news

pieter levels, as a one-person entrepreneurial technical person, has become a beacon for many developers and entrepreneurs with his extraordinary entrepreneurial journey and unique working style.

since 2012, he has relied on his self-taught programming skills to embark on a challenging entrepreneurial journey. so far, he has personally launched an impressive 70 entrepreneurial projects, 40 of which have been successfully operated, and 4 of them have made he has gained huge wealth, with annual profits of only one star project, nomad list, reaching us$2.1 million.

even in the face of a project failure rate as high as 95%, pieter firmly believes in the philosophy of "ship more" - the courage to try and iterate quickly. he firmly believes that as long as he continues to launch new projects, success will eventually come.

pieter's lifestyle is equally desirable. he is a true "digital nomad" who has traveled to more than 150 cities in more than 40 countries around the world. he enjoys traveling the world while immersing himself in technologies like html, jquery, php, and sqlite while sitting unhealthily in his hotel room. he adheres to the principle of "build in public" and presents every attempt he makes to the public without reservation. this open and transparent attitude has won him the love of most people.

in lex fridman's podcast, pieter shared in depth his journey of incubating 12 startups in 12 months, how he learned ai, understood the latest technology trends, started projects from 0 to 1, and learned programming skills. path, technology selection, etc.

the following is an edited transcript of the interview.

Pieter Levels： i have some ideas, because i have a youtube channel and a website called panda mix show for this channel. it mainly contains remixes of electronic music, such as dubstep, techno, and house dance music.

Pieter Levels: yes, my cd album was created using flash. i also sold cds.

Lex Fridman: okay, let me explain to young friends here, flash is a kind of software, which was quite groundbreaking at the time.

Pieter Levels： it's old, but flash was really cool.

Lex Fridman: that's right. and what's it called... oops, i can't remember it. yes, actionscript, a kind of programming.

Pieter Levels： yes, it's in actionscript and flash. it's basically equivalent to javascript at that time.

Lex Fridman: this description is very appropriate. at that time, i thought that the future internet would definitely depend on it, so i spent a long time studying it seriously.

Pieter Levels: later he was killed by steve jobs.

Pieter Levels： i can't tell either. but flash is indeed a relatively closed platform, but the irony is that apple itself is not very open. but at the time, jobs did say, "it's too closed, and people shouldn't use it. flash has security issues." thinking about it now, this was a bit evasive. jobs wanted to use words to make everyone feel that flash was bad. flash is actually pretty cool.

Lex Fridman: yeah, it was really cool at the time. but the thing is, gifs used to be cool. but they eventually came back in different forms, with gifs now becoming an important vehicle for internet memes. anyway, in my impression, gifs are really cool, and i’m not being sarcastic at all. it's really exciting to see something moving like a dancing rabbit on the internet.

Pieter Levels: at that time, there were many personal homepages, and everyone displayed all content in the center. there are also up-and-coming gifs, such as wearing a helmet and dancing under the lights, which are so wonderful in retrospect.

Lex Fridman: so you were just doing the most basic html development...

Pieter Levels: yes, but i have to learn some hardcore coding knowledge, so i might as well start there. it's actually good to launch a startup every month, as it forces yourself to learn programming and master the basics. but my knowledge system is still very fragmented, and there's nothing i can do about it. after all, i don't have much time to explore systematically... i only have one month to do one thing, which means that the time period is already fixed there. i was very strict about my time, and even compiled my schedule into a blog post on hacker news. many people have seen it and commented, "you are too exaggerated." this is actually putting more pressure on myself, because it has been made public for everyone to see, so i have to do it.

Pieter Levels: i remember it was play my inbox. because at the time we were sending cool music to each other via email. spotify didn't seem to exist yet at that time. around 2013, we would send each other music and youtube links. there are quite a few emails on gmail that say something like "this song is great, listen to it now", but they can't be accessed automatically. so i made an app that logs into everyone's gmail, pulls their emails, finds emails with youtube links, and then creates a music library. it's essentially spotify, and my friends love it.

Pieter Levels: no, it uses pop, pop or imap. it actually checks your mail. now it seems that it actually has privacy issues, because it will extract all the user's emails to find youtube links, but nothing will be saved on my side. anyway, it's quite interesting. the first product got press coverage, and i remember it got some attention from the tech press, and everyone was like, "this is cool." it doesn't make money because there is no payment button at all, but people do use it. i estimate that there are probably tens of thousands of users.

Pieter Levels:yeah, we should tell gmail "don't show all emails, just show emails with youtube links or something like that." why doesn't google do that?

Lex Fridman: the whole application ecosystem, but it seems that this has never really...

Pieter Levels:don’t do this, i’ve never seen it before...

Lex Fridman : some people have also done it... for example, i have seen applications like boomerang, and some of them are quite good. but i don’t know why they disappeared later, maybe because it was not easy to make money.

Pieter Levels: yeah, i think deep down people don't want to pay for these extensions and plug-ins because they don't look like real apps. therefore, people generally do not pay too much attention to such products. everyone has an inherent concept that "plug-ins should be free. whether it is in google sheets or elsewhere, i will never pay for the plug-ins i use. what qualifications do plug-ins have to charge money?" but if you visit a dedicated website, everyone’s ideas immediately changed. "okay, i do need this product, and i'm going to pay for it because it's a truly standalone product." so even if the code behind the two is exactly the same, the plug-in is at a disadvantage in terms of attracting payments.

Lex Fridman:clear. the same functionality can be implemented on the browser side through extensions, such as chrome extensions.

Pieter Levels: yes, but who would spend money on chrome extensions? almost no one wants to.

Lex Fridman: absolutely impossible.

Pieter Levels： so no one is doing it probably because they can’t make money.

Lex Fridman: it’s a pity, but i’m afraid this is the reality.

Pieter Levels：chrome extensions can only supplement a startup’s business. for example, let's say you develop a standalone product and a companion chrome extension.

Lex Fridman: i actually hope that chrome extensions can be developed into independent products, and i also hope that chrome can open channels for users to pay easily... in fact, i can imagine that many products can be transformed into extensions, such as many improved functions on social media.

Pieter Levels： such as gpt. various chatgpt extensions are chargeable, and now developers can also receive revenue sharing from openai. i've done quite a few myself.

Pieter Levels: i think the hardest part is figuring out what we don’t need and what we shouldn’t build because our time is so limited. first of all, the login page cannot run, and then it is necessary to build an independent product. after all, only a product can convince people to pay for it. do i need to set up a login system? maybe not. of course we can also develop some simple login systems. for example, for photo ai, after registration, you can use stripe checkout to pay and get a login link. when i first started, i only provided a hashed login link. using such a static link to log in was very simple, but not very secure. after all, there is no verification at all, so the link is leaked and that’s it. now we have a real google login, but at the time it was still a year away. in short, because time is limited, we must ensure that the business is simple and complete enough. i have to focus on things that can be built quickly.

so all it takes is money, stripe, building a product and developing a landing page. i also have to think about how people will find my product, such as posting it on reddit. but how do you promote it on reddit without being seen as a bad person spreading spam? i don’t think there’s any use in telling the truth, like “hi everyone, i’m working on a startup, please try my product”… stop it, no one is interested, and the content will be deleted instantly. the only way that works is if there are a lot of people on a sub-channel on reddit who have encountered similar problems, and you happen to solve this problem, then you can say "hey everyone, i made this tool and it may solve your problem." ". you can also say "temporarily free". this should work well. but in short, you must be targeted and figure out who your development results are suitable for.

Pieter Levels: i think first of all, everyone needs to pay attention to x. x is now very good and has become an information exchange hub for the entire ai industry. one thing about those people is that they all use anime avatars. in fact, a friend asked me this question specifically, "which accounts should i follow to keep up with the latest ai trends?" i said, "go to x and follow accounts that focus on both ai and animation." i sent them some url, made them all laugh and say, "what are these?" but no kidding, those guys are all real practitioners in the ai ​​field. they all work for major technology companies and publish various news on x. many of them are anonymous, and interestingly they prefer to use anime avatars. i don’t know how to use it myself, but those people just travel around the industry and post interesting content they find. for example, they will share research papers. therefore, if you want to keep up with the development trend of ai, go to x.

Lex Fridman: almost all people on my current focus list are engaged in ai. but there is a lot of fun in x, and everyone is noisy. i sometimes go to watch it to relieve my boredom.

Pieter Levels: the reason why i like x is because it is the source of creativity. content posted by others can inspire our own creative juices. it's interesting to see other people share their findings, and then i'll think to myself, "okay, how about i make something of it." anyway, after visiting x, i really want to imitate other people's ideas and try use ai models. for example, manually enter some content, set parameters, etc., and try to see if you can use ai to make an application or website.

Pieter Levels： anyway, i just built a photo ai, and many people have already used it. users generate about 1 million photos every month. i would test parameters, increase the number of steps to generate a photo or change the sampler, somewhat like a scheduler. there are many dpm tools on the market. in fact, i don't understand them at all, but i know that i can choose from them and generate images smoothly. the resulting images vary in quality, and i sometimes couldn't even tell which one was better. so i will do it manually and test it myself. then i thought, "why not let these users test it?" after all, they produce millions of photos every month. so i will randomly do parameter tests on about 10% of users, and then see if they like the corresponding results. this can be judged by whether they adopt and download it. this is how i grasp the direction of optimization. then i would conduct a/b testing to observe the importance of different parameters, which parameter settings are better and which are worse, etc.

Lex Fridman: in this way, you can gradually see which models perform better.

Pieter Levels:yes. then once i find the data to be convincing enough, i can switch all users to the corresponding parameter combination and determine that this will lead to breakthrough quality optimization. to put it simply, it is to use the power of users to improve their products. when they sign up, i tell them "we'll do a sampling that includes testing your photos with random parameters." this works great and i no longer have to do a ton of testing myself because user testing alone is enough to achieve diminishing marginal optimization. with their help, i slowly found the most suitable parameter combination, and that's about it.

Lex Fridman: so, the entire workflow is about parameters, models, and letting users help you explore the model and parameter space.

Pieter Levels： but there is one thing to say, stable diffusion - i was using version 1.5 at the time, and now version 2.0 is here, and there is also an excel version - the newer the version, the worse the quality. so core users are still using version 1.5 because it is not the so-called "castrated version". they effectively neuter functionality in exchange for better security and other features. so most people are still sticking to stable diffusion 1.5. meanwhile, stable diffusion collapsed and the ceo ran away. mainly because there was no money to be made and there was a lot of drama going on in the company. all in all, they gave everyone an open source model that they could use, they raised hundreds of millions of dollars, but so far there hasn't been much real revenue. regardless, they made the open source model freely available to everyone, and they did a great job. this is not easy, thank them.

Pieter Levels： yes, and the amazing thing is that although stable diffusion has raised hundreds of millions of dollars, it is the early-stage startups that have just been born, such as me, that ultimately benefit. everyone developed small programs around the large model that use its functions. now they've charged for the new model, but the actual performance of the new version isn't that great for users, and it's not as open.

Lex Fridman: yes. this is indeed interesting. after all, open source has such a great impact in the field of ai, but the business model behind it is unclear... in any case, companies in the entire ecosystem have adopted the open source model, which is absolutely congratulations.

Pieter Levels： so it is like the previous development framework, except that they did not spend money to buy market share, nor did they charge on the platform.

Pieter Levels: alas, in fact, i can’t tell clearly in most cases. it's probably just about feeling, i will choose the one that feels best for me. after all, there’s no way i’m starting a space company right now, right? although challenges will always exist, i can start with what i can do...

Pieter Levels： haha, how is that possible? i'm very concerned about asteroid mining, like buying asteroids, extracting ore from them, and then bringing them back and selling them. the launch mission can be paid to hire someone else to do it, so the only thing needed is to develop a robot that can operate on the asteroid and send itself back. so, maybe this can happen...

Lex Fridman: but i think asteroid mining and robotics are both difficult.

Pieter Levels：yes, none of them are simple.

Pieter Levels: it’s definitely not possible. but here’s the reality, “we choose a path not because it’s really easy, but because we think it will be easy.” yes, that’s how i view asteroid mining, and that’s why i’m considering it. .

Pieter Levels: yes, maybe my idea makes me look like an idiot, but the actual operation difficulty is already very low. at least compared to making rockets, mining is still much simpler.

Lex Fridman: so you would use services like spacex to complete space launches.

Pieter Levels: yes, spacex will be hired to launch this robot dog or some other form of device.

Pieter Levels: yes, i just named it "trello".

Lex Fridman:trello, yes.

Pieter Levels: until now, i still use telegram to record my thoughts so that they can be easily checked in the future. whenever i have a good idea, i write it down.

Pieter Levels: i usually use whatsapp more, right? in fact, it is like a "send a message to yourself" function, and the whole experience is similar to the previous notepad.

Lex Fridman: so, you will talk to yourself on telegram.

Pieter Levels: yes, it feels like using a notepad to prevent yourself from forgetting something. i will also sort out the content later.

Lex Fridman: i particularly like your habit of trying not to use complex systems or other tools. nowadays, people often use systems such as obsidian and notion, and they actually do things like taking notes. and you can solve it with notepad and don't rely on other tools.

Pieter Levels: yes, and i found that many users on youtube are doing this... there are many so-called "productivity gurus" on the internet, who can use pens to draw beautiful memos on the ipad. i also have an ipad, so i also learned to draw on it with a pen, like a hand-drawn calendar. many people are using it this way, most of them are students, and they can color it as they please. but after doing it for a week, i suddenly asked myself what the hell i was doing... there is no point in being fancy, so why not just organize it into a message and send it to myself, that is enough.

Lex Fridman: speaking of inspiration, you shared a tweet explaining why the first idea that pops into our heads is often the best. you think the reason is that the first reason may have been lurking deep in our consciousness for a long time, actually simmering in the brain for weeks, months, or even years. eight hours of thinking can never compare to these endless thoughts polishing in the background of the brain. so if you are faced with an idea that took eight hours to refine and a thought that suddenly appears in your mind, then this idea is often better. after all, that's what we've been thinking about for years without even realizing it, and it's really interesting.

Pieter Levels: yes, this emergent state is wonderful. our subconscious is like boiling water, and when it's ready it's like a microwave dinging. all of a sudden, we had a great idea.

Pieter Levels: yes, many, many, and i am used to thinking like this.

Lex Fridman: this thing has been hidden in my head for a long time.

Pieter Levels: that's right. but after it appears, i will polish it again, just like sending it back to the kitchen to reheat it.

Lex Fridman: because sometimes ideas are not fully mature.

Pieter Levels: it’s like a pot of creative soup simmering. that’s how my brain works, and i’m sure it’s the same for most people.

Lex Fridman : but it also depends on the timing. sometimes the reason why we have to let it go back to the pot and cook again is not just because we are not ready yet, but the real world is not ready either.

Pieter Levels: yeah, so a lot of times, startup founders come up with ideas that are too far ahead of their time. indeed.

Lex Fridman: one of your best abilities is learning new things. you will first come up with an idea, then try to implement it, and then learn all the skills needed to implement it. in this way, we can explore without having to learn too much knowledge at once. so i can’t help but ask, how do you learn? how do you quickly master the skills you need? for example, you have participated in a 30-day 3d learning course, and you were required to master all the 3d knowledge you need within 30 days.

Pieter Levels: yes, i once took a virtual reality (vr) course because it has also created a craze like ai. in 2016 and 2017, virtual reality suddenly appeared. the first one was htc vive, the heavyweight vr headset that attracted much attention before apple's vision pro. i thought, "this is going to be popular, so i better hurry up and learn." i didn't know anything about 3d at the time, so i installed unity and blender and started learning about it. i feel like this is an emerging technology that will definitely catch on. in this way, as long as you master the skills in this area, you can quickly devote yourself to development. it's fun for me to learn anyway because people always ask me, "how did you learn to code? do i want to learn to code?" and my answer is, "i don't know." i'm learning every day, this may sound like a cliche, but i am literally digesting something new every day.

for example, every day i search on google or ask chatgpt how to solve certain problems. my skill pool is growing every day, so you can never stop learning. and learning itself should be a never-ending process. so the real question is, where do you want to go? how badly do you want to understand a particular current problem? do you have a strong thirst for knowledge? are you willing to live this way for the rest of your life?

so i think the best way to keep learning is to find more things to do and don't let yourself be idle. what needs to be done is determined, but we don’t know how to do it, so what we need to learn will naturally become clear. for example, how to make a website, we can search for relevant terms on the internet, or now we can directly ask chatgpt "how do i make a website? what should be the first step?" it will generate code for us. copy the code, save it as a file, and then open it through google chrome or other browsers. now that we have the simplest possible website, we can make adjustments. so the next question is "how do i add function buttons? how do i add ai functions?" in short, through this sequence of actions, we can master new knowledge faster than reading books or tutorials.

Pieter Levels: yes, it would recommend using wix or squarespace to create a landing page.

Lex Fridman: so if i want to program my own website development, how should i express it in chat? for example, start designing your own website by creating a basic page.

Pieter Levels: yes, that’s fine. you can even ask how to start a website and how to promote it.

Lex Fridman: oh yes, and launching a website, that’s also important.

Pieter Levels: yes, and the whole process is like chatting with someone.

Lex Fridman : just talk about every aspect of website development.

Pieter Levels: the whole usage experience is actually a bit like google analytics.

Pieter Levels: actually you can.

Lex Fridman: then you have to use wix.

pieter levels: in fact, the whole process is not as complicated as you think. today's website building tools are very advanced. what we are faced with is a clickable image grid, and then the pages are opened layer by layer.

Lex Fridman: but how do i learn to program? for example, choose a programming language at random.

Pieter Levels: i personally recommend freecodecamp. you can access considerable learning resources by topic and practice coding skills for 30 to 60 minutes every day. the most important thing is to stick with it for the long term. also join programming communities like reddit...that's a huge treasure trove.

Lex Fridman:understood.

Pieter Levels: anyway, i think this is a good starting point. imagine that for a person with no skills at all, this era has made it possible to build a website or even start a startup. it is precisely because of this that ai technology is so powerful in the field of education. people around the world can ask questions to the big model and then use the guidance to build their own projects.

Lex Fridman: okay, now i understand. then it was just build, build, build. anyway, whether it is a large ai model or a programming tutorial on web development, the most important thing is to actually use it. as long as you have the prototype of the project in your mind, it can eventually be implemented. i really like the idea of ​​launching 12 startups in 12 months, or even more extreme, starting a new project every day. the most important thing is to take action, start building, get results running and improve a little bit every day. this is a really cool experiment.

Pieter Levels: i think that’s what people often call inspiration. there was a girl who created 160 websites in 160 days. of course she was making mini websites, but at least she did learn to code this way. so i think it's good to set yourself a little challenge. you can also go to coding bootcamps, but i think coding training is a bit outdated today. in my opinion, the best way is to break out of your comfort zone, set goals for yourself and force yourself to start learning. this requires strong self-discipline, otherwise you simply won’t be able to persist. and coding... coding has a very steep learning curve, and it can be a real pain in the ass sometimes. because of this, most people have never been able to master this skill.

yes, keep at it, never give up, and getting something done often requires a sense of urgency. so when we record what we have to do, such as organizing it into hood maps or working prototypes, there is an invisible pressure. at first we panic, "oops, what do we do next?" then we go online for help, trying to understand the different options and solutions. but as long as we can survive this step, we will have results that we can try later, and the joy brought by success will quickly stimulate the secretion of dopamine in our brains, making people willing to participate in this positive motivational cycle.

to be honest, this feeling is amazing. and i also live-streamed the whole thing so people can watch it anytime on youtube. so when it works, it feels like a dream. more importantly, i never set my goals too high. i just focus on the problems to be solved next, and if i push forward one by one, i will eventually get a complete application or website, etc. sometimes the reason why people can't do anything is because they set their sights too far. it's like this poster. it's impossible to draw it overnight, but if you can just focus on one thing, the next small challenge, it will slowly and magically appear.

Lex Fridman: and i have to believe in my heart that this matter will not be a problem for me.

Pieter Levels: yes, this is very important. you must stay optimistic, because everyone will encounter very difficult problems, and many big problems have nothing to do with technology, but open choices. maybe people simply don’t like your site, and maybe you’re forced to abandon the site you put so much effort into. in short, anything can happen.

Lex Fridman: what is it like to show your development process publicly in front of everyone? you probably want to get people’s feedback quickly and iterate flexibly, right? this will definitely have a certain urging effect, but people's criticism will also have a serious negative impact.

Pieter Levels: that’s right, so sometimes i feel that those netizens who speak harshly are not all bad people. i find that many people’s opinions are quite valuable. in short, first peel away those purely emotional vents, such as your website is terrible, etc., and try to delete the parts that everyone is most dissatisfied with. if the website sucks, that’s a personal judgment, but why would the other party judge it like that? according to the other party's explanation, you will find that there is actually logic behind it. for example, the registration process is too troublesome, or there is a problem with the data. they'll say the data is bullshit, so i have to improve the quality of the data. nowadays, many people on the internet will simply ignore objections. i think this is a stupid one-size-fits-all behavior. certainly,there are also many people on the internet who mindlessly hate and insult everything. this kind of situation is not uncommon on x, so everyone needs to carefully screen them. all in all, if you want to get along well in x, you must thicken your skin.. in fact, i have blocked a lot of people, and i may have blocked almost 15,000 people by now. i looked it up and i've blocked so many people in the last ten years. this feeling is really...

Pieter Levels:yes.

Lex Fridman: oh my god...

Pieter Levels: this is equivalent to blocking 1,500 people every year. i don't like to block people because the other person will be angry. they'll take a screenshot and post something like, "look, this guy blocked me." but it doesn't matter. i'll just make them disappear for the useless insults. it's nice to have this feature.

Pieter Levels: i’ve been there, oh, it’s really amazing. my server almost went down, and when i checked google analytics, i saw that there were about 5,000 people on the site, which was crazy. it was night time and it was unbelievable to me. to be honest, many apps now become popular by posting promotional short videos on tiktok, youtube reels and instagram reels. so if you make your own app, it's best to make a promotional video, such as "i developed such an app, and this is how it works. i want to explain to you the reasons for making it, and it can solve the problem." this or that question, etc.” as long as the video is good and attractive, the app will become popular. i still remember a video on tiktok that suddenly increased the monthly revenue of the software by about $20,000. the effect is really amazing.

Pieter Levels: it’s someone else, i actually don’t know him. they have written many articles about ai opinion leaders and also demonstrated various ai applications. once a video goes viral, they ask for payment from the people involved. i did some work on it and ended up getting about $4,000. now i do this myself and find that the effect is really good. tiktok is a massive user acquisition and ecological construction platform. i think this random user acquisition method is the best, better than buying advertising space directly. because when we first started, we actually didn’t have any money to buy advertising. so using a more natural approach, or writing popular tweets that will help the app get everyone’s attention, is the way to go.

Lex Fridman: in other words, you must first develop excellent products and have enough followers on your account, so that the results can be noticed by others. in this way, as long as the product is good enough, it will eventually become popular.

Pieter Levels:yes. and on x and various other platforms, we often don’t need a particularly large number of fans, including tiktok and instagram reels, which all have similar recommendation algorithms. so what really matters is the quality of the content, not the number of fans. as long as a small number of people (say 300 people) have watched your content and really like it, the algorithm will push the content to 1,000 people, and so on. such high-quality content will eventually be seen by more and more people. it doesn’t matter whether you have 500,000 followers, 10 million followers or more, this is an era where content is king.

Pieter Levels: well, many start-ups start from free users, so we can open the registration channel and provide applications for free. but for me, this approach doesn’t work very well because i think it’s actually very difficult to convert free users into paid users. in my opinion, if we were starting a business with venture capital, it would be no problem to attract free users, because we can spend the money on advertising and predict how many of these millions of people are willing to pay. in this way, the early free trial and registration guidance are valuable. but in any case, such a process must be carried out step by step, and even then the results will be difficult to develop in the direction we expected.

so i think it's better to charge users money from the beginning. so the most important thing is to show the application and tell everyone how to use the login page, supplemented by demonstration graphics or videos. once you figure out what the product does, users have to pay. it could be 10 yuan, 20 yuan or even 40 yuan. i personally think it should charge at least 10 yuan per month, after all, the monthly subscription fee for netflix is ​​10 us dollars. and since netflix is ​​a big company, they can afford relatively low subscription prices. and if we were just a lone ranger developer developing applications all by ourselves, we would have to charge at least $30 or more from paying users in order to make a return. don't be embarrassed, everyone needs to make money.

Lex Fridman: charging is actually more conducive to building a community that truly cares about the product.

Pieter Levels: you are right, a product community like discord should be created. each ai application has its own discord, which becomes a home base for developers and users to communicate about functional requirements, and ultimately becomes a joint force to promote construction. this has become the norm nowadays. and for entrepreneurs who are just starting out, attracting 1,000 paying users is already quite difficult. and if you charge them $30, you'll get $30,000 in revenue every month, which is quite a lot of money.

Lex Fridman: this money is enough...

Pieter Levels: enough to live a good life.

Pieter Levels: possibly, but that's another story. i have to make sure my profit margin is high enough, so i try to keep costs as low as possible. i don’t hire people, and i will actively negotiate with ai service providers, such as “can it be cheaper?” this is a little trick i discovered. you can directly send an email to the service provider and say, “can you give me a discount?” ? the current service fee is too expensive.” the other party will respond decisively, “of course, do you think it’s 50% off?” really, it’s that easy. i had no idea before that you could directly send an email to negotiate terms. especially during the current economic recession, you only need to submit your demands directly to the service provider. but please note that you still have to be sincere and don't deliberately say "give me a discount, or i will go to another service provider". if the person gets angry, they may say "i'll leave without giving it away." with good communication, a 25% discount or even a 50% discount is possible. many people may think that the api or other service prices published on the website are fixed prices, but this is not the case at all.

Lex Fridman: and you still discuss business live in front of everyone.

Pieter Levels: yes, so the other party is more willing to show sincerity.

Lex Fridman: you spread a vibe of camaraderie and harmony around the world, but also actually try to create cool stuff. with this effort, many companies will definitely be willing to cooperate with you.

Pieter Levels: it kind of means that, although there is also my own unique secret recipe, but even if there is not...

Lex Fridman: simply being honest and friendly can get good results.

Pieter Levels: in addition, it also depends on the scale of our own business. sometimes if you post more promotional posts on x, you will get a bigger discount.

Lex Fridman: that’s right, and for such public projects, direct charging eliminates the need to post promotional posts or spam emails.

Pieter Levels: and it must be admitted that the quality of paying users is higher.

Lex Fridman: higher user quality.

Pieter Levels: i don’t mean any harm, but free users are more difficult to serve. especially for ai startups, they often need to face an audience of millions. because you have no control over how they abuse your app, or even use it to commit hacking attacks, everything will turn into abuse against you.

Lex Fridman: similar situations have been said on the internet. you also mentioned the influx of 4chan into hood maps.

Pieter Levels: yeah, but i have no problem with 4chan. although i can't say i like it, you know what i mean anyway. they're so crazy, and sometimes the things forum users do are really funny.

Lex Fridman: which thing impressed you the most. i remember netflix once released a documentary about anti-social networks. it was actually quite interesting. this is the spirit of 4chan.

Pieter Levels: i feel like i’m not very good at working with other people. i certainly know the benefits of collaboration, but i have a hard time trusting other people.

Pieter Levels: that's right. and i don’t want to have endless meetings to align my thinking. there doesn’t need to be many people, as long as there are three people, the final result will be a compromise. this is how europe is declining. there seems to be a proper term to describe this state. in any case, people are essentially forced to sit in a room and can only leave after negotiating a political agreement acceptable to all parties. but i think that just breeds mediocrity.

if the idea is mediocre, the company is mediocre, and the culture is mediocre, the end result is bound to be failure. there must be someone who has the final say, otherwise it would be better to do everything yourself. that's what i think and do. i do have a few friends that i trust and have started companies with, but that's because we've known each other for a long time. andre, he's one of the very few people i'd love to work with, but that's a very small minority.

Pieter Levels: the basic division of labor is that he is responsible for writing code, and i publish the results on x. that is to say, i am actually in charge of product promotion and formulating product strategies. for example, i am responsible for telling him that here should be better and there should be better. but the coding work is basically handed over to the same person, and he is responsible for it because he can develop in ruby and i can't. i only use php.

Pieter Levels: never in this life.

Pieter Levels: i can’t tell. anyway, i just like to sit in front of my laptop and program, and don’t want to communicate with other people.

Lex Fridman: no, i feel like you have never met like-minded developers...

Pieter Levels: actually, i have actually encountered it. at that time, the photo ai project had an ai developer named philip. at that time, i hired him because of his python skills. ai development is about python, but i don’t know how to do it. there was a lot of work involving the model at that time, especially functional improvements and the like. he worked for 10 months and helped me a lot.

heck, i remember that i also tried to learn to use python, numpy, and package managers, but these things were too difficult for me to learn, and i didn’t have that much time. if it was 10 years ago, i should be able to calm down and figure this out, but it's impossible now. staying up all night just to master a language is no longer suitable for me.

Lex Fridman: because your role has changed, there is no need to master a programming language. in other words, what you are paying attention to is actually ai and new things, not the language itself. in short, you never let other developers participate in the php jquery code you write. at most, communicate occasionally, such as telling them what they should do and what not to do.

Pieter Levels: well, the longest period of close collaboration is one week. after that it was over.

Pieter Levels: because he wants to rewrite all the code.

Pieter Levels: at that time, he was not satisfied with our jquery, so he suggested, "we should rewrite everything in vue.js." i said, "are you sure? can't we just leave jquery alone?" he said, "no, man. "at that time, i felt that there were too many things that needed to be changed. later, we spent a few weeks sorting out the workload and found that it was too time-consuming.

Lex Fridman: i really like working with people like this. when i collaborate with other people, i assume they are the smartest people who ever lived. so i will carefully observe their code or development results and try to discover their genius. anyway, as a part of this, we have to understand other people, really pay attention to other people, and then discuss which approach is better based on that.

Pieter Levels: you are right, but that only applies to a few top developers, the kind of geniuses who have mastered the technology. they can master any technology stack, but there are not many such people. they can be said to be the top 5% of practitioners. no offense here, but most of the developers we come into contact with are not like this... according to my observation, most people who work in regular jobs are not very good at their profession, even many doctors are so.

Lex Fridman: this is so sad.

Pieter Levels: when we realize this, we will find that people often perform very mediocre at work, especially in development and coding.

Lex Fridman: i think a very important skill for developers is to have their own tone and style...

Pieter Levels: that’s so true, really. this represents a code sensitivity.

Pieter Levels: yes, it’s a word i made up, but i think that’s how it is. you need to be able to understand the code, appreciate the subtle differences in the code, and then make suggestions for adjustments based on this. but that's not the case for most people. they just see a problem in jquery and think they should rewrite it.

Lex Fridman: but i think jquery is very good and has very powerful functions. the same goes for php, especially as you said, as versions have evolved, php has become a very serious programming language and extremely fast. the current performance of php is really advantageous.

Pieter Levels: much faster than ruby, yes.

Lex Fridman: so for developers who care about performance, php is absolutely satisfactory. moreover, various programming languages ​​now have their own huge communities and various fully functional frameworks. so the specific language used actually becomes less and less important. if you look at it from the perspective of a collaborator, i think you have been very successful and have delivered so many development results.

so if i were this new person, i would patiently observe and learn after joining, instead of defaulting to treating you as a leader who doesn’t understand the situation. i'm going to assume that you are a genius, the smartest developer ever, and the first step is to learn from you. this way i can understand your thoughts and ideas from the code. after this, if i need to add another small feature, i'll try to write it in a similar style to yours.

my goal is to make you happy and satisfied. only in this way will you start to respect me, trust me, and truly start a cooperative relationship between us. but i really don't know how hard it is to find a good developer these days.

Pieter Levels: i admit that it is not that difficult. excellent developers do exist. maybe i need to try to hire more people and deal with more people.

Lex Fridman: give more opportunities to others and give yourself more opportunities.

Pieter Levels: this will definitely consume a lot of my time and energy. but it's totally worth it, i just haven't figured out whether i want to do it or not. after all, my old methods still solve the problem. yes, many people say that nomad list looks clumsy and poorly designed. well, i will try to improve the design and put this work on the to-do list. as long as you keep doing it, these problems can be solved.

Lex Fridman: but one thing cannot be denied, that is, you are very good at what you do. i have seen the interface of photo ai and you are using jquery. in fact, the entire design is very stylish and the interface is impressive. there is also nomad list, and the interface is also of a very high standard.

Pieter Levels:thank you, i appreciate it.

Lex Fridman: and these are all done by you alone, including every element, every small function, everything...

Pieter Levels: that’s right, just by myself. in fact, many people say that the things i design are a bit suspicious of adhd because there are too many elements crammed into them, which is completely inconsistent with the current minimalist design orientation.

Lex Fridman: i understand what you mean. but the information here is plentiful but useful, delivered in a concise way while still being stylish and interesting. in my opinion, the so-called minimalism is only suitable when you don’t need to convey any information and you want to look cool at the same time.

Pieter Levels: it makes sense, it’s that pretentious kind of beauty, right?

Lex Fridman: no matter whether it is pretentious or not, it must be useless, so that it can be called minimalist. the reason why simplicity lacks the word "extreme" is because it is not so extreme and only emphasizes conveying a large amount of information in a clean way. anyway, in my opinion, your results are excellent.

Pieter Levels: these are all taken from the images in my mind. sometimes my design process is like transcribing things from my head.

Lex Fridmanf: anyway, the final result is very beautiful, all the elements are spaced just right. the font is also nice and very readable.

Pieter Levels: i am definitely satisfied myself, but sometimes i am not sure whether my preferences are universal.

Pieter Levels: not really. i did this because i was too lazy to set up a temporary server on my laptop. so now i will deploy the results directly to the production environment, even if it sometimes causes problems. but the good news is that it works pretty well in most cases. after all, i'm using php, lint, and json, and the system will always prompt me for errors. specifically, i've performed approximately 37,000 git commits in the past 12 months. during this period, i will also make some small modifications and then send the results to github. github makes a web request to the server, and the web server performs the pull and deploys the production environment, and then meets the majority of users.

Pieter Levels: one second, usually no more than two seconds.

Pieter Levels: it's really like this. many people asked me with sighs, "how did you do it? how did your server not crash?" because i am very careful when writing code. in fact, i also know that such an approach will definitely cause big trouble for any large company. but personally speaking, this can greatly improve efficiency and make my development work advance very quickly. if i see a bug report on x, i will even use a stopwatch to monitor how long it takes me to solve it.

for example, i can fix a bug in just two minutes. this is the solution to the problem for me. as for some large manufacturers, bug reports submitted have to wait a full six months to be processed, which is simply terrible. in short, i am willing to help users solve their problems quickly, and they also like the feeling of being valued. but i also understand why large-scale companies cannot do it.

Lex Fridman: i think there must be a better way for large-scale enterprises. after all, whether it is an individual or an organization, software development work must involve the testing phase, preparation phase and development phase, as well as multiple tables and tables used to store status data. database……

Pieter Levels: don’t forget to declare the paperwork.

Pieter Levels: and i seem to represent the other extreme, which is more positive.

Lex Fridman: but it would be great if you could make your job a little safer, just a little bit.

Pieter Levels: maybe it’s because many practitioners focus more on their career advancement rather than cool or more valuable results.

Lex Fridman: i think this is not necessarily out of selfishness, but because the operation of the entire system does not allow it...

Pieter Levels: it’s just that it’s restricted by the system.

Lex Fridman: the system tends to suppress those transformative elements.

Pieter Levels: in fact, i have encountered similar situations when i was in contact with the large companies i cooperated with. i have communicated with many employees there, and they are also working hard to do practical things and have outstanding abilities, but i always like to pour cold water on them, such as "friends, i talked to 20 people for six months, but there was no result at all." "they can only smile and say, "i know, we are doing our best." yes, i know, so this is a kind of systematic rigidity.

Lex Fridman: yeah, i don’t know if there’s a better word, but maybe top-down authoritarian intervention isn’t a bad thing sometimes, like having the ceo directly intervene. take youtube as an example. senior management can forcefully promote the implementation of multi-language audio projects. it should be the highest priority.

Pieter Levels: i think large companies, especially large companies in the united states, put more emphasis on process compliance in many things. the entire process must be completed, so major technology companies will never be able to adopt my lone ranger style. all work there must be carried out step by step and in a manner agreed upon in advance.

Lex Fridman: so my answer is still the same, authoritarian or authoritarian form of power. for example, steve jobs often did this, and the same is true for many courageous leaders i have come into contact with. they ignore legal affairs, ignore regulations, and push forward forcefully.

Pieter Levels: yes, sometimes this is the only way to get things done.

Lex Fridman: ignore public relations and ignore other people’s opinions. they will delegate power to engineers, listen to suggestions from front-line employees, and require certain results to be produced within this week. that's how it was done.

Pieter Levels: regulations can be changed. for example, youtube has released this ai dubbing service, but the lawsuit involves some legal issues, so the laws will change accordingly. after appeals go all the way, the supreme court will consider and rule so that whether the outcome is good or bad, it will be reflected in the law. therefore, the most important thing is to roll out the functions, so that society and the legal framework can be changed. and if you hesitate, take no action, and are always afraid of touching the boundaries of the law, you will never be able to change reality.

disclaimer: this article was translated by infoq and may not be reproduced without permission.