You are working on Apollo 11 mission,

> in 1 year pocket calculator, in 3 years version control, in 10 years UNIX, Bash, SQL, Excel, in 20 years Python

... you do computations by hand, you keep hand written records and charts, you code highly resilient asynchronous systems in assembly without version control

> in 3 years knowledge base, in 20 years website and arXiv, in 30 years Wikipedia, in 40 years StackOverflow

... to get the information you need, you read the printed manuscripts or find someone who read them before you

> in 4 years mobile phone, Ethernet, email, in 20 years IRC, in 40 years Slack

... you share data in printouts by fax, you do all meetings in person or by landline

It worked. We put three man on the Moon and get them back, followed by nine more — the goal so ambitious it has never been accomplished again. What do we do with all the marvelous tech available today?

foreign language

There is something special about medium where you don’t speak language. As you move in, a brief friction is followed by quiet. At its busiest, you remain in your personal space. Your thoughts can reach further, stay sharper.

Ads don’t work on you anymore—you don’t get distracted on them as much. It is surprising how much power they loose without appropriate textual or verbal message. Certainly, vivid imagery still provoke emotions, but, without catchy message, now it is up to you to come up with interpretation. Ads are notoriously crafted to stick in your mind and alter the way you think. They don’t make you smarter. If anything, they plant into your head agenda from people whose intentions go hardly beyond getting your money, by any means. The dirtier the intentions—the stronger ads become. But now, that you break free, wouldn’t you want extra capacity to focus on what you think is really important?

You don’t get involved into surrounding chatter. Of course, it is good to peek sometimes into what other people are up to, but it is your choice now whether to focus on that. Their life gets private and you don’t get any extra cognitive or emotional load. You get an On–Off switch; you get a good sense about people from the way they talk; you observe better; non-verbal signals bolden. And you certainly get the same feeling from the place. Busy Philz in San Francisco, BlueBottle in New York, CoffeeNero in London, CoffeeBean in Seoul—all keep their ambient charm. You are going to get the same vibes wherever you go.

But wouldn't be communication hard? No, it's rather the opposite—it becomes more efficient. Most people around you don’t speak English well, although, given enough opportunity, they understand it perfectly. You start talking more clearly, carefully choosing the right words to convey the exact message in shortest way possible. You strip it to the essence. People who do not get used to this mode of communication try to tell way more than they actually need to, it overwhelms and leads to frustration. You give people more space and become better listener. You may not eloquently converse about ingredients of some dish at restaurant, but you would also never face anything unresolvable. After all, people still know what thumbs-up means. It takes only one good word at the right time to make the day brighter. Make it count.

control your data

1. Back-up your social data. Download everything from your networks and get a private paid blog instead. Be in charge on what happens to your posts and connections. You don't want to be one day locked out of your social circle due to some algorithm blocking your account with no way getting it back. 

2. Use channels that you have control over. Don't let someone nudge you on whom you should connect to and how. Companies optimize for their own objectives, make sure their true agenda aligns with yours. Better yet, control channels yourself. Some good channels are Twitter, Zoom, Signal or even vanilla email.

3. Get control over your news channel, get RSS or Feedly. Choose how you get news yourself, be that in chronological or other way. Choose sources yourself. If you get suggestions, keep the only source of suggestion that you decide is worthy. Control what you see, how you see and when you see.

4. Keep a single cloud service, purge everything else. Use Apple Image Capture to purge Photo library on iPhone. It is easier to manage and reduces attack surface. Use cloud service that allows you to download and delete all your data back in bulk, some providers force you to do it file-by-file. Don't upload to cloud raw data, pack it and encrypt it first by yourself. Don't use cloud as a transport between devices, connect devices directly, it is fast and secure. One day cloud provider can lock you out, get offline back-up. External SSD or HDD is perfect. Apple Photos works well with stand-alone offline libraries and has all the regular nice features. You can also use shared albums without uploading everything. Back-up photos, videos, music and iPhone settings separately, it makes back-up fast and consistent. Apple Photos Import has nice option for "delete after import". Remove duplicated photos and the ones you don't care about, the less data you have to carry around the better. Don't do hoarding of things "you like" or "will check in the future". Chances are, you will not check that at all, it wastes your time and space.

5. Harden your OS.

6. Use email client that respects privacy and doesn't serve ads. Apple's Mail is good option. It has rules and compact interface that is one-to-one match to gmail, before they switched to Inbox. Turn it on at "Mail > Preferences > Viewing > classic layout".

7. Get to know "private mode", Privacy Badger, HTTPS Everywhere, uBlockOrigin. In some cases, you might want to consider Tor and Tails.

8. Get MFA for key services. Get Google Authenticator in case you change phone number. And get YubiKey in case you lose your phone altogether.

9. Use only apps you trust. Check periodically what they are doing in Activity Monitor or even better - Little Snitch. If your music streaming app is downloading and sending tens of MBs per hour of data even if you don't get new songs, there is something wrong with it. Get to know VLC, Transmission, Firefox.

10. If you use CLI, know what you are running. Open source does not automatically mean secure. Use your own judgement™️.

11. Get self-hosted VPN. Make it a default when you connect to the internet. With few exceptions, almost nearly all web is accessible just fine with it. Your network speed may actually improve.

UPD: 2019-07-06

12. Disable loadable content in your email client. 3rd parties can track time, email client (and thus platform), IP address, geo information, fingerprint from your emails. And whether you opened email and how many times.


To begin with, if you driving a car in urban area, you are contributing to traffic congestion. Cars are highly inefficient, they occupy uneven space per person. Think of how many people can be packed in a bus in a rush hour? Twenty? Thirty? What if each of them had four-five meters long metal box? Add to this two meters of gaps between the cars. Why they are taking all that space? Cars don’t accelerate immediately, rather there is a delay spreading in a chain from car to car. Even worse in road junctions – time to change roads comes in fixed time turns. After four minutes it can be either forty people changing their street to continue their journey, or a single one. Why their time is more valuable? We all pay for that nice expensive road you are using as your own to place your personal, more comfortable, space on a wheels, it is only fair if you pay for it appropriately, and not to oligarchs of car-making monopolies, but to people right in front of you, sitting in a bus or an ambulance, whom you delay.

It is weird how not-weird cars are. Did you notice that all cars looks the same? How many cars have you seen with two lights on the sides and one in the middle? How many have the whole horizontal space with lights? How many lights have triangle shape? Why do you always see lightbulbs inside? It does not matter the size, price or type of a vehicle. They all look like some kind of animals. That design is intentional – if people see something big alien moving fast towards them, they will freak out. Did you notice that most of the cars have their windows tinted? You can’t see a person behind a wheel, this completes the image of the car in your mind as a single entity. The cultural footprint of this phenomena is so strong that it is not weird at all for adults and kids alike to enjoy countless movies about cars getting their own will or obtaining humanoid forms. Just look at what advertisers anchor cars with: bulls, horses, jaguars. In reality though, there is always some adult sitting inside, trying to relax from not hitting anybody in search of parking, all while being totally lost from the fact that he is inside of an overpriced metal box that looks like some futuristic tiger.

Now, putting these pricey fetishes aside, let’s consider one important implication – once you behind a tinted window, you feel separated from what is going on outside. After all, nobody can see or hear you. You are becoming anonymous, and, as with any anonymous environment, people becoming the worst of themselves. You can be as big of a jerk as you like when nobody can see you or hold you accountable. This becomes a problem. Think about a pedestrian crossing: these are protected by law designated areas for pedestrians to have priority of crossing the road. They are designed for convenience and safety of walking people. Now, imagine you are crossing the street but all of a sudden there is a huge metal box – in shape of futuristic tiger – rushing at you at 30+ km per hour; flashing bright light; loudly honk. If it hits you, you get damaged pretty badly, maybe even permanently. Your future life is at risk. But for person inside though – unlikely, besides prosecution, if somebody finds him. And even then they may get away with minimal penalties, perhaps taking away their driving license, temporarily. Your unconscious knows this too, and after it does the math, you are likely to stop and let the car pass. Now, the driver can clearly see the pedestrian crossing and knows that you should come first, yet he can just speed-up and get away with it. Nobody can honk back. Nobody will see your face or take a photo of your driver license. Even if they did, what they gonna do with it? I can only imagine what sick feedback loop it reinforces in a driver’s mind. Personal cars is a risk and accountability transfer machines poisoning city life. So what can you do?

There is one common form of action I see people taking – intentionally ignore drivers; look at the smartphone instead of the road. Don’t do that, it is dangerous and makes you look reckless. For once, whenever you feel safe on pedestrian crossing you can just go first and make car stop. There are many factors that will make it scary for you, but try to reason with logic. Cars do regular inspections and are constructed to stop quickly. Nearby to pedestrian crossings there is always maximum speed limit that is just enough for vehicle to stop, unless they are speeding. In most countries, pedestrians have priority protected by law. As long as they are a bit far and can see you – go. And while passing by, make sure to look at the driver, not the front of the car, to place where driver sits – “I see you what you did here”. For evolutionary reasons, human perception got very sensitive to anything that looks at you. However brief, even one-way eye contact breaks sense of anonymity and wakes up the conscious. One word of caution, watch-out cars that are accelerating instead of stopping. There are certain drivers who lost their mind. Try to report them to police. Even simple email can be of use. That person can be just as insane away of driving wheel as while using it. Maybe, one day your report will help investigation to fill the gaps in a more serious case. Of course, all above works if people have respect for the rule of law. If you leave in U.S., maybe you should stay away from the roads. If a person at the wheel is mentally ill or criminal with a gun, meaning they are not afraid of law due to their conditions, then, perhaps, it is wiser to stay away from that car. If that is common in place where you live, then my question for you – what are you still doing there?

But is there a better way to move around? Think about pooling money with other people to buy one big vehicle, so that you can ride with them and utilize public transport infrastructure efficiently. You setup good coverage so that you don’t have to walk too far or wait too long for vehicle to arrive. You appoint dedicated people to drive them all the time, so none of you have to worry about parking on maintenance. Due to high volume you can utilize economy of scale, meaning parts, fuel and associated costs are minimized. Sounds crazy, I know, but hold on, you can also make government to subsidize it and take care of all the operations. One step further, you move roads themselves underground, so that people never have to cross the roads and vehicles have never to wait, even in the rush hour in downtown! And why not making all of them electric by default? I am sure someone in Silicon Valley will re-invent that at some point. The net amount of monetary and time waste that comes with private transport in urban areas is incredible: overpriced vehicles; maintenance of vehicles and infrastructure; fuel. I would not be surprised if only excess of fuel was enough to power extra space program to the Moon or even Mars once a year. However, effectiveness of public transport holds on strong culture of sharing: that people are okay with sharing their space with others, at least for a short-while; that they can be civil and respectful to each other for 40 min; that they feel safe and comfortable in public spaces. Technology is not a bottleneck, it permits good transport for many decades already. You don’t need an AI for that, just a certain level of public trust and manners. This certainly does not hold in U.S., place of the worst public transport to date.

Who is to blame? The U.S. government, who is a corporate sell-out, and public, who can’t find a backbone to fight back. It gives me a bitter feeling to see San Francisco trams used primarily as a museum exhibit in sightseeing tour around this beautiful city. Similar case is in London, where Underground and, to lesser extent, two-deck busses are both museum and transport at the same time. Only example of transport that is working in the Bay Area is CalTrain – old train that goes once an hour (in Asian metropolis that would be once in 10 min) and has single line (not even a network). Sometimes, it is just as packed as trains in famous rush hour in Tokyo in the old days. In the Bay Area pro-car lobbying and trusts killed wide transportation network. If it was build, there would not be a housing crisis. Rather, new urban centers would have emerged and people would be traveling all around the Bay for work and leisure any time. Think of vibrant nightlife of London, but on the hills of San Francisco. It makes one wonder how different Bay Area would look like and what kinds of innovation people would come up here. If you live here today, you live in a strange version of the past, but I wish one day to see this area progress and bloom.


In 2018 I reached 364km across 48 runs, year before that it was 200km. Why do I do that?

It is simple. It is a rare distraction-free time when you have space to collect your thoughts, reflect, plan, think. Quiet, slowly changing environment of a city and background chatter of people passing by has a soothing effect. Or, if you prefer, you can occupy yourself with something more relaxing. Narration in audio books has never been so vivid. Tunes in music have never sounded so well. Things happen all the time. With all the busyness, leave an hour a week for yourself.

If you run in a new place, it is also a good opportunity to connect to the place around you. What it is like in early morning or late night? Life is busy. When you grow up, apart of shopping or commute to work, there aren’t many reasons left to explore area where you live. Running gives you this occasion to see what the place is really like beyond your house or insides of a bus. It is refreshing, once in a while, to step-out and look around.

books of 2018

Another great year. Finished 34 out of 22 planned books. Read a couple I had in back of my head for while and made few great discoveries along the way.

> What is your top pick, Nikolay?
> This year, I would recommend everybody “It does not have to be crazy at work”.

Powerful. Chief talent officer from Netflix shares fresh look at what is important to people at tech company. Recently, it become a cliche that tech company has fancy campuses with chocolate fountains, free food and champagne. Unnecessary perks, weird nudges and intrusion into work and after-work life is a wasteful corporate enterprise that diminishes agency of people working at the company. Instead, companies should focus on what they do the best — plain work. It is time for companies to cut the bullshit, think about what really matters and take people seriously. It was also great to follow story of Netflix from the front row.

Packing to Mars. Going to space is the most grandiose journey anyone can ever embark upon. But what it actually feels like? This book is full of funny and sometimes gross recollections from real witnesses. Here, on Earth, we are humans, and out there, in outer space, we are humans too.

Brotopia. One day, at popular computer science lecture at my alma mater, I noticed that there are 100+ guys and only 2 girls. The sheer scale of imbalance hit me. No matter what is your perception of expected imbalance this is way too extreme. To get some background on what lead to this, I looked into historical account on gender imbalance in U.S. in this field provided by this book.

The lean startup. Famous philosophy of rapid experimentation and growth for building new products and companies. Keep hand on a pulse of feedback from customers. Use tight build-measure-learn loop as an engine of your growth.

A study in scarlet. Beautiful novel from the world of Sherlock Holmes. audiobook 

The ethics of influence. Started this book a while ago and finally finished it this year. Now that we know about nudges, what is the moral grounds of government, or other authority like big corporation, using nudges for their own means? For example, when you walk into a food hall, what products you do see first and which require extra effort? Is it a healthy, ecologically friendly, or domestic option of the same thing? Do you even need it? Key takeaway, if we have choices anyways then why don't someone optimise choice-structure for social good? Another learning is that there are different types of nudges, some are quite subtle. Knowledge on how they work may not grant you immunity, yet the hope is it will tip the scales of power back into your favour.

Skin in the Game. Taleb covers here many interesting mechanisms that arise in case of uncertainty. One is responsibility and how it shapes social structures. Imagine you are seeing a doctor, do his incentives align with yours? Another one that stands out is difference of risk-sharing and risk-transferring. How come some customs and traditions spread and survive and others fade away. Lastly, if you don’t have real stake at what you have to say, perhaps we should not listen to you. And on the other hand, if you survived so far by making your own decisions, then there is some value to your opinion. Of course, that is very crude way to put it. As always, Taleb does extraordinary job articulating ideas, unconventional and profound.

Dreams from my father. Stories from formative years of Barack Obama, written by himself. Here I could follow his story with more detail and no restrictive format of political campaign speeches. The story of his childhood, race, African and American communities, in short his “audacity of hope”. audiobook narrated by Barack himself

Nocturnes. A couple of elegant and subtle novels about nightfall and music. paperback from SF SOMA 🌉

Competing against luck. Keep meaningful metrics and keep shipping. Pretty much same ideas as you would find somewhere else, Lean Startup for example, but less elaborate and in non-remarkable style. This book was a miss this year. audiobook

Structure of Scientific Revolutions. Monumental work on philosophy of science. If you are unfamiliar, in short, it is concerned with what is scientific knowledge and how it evolves. Now I realise, I got incomplete and at times wrong picture on this work when I was studying this topic back in undergrad. This work is very elaborate and acknowledges commonly referenced shortcomings, such as continuity. It was also great to learn how he blends research community and institutions that develop scientific knowledge into his framework. 

Three body problem. Famous science fiction from China. Very entertaining read.

The dark forest. Second part of “Three body problem” series. All the sci-fiction concepts you have heard mixed-up together. Story gradually losses its coherence leading to a complete mess in the end. The very ending was quite witty, but it takes unreasonably long to get there.

Machine, Platform, Crowd. Outlook on trends in tech industry. If you work in tech, you can probably add to this book, but unlikely other way around. audiobook 

Steering the craft. A crash course on writing and story telling. One thing that pops-out right away, keep in mind a persona who is telling the story. There are multiple kinds of voices that you can choose: omniscient observer, non-omniscient 3rd person, witness of story, main character, and others. This helps you to move reader closer to characters in your story. There is also a collection of other useful practical tips.

Bad Blood. Apparently, you could mislead investors, public and government for almost a decade, got a billions in founding without working product and cause physical harm to people. Thrilling story.

Principles. Life and work advices from founder of notorious Bridgewater hedge found. I am not found of the style he has chosen for this book, but looking back more of then not I agree with his advices. paperback

Gentleman. Short and intense novel about a guy who mistakenly sold his wife to devil. Narration gives you vivid pictures, almost like you are watching a play in theatre. paperback from SF SOMA 🌉

Bladerunner. Classic. I was surprised that movies completely missed a couple of elaborate dimensions to this story. That being said, both movies, and book have strong plot and convey unique and rich futuristic mood. paperback from SF SOMA 🌉

All the light we cannot see. Story about german boy and french girl in WWII. Very interesting style. paperback from SF SOMA 🌉

Factfullness. There are many metrics backed by solid facts that world is much better now than before and much better than you are thinking it is. Similarly, there are challenges and problems out there, but they are not exactly what you think they are. Why is it the case? Sensationalism in media and fundamental mechanics of how we learn about world is to blame. It was also great to follow his numerous stories from times he worked as data scientist in healthcare and government. Again, style was not the best, but most of the points make sense. paperback. many thanks to Hoa

All systems red. Good science fiction about life of an android. audiobook

Managing oneself. Rather depressing outlook. Didn’t learn much new from this book. Another miss.

Designing Data-Intensive Applications. Incredible book, very elaborate and coherent. Learned a lot about modern data systems and refined bits I already knew. You might enjoy it a lot if you work in software or just enjoying stretching your mind.

Forever War. What if there was a war that never ended?

On language. One of the most eloquent writings I have seen. Very deep material on fundamentals of language. Interview with Chomsky about politics gives fresh contrarian viewpoint on United States in past century. Especially good if you want to learn how did U.S. get where it is now.

Hagakure: The book of the samurai. Thoughts on life, economics, philosophy and other important affairs of people in ancient Japan.

Cat’s Cradle. Entertaining science fiction from Kurt Vonnegut. audiobook.

The obstacle is the way. Challenges that we face can be milestones that help us progress. Pushing this thought to the extreme, the problems we are solving form the way itself. Here author gives a collection of advices that can help you to build such mindset. Many appropriate references from ancient and recent history. Accompanying bonus interview with author was a pleasant surprise.

All quiet on the western front. Detailed and very personal account on WWI. Today is the most peaceful time in human history, yet we have not been so far in touch with reality as ever before. We see war exclusively through lens of sensationalist news camera, violent movies and games or state propaganda. Our perception of war is wicked. What is war really like? This book gives a glimpse of horrors and despair.

It does not have to be crazy at work. Very short and to the point account on work style. 

AI superpowers. I did not agree with many points Kai Fu Lee is making here about tech nor with his general defeatist attitude towards AI and tech. His paternalistic tone was annoying too. The most thought provoking and even shocking takeaway from this book is how scared of AI and its future can be someone so influential in this field.

Blitzscaling. Interesting read on how to scale-up organisations with lightning fast speed.

The curse of bigness. History of anti-trusts, monopolies and regulation. Originally I did not plan to finish it this year, but material on this topic is so well written in this book that it was a pleasure to finish it on Christmas. ❄️


Imagine you are stranded on a deserted island. You have never been on such island before. To survive you have to learn to do everything from scratch. You discover new plants and over time understand what is edible and what is not, what can heal wounds and what is good for building tools. You are keeping a diary where sometimes you leave notes on plants that you seen so far. Later, another person comes in and reads your records. Even with your help, they will have hard time getting all information about the plants. Some of it you had but did not keep. Some of you did not had in the first place because you lacked the right tools to collect it. Where does it grow? How does it change with seasons? Is there an animal who eats that plant as well? Each new plant in the diary requires an effort bringing back to life memories about it, so that you can answer as much questions as possible. It is certainly not possible to reconstruct it in full without the necessary context from you.

Now you are a robot stranded on an island. You are experimenting with plants and collecting all observations. Similarly, another robot comes in and reads your records. We are not sure how machine imagination works and what machine version of reading would mean, but imagine robot sees a plant identifier and extra information about it encoded right in its word. It sees plant's identifier and at the same time its colour, shape, density, chemical composition, relationship to other entities such animals that feed on it and types of soil it grows on. The robot can also see precise records on how this knowledge has been collected and can judge its reliability.

Today, if you read an article and see word like “business” you are left on your own to understand what it is. If you are interested only in movies, you would imagine "The Wolf of Wall Street". If you have experience in finance or have worked with data about businesses before you may know that: there are X businesses in country Y; businesses are best compared by metrics A,B, and C; each metric varies in ranges [A-A*], [B-B*] and [C-C*]; certain regions in this embedding space means T and others mean F. If so, you see information in article augmented by your prior knowledge. Now, this would be really great if you can also perceive all of it by reading some “hyper”-word that contains that extra information, something like “business:0001100101”. What if all the words are like that? Whenever you read article you get very rich understanding of issue at hand. That would be great.

But there is a problem. Each time you read this “hyper”-word you have to carry extract information. Basically, it is kind of wiki-page with as much important details as possible being shirked into some understandable by human mind embedding. Let's assume each word on in this wiki-page is a normal world and thus page is already has the smallest size. This “hyper”-word now is “business:<raw text from wiki page here>”. Carrying this blob of text is extremely redundant and makes reading slow. But is there a way to keep the very minimal amount of information necessary for identifying object of the word, something like log2 of number of words out there? Thats's just the world itself. There is some redundancy (50% in English) in encoding, but that seems to be for historical reasons. It contains minimum amount of information necessary to identify word out of all other words in language. Turns out languages has developed to minimise amount of information you have to carry around to describe word. This way, once agent has attained necessary starting knowledge, its throughput of communication is very high.

Now, are we learning just statistics of words? No, if so we would be a monkeys typing nonsense words. It is more complex than that. Instead, even the most rare, complex sentences with loose grammar and unseen words can make total sense. To make it even weirder, we learn languages incredibly quickly. More to that, languages that are developed in disconnected groups share similar traits, are equally powerful and easy to learn. Did we develop some innate predisposition to languages or did we develop languages that can be learned quickly? It must be both. The class of languages we have is shaped by our evolutionary constraints. And yet, the power of languages makes it believe that there is value on its own that we just happen to discover. An artifact that goes far beyond scope of functions imposed by constraints and objectives at time of its initial development.

The world we are living in is the experiences of physical world summarised in the language. Once you learn it, you have access to direct experiences of others and achieve temporal awareness yourself. We can find description of external physical phenomena that is able to capture its nature well enough to accurately predict what will happen in future. There is a vast amount of laws and phenomena we can discover in math, pure language. And we are just scraping the surface. What marvels are hidden beyond our reach? The space we discovered so far is as tiny as pale blue dot we call our home in cosmic scale of universe, and language is a highway entrance to it.


I took a very good shot. My friends are going to like it! Hold on, it is really good! Here, let me apply some filter to make it even better. Let's share it to public so others can see it too! But what about all that alcohol, junk food, bad tech, overpriced items, and Russian ads all these people will see?

There is something fundamentally wrong with this social contract.

It strike me today. I have been complaining about ads and social platform for a while. I thought it is enough if I use ad-blocker and will not see them. I also can limit time spent and how I use them. But it is not just me. Whenever I post something I lead others to these platforms too, and there is a good chance that cherished story from my life becomes an intro to a dose of brainwashing sponsored by Heineken or CocaCola.

Yet, passively consuming content on platform is not enough either. I regularly check on my friends' photos, stories, and they can see that. In no time they develop natural expectation that once they post it I will come to check on it. By merely being there I effectively incentivise them to be part of the platform too. Now, even value of such feeble activity compound to activity of your friends. It is easy to miss this externality while discussing ads and individual's agency, but it is a core of the whole business model.

To be fair, twitter and snapchat are doing the same. Instagram, however, has pushed it to the extreme. I only wish to see more companies were brave enough to adopt business model of postheaven (creator pays for the service) or medium (premium users cover costs for everybody) or even traditional subscription.


Two teams of five players select heroes from pool of 110+. Each hero is unique selection of ~20 skills and properties that define it’s durability, resistance, growth speed, etc. Each can carry 6 out of 150+ unique items at a time. The game is 50min long real-time match in big map. Sounds like a lot of possible combinations, why is that then all Dota games look so similar?

All 110+ heroes can be reduced to 5 representative clusters. For example, there are ‘damage dealers’ who can deal a lot of damage in short amount of time under the right conditions (and most of the time they are weak), similarly there are ones who can easily escape (but deal low amount of damage). Higher value in one dimension comes at a cost in others, this balance is done by design. The ways how they deal damage (attacks vs spells) or escape (invisibility, fast movement, teleportation) is quite restrictive. Similarly, all possible spells can be clustered into few groups: protection, damage, obstruction or amplification. Fundamentally, differences in specifics of heroes account only for slight variations from the average of the group they belong to. Items are quite restrictive too. For each bracket of capital only certain set of items can be purchased, e.g. 20 cheap, 20 mid-range, 20 expensive. Moreover, for each category of heroes only selected few items make sense, other bring almost zero value. Generally, items do not augment skills, rather they improve other parameters of heroes -  compensating their weak sides or improving strong ones. Items between heroes do not compose either, items of player A affect player B only via second order effects, such as total damage output of the team or mobility. Thus, items do not increase complexity of the game drastically neither.

Contrary to what it looks on the surface, huge variety per dimension does not lead to huge variety of winning strategies. Specifics aside, each game, even between professional teams, follows predetermined well-defined path. But wouldn’t AI discover something unexpected that works well? Probably no, at least not in a usual sense; it is discouraged by the game design itself. There have been few surprising combos such as fountain hooks at TI3, however one-by-one they have been rooted out of Dota ever since the begging. The game is always under constant scrutiny by millions of players online as well as dev team. Besides of few well known strategies anything would lead to unavoidable gradual failure. Ultimately, the team who has the best draft and execution wins. Thus, it is not surprising that OpenAI Five is undefeatable when allowed to do hero selection as well. On a high level, Dota is closer to tic-tac-toe played in simulation of real world than less visible but complex games of chess and go.

Yet, OpenAI Five is a great step forward, but some refinement is required on what makes it so. Dota does have straightforward structure, however it has to be discovered first, ideally without any supervision. Five has showed remarkable understanding of many fundamental components of the game such as farming, scouting and keeping courier safe. Most importantly, it showed great skill on composing these fundamentals into impressive coherent long-term play. There are still minor issues with long-term games, but they looks more like to-be-fixed bugs rather than underlying issues with model or training. Perhaps, the only thing left is to learn how to train it faster.

Lastly, Dota is very tangible. As opposed to go or chess, you don’t need to be an expert to appreciate some meaningful fundamental block of actions or behaviour discovered by an agent. Advancement of OpenAI Five is very illustrative on how far our capabilities in AI are today.


Recently, while talking about books I have said

“..not everything that is not technical is fiction ..and not everything that is not fiction is technical”

In a moment later, I recollected that these two statements are logically equivalent. Yet, I sensed urge to add second statement to conversation, as if it was different. I have got interested in mechanism behind it and under what conditions it is useful.

Let’s formalise it a bit with t (technical) and f (fiction). Logically both statements are the same
~t && ~f = ~f && ~t

How about conditional probability?

‘probability of not technical if it is not fiction’
p(~t|~f) = p(~t && ~f) / p(~f)

‘probability of not fiction if it is not technical’
p(~f|~t) = p(~f && ~t) / p(~t)

They are not equivalent anymore.

It seems that I had strong inclination to treat both statements as conditional probabilities. Naturally, my next question was “is it only me”? Apparently, people often think of conditional probability when they use material implication in natural language (paper). I expect there are all sorts of gaps between formal logic and our intuition.

When is this useful?

If we posses knowledge about all the books out there then we can check if there are any books that are not technical and not fiction. However, in more realistic setting we don’t know everything at hand, instead we learn gradually about the world via experiments. Moreover, these experiments are costly. For example, if each experiment takes 1h then only 24 can be done daily. With opportunity cost it becomes much worse, each experiment takes slot from all other possible experiments that could have been conducted.

Simple frequentist approach for such tests would be count number of items that are both not technical and not fiction. We check book one by one and see if it is fiction or technical. If we have found at least some, then good — we have proven that such cases exists. If we did not found it, then we can not say anything for sure.

Now, if we know that some pool of books has higher chance of being technical then by knowing conditional probabilities we can prioritise accordingly and find a book of 3rd kind faster, simply because our test are now more effective. If our resources for tests are capped, then it becomes a question whether we can find an answer at all.

Our brain is very nuanced, but there are reasons to believe that it is Bayesian(ish), at least sometimes.