals here. 4 days is the data backed most efficient first interval on Supermemo, but I personally feel longer works for chess, since we’re not optimising for any given puzzle. This schedule allows you to get less bogged down by puzzles you find tough, while taking advantage of the cumulative effect of doing lots of puzzles and making pure memorisation harder.

Update 2 years later

This somehow become the top ranked result on Google for “how to learn chess”. Good news: it still works. I hit 2k on chess.com doing pretty much this, with a bit of calculation added in. OTB is another beast.

Thanks to Anne Marie Droste and Matt Sharp for proof reading. And thanks especially to the language learners, Logozar, Chessable, and Ben Johnson for their ideas and inspiraton.

That's still true, and still exciting. But now I think maybe it gets the future the wrong way around. What's exciting about technology is not the entrepreneurs at the head of the income graph. What's exciting is the creation of an enormous long tail of people who, for the first time in human history, will do their life's work.

Why jobs suck

Doing something you love has mostly been a fantasy since jobs existed. Without technology, by definition, it's almost impossible to get paid to do something you love.

Without technology, if you love doing something weird, you probably can't find someone who'll pay you to do it. If you can find someone who'll pay, your thing can't be that weird after all. But, if it's not weird, you're probably not the best person who does it. And if you're not the best, why would someone pay you to do it?

This is why most jobs suck. It's almost impossible to find people who'll pay you to do what you love. At least, it was.

Why they won’t

With technology, being weird doesn't matter anymore. Even if the thing you love is super weird, the kind of thing only 0.1% of people love, there's still ¬800,000 other people who love it. 800,000 people who are now on the internet.

Because of technology, the weirdest thing you love now has an internet city bigger than Stockholm, entirely populated with weirdos like you. If you could somehow get 0.1% of those 800,000 people to pay you a tad over 5$ a month, you'd already earn the average US income.

Get paid more than most Americans, for doing something you love. If you hate your job - and you probably do - why on earth wouldn't you try this?

Well, you would. With technology, the market for weird gets bigger than the market for normal. After moving online, the market for books that are not even sold in the average book store became larger than the market for books that are. Weird is what Shopify, Patreon, Spotify, Amazon Marketplace, Substack, Twitch, and a hundred other platforms are made for.

Millions of jobs that never existed before, exist. Some positions are already taken - pundits, writers, musicians. But most are vacant, for now.

Sooner or later, millions of workers will stop doing things they hate and start doing things they love. Factory workers will become zany backyard scientists, green grocers will become eccentric modern artists, and checkout staff will become controversial economic historians.

Most won't be rich - they might even earn less than they do now. But they will exist. And, historically speaking, that is magical. They are no longer workers, they are doing their life's work.

The Weird World

A life's work might not make a lot of money, but it can make the world better. Science and art and history and a million things I can't think of, matter. After all, they're the things we love doing! And more people than ever will be doing them.

In the startup world, it’s tradition to talk about being contrarian. I used to live in the world where ‘contrarian’ venture capitalists $100m bought ten startup Series As, a negative rate of return, and a hard lesson. Now, I just want to live in the world where weirdos spend $100m on thousands of other weirdos doing what they love.

I've stopped worrying so much about how to change the world. I just hope you and I take the chance to do our life's work.

Thanks to Matt Clifford, Arnaud Schenk, and Ben Clifford for feedback on this.

Me, Matt & Alice (EF’s Founders), and Anne Marie

Why join? 

I was incredibly lucky to join EF in the first place. Never mind six years, I quit my first big company job after six weeks. And but for that professional organ rejection, I almost certainly wouldn’t have ended up here.

EF wasn’t a career move. Six years ago, EF was technically a not-for-profit, and its only employees were its founders: Matt and Alice. I became one of two unpaid interns, moving back home with the lack of promise endemic to philosophy graduates.

I joined because I was friends with Matt’s brother and because Matt said EF was something to do with McKinsey. It seemed like somewhere I could catch my breath while I applied for other things.

Why stay? 

It turned out EF is unique.

EF fulfils a unique mission. Years ago it seemed implausible, but today thousands of people realise their biggest ambitions because of EF. I can’t imagine working for a regular company anymore.

EF attracts unique people. Matt and Alice are still the most impressive founders I’ve met, and meeting founders has been my job. I am more grateful to them than any blog post can say. The EF Team and alumni have been my best friends, inspiration, and teachers.

EF offers unique opportunities. I joined age 22, with no relevant experience. At 24 I moved to Asia to create EF here. We’ve grown from 0 to 100 team members, from 0 to $150m raised, and from 30 to 800 founders funded each year. My learning curve couldn’t have been steeper.

EF has reached a point where it’s starting to pay off, and the company is going better than ever. There are a lot of reasons to stay.

Why leave?

In business, your resources change your strategy. And in life, your resources change your ambition.

I was lucky to find EF, and I made the most of it. I took the opportunity and put everything aside to make EF succeed. And, for some definition of success, it did. 

In a regular job, doing something for six years often just buys you another six. But making EF successful has given me resources I didn’t have before. For one, I am no longer desperate for money and can, terror free, spend a little time doing things that don’t make any.

More importantly, I used to worry about my future. Less so now. Partly this is because I’m confident of the futures I can create. But partly it’s because I’ve met people to create futures with. I met Anne Marie at EF and, because of her, my futures today are futures I couldn’t imagine before.

Your resources determine which futures are possible and, through EF, my resources have changed. So my ambitions have too.

What’s next?

When I was younger, I told people I wasn’t sure what I wanted to do. What I really meant was, I wasn’t sure I wanted to do anything I could.

But there are so many exciting lives I’d love to live. Here’s a long list. Now I can, I’m going to try and live a few more.

Thank you to the team, investors, and alumni who make EF so special. I’ll be forever grateful to have worked alongside you, and will always cheer for your success.

I’d love to keep in touch, so email me here, do something interesting with me here, and connect with me on all things social.

This is not an accident. It’s an inevitability. Not all Bad tech companies are successful, but almost all successful tech companies are Bad.

Successful tech companies are Bad because to become one, you have to take the Fastest Survivable Path.

The Idea Maze

The Idea Maze maps all your startup’s possible strategies. Every decision you make, deliberate or accidental, forks your path through the maze.

There are many good examples of the maze, but here’s Balaji Srinivasan’s original illustration:

Entering the maze isn’t hard — any idea gets you in — but finding your way through is. Different products solve the same problem, strategies have interdependencies, and it’s impossible to map entirely.

You can’t know in advance whether your path leads to success. If you could, successful paths would become overcompetitive and change the maze. There is an idea observer effect.

What you can know in advance, is that any startup that survives the maze found the Fastest Survivable Path.

Ex Ante

Since you can’t know up front which path will lead to success, without competition, you should simply pick the least risky path. If no-one else is navigating the maze, take as long as you need. Just don’t die.

But competition changes the maze.

If someone builds Spotify before you, you can’t build Spotify. Often paths are single use — owning exclusive rights to distribute music, for example.

A competitor taking a faster, riskier path can turn your safer path into a literal dead end. A definitely legal version of Airbnb might have worked, but the possibly illegal Airbnb took that path away.

With competition, the path with the highest probability of success is the path with the highest probability of succeeding first assuming your competitors pick the other paths.

Fastest Survivable Path

Faster paths are usually riskier, and being Bad is usually risky. You can operate in a legal grey area, you can ignore a toxic culture, you can figure out the business model later. And you may get through the maze faster. But you may die along the way.

With few competitors in the maze, it’s unlikely that high risk paths have a survivor. If a competitor takes a path with a 1% survival rate, they almost certainly won’t make it.

But, the more competition there is, the riskier the winning survivor’s path is likely to be. If 100 competitors take paths with 1% survival rates, most of the time one will make it.

Because so few startups succeed, promising ideas inevitably attract lots of competition. That’s why startups that succeed are Bad. A reckless startup will almost always win, it’s just hard to know in advance which.

Winning startups operate in grey areas until they’re white, with a culture just bad enough not to kill them, and figure out the business model later. They take the fastest, riskiest path that actually works.

They are the survivors of the Fastest Survivable Path.

Ex post

Almost every company that searches for the Fastest Survivable Path will die. But some won’t. And they’re the companies that beat the maze.

That’s why it didn’t matter that Airbnb operated illegally, that Revolut mistreated employees, or that Tesla made a perpetual loss. Or why Uber did all three and can still IPO. They took the Fastest Survivable Path.

Ex ante, the path with the best odds is the path with the highest probability of being first assuming your competitors pick the others. But ex post, the only path that matters is the fastest anyone survives. And that selects out companies that trade being Good for being fast.

Founders will forever accidentally build Theranos, Zenefits, and Napster. Because, for founders, being Good doesn’t pay if it’s not on the Fastest Survivable Path.

Thanks to the Entrepreneur First team for reading drafts of this. If you’d like to take the Fastest Survivable Path, start here.

The most common question founders ask at the start is “how much money should I raise?”. And even good founders get the answer wrong.

At Seed, the answer begins with your Minimal Viable Raise.

Minimal Viable Raise

Your Minimal Viable Raise (MVR) is the smallest amount of money that can get you to Series A metrics in 18 months.

Start with the milestones you need to raise a Series A. Work backwards from those milestones to an 18 month plan, and aggressively cut costs until you reach a number that feels too small. That’s your MVR.

If you’re scared, your MVR is probably right. The test is: if you were offered 10% less, could you revise your plan and still hit Series A? Yes? Not your MVR.

Often, founders try to raise more by asking for more. But, however much you want to raise, your MVR is how much you should tell investors you’re raising.

Your MVR gives you the best chance of raising just enough to make it. But it also gives you the best chance of raising way more.

Easier up than down

Empirically, it’s much easier to increase your round size later than to decrease it. On average, if they raise, EF companies raise 1.5x more than they initially ask for. Some have raised as much as 5x their MVR ask.

If you ask for more money than investors offer you, you’re in trouble. If you said you needed $1m to get to Series A, but you only end up with $500k in offers, the plan to Series A you convinced investors with has to be scrapped. You’ll have to convince the investors you convinced before that you can get to the same point with less money, and now they’ll be wondering why you couldn’t convince anyone else first time around.

But if you get offered as much as you ask for, your options often explode. 

Why you’ll get more than your MVR

Once you have an offer you might accept – your MVR or better– fundraising will often swing in your favour.

And your MVR drives the two most important factors for getting acceptable offers: who’s investing; and competition.

Who’s investing

Generally, investors with big funds need to invest more money per company. Even if your MVR is too small for them, they know they can persuade you to raise more than you’re asking for. After all, if you can do enough with a little, you can do a lot with a lot.

Generally, investors with smaller funds will be put off if you ask for a lot because they can’t fill your round or offer proportional valuations. But, if your small round becomes bigger because you’ve been offered more, it’s a such a good sign that you can often persuade them to stay in.

And since the percentage equity investors take stays roughly constant no matter your round size, your round size implies your valuation. Any investor who might invest at a valuation you’d accept, would invest at your MVR valuation. This is your most competitive possible price.

So, asking for more than your MVR often eliminates investors who might give you an acceptable offer. And the more investors who might give you an acceptable offer, the more ways your round can become competitive.

Competition

Having an acceptable offer is the strongest external signal to other investors that you might succeed. And so few startups succeed that investors must compete if they think you might. Competition is extremely powerful.

If your round becomes competitive, investors will compete by increasing their offer or improving their terms. Your original ask and valuation become irrelevant. Often, you can squeeze everyone in by increasing the round size while improving your valuation and terms.

You can do this because investors can’t afford not to invest your company if they think it’s great, but also can’t afford to dilute you too much by investing at a disproportionate valuation. If you raise double what you ask for, investors won’t dilute you twice as much because they need you to stay incentivised later on.

Paradoxically, starting with your MVR can make raising a bigger round easier than trying to raise a bigger round from the start.

Less is more

Start with your MVR. If you can’t raise your MVR, you can’t raise more. If you can raise your MVR, often you’ll raise more or raise on better terms. Even if you only raise your MVR, you’re still better off: if you ask for more and have to backtrack you might not raise at all.

The exception to MVR is when raising astronomical amounts is your startup’s competitive advantage. But if it is, you’re not reading this post on how to fundraise.

So, how much money should you raise? Start with your MVR.

Thanks to the EF team for reading drafts of this. If you’d like to raise a Seed round, start here.

“Over here, our cucumber water. Here, our collaboration space.

“Desk? No, not a desk. A hotdesk.”

I have arrived. Silicon Valley, 2014. My startup pilgrimage.

From Soylent drunk house parties with YC alumni, to pitching legends of the sport down Sand Hill Road, and cold brew catchups with handcuffed founders who “sometimes see Mark”, it’s all real.

And in my cucumber water, I find the secret of Silicon Valley. The secret is there isn’t one. The Valley is just a bunch of people with laptops, too.

What the hell?

If all you need’s a MacBook and an inclination, how come Silicon Valley's still special?

The Uncanny Valley

The things that seem to make the Valley work, don’t seem to make everywhere else work like the Valley.

Governments spend lots of money on looking like Silicon Valley.

And it looks good. Money can:

• build coworking spaces, incubators, and innovation labs;

• incentivise accelerators, angel investors, and VC funds;

• offer grants, loans, and schemes;

• create, protect, and spin out intellectual property;

• organise conferences, speakers, and parties;

• influence unit economics with rebates and kickbacks.

But no matter how things look, they’re never quite the same. Money can buy the consequences of great startups. It can’t start them.

Cause And Effect

This ecosystem building is a natural confusion. It’s not that ecosystems don’t matter. They do.

But ecosystems manipulate inputs. And manipulation, even done well, can’t get the right outputs without the right inputs.

No investor can fund the great startup that doesn’t exist.

Silicon Valley didn’t have an ecosystem in the beginning. It had great startups. And with enough great startups the ecosystem built itself.

Ecosystems are emergent properties. Trying to build the ecosystem without great startups is like trying to build Facebook without people.

That’s why money is useful once you have great startups, but almost useless when you don’t. Great startups don’t emerge to support ecosystems – ecosystems emerge to support great startups.

Feedback Loops

Once things get going, the ecosystem kicks in.

But an ecosystem isn’t coworking spaces or conferences. It’s a collective memory.

Since most startups fail, the hardest lessons to learn are taught by success. Almost everyone who’s experienced why Silicon Valley works is in Silicon Valley.

So, despite decades of competition and all the money in the world, there’s still no better ecosystem than the Valley.

The Secret

This begs the question – how do you start great startups?

Well, you start by getting great people to start them!

Silicon Valley is just a bunch of people with laptops, too. Its secret is who’s sitting behind the screen.

The Valley exists because world class people started tech companies there. This wasn’t a strategy – it had to happen somewhere, and happened to happen there first.

But because it did, it’s now the only place where the default option for world class people is to start or join tech companies. And you get out what you put in.

This is what makes the Valley special. With enough world class people, the improbable becomes inevitable.

So What’s An Ecosystem To Do?

If you’re targeting Silicon Valley, there’s a simple way to calculate your ammunition:

How many world class people you have

x

How likely they are to start a startup

To win the war, make more people more likely. The ecosystem can wait.

How many world class people you have

The first part of the equation is tough. World class people are by definition hard to make, so it takes a long time to make them.

And no matter how good you are at making them, even by accident the rest of the world will make more than you can. You need people to come to you.

Relax immigration for ambitious people, and aggressively (de)regulate promising areas for them to explore. It works.

Ambitious people are like weeds: they find a way. If you want more weeds, don’t worry about biology, just crack the pavement and watch them appear.

How likely they are to start startups

The second part of the equation is even tougher. Starting a startup is a weird thing to do.

It’s weird when it’s what you should do: if good ideas didn’t sound weird, we’d already be doing them.

But, outside Silicon Valley, startups sound even weirder than they should.

Successful cultures are successful for a reason – they push ambitious people towards what worked before. Culture is history’s way of telling you what to do.

Professional services in London, government in Singapore, startups in Silicon Valley. Once a path looked promising, ambitious people took it.

Soon, a promising path became the default choice. History picked a path, now culture keeps us on it.

But history doesn’t change, even when what matters does. If you want more world class people to start startups, outside Silicon Valley, you have to leave history behind.

You have to change the culture.

Safe House

Changing culture is hard, especially with startups.

Even when world class people start startups, most fail. To get results, you need lots to try. A few defectors aren’t enough.

It’s tempting to make up the numbers, but it’s dangerous to make startups a default path for normal people. Where world class people normally fail, normal people have almost no chance.

This isn’t good for anyone. Soon, normal people will be worse off, and world class people even less convinced.

So what do you do?

You create a safe house. In successful cultures, exceptional people are always waiting to rebel.

Alone, their rebellion fails. And individual failure looks especially bad. Bring people together, however, and individual failures are hidden by group successes.

You don’t have to do anything.

Your job is to legitimise their rebellion, mythologise their journey, and heroise their efforts. Become the default path, one safe house at a time.

That’s what we do at Entrepreneur First. Get hundreds of world class defectors in one place. They start startups together.

First a safe house, second a culture.

Thanks to the EF team for reading drafts of this. If you’re looking for a safe house, it’s here.

In head to head competition, more often than not, Valley startups still win. But if you’re a founder outside Silicon Valley, don’t be afraid.

You can do things they can’t.

Because it’s winning

Competing with Valley companies on their terms is hard. They can access more money, faster growth, and deeper experience.

You don’t want to try to build the best company from the second best place.

That’s how you get OneFineStay when you wanted Airbnb, Hailo when you wanted Uber, or Creative when you wanted Apple.

But, because it’s winning, Silicon Valley can’t win at everything. Winning creates opportunities for arbitrage.

So, how can you build a company Silicon Valley can’t?

I’ve got three suggestions.

1. Arbitrage talent

Talent heavy companies outside Silicon Valley have an advantage.

Because Silicon Valley won, it has a lot of winners. This makes it difficult for new startups to hire.

If you’re a talented person in Silicon Valley, you can probably get a meaningful, highly paid job at a tech company that’s already successful.

New startups can’t compete on meaning or pay so, even if they can convince you to join, they can’t convince you to stay. Especially in exciting areas, last generation’s winners monopolise talent.

But in ecosystems where there are no winners, what do talented people do?

Traditionally, they choose between meaningful work and financial opportunity — poor in academia vs bored in industry. And there’s the arbitrage.

Outside Silicon Valley, if you can offer meaningful work and financial opportunity, you can attract and keep talented people. And there are way more talented people outside Silicon Valley than in it.

Hence the explosion of British machine learning startups. The talent was already there – startups just started doing something with it.

DeepMind is a striking example. Could it have been built in Silicon Valley? Probably not.

Founded by two PhD grads and a philosophy dropout (no offence, they’re my investors), in a few years they managed to assemble the world’s best team of machine learning researchers and do groundbreaking work.

Ironically, you couldn’t have built DeepMind in Silicon Valley because everyone who could've built it was already at Google, getting paid too much not to.

Machine learning was Britain’s talent arbitrage. By definition, it won’t be the same elsewhere.

But, in other fields, talent is global and waiting.

2. Arbitrage business models

North America is a great market for startups. Compared with Europe and Asia, it’s relatively homogenous and therefore easier to scale through quickly.

But, because it’s easier, more has been scaled. In North America, fewer opportunities are left to win.

This makes Valley businesses narrower. And that changes their unit economics. You can use this against them.

Take Uber. In Asia, it has taken a beating.

At first, Uber tried to compete with startups like GoJek. But, after a while, even for Uber(!) the unit economics stopped adding up.

Why? Because Uber is a transportation company, and GoJek isn’t.

Transportation is GoJek’s customer acquisition strategy for internet banking, shopping, and services. Mature markets in North America, but still winnable in South East Asia.

This makes GoJek’s unit economics completely different to Uber’s. Uber’s customers are worth way more to GoJek, and GoJek’s infrastructure costs spread across verticals.

To compete, Uber would have to split itself in two. The GoJek model won’t work back home: the US already has internet banking, shopping, and services.

Valley companies can be beaten elsewhere precisely because Silicon Valley has been so successful back home.

There is an important, but uncomfortable, version of this.

The Valley, especially in venture capital, is infamously white and male. You can do something they wouldn’t by making products specifically for non white non males.

The arbitrage is in investors’ business models. VCs have a habit of funding things they understand. Success without diversity compounds into blind spots.

3. Arbitrage regulation

Nimble governments can make what’s possible, legal. And ambitious governments take regulatory risk.

When governments compete, regulation is their most powerful weapon.

Aggressive regulation was the foundation of Silicon Valley’s fiercest challenger: China.

But regulation doesn’t need to be aggressive to create opportunity. It just needs to be clear. In most places, lack of clarity stops cryptocurrency companies opening bank accounts.

Progressive, clear, or sandboxed regulation can make an inevitable future arrive faster. Self driving car (de)regulation in Singapore lured startups like nuTonomy from the US. Challenger bank licenses in the UK created high growth startups like Monzo.

Sometimes, the US does the dirty work for you. ITAR practically guarantees the Valley won’t win the startup space race.

Whether its immigration, gross ideas, or progressive regulation, ambitious governments can quite literally enforce your arbitrage by law.

Culture

Arbitrages are vacuums for founders to fill. But they don’t last long. Pretty soon, vacuums fill up.

So how do you build an ecosystem, not a company?

In the long run, the one thing that lasts is culture. As our investor, Reid Hoffman, says “Silicon Valley is a mindset, not a location.”

Until the most talented people in a society decide that they’re going to start and join startups by default, before trying traditional routes, world class companies won’t get built.

This means sacrifice. Would be bankers in London, or government officials in Singapore, won’t be.

And just a few is not enough. The more unusual the sacrifice, the harder it is to make. Without momentum, there is no cultural shift.

Momentum is what we’re building at Entrepreneur First.

We find extraordinary people, and bring them into one place. 50, 100 at a time. Small numbers, but big enough to make an extraordinary decision feel ordinary. They start startups together.

With momentum, small numbers become large. EF has 792 alumni now. We’re adding 550 this year.

At some point, it stops being momentum, and starts becoming a culture.

And at that point who do you bet on? Silicon Valley, or the whole of the world?

Thanks to the EF team and Misrab Faizullah-Khan for reading drafts of this. To place your bet, start here.

“Have you ever wondered what humans taste like?

“What if you could buy human meat, lab grown? No humans harmed.

“What about a specific human’s meat? Ronaldo steak. Beyonce bacon. What do you taste like?

“Forget lab grown beef burgers. We’re going to sell lab grown human meat from specific humans’ DNA.”

This is a conversation I had with a founder at Entrepreneur First.

My first response was that their idea was gross. But, the more I think about it, the more interesting gross ideas become.

Gross ideas are open secrets about the future.

Moral Tension

Eating someone’s meat sounds gross the first time you hear it. But it captures your imagination.

It seems wrong, but it’s not obvious why. This moral tension is what makes gross ideas interesting.

The good: No humans are harmed. Maybe there are taste, happiness, or health benefits. Maybe it’ll stop wannabe cannibals from becoming criminals.

The bad: It seems gross. Maybe you need consent from the people being ‘eaten’. Maybe it’s a slippery slope to criminal cannibalism.

I don’t know the answer. But before we can agree what’s right and wrong, gross ideas get built.

New realities

Things that used to be organic are becoming artificial. And this is what makes gross ideas possible.

Artificial experiences mean you can, for the first time, separate experience and reality. In all history, to actually experience something it had to actually happen.

You’ve always had to trade reality for an experience. Now, you don’t. Soon, gross ideas will be everywhere.

As we separate reality from experience, we separate our actions from intent from consequences. And it’s hard to know whether it’s the action, the intent, or the consequences that matter.

This means we’re going to face more and more jarring moral choices:

Virtual reality makes things seem real that aren’t. What if your video game becomes a murder simulator?

Computers can learn how you communicate. What if fake videos of you are indistinguishable from real videos?

Genes can be edited and created. What if you can create new animals or change your own biology?

Generative models can create content that’s never existed. What if you artificially generate the worst kinds of porn?

Brain stimulation can make you feel whatever you want. What if you want to be tortured?

It can be built. And if it can be built, someone will build it.

Moral Dilemmas

Just because something can be built, doesn’t mean it should be built.

Gross ideas mean moral tension. Work on something gross, and you might create something you shouldn’t.

But here’s the thing: moral tension means people feel something. And, awkwardly, this resolves one of the hardest problems startups face.

Finding moral tension means you’ve found an idea people care about. At the very least, you’ll get a reaction. And even a bad reaction is promising — most startups die because nobody cares at all.

So if you’re willing to risk doing something you shouldn’t, you have a head start. Moral dilemmas are open secrets about what might be popular in the future if morality changes.

And morality does change.

Racial and gender equality, slavery and abolition, LGBT rights. Contraceptives, online dating, abortion. Genetic modification, battery farming, fake meat. Immunisation, cloning, stem cell therapy. Marijuana, prohibition, vaping. Pornography, sex toys, prostitution. Weapons, data privacy, assisted suicide.

At some point, to some people, each of these ideas seemed at least gross and at most immoral. A few still might.

Right or wrong, the opportunity is there. Moral tension guarantees people will care. It is the precondition and consequence of many big businesses.

I’m not suggesting it’s good to be immoral. I’m suggesting money gets made advancing both sides of a debate.

Technology is separating reality from experience, our actions from intent from consequences. And that means there are a lot more debates coming.

I can’t predict the future, but I do know it’s going to be gross.

Thanks to Ben Clifford and the EF Team for reading drafts of this. If you’d like to do something gross — but good — start here.

The curious idea is this: technology can eradicate poverty — not by making poor people less poor, but by making it less valuable to be rich.

Instead of trying to eradicate poverty through redistributing resources, we can accept financial inequality in exchange for experience equality.

Financial Inequality

Technology creates financial inequality.

It spreads quickly and cheaply, which creates monopolies. And monopolies create financial inequality.

If products can be quickly and cheaply distributed around the world, most people will use the best product over the second best. That’s why the same mobile apps are often popular everywhere.

It’s also why the winners often end up being orders of magnitude better off than their competitors.

As technology spreads, we should expect more monopolies and more financial inequality.

Experience Equality

Here’s the catch: technology makes its creators rich, but treats its users the same.

240,000,000 Indians use Facebook.

And while Mark Zuckerberg’s bank account contains 10,000,000 times more money than the average Indian’s, his Facebook account works more or less identically.

Quick and cheap distribution goes both ways — it creates financial inequality and experience equality. You can’t pay for a better Facebook experience.

Access doesn’t always start off the same, but technology gets quicker and cheaper to distribute over time. Eventually, it doesn’t matter how rich you are: everyone gets the same experience.

As technology spreads, access to experiences will become more equal.

Companies are incentivised to offer a product if it makes more than it costs. And technology ends up not costing much once you’ve built it.

So, in the end, you charge people whatever they can pay.

And costs scale better than you’d think, even when humans are involved. Uber works in India about as well as it works in San Francisco. People just pay and get paid less.

When tech works, the money is different but the experience is the same.

Artificial experiences

If technology leads to financial inequality but experience equality, so what?

Facebook and Uber won’t get rid of poverty.

But they’re just the beginning of a shift from ‘real’ to digital, artificial experiences. Few things, in principle, can’t be delivered through technology.

Is it possible that, one day, even the richest person would prefer to live in an artificial VR world? To have their senses tricked? To hook themselves up to the experience machine?

Absolutely.

This is a pivotal historical moment. If even the richest person prefers an artificial experience over all that their money can buy, won’t the poorest?

Once even the rich prefer artificial experiences, experience equality becomes possible.

If this technology will ever exist for the richest, one day the poorest will have access to it. Technology gets cheaper and quicker to distribute over time.

We will all have equal access to experiences, whether or not the experiences we choose are equal.

When we prefer artificial experiences, and everyone has access to them, the world will be much less financially equal. But maybe even the poorest won’t experience it that way.

After all, we’ll be charged whatever we can pay. And at that point, why bother getting rich?

Thanks to the EF Team, Ben Clifford, Luke Theobald, and Rob Wiblin for reading drafts of this. If you want to make the world more equal, start here.

Yesterday, I spoke with an Entrepreneur First startup about to start fundraising. This revealed an unfair truth.

They’re good, and this is bad.

Investors don’t invest in good companies. To win, you don’t need to be good. You need to be bold.

This is a story about Tortoises, Hares, and why Uber wins.

Tortoises

Tired of being ridiculed for being slow, the Tortoise challenges the Hare to a race.

The Hare speeds out of sight and, confident of winning, has a nap in the middle of the race. But when the Hare wakes up, his competitor has already reached the finish line. Slow and steady wins the race.

The moral of Aesop’s Fable is don’t go fast if it means you make mistakes.

Good companies are Tortoises. They cover all bases. They address their weaknesses. They avoid mistakes.

There’s nothing wrong with Tortoises, but the best investors never invest in them. They invest in Hares.

Hares often make mistakes. They don’t have all the answers. They have glaring weaknesses. But, whichever way they run, they run fast. Hares aren’t good, but they are bold.

In startups, Tortoises never win the race. And that’s because, in any race worth running, there will be 99 Hares.

99 Hares

In Aesop’s Fable, there’s only one Hare. But startups don’t work like that.

Startups are competing, always. In races worth running, other people run against you. Even if you can’t see your competition, you have to believe they exist, and work like they’re sitting in front of you.

Technology means everyone can access everything, so only the best will do. Second is as good as last — startups are winner takes all.

In competitive winner takes all games, the boldest correct strategy wins. If there are 99 Hares, 98 might make a fatal mistake, but the 99th will find the fastest path, mistakes included.

Most bold strategies fail, but almost all winning strategies are bold.

This is why the best investors don’t like good companies. Good companies are the Tortoise, but the 99th Hare wins the race.

An investor might not be able to explain why your Tortoise will lose but, if you aren’t bold enough, they can still know you won’t win.

Startups can’t spend too much time on their problems — even ones that might turn out to be fatal. You might lose on your weaknesses, but you can only win on your strengths.

Strengths

It’s hard to be bold. It means ignoring problems to focus on a single extreme strength.

The boldest strategies capitalise on one strength to create distance in the race. You make it harder for competitors to catch you, instead of fixing problems you have along the way.

Even the best startups have problems. Management, HR, Sales, Financing — from big to small, companies feel broken inside because these things feel hard even when you’re getting them right.

But what separates winners from losers is how much distance they can create by what they do best, not all the things they do less badly. After all, everything that’s not fatal is fixable after the fact.

Extreme strengths can compensate for horrible problems. Uber’s problems are among the worst: awful culture; haemorrhaging cash; and legal disputes.

But Uber was bold, if not good. They had one extreme strength — growth. And growth afforded Uber resources to withstand the problems they created along the way.

In the beginning, a single extreme strength: an ambitious technology; a pedigreed team; or a compelling vision, is often enough to raise investment. In the end, sometimes growth trumps all.

Like it or not, the winners often look a lot like the 99th Hare.

Thanks to the EF team for reading drafts of this. If you’d like to be the 99th Hare, start here.

Conventional wisdom suggests you should start a company with one of your friends. The relationship between cofounders is crucial. Everyone knows this.

Except it’s false.

Friendship

Ok, so most good companies were started by friends. But why?

It’s not because being friends improves your odds of success. It’s because the kind of people who start companies with strangers aren’t the kind of people who start good companies.

This is obvious once you think about it. Smart people have lots of options. So, if a stranger thinks starting a company with you is their best option, either they aren’t very smart, or they make bad choices.

Picking a cofounder from your friends just filters out dumb people with bad judgement. And this is important: dumb people with bad judgement tend not to make good cofounders.

That’s why good companies are started by friends: only bad founders start companies with strangers.

If pretty much every company started by strangers is bad, most good companies will be started by friends. But that doesn’t mean friendship is what makes good companies good.

It turns out what we mean by ‘start a company with your friend’ is ‘don’t start a company with strangers’.

Dating

Finding a cofounder today is a lot like dating in the ‘90s. Back then, ‘don’t date someone you meet online’ would have been good advice. Why?

Again, it’s obvious once you think about it. Attractive people have lots of options. And if you have options, you’re probably not an early adopter of dating websites.

The kind of people looking for partners online in 1999 weren’t the kind of people you wanted to date. So, in the early days of the internet, most attractive couples still met offline. ‘Don’tdate someone you meet online’ was good advice.

But this is horrible advice in 2017. It turns out that meeting offline isn’t intrinsically better. Today, hundreds of millions of people meet online, and it’s great.

The pool of people you can choose from is huge, awesome, and willing. Relationships that couldn’t happen offline, happen. As surprising as it once seemed, these relationships are often extraordinary. Dating has improved.

Meeting offline isn’t what makes relationships good, just like being friends isn’t what makes companies good. What matters is whether you’re doing it with the right person.

Outcomes

In startups, outcomes trump friendship. It doesn’t actually matter if you’re friends at all.

If it goes well, starting a company is a unique experience. Succeeding is so hard that you can’t help but appreciate your cofounder if you do. They might start off your worst enemy but, if they’re the reason you succeed, they won’t stay that way.

The reverse is also true. If it goes badly, starting a company is horrible. Failing is so painful that you can’t help but resent your cofounder if you do. They might start off your best friend but, if you fail, they won’t stay that way.

The right person is whoever you can get farthest with, fastest. And that person might not be your friend, yet.

I’ve bet my career on this. At Entrepreneur First, we help you find a cofounder. We curate a pool of amazing people ready to start something too, and make it easy to try working with them. If it doesn’t work out, you just find someone new from the pool and start over.

I’ve seen thousands of teams start this way. Some succeeded, most failed. One surprising conclusion is that teams that are friends before they start are no more likely to succeed than teams that aren’t.

The hardest situations at EF are when good people get stuck in bad teams because they’re friends. Both the startup and the friendship end badly. Good teams become friends becausethey’re good. Outcomes trump friendship.

Weird

It still seems weird to start a company with someone you met at EF, just like it used to seem weird to date someone you met online.

But now online dating exists, I can’t believe people try to find a partner, drunk, in clubs and bars. And now EF exists, I can’t believe people try to find a cofounder, aimlessly, in universities and meetup groups.

Turns out, there is a better way. Bring all the good people into one place. Seems obvious in hindsight. And now we’re here, imagine going back to how things were before.

This has some interesting long run effects. Over time, what’s weird will change. Once there are good ways of meeting and filtering through people, partnering with someone the old way stops making sense.

Why would you expect to meet your perfect partner randomly, ‘in real life’? It’s absurd to think that the best possible match is someone you just happened to have bumped into already.

Starting a startup with a stranger used to mean you didn’t have many options. But, in the long run, starting a startup with your friend will be the last resort.

So, should you start a company with your friend? Well, only if you have to.

Thanks to the EF team for reading drafts of this. If you’d like to start a company with a stranger, start here.

On Friday night I had an emotional conversation. The founder I was speaking with felt like quitting. They’re afraid they don’t have what it takes.

But they’re wrong.

They have what it takes. I know because I’ve had an identical conversation with almost every founder I’ve worked with.

The problem is perspective.

Frankenstein’s monster

Starting a startup is like assembling Frankenstein’s monster.

In the beginning, you have literally nothing. It’s a startup — that’s the point. No customers, no product, no team. All you have is the hope that it’ll work out eventually.

Limb by limb you try to breathe life into the beast, praying each new piece works as intended. More often than not, it doesn’t. Customers aren’t keen after all, the product is harder than you thought, and no-one wants to join your ‘project’.

Even when something finally works, the result is never what you expect. You’ve created a monster. Customers pay buttons and expect brilliance, the product is being rewritten again, and your team is simultaneously demanding and incompetent.

A startup going well is almost as hard as one going badly. Inside, every startup I’ve ever seen is a mess. Body parts are not where they should be.

If you want to know how it feels to be chased by an 8 foot monster, the best way is to start a company.

Insecure

Even the best founders become insecure.

How could it be any other way? You’re doing something new. Not just new to you, but new to the world. You couldn’t possibly know what you’re doing.

You can get advice, of course. But memorising cookbooks won’t make you Gordon Ramsey. You only learn by doing.

Besides, often there’s no cookbook. If it’s a good idea, either no-one knows or no-one knows how: otherwise, someone would have done it already.

If you’re feeling insecure, you’re doing it right. The best founders know they’re close to failure, all the time.

You might be smart and your odds might be different to most. But if you’re smart, you know what can go wrong. You’re creating a monster, and you should be terrified by every limb you attach.

Half truth superheroes

This puts founders in an unusual predicament. You need to convince investors to invest, customers to buy, and employees to work.

But how do you get someone to invest in a wretched monster? To buy a product that doesn’t exist? To work for an insecure amateur?

The truth is, you lie.

You know it’s a monster. You know you don’t know. You see the end is near. But they don’t, so don’t bring it up.

Tell them what could be. Tell them your best guess. Tell them what keeps you going. Then keep going. This is what successful founders do: whatever it takes to succeed.

This isn’t misleading, it’s leadership. Stories in the press, stories at events, stories over coffee — not always a lie, but never the whole truth. What you say is whatever’s most useful.

Half truths make founders seem like superheroes. And we like superheroes.

Perspective

Half truths don’t make success easier. They just make it possible.

I used to think hard things came easy to certain people. Now I realise hard things are just, well, hard.

Your perspective makes a difference.

I learned this from my Dad. Growing up, I saw him work hard. He worked endlessly to escape his poor upbringing. Years of 5am starts, financial problems, and personal anxiety took it’s toll: his body started breaking; his friends fell away; and his marriage was jeopardised.

My Dad worked hard for 40 years. In the worst moments, we were scared he didn’t have what it took. I know he almost quit. Almost.

Now, he is a great man. Earlier this year, he sold the company he founded and retired. Ask his team, he’s a confident, fearless leader. Ask his investors, they always knew he could do it. Ask the press, he’s a big company CEO.

He has what it takes, after all.

When I meet a founder who seems like a superhero, I think of my Dad. When I meet a founder who isn’t, I think of my Dad. Now, I know the whole truth.

Founders aren’t superheroes.

I’ll bet even Elon Musk isn’t much like Elon Musk. Just ask his kids.

Thanks to my Dad and the EF team for reading drafts of this. If you’d like to create a monster, start here.