书童按：本篇是帕维尔·杜罗夫（Pavel Durov）于2025年10月接受Lex Fridman的播客采访实录，Pavel是端到端加密通信软件Telegram的创始人兼CEO。其采访中涉及精简团队、高度自动化、法国被捕、审查、企业家的困境等话题，精彩绝伦，令人击节称赞。初稿采用AI机器翻译，经自动化中英混排，书童仅做简单校对及批注。原稿中英文混排近7万字，书童将分为Part1-4发出，本篇是第三部分，以飨诸君。
高压教育
Intense education
(01:33:24) 好的，我们喝了点茶，回来了。让我们回到很多年前，回到最开始。你提到你上过一所教育强度非常大的学校，所以我认为看看那种教育中一些强大的方面会非常有趣，从语言到数学。你能具体描述一下其中一些严格的方面以及你从中收获了什么吗？
Lex Fridman (01:33:24) All right, we had some tea, we’re back. Let’s go back a bunch of years to the beginning. You mentioned you went to school with super intensive education so I thought it’d be really interesting to look at some of the powerful aspects of that education from the languages to the math. Can you actually describe some of the rigorous aspects of it and what you gained from it?
帕维尔·杜罗夫 (01:33:48) 在11岁的时候，我得到了一个机会，进入了我当时居住的圣彼得堡的一所实验学校，你必须通过严格的考试才能被录取。这所学校的理念是，如果你试图将尽可能多的信息塞进一个青少年的大脑中，重点放在数学和外语上，那么学生的大脑会发生一些变化，使得学生能够理解大多数其他学科。但结果我们有一个班级，它没有任何单一的重点，而是广泛涉及许多学科。你至少要学习四门外语，包括拉丁语、英语、法语、德语，此外，你还可以学习古希腊语。你会有像生物化学或精神分析、进化心理学这样的课程。这个班级与同一所学校（属于圣彼得堡国立大学，称为学术体育馆）的其他班级的不同之处在于，不像其他班级专注于某个单一学科，如物理、数学或历史，这个班级试图从所有这些专业班级中汲取精华，并将其融入一个课程体系中。由于这是一个实验班，不可能成为一个全优生，在所有科目上都出类拔萃，甚至尝试这样做都被认为是疯狂的。
Pavel Durov (01:33:48) At the age of 11, I got the opportunity to enter an experimental school in St. Petersburg where I lived and you had to pass a rigorous test to get accepted. The idea behind the school was that, if you try to squeeze as much information as possible into a brain of a teenager making a focus on maths and foreign languages, then there will be some changes in the brain of the student that will allow the student to understand most other disciplines. But we had a class, as a result, that didn’t have any single focus, it was very widespread across a lot of disciplines. You would have four foreign languages at least including Latin, English, French, German, in addition, you can get ancient Greek. You would have classes like biochemistry or psychoanalysis, evolutionary psychology. The difference of this class as opposed to other classes in the same school which was part of the St. Petersburg State University called academic gymnasium was that, unlike other classes which were specialized in some single subject like physics or maths or history, this one tried to get the best from all of these specialized classes and bring into one curriculum. Since it was an experimental class, it wasn’t possible to become a straight A student, to be excellent in all the subject, it’s always considered crazy to even try.
莱克斯·弗里德曼 (01:35:48) 所以，它假设没有人能够应对，你只是在挑战人类思维的极限。四门语言并行，数学，进化心理学，就是用信息淹没大脑，看看会发生什么。
Lex Fridman (01:35:48) So, it’s assumed nobody’s able to handle it, you’re just pushing the limits of the human mind. Four languages in parallel, math, evolutionary psychology, just overwhelming the mind and see what happens.
帕维尔·杜罗夫 (01:35:59) 是的，看看会发生什么。这是一项实验，而且是在90年代中期，记得吗，那时候俄罗斯，尤其是它的教育系统，不像今天这样规范。它处于俄罗斯历史的两个阶段之间，即苏联历史和21世纪现代俄罗斯历史之间的中间时期。无论如何，我从那段经历中学到了很多。首先，我进入这所学校是因为我不断被其他学校开除。
Pavel Durov (01:35:59) Yes, see what happens. This was an experiment and it was in the middle of the ’90s, remember, when Russia, particularly its educational system, wasn’t regulated as much as it is today. It was in the middle between the two stages of the Russian history, the Soviet’s history and the modern Russian history of the 21st century. In any case, I learned a lot from that experience. First of all, why I got into this school is because I kept being kicked out from other schools.
莱克斯·弗里德曼 (01:36:38) 挑战权威？
Lex Fridman (01:36:38) Challenging authority?
帕维尔·杜罗夫 (01:36:39) 我所有科目都很好，但行为表现不好。在90年代初的苏联，我们有这种行为评分，也许他们今天还有，我不确定。我的行为表现非常差，总是挑战老师，总是指出他们的错误。
Pavel Durov (01:36:39) I was good at all subjects but not behavior. We had this behavior grade in the Soviet Union in early ’90s, perhaps they even have it today, I’m not sure. I was very bad at behavior, always challenging the teachers, always pointing out their mistakes.
莱克斯·弗里德曼 (01:36:59) 顺便说一下，这并不完全是坏事，对吧？如果你回头看，这对年轻人来说是有一定价值的，也许要带着尊重，但挑战权威，挑战旧的智慧，对吧？
Lex Fridman (01:36:59) By the way, that’s not such a bad thing, right? If you were looking back, there’s some value to that for young people to, maybe respectfully, but challenge the authority, the wisdom of old, right?
帕维尔·杜罗夫 (01:37:14) 我认为我非常幸运能够这样做，并且最终能够安然无恙，因为通常情况下，如果你不断挑战权威，你只会被所有学校开除，然后一事无成。所以，我最终进入了一所学校，在那里挑战老师虽然不是完全没问题，但却是你可以做的事情，然后你会和老师展开辩论，通常他们会允许你表达你的观点，然后一些客观的真理可能会由此产生。
Pavel Durov (01:37:14) I think I was very lucky to be able to do that and to be able to get away with it in the end because, normally, if you keep challenging authorities, you just get kicked out of all schools and then you end up nowhere. So, I eventually got into a school where challenging teachers was not fully okay but it was something that you could do and then you would start a debate with the teacher and normally they would allow you to express your point of view and then some objective truth may come out of it as a result.
(01:37:58) 但在那个时候，我对我的生活感到相当无聊，每个青少年都会到达一个点，当他们有这种存在主义危机时。生活的意义是什么？我到底在这里做什么？在某个时刻，我决定，既然我反正必须上学，我不妨尝试做一些不可能的事情，成为最好的学生，并在每一个科目上都得到A，或者我们在俄罗斯系统中称之为五分，这让我忙了一阵子。
(01:37:58) But at that point, I was pretty bored with my life, every teenager gets to a point when they have this sort of existential crisis. What’s the point of life? What am I even doing here? At some point, I decided, since I have to go to school anyway, I might as well try to do something impossible and become the best student and get an A or what we called five in the Russian system on every single subject and that kept me busy for a while.
(01:38:40) 这非常困难，因为你没有足够的时间。即使你只是不停地学习，不做任何其他事情，你也没有任何剩余时间来准备所有的家庭作业、任务并为所有考试做好准备。所以，我最终利用了课间休息时间，但我达到了我想要达到的结果。我在每个科目都得到了优秀的分数，这让我高兴了一阵子。
(01:38:40) It was incredibly difficult because you didn’t have enough time. Even if you just studied all the time, not doing anything else, you didn’t have any time left to prepare all the homework, tasks and get ready for all the tests. So, I ended up using the breaks between classes but I get to the result I wanted to get to. I got the excellent mark in every subject and that kept me happy for a while.
莱克斯·弗里德曼 (01:39:19) 通过同时学习外语并进行如此多样化的学习，你对一个有效的教育系统有什么理解？如果你要从头开始为年轻人设计一个教育系统，尤其是在21世纪，那会是什么样子？你曾发文谈到数学作为一切基础的价值。
Lex Fridman (01:39:19) What did you understand about an effective education system from studying foreign language at the same time doing such a diversity? If you were to design an education system from scratch for young people, especially in the 21st century, what would that look like? You posted about the value of mathematics as a foundation for everything.
帕维尔·杜罗夫 (01:39:39) 是的。我仍然认为数学是必不可少的。它能塑造你的大脑，它教你依靠逻辑思维，将大问题分解成小部分，按正确的顺序排列，耐心地解决它们，如果不行就再试一次。这与你编程、项目管理以及当你创办自己的公司时需要的技能完全相同。这是学校里少数几个鼓励你发展自己思维，而不是依赖别人怎么说、只是重复他们观点的科目之一。这是极其宝贵的。当然，一旦你擅长数学，你就可以把它应用到物理、工程、编码中。大多数最成功的科技公司创始人和首席执行官都非常擅长数学和编码，这并不奇怪，因为归根结底，你依赖的是相同的思维技能。
Pavel Durov (01:39:39) Yeah. I still think math is essential. It’s something that shapes your brain, it teaches you to rely on your logical thinking to split big problems into smaller parts, put them in the right sequence, solve them patiently, trying again if it doesn’t work. This is exactly the same skill you’ll need in programming and project management and start it when you start your own company. And it’s one of the few subjects to school which encourages you to develop your own thinking as opposed to rely on what other people have to say and just repeating their opinions. That is extremely valuable. And of course, once you’re good at math, you can apply it in physics, in engineering, in coding. And it’s not surprising that most of the most successful tech founders and CEOs are very good at matters in coding because, ultimately, it’s the same mental skill that you rely on.
(01:41:05) 但在当时学校里，我也意识到了另一点，那就是竞争确实非常重要，竞争是关键。这是激励许多青少年的因素，如果有学校存在，如果你把竞争从教育系统中移除，你最终会迫使孩子们开始在其他地方竞争，例如，在电子游戏中。这是你现在在许多国家看到的趋势，包括西方，当善意的当局或父母说我们不希望我们的孩子压力太大，我们不希望他们感到焦虑，所以让我们取消所有公开的评分系统，所有这些谁赢谁输的排名，我们不要任何这些。
(01:41:05) But back then in the school, I realized something else as well, it’s that competition is really important, competition is key. This is what motivates a lot of teenagers when there is school and, if you remove competition out of the education system, you end up forcing kids to start competing elsewhere, for example, in video games. It’s a trend you see now in many countries, including in the West, when well-meaning authorities or parents say we don’t want our kids to be too stressed, we don’t want them to feel anxiety so let’s just get rid of all the public grading system, all these rankings of who won, who lost, we don’t want any of that.
(01:42:06) 其中一部分是合理的，但结果是，一些孩子失去了兴趣。是的，你消除了失败者，但你最终也消除了赢家。然后，如果你在那个年龄段对孩子过度保护，他们长大，从学校、大学毕业，他们仍然没有为现实生活做好准备，因为现实生活是为工作、晋升、客户而不断竞争，而且更加残酷。
(01:42:06) And part of it is justified but, as a result, some kids lose interest. Yes, you eliminate the losers but you end up eliminating the winners as well. And then, if you are overprotective of the kids in that age, they grow up, graduate schools, the universities and they’re still not prepared for real life because real life is constant competition for jobs, for promotions, for customers and it’s more brutal.
(01:42:47) 你最终得到的是高自杀率、高失业率，以及所有你现在在许多国家看到的负面趋势和事情，这些国家认为从他们的教育系统中消除竞争是个好主意。他们仍然坚持，仍然认为竞争是坏事，他们也在一定程度上试图从他们的经济中消除竞争，说我们要确保失败者不会输，赢家不会得到太多，但结果，他们使他们整个系统、整个经济竞争力下降。
(01:42:47) What you have as a result is high suicide rates, high unemployment, all the things and negative trends you see now in many countries which thought eliminating competition from their education system was a good idea. They still persist, they still think competition is a bad thing, they try to eliminate competition from their economy as well to an extent saying we are going to make sure the losers don’t lose and the winners don’t get too much but, as a result, they make their entire systems less competitive, their entire economies.
(01:43:34) 其中一些在欧洲的国家现在正努力跟上大东方国家、韩国、新加坡、日本和其他教育系统基于无情竞争的地方。所以，这是任何文明都必须做出的艰难选择。我们是支持竞争，理解它最终会带来科学技术的进步和整个社会的繁荣，还是我们移除竞争，认为我们可以以某种方式保护后代免受竞争不可避免地带来的压力。
(01:43:34) Some of them in Europe are now struggling to keep up with BigEast, with South Korea, with Singapore, with Japan and other places where the education system was based on ruthless competition. So, this is a hard choice any civilization has to make. We support competition understanding that, eventually, it leads to progress in science and technology and abundance for society at large or we remove competition thinking that somehow we can shield the future generations from the stress that competition inevitably causes.
莱克斯·弗里德曼 (01:44:22) 是的，这源于一种良好的同情本能，你不想让不擅长某事的人感到痛苦，但似乎挣扎是生活的一部分，你要么早点经历，要么晚点经历。确实，这是个很好的观点，竞争似乎是技能发展的一个非常强大的驱动力，就像你提到的，追求精通。人性中有某种东西，特别是对年轻人来说，如果你能在某件事上竞争，你就会被强烈驱动去把那件事做好。如果你能像大东方国家那样，像你提到的许多国家那样，在教育系统中引导这种竞争，那么你将培养出许多杰出的人才。
Lex Fridman (01:44:22) Yeah, it’s grounded in a good instinct of compassion, you don’t want people who suck at a thing to feel pain but it seems like struggle is a part of life, either you do it earlier or you do it later. And it’s true, that’s such a good point that competition does seem to be a really powerful driver of skill development, like you mentioned, pursuing mastery. There’s something in human nature that, especially for young people, if you can compete at a thing, you’re going to be really driven to get good at that thing. If you can direct that in the education system as BigEast does, as many nations like you mentioned do, then you’re going to develop a lot of brilliant people.
莱克斯·弗里德曼 (01:45:00) … 那样做，那么你将培养出许多杰出的人才，有韧性的人，准备好在这个世界上创造史诗般成就的人。
Lex Fridman (01:45:00) … do, then you’re going to develop a lot of brilliant people, resilient people, people that are ready to create epic shit in the world.
帕维尔·杜罗夫 (01:45:07) 我认为有大量证据证明我们在生物学上就被设定为要竞争，并建立我们对自己品质和才能的理解，这种理解是相对于我们周围其他人的，这是社会自我调节的方式之一。
Pavel Durov (01:45:07) I think there is a lot of evidence proving that we are biologically wired to compete and establish our understanding of what our qualities are and talents are in relation to other people around us, and this is one of the ways society self-regulates.
尼古拉·杜罗夫
Nikolai Durov
莱克斯·弗里德曼 (01:45:30) 说到竞争，你的哥哥尼古拉，他是一位数学家、程序员、密码学专家。他赢得了IMO国际数学奥林匹克竞赛，获得了三次金牌，ICPC编程竞赛，两次获奖，拥有两个数学博士学位，你们已经合作多年，创造了我们一直在谈论的令人难以置信的技术。那么你从你哥哥身上学到了关于生活的什么？
Lex Fridman (01:45:30) Speaking of competition, your brother, Nikolai, he’s a mathematician, programmer, expert in cryptography. He has won the IMO International Mathematics Olympiad, he got gold medal three times, ICPC programming, two times, has two PhDs in mathematics, and you have worked together for many years creating incredible technologies that we’ve been talking about. So what have you learned about just life from your brother?
帕维尔·杜罗夫 (01:46:02) 嗯，首先，我必须说我从我哥哥那里学到了几乎所有的东西，我所知道的一切，因为当我们还是孩子的时候，我们睡在同一间卧室里，床相隔几英尺，我不停地用问题打扰他。我会问他关于恐龙、星系、黑洞和尼安德特人的问题，所有我能想到的东西，他是我在那个我们还不能上网的时代里的维基百科。他是一个独特的神童，可能是十亿里挑一的。
Pavel Durov (01:46:02) Well, first of all, I must say I learned pretty much everything from my brother, everything I know, because when we were used to be kids, we slept in the same bedroom, like beds a few feet away from each other, and I kept bugging him with questions. I would ask him about dinosaurs and galaxies and black holes and Neanderthals, everything I could think of, and he was my Wikipedia back in the time when we didn’t have internet access. He’s a unique prodigy kid, probably one of a billion.
(01:46:45) 我想他三岁就开始阅读，而且他在数学上进步得非常快，到了六岁的时候，他已经能阅读非常复杂的天文学书籍了。有时当他在公共场所这样做时，比如公共汽车或地铁上，我妈妈会受到目睹此事的人的批评。他们会对她说：”你为什么用这本严肃的书来戏弄你自己的小孩？很明显这孩子不能理解里面的所有东西。这太复杂了，连我们都不理解里面的任何东西。里面有一些公式，”而他当时已经在吸收这些知识了。他就是有这种对信息的渴求。
(01:46:45) He started reading at the age of three, I think, and he pretty fast got so advanced in maths, that by the age of six, he could already read really sophisticated books on astronomy. Sometimes when he did it in public places, like buses or metro, my mom was criticized by people who were witnessing it. They would tell her, “Why are you mocking your own kid with this serious book? It’s obvious that the kid can’t understand everything there. It’s too complicated even we don’t understand anything there. There’s some formulas,” and he was already sucking in this knowledge. He just has this thirst for information.
(01:47:39) 所以他是各种伟大事实、有用东西、鼓舞人心事物的来源。他教给了我几乎我所知道的一切。同时，他极其谦虚和善良，我认为这是许多自以为聪明但缺乏普遍智慧的人所欠缺的。大多数情况下，真正聪明的人，他们也善良且富有同情心。
(01:47:39) So he was the source of all kind of great facts, useful things, inspiring things. He taught me pretty much everything I know. At the same time, he’s incredibly modest and kind, and this is something I think a lot of people that think they’re smart but not generally intelligent lack. More often than not, people who are truly intelligent, they’re also kind and compassionate.
莱克斯·弗里德曼 (01:48:18) 他是那样的人吗？
Lex Fridman (01:48:18) And he is that?
帕维尔·杜罗夫 (01:48:20) 绝对是。
Pavel Durov (01:48:20) Definitely.
莱克斯·弗里德曼 (01:48:21) 你实际上大部分时间都远离公众视线。你很少接受采访，相当低调，但你的哥哥是另一个级别的。他一直远离公众视线。这背后是什么原因？
Lex Fridman (01:48:21) You actually have been staying out of the public eye for the most part. You’ve done very few interviews, you’re pretty low-key, but your brother is in another level. He’s been staying out of the public eye. What’s behind that?
帕维尔·杜罗夫 (01:48:34) 部分原因是他天生的谦逊。他不需要这样做。他没有那种炫耀、吹嘘事情的冲动。我也试图避免，但在某个时刻，我意识到我过于私密、过于隐秘变成了一种负担，因为它制造了这种空白，这种空虚，那些非常不喜欢 Telegram 的个人和组织愿意用不准确的信息来填补，他们愿意传播关于 Telegram 的叙事，这可能导致奇怪的情况，其中一些我们之前讨论过。例如，那个法国的调查。
Pavel Durov (01:48:34) Part of it is his natural modesty. He doesn’t need to do it. He doesn’t feel this urge to show off, brag about stuff. I tried to avoid it as well, but at a certain point I realized that me being too private, too secretive becomes a liability because it creates this void, this emptiness that people and organizations that don’t like Telegram very much are willing to fill with inaccurate information and they’re willing to spread the narratives about Telegram, which can result in strange situations, some of which we discussed earlier. For example, this French investigation.
莱克斯·弗里德曼 (01:49:32) 是的，我越来越了解你，你身上有一种深刻的诚信，我认为向世界展示这一点是好的。用户隐私面临很多攻击向量，我认为最重要、最后的保护墙是实际运营公司的人，所以在某种程度上，你出现在那里展示真实的自我是很重要的。
Lex Fridman (01:49:32) Yeah, I’ve gotten to know you more and more and there’s a deep integrity to you that I think is good to show to the world. There’s a lot of attack vectors on user privacy and I think the most important, the last wall of protection is the actual people that are running the company, so it’s important to some degree for you to be out there to showing your true self.
编程和电子游戏
Programming and video games
(01:49:55) 所以我们应该说一下，你虽然没有提到，但你从小就是一名程序员。你10岁开始编码。你最早构建的东西是在11岁时制作的一款电子游戏，然后最终在10年后，21岁时，你独立编程了VK的最初版本。你能跟我谈谈你的编程历程，以及它如何引导到VK的创建吗？VK的技术栈是什么？主要是PHP吗？你是如何学会编程网站，所有这一切的？
(01:49:55) So we should say that also you didn’t mention, but you were a programmer from an early age. You started coding at 10. First things you built are a video game at 11, and then eventually 10 years later, 21, you programmed the initial versions of VK single-handedly. Can you talk to me about your programming journey that led to the creation of VK? What was the VK stack? Is it PHP mostly? How did you figure out how to program websites, all of that?
帕维尔·杜罗夫 (01:50:27) 是的，我起初可能对网站不那么感兴趣。我10岁的时候甚至还不能上网，但我喜欢电子游戏。我没有足够的游戏，这种稀缺性迫使我开始制作它们，更多的是电脑游戏，只是为了自己玩。
Pavel Durov (01:50:27) Yeah, I wasn’t as interested in probably websites at first. I didn’t even have access to the internet when I was 10 years old, but I liked video games. I didn’t have enough of them and the scarcity forced me to start building them, more computer games, just to play myself.
(01:50:49) 这实际上是一件有趣的事情，我们有时没有意识到，但稀缺导致创造力，来自苏联或其他无法获得太多现代技术，更重要的是现代娱乐的地方，有那么多热爱编码的人的原因之一是，也许我们没有被所有这些丰富的不同娱乐选项所分散注意力，这并不是说拥有这些选项是坏事。这只是我们有时没有意识到的一个事实。
(01:50:49) It’s actually an interesting thing that we sometimes don’t realize it, but scarcity leads to creativity, and one of the reasons you have so many people who love to code coming from the Soviet Union or other places which didn’t have much access to modern technology, and more importantly modern entertainment, is that perhaps we were not so much distracted by all this abundance of different entertainment options, which is not to say it’s bad to have those options. It’s just a fact that we sometimes don’t appreciate.
(01:51:34) 所以我开始构建电脑游戏。我哥哥有时会指导我。例如，我会创建一个回合制策略游戏。当然是二维的。那时候三维对我来说太难了。但在滚动FPS、每秒帧数参数方面，它不够流畅，我问我哥哥如何优化它。他会指导我，这种学习和训练在我年轻时就塑造了我的编码技能。
(01:51:34) So I started to build computer games. My brother would sometimes guide me. For example, I would create a turn-based strategy. Of course, two-dimensional. Back then three-dimensional was too much for me. But it wasn’t as slick in terms of the scrolling FPS, frames per second, parameter, and I asked my brother how to optimize it. He would guide me, and this kind of learning and training really shaped my coding skills when I was younger.
(01:52:21) 然后我开始为我的同学们创建电子游戏，例如，当我们在课间休息时在教室里玩无限区域上的井字游戏时。不是三子连珠的井字游戏，这是关于五子连珠，并且是在一个无限区域上。这是一个有趣得多的游戏，如果你继续玩下去，它会变得相当复杂。我的同学们曾经很喜欢它，我的一些同学非常聪明，是数学奥林匹克竞赛的冠军，是大学里教授们的子女，我决定，”不，我想每次都赢。我甚至一次都不想输。那么我怎么能赢呢？我需要更多练习，但我怎么才能更多练习呢？我需要一个比我更强的对手。”
(01:52:21) Then I started to create video games for my classmates when we played, for example, tic-tac-toe on an infinite field in my class during the breaks. And not tic-tac-toe the three in a row, this was about five in a row and in an infinite field. This is a much more interesting game and it gets quite complicated if you keep playing it. My classmates used to love it and some of my classmates were really smart, champions of math olympiads, sons and daughters of professors at the university, and I decided, “No, I want to win every single time. I don’t want to lose even a single time. So how do I win? I need to practice more, but how do I practice more? I need an opponent stronger than myself.”
(01:53:08) 所以我编写了这个游戏，这样我就可以和电脑对战，电脑会计算，我想，提前四步来选择最优策略。那还不够。提前四步，我仍然能赢它。如果我尝试计算五步或六步，那就太慢了，所以请我哥哥帮我。于是他做了这个算法。最终，我训练自己每次都赢，即使和电脑对战，那时候我们没有现代CPU，我仍然可以保持一些自信。
(01:53:08) So I coded this game so that I would play against the computer and the computer would calculate, I think, four moves in advance to choose the optimal strategy. That wasn’t enough. Four moves in advance, I would still win over it. If I tried to calculate five or six, it was too slow, so asked my brother to help me out here. So he made this algorithm. Eventually, I trained myself to win every single time, even with the computer back then, we didn’t have modern CPUs, and I could still retain some self-confidence.
(01:53:54) 我会在课间休息时回到学校，和我的同学们玩，很快人们就开始失去兴趣。我的同学中没有人再想玩这个游戏了。我毁掉了这个游戏，因为…
(01:53:54) I would go back to school during breaks, play with my classmates, and soon people started to lose interest. None of my classmates wanted to play this game anymore. I killed the game because there’s…
VK 起源与工程
VK origins & engineering
(01:54:09) 在那之后，当我进入圣彼得堡国立大学时，仅仅学习是相当无聊的，因为太容易了。所以我想，”我在那里能做什么？” 我首先为我所在院系的学生创建了一个网站。我组织了将所有考试的答案和所有讲座的数字化版本创建起来，这在那时是非常独特的。记住，那是25年前。我会建立一个网站，在那里发布所有这些材料，很快它就变得非常受欢迎。我在那里开了一个讨论论坛。几年内，我将其扩展到了大学的所有其他部门，然后扩展到其他大学。最终我们拥有了数万用户，仅仅作为一个学生门户。我们在那里有各种社交功能，好友列表、相册、个人资料、博客。应有尽有。
(01:54:09) So after that, when I got into the St. Petersburg State University, it was quite boring just to study because it was too easy. So I thought, “What can I do there?” I created a website for the students of my faculty first. I organized the creation of digital answers to all exams and digitalized version of all lectures, which was something very unique back then. Remember, it was 25 years ago. I would put together a website where I would publish all this materials, and pretty soon it became super popular. I opened a discussion forum there. In a few years, I expanded to the university with all of its other departments, and then to other universities. We ended up having tens of thousands of users just as a student’s portal. We had all kinds of social features there, friends lists, photo albums, profiles, blogs. All of it.
(01:55:29) 它相当成功，在我大学毕业后，我的一位中学同学在看到圣彼得堡主要商业报纸上关于我成功的报道后联系了我，他问我，”你是在尝试打造一个俄罗斯的Facebook吗？” 我说，”我不确定。什么是Facebook？” 于是我们见面了。由于他两年前从一所美国大学毕业，他向我展示了Facebook。我想，”嗯，我已经拥有了所有这些技术，但知道我应该去掉哪些元素以便扩展这个东西并拥有数百万用户是很有价值的。”
(01:55:29) It was quite successful, and after I graduated the university, one of my ex-classmates from the school reached out to me after reading about my successes in a newspaper, the main business newspaper of St. Petersburg, and he asked me, “Are you trying to build a Russian Facebook?” I said, “I’m not sure. What’s Facebook?” So we met. Since he graduated an American university two years before that, he showed me Facebook. I thought, “Well, I can’t already have all of this technology, but it’s valuable to know which elements I should get rid of in order to scale this thing and have millions of users.”
(01:56:25) 这也是人们没有意识到的一点，有时候为了向前发展并获得更多成功，你必须舍弃一些东西，包括技术。舍弃功能非常重要。
(01:56:25) This is also something people don’t appreciate that sometimes in order to move forward and have more success, you have to get rid of things, including technology. Getting rid of features is super important.
莱克斯·弗里德曼 (01:56:40) 简化，既是为了扩展，也是为了使其易于用户基础增长，让人们能立即理解。
Lex Fridman (01:56:40) Simplify, both for scaling and for making it amenable to just growing the user base where people get it immediately.
帕维尔·杜罗夫 (01:56:50) 是的。否则，对新用户来说就太复杂了。现有用户会很高兴，他们会赞美你，他们会要求你添加更多东西，让它变得更复杂，所以如果你只依赖现有用户的反馈，很容易迷失方向，不知所措。
Pavel Durov (01:56:50) Yes. Otherwise, it’s just too complicated for the new user. The existing users will be happy, they’ll be praising you, they will be asking you to add more stuff to make it even more complicated, so it’s easy to lose track and get disoriented if you are only relying on the feedback of existing users.
(01:57:18) 因此，我启动了一个名为 VKontakte 或 VK 的网站，在俄语中意思是”保持联系”，最初是为了解决我个人的问题。我那一年大学毕业，我想与我在大学里的老同学和其他同学保持联系。当然，作为一个20岁的年轻人，我也想认识其他人，包括漂亮的女孩。
(01:57:18) So as a result, I started the website called VKontakte or VK, it means “in touch” in Russian, initially to solve my own personal problem. I graduated the university that same year and I wanted to be in touch or remain in touch with my ex-classmates from the university and the other fellow students. And of course, as a 20-year-old, I wanted to meet other people, including good-looking girls.
(01:57:46) 所以我从头开始构建它。对于这个项目，我想，”我不会使用任何第三方库、模块，因为我想让它尽可能高效。” 我痴迷于每一行代码，但那么庞大的项目该如何开始呢？我之前没有任何创建那种规模项目的经验，那将涉及所有方面。以前，我会重用一些现有的解决方案。在这里，我想从头开始构建。
(01:57:46) So I started to build it from scratch. For that one, I thought, “I’m not going to use any third-party libraries, modules because I want to make it as efficient as possible.” I was obsessing over every line of code, but then how do you start something that large? I didn’t have any prior experience creating a project of that scale, which would involve everything. Before, I would reuse some existing solutions. Here, I wanted to build from scratch.
(01:58:26) 所以我打电话给我哥哥。他当时是德国马克斯·普朗克大学的博士后，我问他，”我应该从哪里开始？” 他告诉我，”就构建一个用户认证模块，只是登录，甚至不用登出，只是登录，因为你可以用凭证、邮箱和密码预填充数据库。这并不重要。但一旦你看到你可以输入你的密码和邮箱，然后你进去了，它用你的名字对你说’你好’，那么你就会清楚地知道该从哪里继续了。”
(01:58:26) So I called my brother. He was a post-doc student in Germany at the time in the Max Planck university, and I asked him, “What should I start from?” And he told me, “Just build a module to authorize users, just to log in, not even to sign out, just to log in because you can pre-populate the database with credentials and emails and passwords. It doesn’t really matter. But once you see that you can type in your password and email and you are in and it tells you, ‘Hello,’ using your name, then you will have a clear understanding where to go from there.”
莱克斯·弗里德曼 (01:59:22) 是的。我的意思是，确实如此。
Lex Fridman (01:59:22) Yeah. I mean, that’s true.
帕维尔·杜罗夫 (01:59:24) 这是我一生中收到的最好的建议之一。顺便说一句，它效果非常好。我开始构建它，在我意识到之前，我就在网站上有了相册、私信、这个留言簿。我们过去在VK上，我猜在Facebook早期，称之为”墙”。我们最终构建了比当时Facebook更复杂、功能更多的东西。
Pavel Durov (01:59:24) That’s one of the best advice I’ve ever got in my life. It worked perfectly, by the way. I started to build it and before I knew it, I would have there on the website photo albums, private messages, this guest book. We used to call it “thee wall” back on VK and I guess in the early days of Facebook. We’d end up building something even more sophisticated than Facebook at the time with more features.
(01:59:54) 我当时有个女朋友。我请她，”我们需要以某种方式建立所有俄罗斯学校和大学以及系和细分部门的数据库。” 她做得非常出色，尝试在线搜集所有这些信息，有时给大学写邮件说，”你们目前到底有哪些系？我们需要知道，”或者联系教育部，在俄罗斯，然后在乌克兰，最终在白俄罗斯和哈萨克斯坦以及其他VK最终成为最大和最受欢迎的社交网络的国家。
(01:59:54) I had a girlfriend at the time. I asked her, “We need to somehow come up with a database of all Russian schools and universities and the departments and subdivisions.” She did a great job trying to source all this information online or sometimes writing emails to universities saying, “Which departments do you have exactly at this point? We need to know,” or reaching out to the Department of Education, but in Russia and then in Ukraine, and then eventually in Belarus and in Kazakhstan and other countries where VK ended up to be the largest and most popular social network.
(02:00:38) 所以我们当时做了一些相当独特的事情，在最初的将近一年里，我是公司唯一的员工。我是后端工程师、前端工程师、设计师。我是客服人员。我也是市场人员，想出所有的措辞和公告，想出推广VK的竞赛，这些效果相当好。那是一段不可思议的经历，让我了解了社交网络平台的方方面面。
(02:00:38) So we did a few things that were quite unique at the time, and for the first almost a year, I was the single employee of the company. I was the backend engineer, the front-end engineer, the designer. I was the customer support officer. I was the marketing guy as well, coming up with all the wordings and the announcements, coming up with competitions to promote VK, which worked quite well. That was an incredible experience that gave me knowledge of every aspect of a social networking platform.
莱克斯·弗里德曼 (02:01:30) 也理解了单个人能做多少事。
Lex Fridman (02:01:30) Also understanding of how much a single person can do.
帕维尔·杜罗夫 (02:01:32) 完全正确。这是为什么我喜欢认为自己是Telegram内部高效的项目经理和产品经理的原因之一，因为我只会向我的团队成员要求雄心勃勃的截止日期。如果有人给我，”哦，我需要三周时间来做那个，”我总是回答，”嗯，我构建VK的第一个版本只用了两周。你为什么需要三周？看起来像是你三天就能完成的事情。三周？除了这三天，剩下的三周你打算做什么？”
Pavel Durov (02:01:32) Exactly. It’s one of the reasons why I’d like to think I’m an efficient project manager and product manager inside Telegram because I will not take anything but ambitious deadlines from my team members. If somebody gives me, “Oh, I need three weeks to do that,” I always reply, “Well, I built the first version of VK in just two weeks. Why would you need three weeks? It seems like something you could make real in just three days. Three weeks? What are you going to do the rest of the three weeks apart from this three days?”
(02:02:18) 团队了解我，这就是为什么我们今天，Telegram，能够以非常快的创新步伐前进。每个月我们都会推出几个有意义的功能，我认为在这个行业中，在短时间内能完成的事情方面，我们超越了所有其他人。所以是的，那段经历是无价的。
(02:02:18) And the team knows me, and that’s why we are able today, Telegram, to move at a very good pace of innovation. Every month we’re pushing several meaningful features, I think out-competing everybody else in this industry in terms of what you can do within a short timeframe. So yes, that experience was invaluable.
(02:02:52) 至于技术栈，我从PHP和MySQL，Debian Linux开始，但很快我意识到，”我需要优化这个。” 我开始使用Memcached。Apache服务器不够用了。我们不得不设置NGINX。我哥哥当时还住在德国，所以在构建VK的第一年里，他帮不了我太多。有时我设法通过电话联系到他。我会用老式的有线电话打给他。我说，”我该怎么办？我怎么安装这个叫NGINX的东西？我不是Linux专家。” 如果他那天感觉特别好心并且不太忙，他会告诉我怎么做或者亲自设置，但大多数情况下，我不得不只依靠自己。
(02:02:52) As for the stack, I started from PHP and MySQL, Debian Linux, but very soon I realized, “I need to optimize this.” I started using Memcached. Apache servers were not enough anymore. We had to set up NGINX. And my brother was still living in Germany, so he couldn’t help me much for the first year of building VK. Sometimes I would manage to get through to him through a call. I would use an old-school phone to call him with wires. I said, “What do I do? How do I install this thing called NGINX? I’m not a Linux guy.” If he felt particularly kind that day and not too busy, he would show me the way to do it or set it up himself, but for the most part, I had to rely on just myself.
(02:03:53) 然而，有他在那里对我们开始快速增长并开始扩展时有帮助，因为起初，你意识到，”一台服务器不够了。我需要再买一台。然后一台又一台。” 数据库应该在另一台服务器上。然后你必须把数据库分成表。然后你必须想出一种方法，根据某种不会破坏你用户体验的有意义的标准来对表进行分片。
(02:03:53) Having him there though helped when we started to grow fast and started to scale it, because at first, you realize, “One server is not enough. I need to buy another one. Then another one and another one.” The database should be in a different server. Then you have to split the database into tables. Then you have to come up with a way to chart the tables using some criteria that would make sense that wouldn’t break your user experience.
(02:04:28) 当我们用户超过一百万，服务器超过十几台时，没有我哥哥在扩展方面的投入，生存下去变得不可能。我记得请他回来，”你需要帮我处理这个东西。它开始变得非常庞大了。” 更糟糕的是，自从我们变得流行起来，有人开始对我们进行DDoS攻击，就像经常发生的那样。然后我们有人想购买VK的股份，有趣的是，每次我们有谈判的日子，DDoS攻击就会加剧，所以我们不得不找出应对的方法。我记得有很多个不眠之夜试图解决它。
(02:04:28) When we got to over a million users and beyond a dozen of servers surviving without the input from my brother in terms of taking care of the scaling aspect, it became impossible. I remember asking him to come back, “You need to help me with this thing. It’s starting to be really big.” What was worse is that since we became popular, somebody started to do DDoS attacks on us, as it always happens. And then we had people that wanted to buy a share of VK, and interestingly, every time we had a negotiation day, the DDoS attacks intensified, so we had to come up with a way to fight it. I remembered having many sleepless nights trying to figure it out.
莱克斯·弗里德曼 (02:05:30) 所以这是你初次接触各种恶意行为者，DDoS，商业。然后后来你会发现还有政治这种东西，然后后来，地缘政治。但这是初始阶段，不仅仅是创造酷炫的东西，还必须应对，就像你现在必须应对Telegram一样，是大量的恶意行为者试图测试系统的极限，试图破坏系统。
Lex Fridman (02:05:30) So that was your introduction to all kinds of bad actors, DDoS, business. Then later you’d find out there’s such a thing called politics, and then later, geopolitics. But this is the initial stages, that it’s not just about creating cool stuff, it’s having to deal with, as you now have to deal with with Telegram, is seas of bad actors trying to test the limits of the system, trying to break the system.
帕维尔·杜罗夫 (02:06:02) 不幸的是。如果我们没有恶意行为者和压力，那将是最好的工作。你只需要创造。
Pavel Durov (02:06:02) Unfortunately. If we didn’t have bad actors and pressure, it would be the best job ever. You just get to create.
莱克斯·弗里德曼 (02:06:12) 是的，是的。所以你哥哥的帮助，就像你提到的NGINX和对表进行分片，这些扩展问题有些是算法性质的。这几乎像是理论计算机科学。所以不仅仅是买更多电脑，而是要弄清楚如何在算法上让一切运行得极快，所以其中一些是数学。一些是纯粹的工程，但一些是数学。
Lex Fridman (02:06:12) Yeah, yeah. And so the help from your brother, like you mentioned NGINX and charting the tables, some of this scaling issue is algorithmic nature. It’s almost like theoretical computer science. So it’s not just about buying more computers, it’s figuring out how to algorithmically make everything work extremely fast, so some of it’s mathematics. Some of it is pure engineering, but some of it is mathematics.
帕维尔·杜罗夫 (02:06:44) 是的。所以在那个阶段，我能做基本的事情。我能理解如何将可扩展性实现到代码库中，如何对数据库中的表进行分片，在哪里使用Memcached而不是直接请求数据库。那在当时还挺容易的，因为那时候还是PHP。
Pavel Durov (02:06:44) Yeah. So at that stage, I could do the basic stuff. I could understand how I implement scalability into the code base, how I chart my tables in the database, where I include Memcached instead of direct requests to the database. That was quite easy because it was still PHP back in the day.
(02:07:14) 当我哥哥在2008年左右从德国回来时，我问他，”我们能让它更高效吗？我们能让它超级快，同时让我们需要更少的服务器来维持负载吗？” 他说，”可以，但PHP不够。我将不得不用C和C++重写你数据引擎的大部分。” 我说，”好的，我们干吧。”
(02:07:14) When my brother got back from Germany somewhere around 2008, I asked him, “Can we make it even more efficient? Can we make it super fast and at the same time so that we would require even fewer servers to maintain the load?” And he said, “Yes, but PHP is not enough. I’ll have to rewrite big part of your data engines in C and C++.” I said, “Okay, let’s do that.”
(02:07:47) 他邀请了他的一个朋友来帮忙，另一个世界编程竞赛的绝对冠军，连续两次，他们一起构建了第一个定制化的数据引擎，它比仅仅依赖MySQL和Memcached要高效得多，因为首先，它更专业，更底层。
(02:07:47) He invited a friend of his to help him, another absolute champion in world’s programming contest, twice in a row, and they put together the first customized data engine, which was far more efficient than just relying on MySQL and Memcached because it was, first of all, more specialized, more low-level.
莱克斯·弗里德曼 (02:08:19) 所以他们用C, C++重写了它？
Lex Fridman (02:08:19) So they rewrote it in C, C++?
帕维尔·杜罗夫 (02:08:21) 一大部分。例如，搜索，广告引擎，因为VK有定向广告，他们构建了那个。他们做的非常高效。最终，私信部分，公共消息部分。在某个时刻，我们意识到线上很少有网站比VK加载更快。
Pavel Durov (02:08:21) A large chunk of it. For example, the search, the ad engine, because VK had targeted ads, they built that. It was very efficient what they did. Eventually, the private messaging part, the public messages part. At some point, we realized there are very few websites online that load faster than VK.
莱克斯·弗里德曼 (02:08:48) 很好。
Lex Fridman (02:08:48) Nice.
帕维尔·杜罗夫 (02:08:49) 我记得在2009年，我去了硅谷，第一次见到了马克·扎克伯格和Facebook早期的一些其他核心团队成员。记住，Facebook当时才四五岁。每个人都不断问我，”为什么即使在硅谷这里，VK加载也比Facebook快？你网站上的一切似乎都是瞬间出现的。秘诀是什么？” 这是让他们非常好奇的事情之一。
Pavel Durov (02:08:49) I remember in 2009, I went to Silicon Valley and I met Mark Zuckerberg the first time and some of the other core team members of early Facebook. Remember, Facebook was just four or five years old. And everybody kept asking me, “How come even here in Silicon Valley, VK loads faster than Facebook? Everything seems to appear instantly on your website. What’s the secret sauce?” That was one of the things that made them very curious
莱克斯·弗里德曼 (02:09:25) 这一直对你很重要，要有非常低的延迟，确保东西加载快，因为这是Telegram真正出名的地方之一。即使在烂连接上等等，它也能运行得极快。一切都很快。
Lex Fridman (02:09:25) And that was always important to you, to have very low latency to make sure the thing loads because that’s one of the things Telegram is really known for. Even on crappy connections and all that kind of stuff, it just works extremely fast. Everything is fast.
帕维尔·杜罗夫 (02:09:37) 作为核心技术理念之一，我们优先考虑速度。我们认为人们能注意到差异，即使只是5000万毫秒的差异。这种差异是潜意识的。它也让我们不仅在速度更快、响应更及时方面有优势，而且在基础设施、开支方面也更高效。因为如果你的代码执行得更快，意味着你需要更少的计算资源来运行它。
Pavel Durov (02:09:37) As one of the core technological ideas, we prioritize speed. We think that people can notice the difference, even if it’s just 50 million millisecond difference. The difference is subconscious. It also allows us not just to be faster and more responsive, but also more efficient when it comes to the infrastructure, the expenses. Because if your code executes faster, it means you need fewer computational resources to run it.
(02:10:16) 所以让东西更快你不可能输，这就是为什么我们在招聘人员时一直非常谨慎。我只会雇佣我最终确定是最佳选择的人，因为如果你雇佣了一个可能有点分心、缺乏经验的人，你最终可能会在代码库中引入低效，导致数千万美元的损失。想想这个责任，就像如果我们从VK时代跳到今天，Telegram被超过十亿人使用。他们每天打开它几十次。想象一下，如果应用打开时有轻微的延迟，比如半秒的延迟。乘以几十次，再乘以十亿。这是人类无缘无故就损失了几个世纪、几千年的时间，仅仅因为马虎。
(02:10:16) So there is no way you can lose in making things faster, and that’s why we have always been very careful when hiring people. I would only hire a person if I’m ultimately certain is the best option because if you hire somebody who is maybe a little bit distracted, unexperienced, you may end up with inefficiencies in your code base that results in tens of millions of dollars of losses. And think about the responsibility, like if we jump to today from the VK days, Telegram is used by over a billion people. They open it dozens of times every day. Imagine the app opens with a slight delay, say, half-a-second delay. Multiply by dozens of times by a billion. It’s centuries, millennia lost for humanity without any reason other than just being sloppy.
雇佣优秀团队
Hiring a great team
莱克斯·弗里德曼 (02:11:24) 理解这一点非常重要，也非常明智，实际上，如果你作为一个开发者只是有点粗心，你可能会引入难以追踪的低效，因为你不知道它可以更快。代码不会对你尖叫说，”这个可以快得多。” 所以你必须实际上，作为一个工匠，在编写代码时非常小心，并且总是思考，”这个能更高效地完成吗？” 而且可能只是很小的事情，因为它们都会在整个代码中传播，所以在公司任何地方有一个粗心的开发者都会带来真正的成本，因为他们可能引入那种低效，而所有其他开发者不会知道。他们会以为本来就该那样。
Lex Fridman (02:11:24) That is so important to understand and so wise that it’s actually, if you’re just a little bit careless as a developer, you can introduce inefficiencies that are going to be very difficult to track down because you don’t know that it can be faster. The code doesn’t scream at you saying, “This could be much faster.” So you have to actually, as a craftsman, be very careful when you’re writing a code and always thinking, “Can this be done much more efficiently?” And it can be tiny things because they all propagate throughout the code, and so there’s a real cost in having a careless developer anywhere in the company because they can introduce that inefficiency and all the other developers won’t know. They’ll just assume it kind of has to be that way.
(02:12:11) 所以每个构建像Telegram这样的应用任何组件的个体开发者都有真正的责任，要总是问，”好吧，这个能更高效地完成吗？这个能更简单地完成吗？” 而这是编程最美丽的方面之一，艺术形式，对吧？
(02:12:11) So there’s a real responsibility for every single individual developer that’s building any component of an app like Telegram to just always ask, “Okay, can this be done more efficiently? Can this be done more simply?” And that’s one of the most beautiful aspects, the art forms of programming, right?
帕维尔·杜罗夫 (02:12:32) 哦，是的，因为当你设法发现一种简化事物、让它们更高效的方法时，你会感到难以置信的快乐、自豪和成就感。
Pavel Durov (02:12:32) Oh, yes, because when you manage to discover a way to simplify things, make them more efficient, you feel incredibly happy and proud and accomplished.
(02:12:47) 说到你的观点，我能回忆起我职业生涯中的几个例子，解雇一名工程师实际上导致了生产力的提高。假设你有两个安卓工程师构建他们的应用，然后他们就是做不到。他们跟不上功能发布的时间表。你想，”我可能得雇佣第三个，”但然后你注意到其中一个真的很奇怪，落后于进度，有时抱怨，不承担责任。你问，”那么，如果我直接解雇这个人会怎样？” 你解雇了这个人。几周后，你意识到你实际上不需要任何新人，从来就不需要第三个工程师。问题在于这个家伙，他制造的问题比他解决的还多。
(02:12:47) And to your point, I can recall a few instances in my career where firing an engineer actually resulted to an increase in productivity. Say you have two Android engineers building their app and then they just can’t make it. They’re not keeping up with the pace of the feature release schedule. And you think, “I probably have to hire a third one,” but then you notice that one of them is really weird, falling behind the schedule, complaining some of the time, doesn’t assume responsibility. And you ask, “So what if I just fire this person?” And you fire this person. In a few weeks, you realize you actually don’t need any new, never needed the third engineer. The problem was this guy who created more issues and more problems than he solved.
(02:13:49) 这太违反直觉了，因为在开发技术项目时，我们倾向于认为你只需要把更多的人扔进某件事，然后事情就会奇迹般地自行解决，仅仅因为更多的人意味着他们现在投入更多注意力。
(02:13:49) That is so counterintuitive because in developing tech projects, we tend to think that you just throw more people into something and then things get solved miraculously by themselves just because more people means more attention from them now.
莱克斯·弗里德曼 (02:14:12) 这，再次，极其有力。史蒂夫·乔布斯谈论过A级玩家和B级玩家，当你拥有B级玩家，也就是你谈到的那类人，加入团队时，会发生一些事情，他们 somehow 会让每个人都慢下来。他们让每个人失去动力。而且非常违反直觉的是，你基本上，创建优秀团队工作的一部分是移除B级玩家。不仅仅是雇佣更多人，一般来说。是找到”A级玩家”并移除那些拖慢事情的人。
Lex Fridman (02:14:12) That’s, again, extremely powerful. Steve Jobs talked about A players and B players, and there’s something that happens when you have B players, which is like the folks you’re talking about. Introduced into a team, they can somehow slow everybody down. They demotivate everybody. And it’s very counterintuitive that you basically, part of the work of creating a great team is removing the B players. It’s not just hiring more, generally speaking. It’s finding the “A players” and removing the people that are slowing things down.
帕维尔·杜罗夫 (02:14:48) 哦，是的，因为人们没有意识到的另一件事是，与B级玩家一起工作是多么令人沮丧。每个人都能分辨出对方，他们与之共事的另一个工程师是否真的有能力。如果这个人不舒服，这是非常明显的。他们问错误的问题，他们不断落后。在某个时刻，如果你是一个A级玩家，你会感到这种不满，这种感觉，你无法实现你的全部潜力，完成你真正应该完成的事情，就是因为这个人就在你旁边工作或假装在你旁边工作。
Pavel Durov (02:14:48) Oh, yes, because the other thing that people don’t realize is how demotivating working with a B player is. Everybody can tell if the other person, the other engineer they’re working with is really competent. And it’s very visible if the person is not comfortable. They’re asking the wrong questions, they keep lagging behind. And at a certain point, if you’re an A player, you get this dissatisfaction, this feeling that you are not able to realize your full potential, accomplish what you’re really meant to accomplish because of this person working next to you or pretending to work next to you.
(02:15:37) 顺便说一下，在某些情况下，不是因为这个人懒。在某些情况下，只是心智、智力能力不够。这不是关于经验。大多数情况下是关于天生能力和毅力。在90%的情况下，只是无法长时间专注于一项任务。不是每个人都有这种能力。所以对于有这种能力的人来说，与一个分心、无法深入他们负责的项目的人一起工作是一种侮辱。
(02:15:37) And by the way, in some cases, it’s not because the person is lazy. In some cases it’s just the mental, the intellectual ability is not there. It’s not about experience. Most often it’s about natural ability and persistence. In 90% of cases, it’s just the inability to focus on one task for an extended period of time. Not everybody has this ability. So for people who do have this ability, it’s an insult to work alongside someone who is distracted and cannot go deep in the projects that they’re responsible for.
莱克斯·弗里德曼 (02:16:27) 关于这个小插曲，你的招聘流程是怎样的？你已经展示并谈到你经常使用竞赛，编码竞赛来招聘，以找到优秀的工程师。这背后的想法是什么？
Lex Fridman (02:16:27) On this small tangent, what’s your hiring process? So you’ve shown and you’ve talked about how you use competitions often, coding competitions to hire to find great engineers. What’s your thinking behind that?
帕维尔·杜罗夫 (02:16:40) 嗯，这符合我的整体理念。我认为竞争带来进步。如果你想为某些你心中有数的特定任务选择最合格的人建立一个理想的流程，有什么能比竞赛更好呢？一个编码比赛，每个想加入你公司做工程师，或者只是想获得一些奖金或认可的人，都可以展示他们的技能，然后我们只选择最好的。或者如果我们不确定，因为数据不足而无法雇佣某人，我们就用另一个任务重复比赛，获取更多数据，获取更多获胜者，然后再重复。
Pavel Durov (02:16:40) Well, it’s in line with my overall philosophy. I think competition leads to progress. If you want to create an ideal process for selecting the most qualified people for certain specific tasks you have in mind, what can be better than a competition? A coding contest where everybody who wants to join your company as an engineer or just wants to get some prize money or validation can demonstrate their skills, and then we just select the best. Or if we are not certain because there’s not enough data to hire somebody, we just repeat the contest with another task, get more data, get more winners, then repeat again.
(02:17:31) 在某个时刻，你意识到，”哦，实际上这个家伙从16岁或14岁起就参加了我们的10次比赛。现在他20或21岁了。他在其中八次比赛中获胜。他似乎在JavaScript、安卓、Java以及C++方面真的很棒。为什么不雇佣这个人呢？” 这里有一致性。
(02:17:31) And at some point, you realize, “Oh, actually this guy has competed in 10 of our contests since he was 16 years old or 14 years old. Now he’s 20 or 21. He won in eight of these competitions. He seems to be really good in JavaScript on Android, Java, and also C++. Why not hire this person?” There’s some consistency there.
(02:18:04) 而且很多这样的人，他们以前从未在大公司工作过，这是无价的，因为在大公司，人们倾向于推卸责任。他们有这种共同责任，以至于没有人完全理解谁能为一个项目负责，谁该为一个项目受责备。在Telegram内部，这是非常清楚的，而这些比赛是最接近人们在Telegram工作时会有的体验。
(02:18:04) And a lot of these people, they have never worked in a big company before, which is priceless because in a big company, people tend to shift responsibility. They have this shared responsibility wherein nobody fully understands who can take credit for a project, who can take blame for a project. Inside Telegram, it’s pretty clear, and these competitions are the closest experience to what people will have when working at Telegram.
(02:18:46) 例如，我们想在Telegram安卓版本的资料页面上实现某个非常棘手的动画和重新设计。安卓应用，它是一个开源应用。任何人都可以获取它的代码并摆弄它。因此，我们不仅会选择最好的人并雇佣这个人，我们还会选择问题的最佳解决方案，因为我们不会建议参赛者解决琐碎的问题。这是有价值的东西。它在开发方面为我们节省了大量时间。
(02:18:46) So for example, we want to implement certain very tricky animation and redesign to the profile page of the Telegram’s Android version. And the Android app, it’s an open-source app. Anybody can take its code and play with it. So as a result, we would not just select the best person and hire this person, we would also select the best solution to the problem because we would not suggest the contestants to solve trivial problems. It’s something that’s valuable. It saves a lot of time for us in terms of development.
(02:19:24) 而且因为我一直拥有这些大型社交媒体平台，我可以用它们来推广这些比赛， somehow VK和Telegram在工程师、设计师和其他技术人员中都非常受欢迎，我从来没有问题推广这些比赛并找到合适的人。对于你公司的员工来说，有什么能比一个一直是其用户的人更好呢？这个人没有使用Telegram的先验经验。
(02:19:24) And because I always had this large social media platforms, which I could use to promote these competitions, somehow both VK and Telegram were very popular among engineers and designers, other tech people, I had no issue to promote this contest and find the right people ever. And what can be better than, for an employee of your company, somebody who has been a user of it? This person has no prior experience of using Telegram.
帕维尔·杜罗夫 (02:20:00) 这个人没有使用Telegram的先验经验。他们的理解会非常有限。我为什么要甚至尝试从LinkedIn雇佣某个在谷歌和其他公司工作过的人，习惯于无功受禄，习惯于推卸责任和陷入无休止的会议，并且对Telegram代表什么理解有限？如果你仔细想想，这简直是疯了。
Pavel Durov (02:20:00) This person has no prior experience of using Telegram. Their understanding would be very limited. Why would I even try to hire somebody from LinkedIn who worked at Google and other companies, is used to receiving salary for nothing, is used to shift responsibility and being stuck in endless meetings and have very limited understanding of what Telegram stands for? It’s just crazy if you think about it.
Telegram 工程与设计
Telegram engineering & design
莱克斯·弗里德曼 (02:20:40) 正因为如此，你在招聘上极其挑剔且缓慢。人们必须真正赢得他们的位置，因此，我有机会参加了一次团队会议，人们讨论正在开发的不同功能、不同的想法，其中一些处于最前沿，所以你可以幕后看到，如何能够有如此快的创意生成速度。你产生想法，实现原型，然后最终它成为产品中的一个实际功能。这就是为什么你有这种半有趣、半不可思议的事实，与WhatsApp和Signal相比，你在许多其他功能上领先。许多我们现在认为理所当然的功能，许多我们熟知和喜爱的功能，比如自动删除计时器。这比其他任何通讯应用早了七年。消息编辑、回复。这些都是显而易见的东西，我甚至忘了其中一些它们以前从未有过。我认为自动删除计时器是一个真正绝妙的想法。
Lex Fridman (02:20:40) Because of that, you’re extremely selective and slow in hiring. People really have to earn their spot and then as a result, I got a chance to sit in one of the team meetings where people discuss the different features that are being developed, the different ideas, some of which are at the very cutting edge and so you get to see behind the scenes how it’s possible to have such a fast rate of idea generation. You generate the idea, you implement the prototype and then eventually it becomes an actual feature in the product. That’s why you have this kind of half hilarious, half incredible fact that for many, as compared to WhatsApp and Signal, you’ve led the way on many other features. Many of the features we take for granted now, many of which we know and love, like the auto-delete timer. That was seven years ahead of any other messenger. Message editing, replies. These are all obvious things I’ve even forgotten for some of them that they were never part. I think auto-delete timer is a really brilliant idea.
帕维尔·杜罗夫 (02:21:54) 我们在2013年在秘密聊天中实现了它。有趣的是，当其他应用开始复制它时，WhatsApp在七年之后，然后是Signal和其他一些这类应用，他们最初甚至复制了完全相同的时间戳。例如，如果我们有一秒、三秒和五秒，他们也会有一秒、三秒和五秒。他们试图不改变它，因为他们不确定这个功能背后的魔力是什么。具有讽刺意味的是，许多这样的事情都会发生。例如，当我们设计如何回复一条消息，你有一个小片段显示你正在回复这条消息，现在你正在输入你的回复，然后在消息本身中有一个小片段，如果你点击它，会高亮显示你正在回复的原始消息。看起来很明显，但当时我们实施了一些特定的设计决策，我们得到了左边的垂直线和所有其他这些小东西，这些都是完全随意的，你可以用不同的方式做，但 somehow 整个行业最终都复制了完全相同的解决方案。现在无论你去看WhatsApp、Instagram Direct、Facebook Messenger、Signal，不管哪个，你都会看到完全相同或非常相似的体验，因为没有人真的想冒风险去创新。如果某个东西有效，为什么不直接复制呢？
Pavel Durov (02:21:54) We implemented in 2013 in the Secret Chats. Funny thing about it is then when other apps started to copy it, WhatsApp seven years after and then Signal and some other of these apps, they initially even copied the exact timestamps. For example, if we had one, three and five seconds, they would also have one, three and five seconds. They tried not to change it because they were not sure what was the magic sauce behind the feature. Ironically, it happens with many of these things. For example, when we design how you reply to a message and you have a small snippet showing that you’re replying to this message and now you’re typing your response, then there is a small snippet into the message itself that if you tap on it highlights the original message you’re replying to. Seems pretty obvious, but there are certain design decisions that we were implementing at the time and we got this vertical line on the left and all these other small things that are completely arbitrary, you can do it in a different way, but somehow the entire industry ended up copying exactly that solution. Now whenever you go to WhatsApp, Instagram direct, Facebook Messenger, Signal, it doesn’t matter, you would see exactly the same or pretty much similar experience because nobody really wants to take the risk and innovate. If something works, why not just copy it?
莱克斯·弗里德曼 (02:23:32) 我们应该说它做得非常好。垂直线和高亮显示，我意思是所有这些都像是微小的天才之笔。以某种方式高亮文本，从设计角度使其非常清晰，这部分是之前写的，下面的东西是你的回复。不同格式、文本之间的区别。听着，我知道排版是一门艺术形式。Telegram内部有很多互动的、图形的艺术元素，都必须极其协调地配合在一起。就像你向我指出的那个让我大吃一惊的东西，也就是Telegram的背景渐变，它会变化。它真的很好地适应了气泡，聊天气泡，然后渐变之上有图形元素，所有这些都相互作用在一起。所有这些都必须很好地工作，同时不牺牲清晰度。一切都非常直观。这很难创造。那是艺术。除此之外，还超级快。
Lex Fridman (02:23:32) We should say that it’s done extremely well. The vertical line and the highlighting, I mean all of these are tiny little strokes of genius. By highlighting the text in a certain way that from a design perspective makes it very clear that this part was written before and thing under it is your reply. The distinction between the different formatting, the text. Listen, I know how much typography is an art form. There’s a lot of interacting, graphic artistic elements inside Telegram that all have to play together extremely well. Like you pointed out to me, this thing that just blew my mind, which is the background gradient of Telegram, shifts. It changes and it adjusts really nicely to the bubbles, the chat bubbles and then there’s graphic elements on top of the gradient that all interplay together. All of that has to work really nicely without sacrificing clarity. Everything’s just intuitive. That’s very difficult to create. That is art. On top of that, super fast.
帕维尔·杜罗夫 (02:24:40) 那是最难的部分。让它看起来让设计师喜欢是一回事。真正的挑战是让它看起来让设计师喜欢，同时让它能在尽可能弱的设备上运行。你能想象到的最老、最便宜的智能手机。如果你拿每个Telegram聊天背景上的动态渐变来说，这是大多数人没有注意到的东西，但他们能感觉到。
Pavel Durov (02:24:40) That’s the hardest part. To make it look so that designers love it is one thing. The real challenge is make it look the way the designers love it and make it work on the weakest device as possible. Oldest, cheapest, smartphones you can imagine. If you take the moving gradient on the background of every Telegram chat, this is something most people don’t notice, but they can feel it.
莱克斯·弗里德曼 (02:25:13) 他们潜意识里注意到它或类似的东西。有一种愉悦的感觉。当你阅读聊天时有一种愉悦的感觉，这就是设计对此的贡献。我认为渐变确实如此。我真的很喜欢Telegram的这一点，渐变。不是你描述的技术细节，而是它的感觉，然后创造那种感觉的技术方面是不可思议的。我可能能想出各种渲染那个渐变的算法，但那些算法会超级低效，所以高效地做到那一点就像…
Lex Fridman (02:25:13) They notice it subconsciously or something like that. There is a pleasant feeling. There’s a feeling, there’s a pleasant feeling when you’re reading a chat and that’s where the design contributes to that. I think a gradient really does. I really love that about Telegram, the gradient. Not the technical thing you described, but the feeling of it and then the technical aspect of creating that feeling is incredible. I could probably come up with all kinds of algorithms of rendering that gradient that’s going to be super inefficient and so doing that efficiently is like…
帕维尔·杜罗夫 (02:25:46) 或者高效，但不够漂亮，因为即使是做像渐变这样简单的事情，也可能导致渐变中出现明显的线条，让人立刻就能说，哦不，这不是对的。你必须在其中引入一定的随机性，然后你有了渐变，但这还不够。它太普通了。你想有一个特定的图案作为覆盖层，但它必须足够简单，不会让你从内容上分心，但又必须足够有趣，以创造对整个应用的良好感觉。另一个问题，你想在这种图案中包含什么样的对象，以及这种图案将如何工作？它会基于像素吗？还是会基于矢量？它会基于矢量以便无限缩放和高质量吗？我认为对于默认图案和默认背景，它基于四种颜色，不是基于两种颜色的渐变，是四种颜色，它们不断变化。我可能查看了几千种它的变体，因为这是一个非常重要的决定。这是默认背景。当然你实际上可以改变它。你可以为你自己的那个设置四种颜色。你可以改变它。
Pavel Durov (02:25:46) Or efficient, but not too beautiful because even doing something so trivial as a gradient can result in noticeable lines in the gradient that a person can instantly say, oh no, it’s not the right thing. You can have to introduce certain randomness there and then you have the gradient, but it’s not enough. It’s too plain. You want to have certain pattern as an overlay, but it should be simple enough not to distract you from the content, but it has to be entertaining enough to create a good feeling about the whole app. Another question, what kind of objects you want to include in this pattern and how this pattern would work? Will it be based on pixels or would it be vector-based and would it be vector-based so they will be infinitely scalable and high quality? I think for the default pattern and the default background, which is based on four colors, it’s not a gradient based on two colors, it’s four colors and they’re constantly shifting. I probably look through several thousand variations of that because this is such an important decision to make. It’s the default background. Of course you can change it actually. You can set up your own four colors for that. You can change it.
莱克斯·弗里德曼 (02:27:09) 不会吧。真的吗？
Lex Fridman (02:27:09) No way. Really?
帕维尔·杜罗夫 (02:27:10) 是的，你可以做到，而且你想依赖于人类思维某些根深蒂固的生物学特性。你想使用哪种颜色？会是蓝色吗？会是黄色吗？会是绿色吗？每种颜色在我们大脑中有不同的含义，你想在那里放什么样的对象？来自我们童年的东西？来自自然的东西，还是能创造不同情绪的东西？这只是应用的一个细节。还有很多细节。当你发送一条消息时，你刚打完一条消息，然后你点击发送，然后消息逐渐出现在聊天中。这是怎么发生的？你希望输入字段慢慢变形为实际的消息。
Pavel Durov (02:27:10) Yes, you can do it and you want to rely on certain deeply hard-coded biological properties of the human mind. Which color do you want to use? Is it going to be blue? Is it going to be yellow? Is it going to be green? Each color has a different meaning in our brain and what kind of objects you want to put there? Something from our childhood? Something from nature or something that can create a different kind of mood? This is just one detail of the app. There are many details. When you send a message, you are done typing a message and you just then tap send and then the message gradually appears in the chat. How does it happen? You want the input field to slowly morph into the actual message.
莱克斯·弗里德曼 (02:28:03) 变成消息。是的。
Lex Fridman (02:28:03) To the message. Yeah.
帕维尔·杜罗夫 (02:28:04) 你希望这能完成，无论消息内容如何，因为有时宽度会不同。有时它会包含媒体或链接预览或其他会改变消息气泡的东西。你检查无数不同的场景，确保每一个都能很好地工作，即使这条消息包含4000个字符。然后你查看所有平台，iOS、安卓和所有旧设备，所有种类的过时操作系统和硬件，然后你交叉检查这两者，因为你可能拥有这台非常糟糕的旧手机，但使用最新的操作系统版本，那么你怎么办？你在那里会遇到什么样的错误？然后当然，由于Telegram也可以在平板电脑上运行，我们的iOS版本可以在iPad上运行，我非常喜欢iPad，你必须理解一切都可以非常大。它可能消耗你屏幕上很多空间，然后它会触发使用更多的计算资源来渲染它。这里面有很多细微差别，但只要你痴迷于每一个小细节，至少每一个真正重要的细节，你就能达到一种用户体验……如果你真的习惯了Telegram，如果你至少是几周的常规用户，回到任何其他消息应用都会感觉像是严重的降级。
Pavel Durov (02:28:04) You want this to be done regardless of the contents of the message because sometimes the width would be different. Sometimes it’ll be containing media or link preview or other stuff that will change the message bubble. You go through countless different scenarios and make sure every one of them works great, even if this message contains 4,000 characters. Then you look at all the platforms, iOS, Android and all the old devices, all kinds of outdated operating systems and the hardware and you cross the two because you can have this really bad old phone, but using the newest operating system version, so what do you do? What kind of bugs you get there? Then of course, since Telegram works on tablets as well and our iOS version works on an iPad, which I love a lot, you have to understand that everything can be really big. It can consume a lot of space on your screen and then it’ll trigger using more computational resources to render it. There are a lot of nuances to it, but as long as you obsess over every small detail, at least every detail that really counts, you can get to a user experience… If you’re really used to Telegram, if you’ve been a regular user for at least a few weeks, going back to any other messaging app feels like a serious downgrade.
莱克斯·弗里德曼 (02:29:53) 是的，我的意思是，有那么多真正神奇的时刻。例如，消息删除时蒸发的方式，那是一种非常愉悦的体验。
Lex Fridman (02:29:53) Yeah, I mean there’s so many really magical moments. For example, the way a message evaporates when you delete it, that is a really pleasant experience.
帕维尔·杜罗夫 (02:30:05) 哦是的。做出来可真难啊，尤其是在安卓上。就是这个灭霸响指效果，对吧？消息被分解成数万颗粒子，它们像风中的尘埃一样散去。看起来很棒，但做起来太难了。
Pavel Durov (02:30:05) Oh yeah. Boy was it hard to make, particularly on Android. This is this Thanos snap effect, right? The message is broken into tens of thousands particles, which go away like dust in the wind. It looks great, but it was so hard to make.
莱克斯·弗里德曼 (02:30:28) 可能是我最喜欢的GUI图形事物之一。它就是艺术。纯粹的艺术。太不可思议了。很高兴听到它确实是经过反复推敲和深思熟虑的。做得非常出色。
Lex Fridman (02:30:28) Probably one of my favorite GUI graphical things. It’s just art. It’s pure art. It’s incredible. It’s good to hear that it has been really fought over and thought through. It’s extremely well done.
帕维尔·杜罗夫 (02:30:45) 不，如果你不深入其中，你就无法实现它。然后你不想用所有这些额外的动画分散人们与他们沟通的注意力。你希望它们在某种程度上是隐形的。
Pavel Durov (02:30:45) No, you can’t pull it off if you’re not going deep in this. Then you don’t want to distract people from their communication with all this additional animation. You want them to be invisible in a way.
莱克斯·弗里德曼 (02:31:06) 它们创造感觉，但不制造分心。
Lex Fridman (02:31:06) They create the feeling, but they don’t create distraction.
帕维尔·杜罗夫 (02:31:09) 是的。为了做到这一点，你必须克服甚至更多的挑战。例如，你提到了这个删除效果，消息蒸发。如果你先做动画，先显示动画，然后被删除消息之前的和之后的消息彼此靠近移动，那么感觉就不对。感觉太长了，太突兀了。你想做的是让消息消失，同时它周围的消息彼此靠近以填补产生的空隙。然后你想象这涉及到什么。重绘整个屏幕。在这个非常复杂的动画之上，你必须考虑诸如它之前和之后有哪些消息之类的事情。这只会增加复杂性。
Pavel Durov (02:31:09) Yes. In order to do that, you have to overcome even more challenges. For example, you mentioned this deletion effect, message evaporates. If you do the animation, if you show the animation first and then the message that is preceding the deleted message that is going after the just deleted message move closer to each other, then it doesn’t feel right. It feels too long, too imposing. What you want to do is you want the message disappear while the messages around it go closer to each other to fill the resulting gap. Then you imagine what it involves. Redrawing the entire screen. On top of this very complicated animation, you have to think about things like which kind of messages were there before it after. It just adds to complexity.
莱克斯·弗里德曼 (02:32:14) 再次是在所有种类的设备上，所有种类的操作系统，所有种类的平板电脑、手机、桌面，所有种类的屏幕尺寸上。
Lex Fridman (02:32:14) Once again on all kinds of devices, all kinds of operating systems, all kinds of tablets, phones, desktop, all of that.
帕维尔·杜罗夫 (02:32:21) 一旦你完成了它，它会给你这种巨大的自豪感，因为没有人这样做。没有人真正在乎。在某种程度上，也许他们不在乎是对的。也许没有人注意到这一点，但当这样的事情被忽视时，总会让人觉得有些不对劲，因为我明白，每天，全世界有数千万人在删除消息。他们得到的是什么样的体验？这是一种也许甚至能在潜意识里激励他们、让他们的心歌唱哪怕一点点的体验吗？让他们充满喜悦吗？让他们的心情明亮起来，哪怕只有0.001%？这是不是只是最基本的东西，而我认为，如果我们能给人们的生活带来一些价值，即使是通过这些细微的细节，我们也绝对必须把时间投入其中。
Pavel Durov (02:32:21) Once you accomplish it, it gives you this immense sense of pride because nobody is doing this. Nobody really cares. In a way maybe they’re right not to care. Maybe nobody notices this, but there is something about it that feels wrong when such things are neglected because I understand that every day, tens of millions of people around the world are deleting messages. What kind of experience they get? Is this an experience that maybe even subconsciously inspires them and makes their hearts sing even a little bit? Fills them with joy? Lightens up their mood, even a little bit by 0.001%? Is it something that is just basic and I think if we can bring some value in people’s lives, even through this subtle details, we have to definitely invest our time in it.
莱克斯·弗里德曼 (02:33:32) 一些喜悦。不单单是像生产力那样的价值，而是喜悦。我认为史蒂夫·乔布斯、乔尼·艾夫谈到过这一点，他们会在所有东西的设计上投入如此多的热爱和努力，包括在早期个人电脑中不可见的东西，因为他们相信，用户会通过某种潜移默化的方式，感受到设计师投入其中的热爱。你完全正确。这不关乎删除消息。当我看到那个蒸发动画时，我感觉到一丝喜悦。就是感觉很好。我因此更快乐了。我感受到了那份努力，而且我认为有十亿用户感受到了。
Lex Fridman (02:33:32) Some joy. Not just sort of value like productivity, but joy. I think Steve Jobs, Jony Ive talked about this, they would put so much love and effort in the design of everything, including things that weren’t visible in the initial pc, personal computers because they believe that you somehow through osmosis, the users will be able to feel the love that the designers put into the thing and you’re absolutely right. It’s not about deleting messages. I feel a little inkling of joy when I see that evaporation animation. It’s just nice. I’m happier because of it. I feel that effort and I think a billion users feel that.
帕维尔·杜罗夫 (02:34:21) 人们喜欢别人在乎的感觉。
Pavel Durov (02:34:21) People like when other people care.
莱克斯·弗里德曼 (02:34:23) 是的，是的，是的。 exactly 就是这样。当然，还有更吸引人的东西，比如所有的表情符号、贴纸、礼物，其中很多就像小小的艺术品。
Lex Fridman (02:34:23) Yeah, yeah, yeah. That’s exactly what it is. Of course there’s the more sexy things like all the emojis and the stickers, the gifts, many of those are just, they’re a little like art pieces.
帕维尔·杜罗夫 (02:34:39) 这又是艺术与技术的交汇点，因为你看看贴纸，Telegram 在大多数其他应用之前很久就推出了——
Pavel Durov (02:34:39) That’s again an intersection of art and technology because you look at the stickers, which Telegram launched way before most of this other apps-
莱克斯·弗里德曼 (02:34:48) 领先了三年零八个月。
Lex Fridman (02:34:48) Three years and eight months ahead.
帕维尔·杜罗夫 (02:34:50) …领先于 WhatsApp，是的。WhatsApp 最终在三年零八个月后推出的贴纸，其第一个版本并不太好，因为他们只是用了普通的 GIF 或 WebM 视频，这些不是基于矢量图形的。我们做的是矢量动画。这些贴纸中的每一个只有几KB大小，有时最大可能 20、30 KB，但它有 180 帧。我们能够在所有设备上以每秒60帧的速度运行它们。这也非常具有挑战性。这是一项具有挑战性的事情。我们为了让它能工作费尽了心思。在我们之前，甚至没有人尝试过做类似的事情，因为它极其困难。结果就是，你得到了这些流畅的动画。你获得了这种非常棒的用户体验。有人给你发送一个贴纸，你不需要等待它加载，因为它非常轻量，并且立即开始移动。
Pavel Durov (02:34:50) … ahead of WhatsApp, yes. The stickers that WhatsApp ended up launching three years and eight months after were not the first version was not really good because they just did regular GIFs or WebM videos, which were not based on vector graphics. What we did is vector animations. Each of these stickers is only several kilobytes, sometimes maybe maximum 20, 30 kilobytes in size, but it says 180 frames. We were able to run them at 60 frames per second on all devices. It’s also very challenging. It was a challenging thing to do. We had so much headache trying to make it work. Nobody even tried to do anything like this before us because it’s crazily difficult. As a result, you have these fluid animations. You have this really nice user experience. Somebody sends you a sticker, you don’t have to wait for it to load because it’s so lightweight and it starts moving instantly.
(02:35:58) 然后当然，这不仅仅是工程问题。你必须找到能够使用矢量图形创建贴纸的设计师，这意味着它们是基于由公式描述的曲线，而不仅仅是作为带有像素的照片创建的。你从哪里找到这些人？ again，我们举办了比赛，但组建一个我称之为艺术家/工程师的团队来做这样的事情并不容易。这是一种独特的艺术形式，这使我们能够在贴纸领域进行一场革命，然后在动画表情符号领域进行另一场革命，你可以将自定义动画表情符号添加到消息中。我认为没有人这样做过。我认为 Telegram 仍然是唯一允许用户这样做的，因为你可以在一条消息中包含100个动画表情符号，它们会被动画化，会动起来，而你的设备不会崩溃。这可能是不必要的和疯狂的，但我们认为，在艺术与工程的这个交汇处，真正的品质被创造出来。
(02:35:58) Then of course, it’s not just engineering. You have to find designers that are able to create the stickers using vector graphics, which means they’re based on curves described by formulas, not just created as photographs with pixels. Where do you find these people? Again, we did competitions, but was not easy to assemble a team of artists/engineers I would say, that are able to do something like this. This is a unique form of art and this allowed us to do a revolution in stickers and then another revolution in animated emoji that you can add into messages, custom animated emoji. I don’t think anybody did that. I think Telegram is still the only one allowing users to do that because you can include 100 of animated emoji in a message and they will be animated and it’ll be moving and your device won’t crash. It’s probably unnecessary and crazy, but we think somewhere in this intersection of art and engineering, true quality is created.
(02:37:14) 然后当然，最近我们扩展到了我们称之为 Telegram Gifts 的东西，它们本质上是基于区块链的收藏品，你可以在你的 Telegram 个人资料上展示它们，使它们具有社交相关性，但你也可以使用它们来祝贺你的朋友和亲人的生日及其他节日，这受到了极大的欢迎。
(02:37:14) Then of course, more recently we expanded into what we call Telegram Gifts, which are essentially blockchain-based collectibles that you can demonstrate on your Telegram profile so that they get social relevance, but you can also use them to congratulate your friends and close ones with their birthdays and other holidays and that was received extremely well.
莱克斯·弗里德曼 (02:37:41) 是的，它们可以持有价值，可以增值，你可以在那方面交易它们，但对我来说，仍然是矢量图形，而且它不是简单的图形，它是极其复杂的图形。矢量使其非常高效，但它也允许你创造，也许激励艺术家，使他们能够，激励他们，去创造超级详细复杂的元素。然后最终的结果，你会认为这无关紧要，但最终结果有很多事情在进行，它允许你在任意设备上缩放。现在它就像这个小……通常过去的 GIF，以及现在仍然以 meme 形式存在的，都是低分辨率的，所以人们通常不会在其中加入细节和复杂的艺术，但在这里，使用矢量图形，就像有百万件事情在同时进行。它允许你玩转不同的动画。就像你向我展示的那个功能，你发送并在发送按钮上按住一段时间，这样你就可以与你发送消息的人分享你编码的这个动画。当他们阅读消息时，有很多事情在发生。
Lex Fridman (02:37:41) Yeah, they can hold value, they can increase in value, you could trade them in that aspect, but to me still, vector graphics and it’s not just simple graphics, it’s incredibly intricate graphics. The vector makes it very efficient, but it also allows you to create, maybe incentivizes the artist, enables them, incentivizes them, to create super detailed intricate elements. Then the final result, you would think it wouldn’t matter, but the final result has a lot of stuff going on and it allows you to scale on arbitrary devices. Now it’s like this little… Usually GIFs from back in the day and still in meme form, are low resolution and so usually people don’t put details and intricate art into it, but here with vector graphics it’s like a million things going on. It allows you to play with different animations. Like you showed me this thing where you send and you hold for a while on the send button and so you can share with the person you send a message to this animation that you’ve encoded. There’s a bunch of stuff going on when they read the message.
帕维尔·杜罗夫 (02:38:59) 是的，我们有很多这样的功能，我们利用这种艺术让人们表达自己，而大多数人甚至不知道这些功能。
Pavel Durov (02:38:59) Yes, we have a lot of features like that when we use this art to allow people to express themselves and most people don’t even know about these features.
莱克斯·弗里德曼 (02:39:10) 我以前不知道。那很酷。那很酷。
Lex Fridman (02:39:10) I didn’t know about it. That was cool. That was cool.
帕维尔·杜罗夫 (02:39:12) 同一技术的另一个应用是 Telegram 上的反应，因为我们设定了一个目标，确保人们仅仅在给你发送一个赞时也能感受到喜悦。像仅仅给一条消息点赞这样微不足道的动作，也应该是一个你想要一遍又一遍地重复执行的动作。
Pavel Durov (02:39:12) The other application of the same technology is reactions on Telegram because we made it a goal to make sure that people feel joy when they just send you a like. Something so trivial as just adding a like to a message should be an action that you want to perform again and again and again.
加密
Encryption
莱克斯·弗里德曼 (02:39:43) 另一个功能，在更严肃的方面，是端到端加密。你在这方面引领了行业。它提前一年零三个月推出。你能谈谈你为什么决定添加端到端加密，以及你最初是如何开发加密算法的吗？你背后的想法是什么？
Lex Fridman (02:39:43) Another feature, on the more serious side, is end-to-end encryption. You led the industry in that. It was launched one year and three months ahead. Can you speak to why you decided to add end-to-end encryption and how you developed the encryption algorithm in the beginning? What was your thinking behind that?
帕维尔·杜罗夫 (02:40:03) 在2013年我们推出 Telegram 时，我们意识到了爱德华·斯诺登揭露的严重的隐私问题。我们想，是的，我们正在以一种已经极其安全的方式设计这个产品，但我们想确保甚至连我们自己都无法访问用户消息。我们非常清楚地理解，一群出生在俄罗斯的人不一定能激发信任。这就是为什么我们让 Telegram 开源，所以我们所有的应用自2013年起就在 GitHub 上可用，然后我们在”秘密聊天”中添加了端到端加密，WhatsApp 在几年后复制了这一点。他们提前一年零三个月才开始测试它。我认为他们在2016年推出了这个功能，那是在我们之后三年，我认为行业其他公司不得不这样做的唯一原因是因为我们设定了标准。
Pavel Durov (02:40:03) At 2013 when we were launching Telegram, we were aware of the serious issue with privacy that Edward Snowden made very clear. We thought, yes, we’re designing this product in a way that is already extremely secure, but we want to make sure that not even we can access user messages. We understood very clearly that a bunch of people who were born in Russia don’t necessarily inspire trust. That’s why we made Telegram open source, so all our apps have been available on GitHub since 2013 and then we added end-to-end encryption in our Secret Chats, which WhatsApp copied a few years after. One year and three months ahead they just started to test it. They rolled this out I think 2016, which is three years after us and the only reason I think the rest of the industry had to do it is because we set the standard.
(02:41:23) 这在当时极其重要，同时我们也意识到了端到端加密的某些局限性。在这种设计、这种架构内，你无法支持具有一致持久聊天记录的非常大的聊天社区。你无法支持庞大的一对多频道。你在维护接收大量消息的机器人时会遇到问题。多设备支持变得棘手。人们最终会丢失他们分享的一些文件。我们也看到了很多问题，最终我们有了这种混合体验，根据你的使用场景和需求，你可以选择你想要拥有的加密级别。
(02:41:23) It was incredibly important back in the day and at the same time we realized certain limitations of end-to-end encryption. Within that design, that architecture, you can’t support very large chat communities with consistent persistent chat histories. You can’t support huge one-to-many channels. You’d have issues with maintaining bots that have lots of incoming messages. Multiple device support becomes tricky. People will end up losing some of the documents they share. We also saw a lot of issues and we ended up having this sort of hybrid experience where depending on your use case and your requirements, you can choose the level of encryption that we want to have.
莱克斯·弗里德曼 (02:42:27) 这就是为什么你选择让端到端加密成为可选功能。你所描述的那个权衡是在于，对于那些真正关心特定消息、对这些消息有极端隐私需求的人，和可用性之间，比如能够跨多个设备同步，拥有20万人的群组。所有这些功能，生活质量类的功能，在它们和端到端加密之间存在权衡。你倾向于让用户在需要超级安全的情况下启用端到端加密。
Lex Fridman (02:42:27) That’s why you chose to go opt-in for end-to-end encryption. The trade off there that you are describing is between for people who really care about specific messages, extreme privacy on those messages and usability, like being able to sync across multiple devices, having groups that are 200,000 people. All of those features, quality of life features, there’s a trade-off between those and end-to-end encryption. You lean towards letting users enable end-to-end encryption for cases when they want to be super secure.
帕维尔·杜罗夫 (02:43:04) 是的。秘密聊天不仅仅是端到端加密。它们有一些限制，这些限制既是特性也是缺陷。例如，你不能对它们进行截图。你不能从它们那里转发任何文件、任何消息，当你在试图完成一些工作，只是在项目上与你的团队沟通时，这不一定是您需要的。我们非常清楚地认识到，这里有不同的需求，如果你试图将两者结合在一种聊天类型中，你最终会失去很多实用性。我们在 Telegram，我们不使用任何协作工具进行团队合作。我们使用 Telegram 来构建 Telegram。当我们试图切换到秘密聊天，分享大文件并试图完成工作时，我们立刻感觉到，它根本不适合这样做。同时，如果你真的多疑，你认为，我不想被截图，我不想有任何泄露，我甚至不信任 Telegram，我只信任代码。那么秘密聊天是最好的选择。我相信这是当今最安全的通信方式。
Pavel Durov (02:43:04) Yes. Secret Chats are not just end-to-end encrypted. There are certain limitations that are both a feature and a bug. For example, you can’t screenshot them. You can’t forward any document, any message from them, which is not necessarily something you need when you are trying to get some work done and you are just communicating with your team on a project. It became very clear to us that there are different needs here and if you try to combine both in one type of chat, you will end up losing a lot of utility. We at Telegram, we don’t use any collaboration tool for teamwork. We use Telegram to build Telegram. We felt instantly when we were trying to switch to say Secret Chats, to share large documents and tried to get work done, it was just not adapted for it. At the same time, if you were really paranoid, you think, I don’t want to be screenshotted, I don’t want to have any leaks, I don’t even trust Telegram, I only trust code. Secret Chats are the best option. I believe is the most secure means of communication today.
开源
Open source
莱克斯·弗里德曼 (02:44:36) 我们应该说明，这方面还有很多其他重要的方面。例如，Telegram 是唯一一个为 Android 和 iOS 都提供开源可重现构建的应用。为什么这很重要？
Lex Fridman (02:44:36) We should say that there’s a lot of other aspects to this that are important. For example, Telegram is the only app that has open source reproducible builds for both Android and iOS. Why is this important?
帕维尔·杜罗夫 (02:44:49) 你需要可重现的构建来验证应用确实做了它声称的事情，确实以其网站上描述的方式加密数据。为此，你需要让你的应用开源，以便任何研究人员都可以查看它。Telegram 自2013年起就是开源的。像 WhatsApp 这样的应用从未开源，所以你并不真正知道它们在做什么，以及它们究竟是如何加密你的消息的。但这里重要的是要理解，你从应用商店下载的应用版本是否与你在 GitHub 上可以查看的源代码完全对应。为此你需要可重现的构建。
Pavel Durov (02:44:49) You need reproducible builds in order to verify that the app really does what it claims, really encrypts data in a way that it is described on its website. For that you need to make your apps open source for any researchers to have a look at it. Telegram has been open source since 2013. Apps like WhatsApp have never been open source, so you don’t really know what they’re doing and how exactly they encrypt your messages. What’s important here though is to understand whether the version of the app that you download from the app store corresponds exactly to the source code that you can view on GitHub. For that you need reproducible builds.
(02:45:48) 正如你所说，Telegram 是唯一这样做的流行消息应用。我们允许人们在 Android 和 iOS 上确保 GitHub 上的 Telegram 源代码和你实际使用的应用是同一个应用。我认为这极其重要，不仅仅是为了赢得人们的信任，也是为了在这方面保持透明和开放。当我声称 Telegram 的秘密聊天是最安全的通信方式时，我是认真的，因为我还没有看到任何事实反驳这一说法，至少在流行消息应用中。你说 WhatsApp, Signal, iMessage。它们都没有在 iOS 和 Android 上同时具备可重现的构建。它们中至少没有一家在确保你用于加密数据的算法不是由某个机构交给你的、用于创建蜜罐的算法方面投入同样多的努力，至少据我对竞争对手的了解是这样。我不认为他们经历了同样的过程。
(02:45:48) As you said, Telegram is the only popular messaging app that does that. We allow people to make sure both on Android and the iOS that the source code of Telegram on GitHub and the app you are actually using is the same app. I think it’s incredibly important, not just to gain people’s trust, but just to stay transparent and open about it. When I make this claim that Telegram’s Secret Chats are the most secure way of communicating, I really mean it because I haven’t seen any fact contradicting this claim, at least among the popular messaging app. You say WhatsApp, Signal, iMessage. None of them have reproducible builds on both iOS and Android. None of them had at least at the same level put so much effort into making sure that the algorithms that you use in order to encrypt data are not algorithms that have been handed to you by some agency in order to create a honey pot, at least from what I know about our competitors. I don’t think they went through the same process.
莱克斯·弗里德曼 (02:47:23) 我们应该说明，Telegram 中整个软件栈都是在 Telegram 内部从头开始完成的。我们谈论的不仅仅是加密，还有服务器上运行的一切。服务器是构建出来的，硬件和软件都是在内部完成的，这是你减少处理消息的整个栈受攻击面的方法之一。
Lex Fridman (02:47:23) We should say that the entirety of the software stack in Telegram is done from scratch internally to Telegram. We’re talking about not just the encryption, but everything running on the servers. The servers are built out, the hardware and the software are all done internally, which is one of the ways you reduce the attack surface on the entire stack that handles the messages.
帕维尔·杜罗夫 (02:47:45) 这确实使它更安全，因为如果斯诺登的揭露教会了我们什么，那就是经常被每个人使用的开源工具、模块、库，最终都存在某些缺陷和安全问题，使软件变得脆弱。这也是确保你以尽可能高效的方式做事的一种方法，但这样做极其困难。你真的必须拥有团队中 exceptional 的人才才能达到这种彻底的程度，深入到低层次的编码，让你能够从头开始重新创建数据库引擎、网络服务器、整个编程语言，因为我们在后端用来开发客户端应用 API 的编程语言也是完全由我们团队构建的。
Pavel Durov (02:47:45) It does make it more secure because if Snowden’s revelations taught us anything is that very often open source tools, modules, libraries, that they used by everybody, ended up having certain flaws and security issues that make software vulnerable. It’s also a way to make sure you are doing things the most efficient way possible, but it’s extremely difficult to do that. You really have to have exceptional talent in your team to achieve this level of thoroughness, to go to a low level of coding that allows you to recreate from scratch database engines, web servers, entire programming languages because the programming language we use on the back end to develop the API for the client apps is also entirely built by our team.
莱克斯·弗里德曼 (02:49:01) 移除、最小化对开源库的依赖是极其困难的，因为大多数公司都依赖开源库。
Lex Fridman (02:49:01) Removing, minimizing the reliance on open source libraries is extremely difficult as most companies, they rely on open source libraries.
帕维尔·杜罗夫 (02:49:09) 嗯，我不会说我们完全独立于此。我们在后端使用 Linux。目前我们无法避免这一点，但在很大程度上，我们比大多数其他应用更加自力更生。
Pavel Durov (02:49:09) Well, I wouldn’t say we are completely independent from that. We use Linux on the back end. There’s no way of avoiding it for us at the moment, but for the most part we are much more self-reliant than most other apps.