Before college, apart from my studies, I was mainly involved in programming and writing. I didn't write an essay. I wrote what should be (and probably should be these days) for aspiring writers: short stories. My stories were terrible. They had almost no plot, only characters with strong feelings (it seemed to me that these feelings made the characters deep).
I tried to write my first programs on the IBM 1401, which was used in our district for what was then called "data processing". It was in 9th grade, so I was 13 or 14 years old. This 1401 was in a high school basement and my friend Rich Draves and I got permission to use it. That basement was like a Bond villain's lair, containing a bunch of alien devices - processors, hard drives, a printer, a card reader, all under bright fluorescent lights.
We wrote on one of the first versions of Fortran. Programs had to be typed on punched cards, then folded into a card reader and pressed on a button to load the program into memory and run. Usually, as a result of her work, something was printed on an incredible loud printer.
IBM 1401
That 1401 was a mystery to me. I did not understand what to do with it, Looking back, it is clear that I really couldn't do anything. The only form of input data transmission was punched cards, and I did not store anything on them. Alternatively, it was possible to write programs that did not rely on input (like calculating pi around the code), but I didn't know the math very well for that. That is why I am not surprised that I cannot remember any of my programs from that time - because they did not know much. My best memory is when I learned that programs might not end (when one of mine didn’t). Given that that computer had no time limits, it was not only a technical, but also a social error - the data center manager informed me about this.
Everything changed with the advent of microcomputers. We got computers, standing on the table right in front of us, they reacted to pressing buttons right during work, and did not go through a stack of punched cards and turn off. [1]
The first friend of mine who got a microcomputer built it himself. Back then, self-assembly kits from Heathkit were on sale. I vividly remember how impressed and jealous I was, watching my friend sit and type programs right on the computer.
Heathkit H8 Digital Computer Kits
Computers were expensive back then, and I had to persuade my father for years before he bought the TRS-80 around 1980. The Apple II was the gold standard back then, but the TRS-80 was pretty good too.
TRS-80
Then I really started programming. I wrote simple games, developed a program to estimate the flight path of my rocket models, and also wrote a word processor that my father used to write at least one book. The computer only had enough memory for 2 pages of text, so my father wrote two pages and printed them at once, but that was much better than working with a typewriter.
Even though I liked programming, I had no intention of studying it in college. In college I was going to do philosophy, which sounded cooler then. To the schoolboy, it seemed to me that philosophy is the study of the ultimate truth, while in other areas they are engaged in subject knowledge. When I entered college, I saw that there were so many ideas in other fields that there was no room for absolute truths. It seemed that philosophy was left to deal with edge cases that were simply ignored in other fields.
I couldn't put it into words when I was 18 years old. I kept on taking philosophy courses and they were boring. This is why I started doing AI.
AI was talked about a lot in the 1980s, but I was particularly motivated by two things: Heinlein's novel, The Harsh Mistress Moon, which used a smart computer named Mike, and a PBS documentary that used SHRDLU by Terry Winograd.
SHRDLU
I did not try to re-read Heinlein's novel because I understand that it is out of date, but in those years I plunged into the world of this book. It seemed like it was only a matter of time before Mike appeared, and when I saw Grape using SHRDLU, it seemed to me that it would take several years to create it. It seemed that it was only necessary to teach SHRDLU more words.
There were no AI courses at Cornell back then, so I had to learn myself. This meant I needed to learn Lisp, because it was an AI language back then. At that time, most programming languages were primitive, which means that the ideas of programmers were the same. By default, everyone in Cornell wrote in the Pascal-like language PL / I, which was used almost everywhere. Learning Lisp expanded my understanding of the concept of programs so much that it took years for me to understand the limitations. This is exactly what I expected from college. This effect did not come from classroom activities, but that's okay. For the next couple of years, I was on fire. I understood what I was going to do.
For my undergraduate thesis, I was doing SHRDLU reverse engineering. God, how I enjoyed doing this program. It was nice to write that code, but it was even more exciting to think about the fact (it's hard to believe now, but in 1985 so many thought) that those programs were really smart and conquered the heights of AI.
I studied at Cornell on a program that did not imply a choice of specialty. To obtain a degree, it was necessary to attend any classes that I liked. Of course, I chose Artificial Intelligence. When I received my diploma, I was alarmed by the presence of these quotes. It bothered me then, but now it seems funny and accurate (for reasons I just had to find out).
I applied to 3 graduate schools: MIT, Yale, which was then famous in the field of AI, and Harvard - I visited it because Rich Draves went there, and Bill Woods also lived there, who developed a parser for my SHRDLU clone. I was admitted only to Harvard, so I went there.
I don't remember when it happened and if it was at any particular moment, but in the first year of study, I realized that the concept of AI is a lie. I'm talking about this type of AI, in which the program is told that "the dog is sitting on a chair," and the program translates this information into some kind of formal representation and adds it to its knowledge base.
All these programs showed that there are subsets of natural language - formal languages. Very useful subsets. It was clear that there was a huge gap between their capabilities and real understanding of natural language. Overall, simply teaching SHRDLU more words was not enough. Techniques for creating AI based on data structures representing concepts didn't work. His collapse, as it often happens, gave rise to opportunities to write articles about all kinds of patches for him, but none of this would have allowed us to create Mike.
I began to think that it was possible to save from the fragments of my plans, and I remembered about Lisp. From my experience of working with it, I knew that this language is interesting in itself, even in isolation from AI (although at that time people were studying it only in this context). This is why I decided to focus on Lisp. I decided to write a book about hacking in Lisp. It's scary to think how little I knew about this when I started writing the book. However, there is nothing better than writing a book to understand a topic. The Lisp book was published in 1993, but I wrote most of it in graduate school.
Computer Science is a complex union of theory and systems. Theories allow you to build evidence, and with the help of systems people build and create. I wanted to create. I had a lot of respect for theory (in fact, I suspect that this half is more beautiful), but it seemed that creating something would be more interesting.
The problem with working on systems was that it didn't take long. Any program written today, no matter how good it is, will become obsolete in a couple of decades at best. People may think of your software in footnotes, but no one will use it. It seems like a very bleak job. Only people who understand the history of this area can understand that this is actually great.
Xerox Star 8010 “Dandelion”
At some point in the laboratory, several extra Xerox Dandelion computers appeared. Anyone who wanted to play with it could take it. I myself was tempted for a while, but they were very slow by today's standards, so what's the point? No one needed them, and therefore disappeared. This is exactly what happened with systems.
I wanted not just to create something, but to create something that will last a long time.
Unsatisfied, I went to see Rich Draves at the Carnegie Mellon Institute, where he was in graduate school. Once I went to the Carnegie Institution, as a child I spent a lot of time there. I looked at the picture and a thought came to me that seems obvious, although then it surprised me. Right there on the walls there were things that were done for a long time. The paintings are not outdated. Some of the best were hundreds of years old.
Moreover, painting could make a living. Of course, it's not as easy as writing programs, but I thought that a hardworking and humble person would be enough. Moreover, the artist can be independent. He doesn't have a boss and doesn't need research funding.
I've always enjoyed looking at pictures. Could I write them myself? I had no idea then. I couldn't even imagine that it was possible. Intellectually, I understood that people create art - that it did not arise by itself, but these creators seemed to belong to a different kind. They either lived a long time ago, or they are mysterious geniuses, about whose eccentricities they write in Life. The thought of taking up art seemed incredible.
That fall, I started taking art classes at Harvard. Graduate students could attend classes in all faculties, and my scientific advisor Tom Cheatham was a very calm and simple person. Even if he knew about the strange courses I was taking, they never said anything.
So, I was doing graduate school in computer science, planned to be an artist, loved to write in Lisp and wrote a book about it. In other words, like many other graduate students, I have vigorously pursued anything other than my dissertation.
I saw no way out of this situation. I didn't want to drop out of graduate school, but what was left for me? I remember my friend Robert Morris getting kicked out of Cornell for writing an internet worm in 1988 - I was jealous that he had found such an exciting way to quit graduate school.
One day in April 1990, everything started to budge. I ran into Professor Cheetham - he asked if I could graduate in June. By that time I hadn't written a word, but at that moment I made the fastest decision in my life - I decided to write my thesis about 5 weeks before the deadline, reusing fragments of my book "On Lisp" whenever possible. That is why I immediately replied, “I think so. I'll give you reading materials in a few days. "
I chose continuation apps as my theme to work with. In hindsight, I realize that I should have written a paper on macros and embedded languages. There is a whole world in this area that has hardly been explored. However, all I wanted was to finish my graduate school, and my hastily written dissertation was barely enough for that.
Along the way, I tried to go to art school. I applied to the Rhode Island School of Design and the Venice Academy of Fine Arts (as I thought it was the oldest good school). I was admitted to Rhode Island, and I never got an answer from Florence, so I set off from Providence. I went to the Bachelor of Fine Arts (BFA) program, which meant that I basically went back to college. It wasn’t as weird as it sounds, since I was 25 and art schools are full of people of all ages. At the School, I was considered a sophomore and was told that I needed to prepare the foundation. The foundation was defined as basic courses in drawing, color and design.
Toward the end of the summer, a big surprise awaited me: a letter came from the Venice Academy (it was delayed due to the fact that it was sent to English Cambridge, not American) with an invitation to pass the entrance exams in the fall. Several weeks remained until the fall. My dear hostess allowed me to leave things in her attic. I put some money aside from the consulting I was doing in graduate school, so I could have enough for a year of humble life. I just had to learn Italian.
Only strangeri (foreigners) had to take entrance exams. Looking back, I understand that apparently this was a way to weed out foreigners, because otherwise Italians would be in the minority. That summer I was in good shape in terms of drawing and painting, but did not understand how to take the written exam.I remember answering a question in an essay by writing about Cezanne - I pulled out the intellectual level to the maximum that my limited vocabulary allowed. [2]
I was only 25, and interesting patterns had already appeared in my life. I was again eager to enroll in a prestigious educational institution with the aim of studying some prestigious will come, and again I was disappointed. The students and faculty at the academy were great, but they had a tacit agreement - the students did not demand to be taught, and the academy did not require the students to learn anything. At the same time, everything happened with the conventions of a studio from the 19th century. We actually had a wood-burning stove, and a nude model, who sat as close as possible to it so as not to get burned. Except for me, almost no one drew it. The rest of the students chatted or tried to imitate what they saw in American art magazines.
It turned out that our model lived on the same street with me. She made a living posing and counterfeiting for a local antiques store. She copied whimsical old pictures from books and then aged them. [3]
During my studies at the academy, at night in my room I painted still lifes. These paintings were small: firstly, the room itself was small, and secondly, I painted them on scraps of canvas - I could not afford more. Drawing still lifes is different from drawing people (because, as the name suggests, objects cannot move). People cannot sit still for more than 15 minutes, and even at this time they do not completely freeze. The standard method for drawing people is to know how to draw a typical person, and then adjust that knowledge to suit the person you are drawing. Still life can be copied pixel by pixel from what you see. Of course, I don't want to stop there, otherwise you will get photographic accuracy - still lifes are interesting precisely because they pass through the artist's head.You will want to emphasize visual features such as telling that a sharp change in color at a point describes the edge of an object. By subtly emphasizing such moments, you can create paintings that are more realistic than photographs - not only in a metaphorical, but also in a strict information-theoretic sense. [five]
I loved painting still lifes because I was curious about what I was seeing. In everyday life, we do not think about what we see. Most of the visual perception is related to low-level processes - they tell your brain that "this is a drop of water" without giving details about its lightest and darkest areas, or they allow you to identify a bush without telling its shape and position of all leaves. This is a feature of the brain, not a mistake in its work. In everyday life, it would be inconvenient to be distracted by every single leaf. But when you want to draw something, you need to take a closer look, and when you do, you have something to look at. You may notice new and new details after trying to draw something that others take for granted - just as you may notice new thoughts after several days of trying to write an essay aboutwhat everyone thinks is obvious.
This is not the only way to draw. I'm not 100% sure he's good at all. However, it seemed worthwhile to me, which means I had to try.
Our teacher, Professor Ulivi, was a good person. He saw that I worked hard and gave me good grades in the diary that each student had. However, the academy did not teach me anything other than Italian, and I also began to run out of money, so at the end of the first year I returned to the USA.
I wanted to go back to Rhode Island School of Design, but I was broke and it was expensive to study there. Because of this, I decided to take a job for a year and then continue my studies next fall. I got a job at Interleaf, this company was developing software for creating documents. Like Microsoft Word? Yes exactly. That's when I realized that cheap software can absorb high-level software. However, Interleaf had a few more years to live. [five]
At times, Interleaf did bold things. The company was inspired by Emacs and created its own scripting language - it was a dialect of Lisp. The company needed a Lisp hacker who could write in this language. What I did there was most like a normal job out of everything I did (I apologize to my boss and colleagues - I was a bad employee). Their Lisp was a thin layer of icing on a huge C cake, and since I didn't know it and didn't want to learn, I didn't understand most of the company's software. Plus I was terribly irresponsible. Back then, working as a programmer meant coming to work at a certain time. It seemed unnatural to me, and now the whole world is converging on my way of thinking, but then because of this there were many conflicts.Toward the end of the year, most of the time, I was secretly writing On Lisp — the book got a publishing contract.
Fortunately, I was paid a lot of money, especially by the standards of art students. In Florence, after paying the rent, my budget was $ 7 per day. Now I was paid 4 times more, even if I was just sitting in the meeting. Living frugally, I was able to not only save up to return to Rhode Island School of Design, but also pay off my student loans.
I learned a lot at Interleaf (although I mostly figured out what not to do). I've learned that tech companies are better led by people who specialize in products, not sales (although understanding sales is an important skill, and people who have it are really cool), that if too many people work on the code, then bugs accumulate in it, that a cheap office would be unprofitable if everyone is discouraged, that scheduled meetings are giving way to hallway conversations, that large bureaucratic clients can be a dangerous source of money, and that work hours and coding space do not necessarily overlap with optimal ones.
But the most important lesson I have learned (it came in handy at both Viaweb and Y Combinator) is that there is nothing wrong with working at the entry level, even if it is not so prestigious. Otherwise, someone might just push you into the ceiling. All of this means that prestige can be dangerous.
The next fall, I was returning to Rhode Island School of Design, and I arranged a freelance job at an office developing various projects for clients - which is what I survived for the next few years. When I came back to one of the projects, someone told me about a new language, HTML - sort of like it was derived from SGML. The enthusiasm for markup languages was a cost of the profession at Interleaf, and I ignored it, although later this very HTML became an important part of my life.
In the fall of 1992, I returned to Providence to continue my studies at the School of Design. I had just started to delve into the basics, and studying at the academy was just ridiculous. Now I was going to see what a real art school was like. Alas, it was more like the Venice Academy. Of course, everything was much more organized (and much more expensive), but it became clear that the art school does not have the same relationship to art as medical to medicine. At least in the art department. At the designers (my neighbor studied for him), it seems, everything was much stricter. The same went for illustrators and architects. But in painting, everything was not very strict. Art students had to express themselves, which for ordinary people meant finding their own style.
A corporate identity is the visual equivalent of what is called a "chip" in business: it is thanks to it that people can understand that this work belongs to you and not to someone else. For example, when you see a cartoon style painting, you know that Roy Lichtenstein painted it. If you see a picture like this in a hedge fund manager's apartment, it's clear that he paid millions of dollars for it. Not all artists have their own corporate identity, although clients usually pay for it. [6]
There were also quite a few serious students: the guys who "knew how to draw" in high school, and now they came to the best art school in the country to learn how to draw even better. They were confused and demoralized by what they saw at the Rhode Island School of Design, but they kept going there because they were painting.I was not the type to draw in high school, but I was closer to these people than to seekers of the author's style.
I learned a lot in color lessons, but otherwise I learned to paint on my own and I could do it for free. In 1993 I dropped out of school. I took a short walk around Providence, and then my college friend Nancy Parmet did me a great favor. The moderate-rent apartment in her mother’s house in New York was empty. Did I want to go there? It was not much larger than my own apartment, and there were many artists in New York. So yes, I did! [7]
The Asterix comics begin in a tiny corner of Gaul that doesn't appear to be controlled by the Romans. There is something similar in New York: if you zoom in on a map of the Upper East Side, you will see a tiny, poor area (at least it was like that in 1993). It's called Yorkville and it was my new home. I became a New York artist (technically, I painted and lived in New York).
I was nervous about money because I felt that Interleaf was going down. Freelancing in Lisp was rare, and I didn't want to write in another language - in those days it would be C ++ if I was lucky. I had an unmistakable nose for financial ability, so I decided to write another book on Lisp. It was a simpler and more popular book that can be used as a textbook. I imagined myself living on royalties frugally and spending all my time drawing (for the cover of this book, ANSI Common Lisp, I drew around that time).
What I liked most about New York was that Idel and Julianne Weber lived there. Idel Weber was an artist, one of the first to start working in the style of photorealism, I attended her classes at Harvard. I have never seen teachers be so fond of students. Most of the former students kept in touch with her, including myself. After moving to New York, I became her de facto studio assistant.
She liked to paint on large square canvases, 4 to 5 feet wide. One day in late 1994, when I was stretching one of these monsters, they were talking on the radio about a famous stock manager. He was not much older than me and was very rich. A thought suddenly occurred to me: why shouldn't I become rich? Then I can work on whatever I want.
Along the way, I learned more and more about the recently appeared World Wide Web. Robert Morris showed it to me at Cambridge, then he went on to graduate school at Harvard. It seemed to me that the Internet would be very important. I've seen what graphical user interfaces have done for the popularity of computers. The web seemed to do the same for the internet.
If I wanted to get rich, it was the train coming to the station. I was right in essence, but wrong with the idea. I decided I needed to start a company to host art galleries on the internet. After reading a lot of applications at Y Combinator, I can say that this is the worst startup idea. Art galleries did not want to go online and still do not want to, even the most fashionable ones. Their sales work differently. I wrote software to create sites for galleries, and Robert wrote several programs to resize images and configure the HTTP server that served pages. Then we tried to conclude contracts with galleries. To say that it was difficult is to say nothing. Even handing out our product was difficult. Several galleries allowed us to make sites for them for free, but no one paid us.
Then online stores began to appear and I realized that they differed from gallery sites only by the presence of an order button. We already knew how to create all those impressive online stores.
So, in the summer of 1995, after I handed the finished ANSI Common Lisp manuscript to the publishers, we started trying to write software for building online stores. It should be desktop software first, and therefore Windows. It was a disturbing prospect — we didn't know how to write Windows programs and we didn't want to learn. We lived in the Unix world. But we nevertheless decided to write a prototype of a store constructor for Unix. Robert wrote a shopping cart, and I wrote a website generator - of course, in Lisp.
We worked at Robert's apartment in Cambridge. His neighbor was not there for a long time, I was sleeping in his place at that time. For some reason, there was no bed frame, no sheets, only a mattress on the floor. One morning, as I was lying on this mattress, I got an idea that caused me to curl up into the letter "G". What if we run the software on a server and let users control it by clicking on links? Then we wouldn't have to write anything for client computers. We could create sites on the same server from which they were served. Users don't need anything other than a browser.
Such software is now called a web application and is ubiquitous, but then it was not clear whether it was possible at all. To find out, we decided to create a version of our online store builder that could be controlled via a browser. A couple of days later, on August 12, we had a working version. The user interface was terrible, but it became clear that it was possible to create a store through a browser - without the client software or entering commands in the server terminal.
We felt that we were doing something. We had a vision for next-generation software that would work that way. Versions, ports, and all that stuff were no longer needed. Interleaf had a whole team of release engineers, and they worked no less than developers. Now the software could be updated directly on the server.
When we managed to deploy our software on the network, we founded the company. It was called Viaweb and we got our first official funding - $ 10,000 from Julianne, Idel's husband. In exchange for money, legal assistance and business advice, we gave him 10% of the company. A decade later, the deal became a model for Y Combinator. We knew that the founders needed something like this because we ourselves needed it.
At the time, my balance was negative because the $ 1,000 or so I had was counterbalanced by my tax debts (did I save the money I earned from consulting Interleaf? No, I didn't). So even though Robert was getting a graduate scholarship, I needed seed funding to live on.
We originally planned to launch in September, but as we worked on the software, the ambition grew. In the end, we were able to create a WYSIWYG site builder that was supposed to look like static after generation (except that all links did not lead to static pages, but closures in a hash table on the server).
Studying the arts helped because the main task of an online store developer is to make everything look solid to the user, and high standards are the key to that. With the right page layout and colors and fonts, you can make the guy running the store from his bedroom more respectable than a big company.
(If you're curious why my site looks so old-fashioned, it's made with this very software. It may seem awkward today, but in 1996 it was cutting edge.)
In September, Robert revolted. “We have been working on this project for a month now, but it is still not completed,” he said. In retrospect, this is especially funny because he worked on it three years later. But I decided it was worth hiring more programmers and asked Robert about the cool guys in his graduate school. He recommended Trevor Blackwell, which surprised me at first, as he was keen to put everything in a pile of notes that he carried with him everywhere. But Robert, as always, was right. Trevor was incredibly efficient at writing code.
Working with Robert and Trevor was a lot of fun. These are two people with the most independent minds that I know, and they are completely different. If you look into Robert's head, then everything looks like in a church in New England, and Trevor has excesses of Austrian Rococo there.
We opened with 6 stores in January 1996. It's good that we waited a few months, because even though we were afraid we were running late, it was actually too early. At that time, the press wrote a lot about e-commerce, but not many wanted to start their own online stores. [8]
Our software consisted of three main parts: a website editor, which I wrote, a shopping cart, which was written by Robert, and a manager for tracking orders and statistics, which was written by Trevor. At one time, our product was one of the best universal website builders. I wrote the code concisely and did not have to connect my programs to anything other than the projects of Robert and Trevor, so it was quite fun to work on all this. If for the next 3 years I only needed to work on this software, it would be the easiest time of my life. Unfortunately, there were many other things to do, which turned out worse for me than writing code, and therefore the next three years were the most stressful.
In the second half of the 90s, there were many startups developing e-commerce software. We were determined to create Microsoft Word, not Interleaf. To do this, our product had to be easy to use and inexpensive. We were lucky that we were poor ourselves, which made us reduce the cost of Viaweb even more. We charged $ 100 a month for a small store and $ 300 for a large one. This low price was both a temptation and a burden on competitors, but we did not set it out of reasonable considerations. We had no idea what businesses were paying us and how they did it. $ 300 a month seemed like a lot of money to us.
We made a lot of the right decisions by accident. For example, we did what are now called " non-scalable solutions"even though we would then describe it as" something so flimsy that we go to desperate measures to attract users. "The most common such manifestation was creating stores for customers. This seemed especially humiliating to us, because the whole point of our software was to provide people the opportunity to create their own stores. But we were ready to do anything to get users.
We learned more about retail than we would like, For example, if we had a small photo of a shirt (and by today's standards, all the pictures were small then), then it was better to have a close-up of the collar than a shot with the whole shirt in. I remember finding out about this because it meant that I had to redo 30 pictures of the shirts. However, their first versions were beautiful too.
Even though it seemed wrong, it was the right decision. Building customer stores has allowed us to learn a lot about retail and the use of our software. At first I was pushed away by the "business" itself - I thought that we would need a "person from business" who would be in charge of everything, but as soon as we started attracting users, I changed the same way as after the birth of children . Whatever the users wanted, I was at their mercy. Maybe one day so many users would come to us that I could no longer make pictures with shirts for them, but at the same time there was nothing more important.
There was another thing that I did not understand then. I didn’t realize that growth is the biggest challenge for a startup.... We had about 70 stores at the end of 1996 and about 500 at the end of 1997. I mistakenly thought that the absolute number of users decides. This is important in terms of your income - if you don't have enough, you can go out of business. But in the long term, the growth rate straightens out the absolute number of users. If we were a startup that I consulted at Y Combinator, I would say this: stop being nervous, you're fine. You have sevenfold growth every year, just don't hire too many people, then your business will become profitable, and you will be able to control your destiny.
Alas, I have hired a lot of people. Partly because this was the desire of our investors, and partly because many did so during the Internet bubble. So we didn't break even until the 1998 deal with Yahoo. This, in turn, meant that we were at the mercy of investors throughout the life of the company. And since we and our investors were both new to startups, the result was a mess, even by startup standards.
When we left for Yahoo, a great relief came. Overall, Viaweb's stock was valuable. They represented a stake in a fast growing business. But for me all this was not very valuable. I had no idea how to value a business, but I felt all too well the near-death experiences that seem to come every few months. Since we started, I have not substantially changed my postgraduate lifestyle. So when Yahoo bought us, it was going from rags to riches. Since we were moving to California, I bought a yellow 1998 VW GTI. I think the leather seats of this car were my most luxurious possessions.
The next year, from the summer of 1998 to the summer of 1999, was probably the least productive part of my life. I didn't realize it at the time, but I was very tired of the effort and stress associated with launching Viaweb. For a while after moving to California, I tried to act the same way. I programmed until three in the morning, but fatigue, along with the prematurely aging corporate culture of Yahoo and the gloomy office in Santa Clara, gradually finished me off. After a few months, I felt very unpleasant, like when I was working at Interleaf.
Yahoo gave us a lot of options after the purchase. At the time, I thought Yahoo was very overvalued, so I was surprised when I found out that the stock price had increased 5 times by the next year. I held out until the first options were received and left in the summer of 1999. I haven't drawn anything for so long that I almost forgot why I was doing all this. For 4 years my brain was clogged with software development and shirts. I reminded myself that I did all this to get rich and be able to paint. I got rich, which means I had to do drawing.
When I said I was leaving, my boss at Yahoo had a long conversation with me about my plans. I told him about the pictures I want to paint. Then I was moved by his interest. Now I understand that he just thought I was lying. Then my options were worth about $ 2 million. If I just left this money, it would be enough to start a new startup, and for this I could take other people with me. During those years, the Internet bubble was at its peak, and Yahoo was the epicenter of that era. My boss at the time was a billionaire. Leaving Yahoo in order to start a new startup seemed to him a crazy, but ambitious plan.
But I did go away to paint, and I took it straight away. There was no time to waste. I already burned 4 years to get rich. Now when I talk to founders who are leaving after selling their company, I always give one piece of advice: Take a vacation. That's what I had to do, just go somewhere and a month or two and do nothing, but this idea never crossed my mind.
I tried to paint, but it looks like I had neither energy nor ambition. Part of the problem was that I hardly knew anyone in California. I made it worse by buying a lovely home in California with a wonderful view, miles from my favorite spots. I held out for a few months and then returned to New York in despair. You will be surprised if you don’t know how rent control is organized in New York, because the apartment was sealed and looks like the tomb of my old life. At least Idel was still in New York and there were other people who tried to paint, although I didn't know any of them.
When I returned to New York, I began to live as before, only now I was rich. It was just as weird as it sounds. I did everything as before, although there were new opportunities. If I was tired of walking, all I had to do was raise my hand (if it wasn't raining) for a taxi to pick me up. Now that I was passing by charming restaurants, I could stop by and order lunch. It was exciting for a while. Drawing has gotten better. I experimented with a new method of painting still lifes: first I painted the painting as usual, then I photographed it, printed it out, placed it on the canvas, and then used it as an underpainting for a second painting, painted with the same objects (which I hoped did not rot for this time).
Along the way, I was looking for an apartment that I could buy. Now I really could choose which area to live in. I tried to find out where Cambridge is in New York. After several visits to real Cambridge, I realized that he was not there. Eh.
Around this time, in the spring of 2000, I had an idea. From our experience with Viaweb, it became clear that the future is with web applications. Why not create a web app to build web apps? Why not give people the ability to edit the code on our server through a browser and then host the resulting applications on the web? [9] Users could run all sorts of services on servers that these applications could use simply through API requests: making and receiving calls, processing images, accepting credit card payments, etc.
I was so excited by this idea that I could not think of anything else. The fact that this was the future seemed obvious. I didn't really want to start a new company, but it was clear that this idea would need to be implemented as a whole, so I decided to move to Cambridge and take on it. I had hoped to lure Robert into working with me on this project, but he was now a postdoc at MIT. Despite the fact that when I invited him last time, he made a lot of money, he also lost a lot of time. Although he agreed that the idea might work, he flatly refused to work on it.
Hmm. Well, then I had to do everything myself. I hired Dan Giffin, who worked at Viaweb, and two students who were looking for work for the summer, and we got down to business. Now it is clear that our project can be divided into 20 companies and several open source projects. The language for developing applications was, of course, a dialect of Lisp. However, I was not naive enough to believe that I could promote Lisp to the general public. We hid the brackets like Dylan did.
By then, Viaweb could be called an Application Service Provider (ASP). This name did not live long, it was replaced by "Software as a service" (software as a service), but I still called the new company Aspra.
I started working as an app builder, Dan was in the networking infrastructure, and two senior students worked on the first two services (for images and calls). About halfway through the feed, I realized that I didn't really want to run a company - especially a small one, which seemed to be the way it should be. I didn't need money anymore, so why did I do it? If my vision was to be realized as a company, then to hell with that vision. I would create a small part of it as an open source project.
To my surprise, no time was wasted on this project. Since launching Y Combinator, I've often seen startups working with parts of this new architecture, it was good to take the time to think and even try to write something about it.
As a result, the open source project I was working on became a new version of Lisp, the brackets of which I did not want to hide. Many people who write Lisp code dream of creating a new Lisp. This is partly because one of the distinguishing features of this language is its many dialects, and partly because somewhere in our minds there is a Platonic form of Lisp, from which all dialects come. Towards the end of the summer, Dan and I were working on a new Lisp dialect, which we called Arc, in a new home I bought in Cambridge.
The next spring thunder rumbled. I was invited to speak at a Lisp conference, I told how we wrote on it in Viaweb. I posted a postscript of this talk on paulgraham.com, which I created long before Viaweb but never used it at all. The page with the performance once got 30,000 views. What the hell happened? It was clear from the URL links that someone posted my talk on Slashdot [10]
Wow, I thought, I have an audience. If I write something and put it on the web, anyone can read it. Now it seems obvious, but then it was amazing. In the era of printing, there was a narrow channel for conveying information to readers, guarded by fierce monsters - editors. It was possible to attract the audience to your text only by publishing it in the form of a book, in a newspaper or in a magazine. Now, anyone could publish anything.
All this has been possible since 1993, but no one thought about it. I was closely connected with the development of the Internet infrastructure, I wrote texts, but even it took me 8 years to come to this idea. Then it took me several more years to realize the consequences. This meant that a new generation of essays was coming . [eleven]
In the era of printing, there were very few channels for publishing essays. With the exception of a few well-known thinkers who attended the right parties in New York, essays were only allowed to be published by specialists writing about their activities. Many essays were never written due to the lack of channels for their publication. The channel appeared and I was going to write. [12]
I worked on several different things, but when I started posting essays on the internet, there was a turning point. Since then, no matter what I did, I knew that I would always write essays.
I knew the scope of online essays would be marginal at first ... Socially, these lyrics sounded more like the GeoCities psycho tirades than the noble and beautifully typed lyrics from The New Yorker. But at this point I knew enough to be reassuring, not frightening.
One of the most noticeable patterns I've noticed in my life is how good it is (for me at least) to work on something that isn't considered prestigious. Still life has always been the least prestigious type of painting. When we started, it seemed to everyone that Viaweb and Y Combinator were of no use to anyone. Strangers are still surprised when I say that I am writing an essay and am going to publish it on my website. Even Lisp, which is considered to be intellectually prestigious (like Latin), seems to be just fashionable.
It's not that low-profile jobs are good in and of themselves. But if you understand that you are attracted to some kind of work, despite its problems with prestige, this suggests that there is something right in it, and that you have the right motives. Wrong motives are a big problem for ambitious people. If anything can lead you astray, it's the desire to impress people. So while working on low-profile stuff doesn't guarantee that you're on the right track, at least it ensures that you're not on the popular wrong path.
Over the next few years, I wrote many essays on a wide variety of topics. O'Reilly published their collection, Hackers and Artists, after one essay about them. I also worked on spam filters and kept painting. I invited friends to dinner on Thursday nights, which taught me how to cook for a group of people. Later, I bought another building in Cambridge, a former candy factory (and as it later turned out, a former porn studio) to use as an office.
One day in October 2003, I had a big party at my house. It was a great idea for my friend Maria Daniels, she came to dinner one Thursday. Three different hosts invited guests to parties. Thus, for every guest, two-thirds of the other guests will be strangers they would like to know. One of the guests was a girl I didn't know who I really liked: Jessica Livingston. A couple of days later, I asked her out on a date.
Jessica was in charge of marketing at the Boston Investment Bank. The bank was convinced that they knew about startups, but the next year, when Jessica met with my friends in the startup world, she was surprised at how different things were from reality. She was also surprised by their bright and colorful stories. So she decided to write a collection of interviews with startup founders .
When the bank began to struggle financially and had to cut half of the staff, Jessica began looking for a new job. In early 2005, she was interviewed for a marketing position at a Boston venture capital firm. It took them several weeks to make a decision, during which time I began to tell her about everything there is to know about venture capital. That the company needs to make a lot of small investments instead of a few gigantic ones, that they should fund younger and tech-savvy founders rather than MBAs, that they should keep the founders in the CEO position, and so on.
One of my techniques for writing essays is giving a talk. The prospect of standing in front of a group of people and telling them something that won't waste their time was a great stimulus to the imagination. When the Harvard Computer Society (Computer Bachelor's Club) asked me to give a talk, I decided I would tell them about building startups. Maybe they can avoid the worst mistake we've made.
So, I made this report. I said that the best sources of funding for startups might be founders of successful startups, as they can also provide advice. Then it seemed that all the listeners were looking at me expectantly. Horrified at the prospect of my inbox being littered with business plans (if only I knew), I blurted out, "Not me!" and continued the report. But then it occurred to me that I should stop postponing this business and become a business angel. I wanted to do this back in the days of our deal with Yahoo, since then 7 years have passed, and I have not made a single investment.
Along the way, Trevor and Robert and I were planning projects that we could work on together. I missed working with them and it seemed like we could find something to collaborate on.
As Jessica and I walked home after dinner on March 11 at the corner of Garden and Walker Streets, the three strands came together. Fuck the venture capitalists who took so long to make a decision. We decided to open our own investment firm and implement the ideas we talked about. I would finance this company, and Jessica could quit her job and start working with us, and Robert and Trevor would become our partners. [13]
Ignorance worked in our favor again. We had no idea how to be business angels, and in 2005 there was still no Ron Conway to learn from. We just made the obvious choice, and some of our solutions were new.
Y Combinator has several components, and we didn't immediately articulate all of them. First of all, we became an angel firm. Then these two words did not fit together. Then there were venture capital firms whose job was solely to invest, but they only made large investments of millions of dollars. And there were angels who made small investments, but these were people who were usually focused on other things and invested on the side. None of them could help the founders enough in the beginning. We knew how helpless the founders could be because we remembered how helpless we ourselves were. For example, Julian once did what seemed like magic to us - he founded a company. We wrote software beautifully, but legal design, promotions and so on - what is it all about? We were going to engage in more than just seed investment,we wanted to do everything Julian did for us.
YC was not originally a foundation. The company was not too expensive, so we financed it with our own funds. 99% of readers will not have any questions about this, and professional investors thought, "Wow, so they took all the profits for themselves." Again, this was not due to discernment on our part. We didn't know how venture capital firms were. It never occurred to us to collect a fund, and if it did, we would not know where to start. [14]
The main distinguishing feature of YC is a batch model, we financed several startups at once twice a year, and then we tried to intensively help them for three months. This happened not only implicitly, but also explicitly due to the fact that we knew little about investing. We needed experience. We thought - what could be better than funding several startups at once? We knew that during the summer, students get temporary jobs at tech companies. Why not organize a summer program where startups are launched instead? We did not feel guilty that we were, in a sense, fake investors, since they were, in the same sense, fake founders. We probably didn't have to make a lot of money from this, but we could practice investing, and the guys with whom we will work,spend their summer more interesting than working at Microsoft.
We used a building that I owned in Cambridge as our headquarters. We all had dinner there once a week (on Tuesdays, because I cooked dinners on Thursday), and after dinner we invited startup experts to speak.
We knew students could solve summer jobs in days, so we came up with a Summer Founders Program. I posted an ad about her on my websiteand invited students to apply. I never thought that writing an essay would create a "stream of deals" like investors do, but it worked. [15] In total, we received 225 applications for the Summer Founder Program and found that many of those who applied were either graduating or about to graduate. This whole story with the summer program began to seem more serious than we expected.
We invited 20 out of 225 groups for face-to-face interviews and we decided to invest in 8 of them. It was an impressive group. The first stream included Reddit, Justin Kahn and Emmet Sheer, who later founded Twitch, Aaron Schwartz, who had already helped write the RSS spec and would become an "open access" martyr a few years later, and Sam Altman, who later became the second president of YC. I don’t think the first stream was good just because of luck. It took the courage to sign up for such a program instead of working in a reputable place like Microsoft or Goldman Sachs.
The start-up deal was based on a combination of a deal we did with Julian ($ 10,000 for 10%) and what Robert said was given to MIT graduate students for the summer ($ 6,000). We invested $ 6,000 per founder, which meant an average of $ 12,000 in exchange for 6%. It had to be fair, because this is twice the deal we made at the time. Plus that summer is really hot and Jessica arranged free air conditioning for the founders. [16]
I soon realized that we had stumbled upon a way to increase funding for startups. Batch financing was more convenient for us, since it allowed us to do something for a large number of startups at the same time, and it was also more useful for them to work in groups. It also addressed one of the biggest challenges founders face: isolation. Now the founders had not only regular colleagues, but also colleagues who understood their problems and shared methods of solving them.
As YC grew, we began to see other growth benefits as well. The alumni became a close-knit community, they were eager to help each other and the startups in their current groups, because they remembered what it was like for themselves. We also noticed that startups were becoming clients of each other. We used to joke about "GDP Y Combinator", but now there is less and less joke. Now, many startups are getting their first customers among their group mates.
Initially, I did not plan to make YC my main and full-time job. I was going to do three things: write code, write essays, and work at YC. As YC grew, I became more and more involved in his affairs, and as a result, he began to take more than a third of my attention. But for the first few years I could do other things calmly.
In the summer of 2006, Robert and I began work on a new version of ARC. It was pretty fast as it compiled to Scheme. To test the functionality of this language, I wrote Hacker News in it. Initially, it was supposed to be a news aggregator for startup founders, it was called Startup News, but after a few months I got tired of reading exclusively about startups. Plus, we didn't want to contact the founders of startups. We wanted to reach out to future founders. So I changed the name to Hacker News, and the topic could be anything that interested intellectual curiosity.
HN has certainly been good for YC, but it has become a big source of stress for me. If I only had to choose founders and help them, life would be very easy, But that would mean that HN was a mistake. At the time, I was like someone in pain during a marathon, not from fatigue and stress, but from a blister from inappropriate shoes. When I ran into urgent problems while working with YC, there was a 60% chance that they were HN and 40% for everything else combined. [17]
Besides HN, I wrote all of the internal software for YC on Arc. But while I continued to write in Arc, I gradually stopped working on the language itself. Partly because of the lack of time, and partly because the infrastructure depended on it. So now I had two projects: essay and YC.
YC was not like the job I had done before. I could no longer choose what to work on, problems came to me on their own. Every 6 months a new batch of startups appeared, and their problems (whatever they were) became ours. It was exciting because their tasks were varied and good founders were very productive. If you wanted to learn as much as possible about startups in a short time, there is simply no better way.
There were also parts of the job that I didn't like. Disputes between co-founders, trying to expose lies, fighting people who are rude to startups, and so on. But I also worked hard on what I didn't like. I was haunted by the thought of Kevin Hale: "Nobody works harder than the boss." It was both a descriptive thought and a prescription, and I was afraid of its second essence. I wanted YC to be successful, and if the way I work sets a high bar for everyone else, then I better work really hard.
Robert Morris came to California one day in 2010 and he did something amazing - he gave me unsolicited advice. I remember this from him only once. Once in Viaweb, I doubled over with a kidney stone and Robert decided he should take me to the hospital. These are the reasons Robert needed to give unsolicited advice. So I remember his words very well: "You know, you have to make sure that YCombinator is not your last cool thing."
Then I did not understand what he meant, but it soon dawned on me that he was advising me to leave. This advice seemed odd since YC was doing great. If anything happened less often than unsolicited advice from Robert, then it was his mistake. It got me thinking. Indeed, on that trajectory, YC would be my last business, because he himself took so much of my attention. He had already consumed Arc and began to consume essays. Either YC had to be my life's work, or I had to leave. And I made up my mind.
In the summer of 2012, my mom had a stroke caused by a blood clot caused by rectal cancer. The stroke destroyed her balance and we placed her in a nursing home, but she wanted to get out of it and return home. My sister and I were determined to help her with this. I often flew to Oregon to visit my mom, and on these flights I had a lot of time to think. At one of them, I realized that I was ready to transfer YC to someone else.
I asked Jessica if she wanted to be president, but she refused, so we decided to hire Sam Altman. We spoke with Robert and Trevor and agreed on a complete change of guard. Up to this point, YC was controlled by an LLC founded by the four of us. We wanted YC to live for a long time, which means it was impossible to give control to the founders. So if Sam agrees, then we would give him the opportunity to reorganize YC. Robert and I would have retired, and Jessica and Trevor would become regular partners.
When we asked Sam if he wants to become president of YC, he initially replied in the negative. Oh, I wanted to start a nuclear reactor startup. I kept pushing, and in October 2013 he finally agreed. We decided that he would take over from the winter 2014 stream. Until the end of 2013, I was giving Sam more and more authority - partly so he could master the job, and partly because I was focused on my mother, who had her cancer back.
Mom died on January 15, 2014. We knew this moment was close, but when it happened it was hard.
I continued to work at YC until March to help startups get to the demo day and then check the results (I still talk to alumni and newbies working on what interests me, but it only takes a few hours a week).
What should I do next? Robert's advice didn't say anything about it. I wanted to do something else, so I went back to drawing. I wanted to see what I could achieve if I focused on it. So the next day after leaving YC, I started painting. I was out of shape and it took me a while to get her back, but it was fun. [eighteen]
I spent most of 2014 drawing. I have never been able to work like this continuously, and I had to be better than before. I wasn't too cool, but still better. Then in November, right in the middle of painting, I fizzled out. Up to this point, I was always interested to see how the picture I was working on would turn out, but suddenly the completion of this seemed to me a routine work. I stopped working on this painting, cleaned my brushes, and didn't do it anymore - at least not yet.
I understand that it sounds weak. But remember, this is a zero-sum game. If you can choose what to work on and chose not the best (or just not very good) project for yourself, then it will interfere with another project. And at 50, there was a nice missed opportunity to mess around.
I started writing essays again and wrote a bunch of new texts over the next few months. I even wrote a couple of articles that weren't about startups. Then in March 2015 I started working on Lisp again.
A distinctive feature of Lisp is that its core is a language interpreter written in it. Initially, it was not conceived as a programming language in the usual sense. It was supposed to be a formal model of computation, an alternative to the Turing machine. If you want to write an interpreter for a language, what is the minimum set of predefined operators you need? Lisp, invented (or rather discovered) by John McCarthy, is the answer to this question. [19]
McCarthy didn't realize that this Lisp could be used for computer programming until his student Steve Russell suggested it. Russell translated the McCarthy interpreter into the IBM 1704 machine language, and from that moment Lisp became the programming language in the usual sense of the word. But its origins as a model of computation gave it a strength and elegance that other languages could not match. This is what attracted me to college, although I did not understand why.
Lisp McCarthy in 1960 did not know how to do anything other than interpreting expressions. He lacked many things that one wants to have in a programming language. They had to be added, and when they appeared, they were not defined using McCarthy's axiomatic approach. It was not possible then. McCarthy tested his interpreter by hand by simulating program execution. He was approaching the limit of interpreters that could be tested in this way - there was a bug that McCarthy did not notice. To test a more complex interpreter, you had to start it, and then computers were not powerful enough.
Now they are powerful enough. You can use McCarthy's axiomatic approach until you define a complete programming language. And as long as your changes to McCarthy's Lisp preserve the principle that it was discovered, not invented, you can create a complete language with that quality. It is certainly more difficult to do it than to say it, but if it is possible - why not try it? I decided to give it a try. It took 4 years, from March 26, 2015 to October 12, 2019. Fortunately, I had a clearly defined goal, otherwise it would be difficult to do this for so long.
I wrote a new Lisp called Bel in Arc. It might seem like there is a contradiction, but it is an indicator that I had to resort to trickery to get everything to work. With the help of a number of tricks, I managed to write something as close as possible to a workable language interpreter written in the same language. It's not very fast, but fast enough for tests.
Most of that time, I forbade myself to write essays, otherwise I would never have finished.At the end of 2015, I set aside 3 months to write an essay, and when I returned to work on Bel, I could barely understand the code. Not because it was poorly written, but because the program was confusing. When you are working on an interpreter for a language written in the same language, it is difficult to keep track of what is happening at what level, errors can be encrypted before they are received.
So I didn't write any more essays until I finished working on Bel. During these years it may seem that I am not doing anything, although I worked harder than ever. Occasionally, after hours of struggling with creepy bugs, I would go to HN or Twitter and see posts like "Is Paul Graham still writing code?"
Bel was challenging but rewarding. I worked on it so intensely that I constantly had chunks of code in my head, and I could write more and more. I remember walking with my boys to the beach on a sunny day in 2015 and thinking about how to solve the problem with the sequels, watching the guys play in the waves. I remember this moment because I was alarmed by its novelty. It's good that over the next few years I have had more moments like this.
In the summer of 2016 we moved to England. We wanted our children to have the experience of living in another country, and since I was a British citizen by birth, this choice was obvious. We wanted to stay there for only a year, but we liked it so much that we decided to stay. Most of Bel was written in England.
In the fall of 2019, I finished work on Bel. Like the original McCarthy Lisp, it is more a specification than an implementation, and like McCarthy's Lisp, it is a specification expressed in code.
Now I could go back to writing the essay again. I've covered a bunch of interesting topics to me. I continued to write essays until 2020, and then again began to think about what I could work on. How to choose what to do? How did I make this choice before? I wrote an essay for myself to answer these questions and was surprised at how long and confusing the answer was. I thought that if he surprised me, the person who lived it all, how interesting would it be for other people? Maybe this text will inspire other people whose lives are so chaotic? I wrote a more detailed version for them so that other people can read it - and this is her last sentence.
Notes
[1] In my experience with PCs, one era is missing: time-sharing machines and interactive operating systems. I switched from punch cards directly to microcomputers, which made the latter less exciting.
[2] The meaning of Italian words for general concepts can be predicted from their English equivalents (with the exception of traps like Polluzione). Only everyday words are different. If you connect a number of common concepts and simple verbs, you can make some progress in your Italian learning.
[3] I lived in Piazza San Felice 4, so my walks to the Accademia took place all over old Florence: past Pitti, across the Bridge, past Orsanmichele, between the Duomo and Baptistery, then along Via Ricasoli to Piazza San Marco. I've seen the streets of Florence in a variety of states, from dark winter evenings when they were empty to hot summer days when the streets are packed with tourists.
[4] Of course, you can paint people as still lifes if you want (and if they are ready). Such a portrait is probably the pinnacle of still life, although the long immobility causes painful expressions among the sitters.
[5] Interleaf was one of the many companies that had smart people who built cool technology, but it was all crushed by Moore's Law. In the 1990s, the exponential growth in processor power (from Intel, for example) fettered companies producing specialized software and hardware.
[6] Rhode Island School of Design brand identity seekers are not necessarily mercenaries. Anything expensive gets cool, and anything that looks cool will soon become expensive.
[7] Technically, the rent at the apartment was not controlled, but was stabilized, but all these things are clear only to New Yorkers. The bottom line is that it was very cheap, below half the market value.
[8] Most of the software can be released after development is complete, but if you are working on an online store builder and you have no users, then it's not that simple. Before the public launch, we had to do a private launch, that is, recruit a limited group of users and make sure they get decent stores.
[9] We had a code editor in Viaweb that allowed us to create our own page styles. They didn't know it, but under the hood, they edited Lisp expressions. But it was not an application editor, because the code was run when the sellers created the sites, not when the buyers visited them.
[10] This was the first example of such an experience, which later became familiar. The same thing happened when I read the comments and found a lot of angry people there. How could I say that Lisp is better than other languages? Weren't they Turing-complete? People who see reactions to my essays sometimes say that they feel sorry for me. I am not exaggerating when I say that this was the case from the very beginning. It all comes with distribution. Essays tell people about things they don't know yet, and people don't like it.
[11] Of course, in the 90s people posted a lot of things on the Internet, but putting them on the Internet and publishing are two different things. The publication assumes that you consider the Internet version as the main one.
[12] There is one general lesson we have learned from our experience with YCombinator: customs will restrict you long after the conditions that caused them are gone. Once upon a time, common venture investment practices, like essay writing techniques, were based on real-world constraints. Launching startups was more expensive, and therefore rarely happened. Now they might be cheap and widespread, but the customs of venture capitalists reflected the order of the old world, just as the customs of essay writing still reflected the customs of the old print era.
All this means that people with independent thinking (that is, less influenced by custom) will have an advantage in rapidly changing areas (where custom is more likely to become outdated);
An interesting point: it is not always possible to predict which fields will be affected by rapid changes. Obviously this is true for software and venture capital investments, but who would have thought of writing an essay?
[13] The name YCombinator was not the original name. We first named the company Cambridge Seed. We wanted to get rid of the region-specific name in case someone in Silicon Valley copies us, so we renamed the company after one of the coolest lambda calculus tricks: the Y combinator.
I chose orange as our main color because it is warm and because it hasn't been used by other VCs. In 2005, all venture capital funds used austere colors of maroon, navy, and green because they were trying to get the attention of limited partners, not founders. The YC logo is an inside joke. The Viaweb logo was a white V in a red circle, so my YC logo is a white Y in an orange square.
[14] Starting in 2009, YC was a fund for a couple of years, but then it grew so much that I could no longer fund it myself. However, after purchasing Heroku, we had enough money to go back to self-financing.
[15] I never liked the term "deal flow" because it implies that the number of new startups is fixed at any given time. This is a lie, and YC's goal is to refute this claim by helping to found startups that would not otherwise exist.
[16] Jessica said that they were all different shapes and sizes, because there was a huge demand for air conditioners, and she needed to find everything she could. They were all heavier than she could have carried.
[17] Another problem with HN is the weird edge case that occurs when you write an essay and host a forum. When you host a forum, it is assumed that you can see, if not all conversations in general, then all conversations with your participation. When you write an essay, people post on forums with very loose and incorrect interpretations. Separately, these phenomena are tiring, but bearable, but together they are destructive. Misinterpretations need to be responded to, because assuming you are in the conversation implies rejecting the popular misinterpretation and recognizing it as correct. On the other hand, it is reassuring: anyone who wants to fight you will feel like they have a chance.
[18] The saddest thing about leaving YC was that we didn't work with Jessica anymore. We have been working on YC almost all the time that we knew each other, and we did not try or want to separate this work from our personal life. This departure was like pulling out a deeply rooted tree.
[19] One way to separate the concepts of invention and discovery is to talk about aliens. Any sufficiently advanced alien civilization probably knows about the Pythagorean theorem. I believe (albeit with less certainty) that they heard about Lisp from McCarthy's 1960 article.
But if this is so, then there is no reason to assume that this is the limit of the language they know. The aliens will probably need numbers, errors, and I / O. So there is at least one possibility that Lisp McCarthy was a discovery.
Thanks to Trevor Blackwell, John Collison, Patrick Collison, Daniel Huckle, Ralph Hazell, Jessica Livingston, Robert Morris and Harge Taggar for reading the drafts of this text.
Follow the YC Startup Library news in Russian on the Telegram channel or on Facebook .