This kinda stuff is exactly why I keep thinking of giving up on being a programmer. I still struggle at easy lc problems after 100 problems or so. Should I even keep at it?
You are absolutely good enough to be a programmer, so long as you enjoy programming at least a little bit and have the desire to put in a consistent amount of effort. And yes, there are definitely companies that won't give super difficult LC interviews. I know it's demotivating, but don't think of it as being smart or stupid. It's more that everyone's brain is wired differently. What makes you bad at one thing makes you better at others. The beauty of software engineering is that it's more than just about coding. You can differentiate yourself in many different ways.
You either like to code or you don't. That is all that matters. Yes, you do need to grow as a programmer and learn your craft but if you're not doing it as a job then it does not matter how long it takes you to learn. Here's a secret. Only about 1% of programmers, if that many, are as talented as the people in this video for one simple reason. They happen to be math geniuses. Learning a programming language is only a small part of the package. A far larger part is gaining an understanding of algebra, linear algebra, trigonometry, calculus and geometry amongst other things. Someone who is born good at this will always be better than you as a programmer. Instead of looking for back pats and head rubs on RUclips decide if you like programming enough to stick with it and take it from there.
I know I am never going to produce singularity changing code like these guys, but I work from home, deploy regularly, have a girlfriend, family and friends that love me and make enough each month to do whatever I want and save. I am not going to change the world, but I really am making the most out of it. Count your blessings, don’t compare yourself, be humble and grateful.
That is a good life, you are right. But I wonder if it’s possible to do both without sacrificing one or the other to some extent. Probably not, but that would be ideal
im being completely real here. if he is actually so good that he can invent software that takes your job without him lifting a finger, does he not deserve that? in my opinion he deserves it
these guys aint geniuses. they just had the growth mindset and worked on it over a long period of time. ANYONE has the capability to become as smart as these people. you just need to spend a lot of time and really enjoy what you do, and you will be as smart as these people. its time spent on something more than talent, even the talented have to study years for legendary grand master on cf. even if you are late to it, you can definitely become really strong after spending 2 years working on it, which is most important.
I know a guy with two gold medals in the IOI. He contributes code every single day, and runs a weekly reading group where he covers really deep technical subjects. He's obviously very intelligent, but more than anything he is just _driven_ and I think we don't talk about that much.
Exactly. That part is almost always left out when talking about high performers and highly intelligent people. You have to practice most things in life that aren’t innate (like breathing). This stuff isn’t magic.
Every single person in the euro league, G league, NCAA, NBL are driven. A fraction of them make it to the NBA. At a certain point, you're the bottom percentile of the top 1 percent, and then there's the 99% who never could even if they wanted to.
The way he solved the first problem is using the formula a^2 - b^2 = (a + b)(a - b). To recognize it that fast and do the computation on the fly is very impressive.
It’s one of the first things you learn in high school pre-calc. Impressive he knew about it, but basically every math student will know about it immediately.
I think people give AI more credit than it deserves and human genetics less credit. The things we get out the box are lowkey cracked- facial recognition, motor movement, self-training etc. Sure an AI can consume hundreds of years of data and become a glorified search engine and there is more potential output but unless it can utilize an incredibly accurate simulation and discover new things, it is still bound by human time in it's discoveries through good ol' trial and error.
Most people don't really realize how technically strong these guys are and the potential they can build at. Maybe they don't know how to code in react or whatever , but the point is that their fundamentals are so strong, especially the concrete mathematics they work with, its really the core of all computer science and they nail it. All this stuff we talk about in the industry are really just high level abstractions originally built by people who were really solid at math. So its guys like these, the so called nerdy guys as he says, that have the potential to create better solutions for the world. Its just that we need to push competitive programmers in a direction where they can apply their skills in the real world.
Sorry but no. I have enormous respect for John Von Nuemann who solved sorting in the most mathematically optimal way in 1945 (and other things). But he wasn't an engineer. He was a mathematician. Computer science is a field of mathematics. You don't see mathematicians building bridges. In fact they are notoriously bad at engineering. The same applies here. The manhattan project was made up of a few flashy physicists and thousands of engineers. They could very well be good software engineers, but your conclusion is based on nothing but conjecture. And there are a lot of people that will tell you that Leetcode does not make you a good engineer. The fundamentals will help you a LOT. They will make you a BETTER engineer. But they are one piece of a very large puzzle of skills. Especially for a senior software engineer in a FAANG company (for example). Coding ability isn't the factor that gets you to that level. They have those tough interviews because coding can't be what you struggle at. The design, management, how to sell, work with enormous ambiguity, get people to work well together and so on. Those are what get you beyond base level in those companies. The coding was the most basic requirement like being read/write is to be considered literate.
It's not that they don't know React, they did not have good error handling lmao💀, also this is just a chatgpt wrapper. A small company cannot even afford to make their own llm that is more powerful than tech giants. It was fine, they used tech giant product, but guess what they are lying too lmao. Doesn't really give off a good software engineer vibe, when all they are doing is scamming and lying for investor money
@@jvandervyvercorrect. they have potential but that doesn’t mean anything until they make a good product. CS, let alone math, and software engineering are different
Engineering is not a sub discipline of math, which I believe is what you're implying. You can see this most clearly in a bicycle. We just maybe a few years ago began to understand and model the way a bike works mathematically, but bikes are old as hell. Bill Hammock's excellent book The Things We Make elaborates on this point.
4:16 is simple when u know the trick to too. He does the operation quickly because he trained well and did his reps. (255-245)(255+245). He immediately got the ten, so his attention was at the addition right away since he would just need to put the 0 on the end once he got that. With enough training in addition, he recognized perfect additions to 100 very easily. 63 and 27, 55 and 45, etc. so he saw it was 500 quick too. Add the zero, 5000 11:15 also easy. Division rules: 2 check last digit, 4 check last 2 digits, 8 check the last 3 digits. Since 8 letters long 10/8 remainder is 2, so A.
@@darekmistrz4364 most people can become smart if they are in the right environment. Maybe some point in the future, what we see as impressive will become the standard once education improves enough
Could you explain what do you mean by "2 check last digit, 4 check last 2 digits, 8 check last 3 digits"? What does 2,4,8 and " checking the digits" refer to?
I feel bad for the girl who was competing against him in that mathcounts competition lol. Of course that for reaching that point she is insanely smart, but she got completely overshadowed by scott
@@mohammadhassan1649 sorry this isn't true. the question itself is a pretty generic math competition question. all it required was the identity x^2-y^2=(x+y)(x-y). anyone who did competition math enough to get to that point in mathcounts would know that identity like the back of their hand. what made him answering these questions impressive was the speed he had. i have 0 doubt the girl could also solve them in a few seconds, but not nearly as fast.
It is a well-known fact though that interview performance (i.e. LC solving) does not correlate with job performance. There is a huge difference between solving 10-30 minute puzzles and working in codebases with 10s of millions of lines of code with decades of unwritten assumptions in underspecified problems, emergent behaviors, etc. It is an industry-changing tech for sure, but not yet worried about myself as it is currently making me more effective. (Btw. Yes on math importance being underestimated, speaking as a CS/math double major with two masters and a PhD)
6:29 I’m not too sure about they explanation, but I think of it as 5 choose 2 (5C2) that represents the placement of 1 and 2, since they will always be in one order from left to right, and fill in the rest of the empty spaces with 3! permutations of numbers. It comes out to be 10 times 6 which is 60.
While he may be good at mental sports, I’m not convinced the team is made of genius programmers. Currently devin is just GPT wrapped in a sandbox. Nothing technically groundbreaking. This is all superficial marketing, and the substance has yet to be proven
you're right, the only innovation they have shown so far is a nice UI, they're very late to the party for their project, nothing unique or special yet.
@@ayushshshsh Why would you troll him when he is in fact stating facts? Mathematicians and good competitive programmers who are computer scientists have an ego problem, which is exacerbated by the size of the field (everyone wants to be in Tech and AI). Just because you can solve olympiad-level problems does not mean you can build AGI. It takes more than that, a variety of people from multiple disciplines and backgrounds.
I take 6 min to solve this problem in coding. But this guy is mind blowing. var a= "MATHLETEMATHLETEMATHLETEMATHLETEMATHLETEMATHLETEMATHLETE"; var position = 2010; const b = (Math.ceil(position/a.length) - 1) * a.length; console.log(a.charAt((position-b)-1));
there are 8 letters in that sequence so you just divide 2010 by that. Since there is a remainder 2, that'll be the letter where it stops; which is the 2nd letter, "A".
Scott Wu is a legendary grandmaster on Codeforces. The difference between him and Neetcode (who is probably an Expert/Candidate Master on Codeforces) is like the difference between Neetcode and someone who's only solved 10 LC problems 😅 I've been programming competitively and studying DSA for the better part of three years now, and I'm just breaking Specialist on Codeforces. They really are that cracked. Plus, they went to Harvard lol.
Not saying this to discourage anyone though. In the real world, you aren't competing with Scott Wu or his brother. If he wanted to work at FAANG he'd have been hired ages ago 😅 Keep pushing guys 🎉
@@eti-iniERThese people spent their ENTIRE childhoods on this where they basically had unlimited time. There are plenty of people competing with scott wu and his brother. You are being too delusional over their ability to complete one genre of cognitive tasks. Where is this same energy for the putnam fellows who are honestly probably "smarter" than these kids.
side not but I hate people like you lmfao: any competitive domain has ppl like this guy that glaze the fuck out of the top 1% by stating the skill differential as if its some kind of insurmountable gap, discouraging people that haven't hit their peak from trying by saying "oh btw not saying this to discourage anyone tho guys !!" just because you took 3 years to hit specialist doesn't mean everyone has to be measured by your (frankly below avg) progress. Stop discouraging people
@@jepper6140 Competitive Programming is mostly practice, literally no one ever solved a few problems and got to Candidate Master. Most people who start Competitive Programming at a young age and keep at it with IQ > 90, will become red eventually.
For those intimidated by these test geniuses - mastering a boxing bag and mastering boxing in a ring with a real opponent are two very different things. These tests are the former. Yes, the bag's useful, you should probably spend time getting competent with it, but it's only one piece of the puzzle and is no substitute for experience in the ring. And sometimes the skills that count in the actual fight are quite tangential to the ones that applied with the bag.
14:00 actually Sam Altman said the same thing. There's gonna be a point in this AI bubble where we are gonna be forced to find better algorithms. Or try to find better ways to make smaller more accurate models, rather than general LLMs. One more thing I'd like to add is what Primeagen said "AI is good with generating media and language, since you do not have to be 100% accurate. Math and Coding needs the 100% accuracy"
"AI is good with generating media and language, since you do not have to be 100% accurate. Math and Coding needs the 100% accuracy" -- Wow, pretty thought-provoking
The first question is a difference of squares, in case you thought he was a computer(still is but not a savant just very smart dude). X^2-y^2=(x-y)(x+y) = (255-245)(255+245)=(10)(500)=5000 Most of these Olympiad question have tricks that these kids memorize and yes their math foundation is very good to be able to recall difference of squares and do the mental math in their head lol
You can be slightly more rigorous with the math problem. A 5 digit number has 5 slots. First realize there are exactly 5 choose 2 pairs of slots. For each pair of slots it’s possible for the numbers 1 and 2 to be arranged in 2 ways. In only half of them 1 is to the left of 2. So there is 5 choose 2 valid ways for the numbers 1 and 2 to be arranged. For the rest of the 3 numbers, there are 3 factorial ways for them to be arranged given each 5 choose 2 arrangement of 1 and 2. Therefore 3 factorial times 5 choose 2 is the answer. (Simplifies to 60)
Am I the only one who doesn’t get the “just divide by 2” solution given for the permutation problem? I came up with a solution that is 3! * (1+2+3+4) which is indeed 60, and you can generalize this up to ten digits and even then it’s still just half of the total permutations. What is the intuition for the solution he presented?
Think of it this way : You can separate the 120 combinations into two groups : one where 1 is before 2, and the other is where 2 comes before 1 And then ask yourself: is there a reason why one group should be bigger than the other ? NO, because here the digits 1 and 2 play the EXACT same role. Thats why the 2 groups are the same size, so each of them is size 120/2 = 60
I think I get it now, there are two groups which are exhaustive, 1 coming before 2 and 2 coming before 1. Since they are essentially saying the same thing, they must be equal and therefore it must be 2*x = 120 and solving for x is 60. That makes a lot of sense now, thanks for the respose!@@zakariaelghazi
There's so much more into developing software it's why someone like Steve Jobs can be so successful. This is more of a relief than a concern for competition. Sounds like a lot of IQ but you need high EQ to lead. That's what ppl like Steve Jobs have, a deep understanding of human beings and your customers. Time will tell
@@darekmistrz4364 he wasn't one month old in the clip where he used permutation i'm not saying that you can teach general relativity to a one month old but there are certainly ways of educating children that dramatically change their behaviour which will affect how "smart" he is
you were very sympathetic throughout this video. even if I disagreed with you on some your hypotheses, I still have liked your reasoning, ideas and character. a primagen-lvl streaming commentery and interaction. stay safe, keep on improving, we are in great times
I heard about the wonder kid years ago, similar background to Larry Page, both Korotevich's parents are Computer Science professors at Francysk Skaryna Homiel State University, and both of Page's parents were professors in Computer Science at Michigan State University.
I think I can't agree with the last bit of the video. Machine learning system indeed do the system-2 thinking. They apply all the logics formulatically based on the given data. at least 100X faster than human. If you are given a choice between A and B, you would make some arguments and then choose 1. Machine learning does the same thing except it can do 1 million arguments in a second and choose A or B. LLM's looks like they are on steriod, I get it. But that is because human perception is shallow. You can't really comprehend the fact that every data point is going to a very sophisticated machine learning model (those transformer architecture) and a giant book of logic was made when the LLM is trained with massive data. It's just the retreiving process is so fast that it looks like it is on steriod but what it really does is, if you give choice A and B, LLM looks at the giant logic book and start at page 1. it make some decision and that page take it too say page 245. it looks up there and make some decision, and then that page hints it to page 812 and it make some more decision. And just like that it travels the whole logic book to finalize the prediction and gives it to you. Now again this analogy is really high level and what happens inside is not known to any human and also this is not comprehendable by humans either.
Quick observation: Scott Wu has a brother (Neal Wu). Both had impressive results in competitive programming. The video presents some information about Neal as being about Scott (e.g. the Leetcode #1 ranking and the Twitter profile).
0:18 1. Introduction to Devon AI Team Overview of the talent-dense team behind Devon AI 0:53 2. Competitive Programmers' Fitness Surprising revelation about competitive programmers' physical fitness 1:16 3. Insights into Competitive Programming Exploring the skills and approach of competitive programmers 2:15 4. Founder's Exceptional Skills Highlighting the exceptional skills of Scott Wu, the founder 3:14 5. Importance of Problem-Solving Skills Discussion on the significance of problem-solving skills in software engineering 4:03 6. Math Prodigy behind Devon AI Unveiling the math prodigy driving the innovation of Devon AI 5:06 7. Coding Olympics and Talent Exploring the competitive coding landscape and early coding skills 5:21 8. Permutations Problem Explanation Brief explanation of a permutations problem with five digits 5:43 9. Permutations and Symmetry Calculating permutations and understanding symmetry in numbers. 6:35 10. Early Exposure to Math Importance of early math education and competitions. 7:18 11. Launch of Coignition Labs Introduction of Coignition Labs and the creation of Devin AI. 8:16 12. Investing in Disruptive Technology Discussion on investing in disruptive technologies and innovation. 8:20 13. Debate on AI vs. Programmers Exploring the potential of AI replacing programmers. 8:31 14. Cognition Lab Team Overview of the team behind Cognition Lab and Devin AI. 9:03 15. Tech Innovation by Young Minds Achievements of young developers in tech innovation. 9:51 16. Utilizing LULMs in AI Discussion on the use of existing LULMs in AI development. 9:59 17. Skills in the Age of AI Importance of fundamental skills in the era of AI. 10:15 18. Deciphering the 2010th Letter Solving a coding problem through mathematical reasoning. 11:14 19. Competing with AI in Coding Exploring the advantage of using AI in competitive programming. 12:03 20. The Mission Beyond Software Engineering Discussing the broader goal of solving global challenges with AI. 12:50 21. Importance of Reasoning in Robotics Highlighting the role of reasoning in mathematics and science. 13:39 22. Challenges in AI Reasoning Addressing the complexity of reasoning abilities in AI development. 14:35 23. Efficiency vs. Replacement in AI Debating the impact of AI on software engineering efficiency. 14:58 24. The Impact of AI on Engineering Discussing the challenges and benefits AI brings to engineering and problem-solving. 15:21 25. AI Interacting with the Physical World Exploring the need for AI systems to interact physically, such as through robots. 15:52 26. Data Companies vs. Self-Driving Cars Analyzing the shift from data companies to the challenges of achieving fully self-driving cars. 16:42 27. Embodiment in Intelligent Agents Understanding the concept of creating intelligent agents capable of navigating the real world. 17:15 28. Human vs. Computer Learning Comparing the fundamental differences between human and computer learning processes. 18:15 29. System 1 and System 2 Thinking Explaining the two systems of thinking in the brain and their functions. 18:35 30. Large Language Models and Original Thinking Examining the limitations of large language models in terms of original thinking compared to human brains. 19:08 31. Evolutionary Differences in Reasoning Discussing the evolutionary advantages of human brains in reasoning abilities. 19:38 32. System 1 Coding Mastery Exploring the realm of fast and efficient coding. 19:43 33. System 2 Coding Complexity Delving into the intricacies of problem-solving and human ingenuity. 19:52 34. The Billion-Dollar Question Contemplating the future of coding and AI development. 19:57 35. Uncertainty and Speculation Examining the unpredictability of AI advancements and expert opinions. 20:06 36. Elon Musk's Predictions Analyzing Elon Musk's forecasts on AI surpassing human intelligence. Generated with Tubelator AI Chrome Extension!
That math question was pretty easy if you're good in school level algebra 255^2-245^2 = (255+245)×(255-245) =500×10 =5000 If you're focused then you can solve it within 3 to 5 seconds.
maybe i'm insane and coping but the fact that devin has "competitive developers" is like suuuper indicative that devin is meant to capture VC funding. like what kind of VC could resist a team of geniuses putting their brains together
Remember how everyone was singing praises for Sam Bankman-Fried, calling him a genius and all? And look how that turned out. Not saying this is the same deal, but it does make you think twice. Everything seems a bit too perfect-like, the team, what they're selling, their track record. Kinda makes you wonder, doesn't it? Always good to keep your eyes peeled for stuff like this.
SBF? Never showed any skill just a scammer. Of course devin is overhyped for max funding but that doesn't mean they are scamming people. There are already quite some people have early access and no it of course it is still limited. It also still depend on the LLM under the hood. Complex "wrapper" will be more common soon.
I’d argue that the quickness of his answer of 60 was simply due to him seeing this question before. Not saying he isn’t extremely intelligent but it seemed a little too fast for him to read and compute that in a matter of seconds.
If you've thought of giving up, please do give up, straight up. People who love/like doing what they're doing never think of giving up what they're doing, ever. People who love/like what they're doing will conjure endless reasons why they should continue. People who aren't sure, or don't really like what they're doing will dance around excuses to stop - this activity will drain and harm you. So, give up. There's no shame in it. It just means it's not for you, it just means you won't be happy with it and it will burn you out. Find your own happiness elsewhere. To each their own.
Cracking a job interview using AI or chatgpt is prohibited for humans but when Devin AI clears the software Engineer interview , it was considered good....what an irony a machine is replacing human but human with machine is not allowed !!!
Except Devin is a joke. It does not work. You can be a math genius (Although I would not say so just because he is quick. Being quick is different from solving difficult problems. There are people who won the Field's medal who would not win that competition.), and still do dumb things.
Have you ever seen a codeforces question? Especially at the higher levels, there's no way the average development "difficult" problem is anything compared to competitive programming at that level. And it doesn't work yet. It'll get better. (And actually it does work now with o1)
I really don't think constantly making comparison to others is healthy as a Software Engineer. What these young people is doing is impressive but how does it translate to real world software? There are so many other attributes you need as a SWE, speed is the death of quality.
Agree with your point about comparisons, but What do you mean "how does it translate to real world software?" They literally bootstrapped and shipped a functioning product in 5 months that allegedly has a decent improvement over baseline metrics of other top research labs, and they raised several million for it. If that doesn't quantify what "good software" is, then I dont know what to say. And before you say "Yea but they shipped a subpar product after speedrunning it in 5 months!" Yes, but your "real" software engineers at google were twiddling their thumbs for 6 years sitting on the transformer paper and still messed up the release of their premier generative AI bot. These people are not just speed coding one tricks. They also know how to run a business and they know how go to have a good go to market and product strategy. Thats honestly better than most SWEs I know who barely have any product sense. Good software is something that has a product market fit, generates traction and revenue, and solves a problem. They've arguably hit all those criteria.
there was a research paper from google or someone big that compute increase is proportional to solving more complex problem. they might've proven that the plateau only comes when you can't exceed compute power (I might be 100% misremembering the conclusion). anyone remembers that paper?
Hard skills: top of leetcode Soft skills: Convince top of leetcode to work for you. I mean, i always try to get better, better people serve as inspiration. But i started to think that the real skills are not hard...........
@@CoconingaBut can we be serious here? Very soon, robots will be able to replace most doctors, they will perform all known surgeries. It will no longer be one doctor performing one surgery and then maybe three other surgeries that day, it will be 30 robots doing it simultaneously on 30 patients. People who make and maintain these robots are the ones who will be making money.
Compute does matter a lot, as seen by what SORA does from 1-16x compute. But the underlying algorithm also needs to be correct and efficient for "AGI" level of reasoning to be feasible, imo. Also the bit about humans, being able to do "original" thinking, is contentious for me. Lot's of original ideas come from the intersection of ideas from other sectors, or a combination of what others have thought about before. Like how piano playing led to the idea of frequency hopping for wifi networks. LLMs ability to have access to so much data, and the possibility of it linking ideas and data across different fields can lead to original ideas, that humans would take decades to learn and finally put together, in an eureka moment. LLMs also have a disadvantage, as the have no knowledge of the real world, only what is available to it through text.
Keep going! You will only get better over time the above comment is stupid because who needs to be him when you can be you and if you are superior to him then therefore become HIM
@@stoppls1709 its a joke ,I know that 3 leetcode questions is pretty booboo lol. Ive done more in the past in a single day but this time I was being consistent 3 per day. "your are"
The AI revolution is akin to the agricultural revolution, where technology reduced manual labor but created specialized jobs like harvester drivers, repairmen, and engineers, signifying not job loss but human development and occupational evolution. even this comment I regenerated with GPT for you fellas, to make it shorter and save you time, cheers😉
I'd rather be a Carmack or a Tim Sweeney than whatever these "competitive programmers" are. I'm sure there's value in getting some degree of aptitude with those kinds of tests, but past about 90% I'd expect severely diminishing returns, at which point you're doing it to what... impress other nerds?
Yeah, honestly I wonder if the competitive guys actually end up producing things of note. I mean just looking at Devin, it just isn't impressive to me in any way. It's a wrapper over ChatGPT and that's basically it.
People don't understand, your environment and childhood matters a lot. Chances are he'd be average kid if born into a family that didn't encourage him academically. Don't feed bad, if you weren't pushed to learn,
True. Most people don't realize it's not about the talent but the environment and the upbringing these kids got that made them who they are today. For me, I didn't had the best environment nor the best support from my family to do better in academics. I was emotionally neglected in my family. Constant quarrels betweens parents everyday have traumatized me to this day.
Here is a question i asked gemini, "hey i earned 150000 last year and spent some money on rent - 2000$, grocery 350$ and phone bills 5000$. How much money i have left?" a human would assume 150000 was probably yearly or ask a clarifying question. whereas gemini gave me this answer. "Great! It sounds like you are financially responsible by keeping track of your income and expenses. After accounting for rent ($2,000), groceries ($350), and phone bills ($5,000), you have $142,650 left." if i have to give correct input and context there is no difference between database and AI. there is a long way to go... We don't know when AI will take over but we know for sure that there is plenty of hype going around that is going to make many people rich.
Thank you, bias af because he grinded a skill that will be made irrelevant in the next decade and now he equates software engineering to generating piss code on a timer.
@@cachestache2485 competitive programming is supposed to be a fun mind sports. Calling it generating "piss code" is a good way to call yourself stupid in public tho.
@@rohakdebnath8985 It is, what use is it to anyone? It's like solving a crossword puzzel, if you don't get that you clearly have no idea what you're talking about. To double check what have you built? Where have you worked? What interesting and unique problems have you solved. I'll tell you, nothing. You have less depth then the joke you're defending. Open your mouth just like him with nothing behind it. You grind leetcode let the professionals actually make things sheep.
@@cachestache2485 Its called a mind sport, calling it piss code is stupid and you are probably one. I cant waste my time arguing what the purpose of a mind sport it.
Maybe I should quit university (I will finish bachelor's degree at SE in 2027) and go learn plumbing instead because I has do plumbing for my family quite often earlier...
@@SandraWantsCokewhat?? You sound a little over the top, the cost to develop technologies to make energy efficient, 3d coordination, movement and price won't be worth for a very long time
@@Sg6CrossOver Interesting points, but I don't agree, since self driving vehicles already exist, and modern factories are a bunch of robots doing similar things to surgeries.
@@SandraWantsCoke They've should be autonamus 10 years ago, geforce was trying to do it nearly 20 years ago, all of that just to put on perspective how long it takes to advance those technologies. We're having advancements based on how much computer power and data in the internet increased in the last 10 years. It will take longer to achieve things that had been idealized 40 years ago.
With how bad some AI data sets can be, I wouldn't be surprised if their market advantage was to just write their own (albeit world class) coding data sets for 5 months straight.
I also feel like a plateau is coming. I can't explain why my gut tells me this, but I think we're going to stall a bit with transformers/LLMs/current gen. They will keep improving, but the "exponential" progress everyone things is/should be/will continue happening will turn out to be linear by next year. Just like we were many years between "ML" and "transformers," I feel there will be several years before transformers give way to the next thing. The hype cycle will probably fade before the next "AI Winter." LLMs are really, really great at a bunch of things. Applying them to problems they are good it will be worth a lot of money to the likes of NVidia and ServiceNow and domain-specific automation. But most of the "moonshot" projects will probably fail to significantly improve on what we already have in the next couple of years ... and people will feel disillusioned at that point.
not that my opinion should mean anything, but 10/10 video, dude informative, hilarious, and super entertaining I had watched Fireship’s video about Devin a while ago, but Neetcode really had the scoop on this - I totally thought some random 28 year old in SF made it instead of some brilliant world-class competitive programmer
Man, Devin better be a fucking antichrist-level event the way people are fawning over these math medals. If they didn't create their own AI, then they're bound to whichever product they are building on. I think that Devin all on its own is a great idea and really cool, just like AutoGPT was, but for some reason people seem to now think that the missing piece was math medals?
Neal Wu actually puts up videos in here, youtube. I've been wondering where he works since I found out his leetcode score (oh, my god!). Now I know. It makes perfect sense, doesn't it?
I rather don't believe in this project, but actually, I am trying to build something similar. I think that guys from the company build some pipelines manually to use output from some LLMs... The best solutions for coding will come from using the specialized neural network in the most direct way of doing it. I build something similar to them, but it's amateur-like if-ology.
Gotta learn more about human spirit and the fight against odds... Stats means a lot in a paper, but the human history is filled with people with no stats being extremely important and defeating giants of their fields. Dont let that worldly lie get into you.
Are you sure about that? The history is filled with smart people only. Every chip in your microwave and phone, the materials in your car, are all made by the smartest this world has to offer. Do you think it's some idiots who keep improving LCD panels? It's the best of the best.
If you are founder and are building a product, remember, being amazing at programming is cool, but people buy your product not because it is a great code, but because you are solving their problem. Things like design, marketing and sales matter just as much if not more when it comes to building a great product. So don't compare yourself, we all have our advantages and disadvantages :)
The answer is in your rambling…. One might be smart to demonstrate their reasoning but it doesn’t necessarily mean the reason is perceived by someone else the some way. The reason people like me shut down when encountering leet coders
This kinda stuff is exactly why I keep thinking of giving up on being a programmer. I still struggle at easy lc problems after 100 problems or so. Should I even keep at it?
Search for companies that don't interview with useless lc problems and make sure you can build real world stuff.
You are absolutely good enough to be a programmer, so long as you enjoy programming at least a little bit and have the desire to put in a consistent amount of effort. And yes, there are definitely companies that won't give super difficult LC interviews.
I know it's demotivating, but don't think of it as being smart or stupid. It's more that everyone's brain is wired differently. What makes you bad at one thing makes you better at others.
The beauty of software engineering is that it's more than just about coding. You can differentiate yourself in many different ways.
I would suggest to not compare yourself with others, there is always someone way better than you , and someone way worser than you in any domain.
You either like to code or you don't. That is all that matters. Yes, you do need to grow as a programmer and learn your craft but if you're not doing it as a job then it does not matter how long it takes you to learn. Here's a secret. Only about 1% of programmers, if that many, are as talented as the people in this video for one simple reason. They happen to be math geniuses. Learning a programming language is only a small part of the package. A far larger part is gaining an understanding of algebra, linear algebra, trigonometry, calculus and geometry amongst other things. Someone who is born good at this will always be better than you as a programmer. Instead of looking for back pats and head rubs on RUclips decide if you like programming enough to stick with it and take it from there.
whyyyyyyyyyy thats's stupid
I know I am never going to produce singularity changing code like these guys, but I work from home, deploy regularly, have a girlfriend, family and friends that love me and make enough each month to do whatever I want and save. I am not going to change the world, but I really am making the most out of it. Count your blessings, don’t compare yourself, be humble and grateful.
Perfect comment!
That is a good life, you are right. But I wonder if it’s possible to do both without sacrificing one or the other to some extent. Probably not, but that would be ideal
@@darcash1738 sacrificing the latter isnt even an option if ur not a genius in the first place
You already are changing the world of at least some people around you. And that's more than enough.
Keep winning dude
Becoming the smartest genius software engineer to destroy the career of every other software engineer ....
Modern day alpha male. Strongest in the tribe beats weaker males to death
"It is not enough that I succeed. Others must fail."
@@dannyhantx lol 😅😅😅
im being completely real here. if he is actually so good that he can invent software that takes your job without him lifting a finger, does he not deserve that? in my opinion he deserves it
these guys aint geniuses. they just had the growth mindset and worked on it over a long period of time. ANYONE has the capability to become as smart as these people. you just need to spend a lot of time and really enjoy what you do, and you will be as smart as these people. its time spent on something more than talent, even the talented have to study years for legendary grand master on cf. even if you are late to it, you can definitely become really strong after spending 2 years working on it, which is most important.
I know a guy with two gold medals in the IOI. He contributes code every single day, and runs a weekly reading group where he covers really deep technical subjects.
He's obviously very intelligent, but more than anything he is just _driven_ and I think we don't talk about that much.
Exactly. That part is almost always left out when talking about high performers and highly intelligent people. You have to practice most things in life that aren’t innate (like breathing). This stuff isn’t magic.
100% agree
@@tasheemhargrove9650but then you got the crack heads who do practice breathing and they are on a whole 'nother level. that classic bruce lee saying.
Every single person in the euro league, G league, NCAA, NBL are driven. A fraction of them make it to the NBA. At a certain point, you're the bottom percentile of the top 1 percent, and then there's the 99% who never could even if they wanted to.
@@tasheemhargrove9650 You enjoy what you are good at. If you are a coding genius and find everything easy, you would be motivated and driven too.
“I’m not saying you’re stupid” then proceeds to list all the reasons why I’m stupid.
The way he solved the first problem is using the formula a^2 - b^2 = (a + b)(a - b). To recognize it that fast and do the computation on the fly is very impressive.
It’s one of the first things you learn in high school pre-calc. Impressive he knew about it, but basically every math student will know about it immediately.
yea that's why they don't usually ask square. They ask year^another_year - another_year2^another_year or solve for power m in a^m - b^m = y
they are mostly good marketing strategists.
10:12 you were definitely about to say lazy there lol
Surprised you caught it 😅
😂
I think people give AI more credit than it deserves and human genetics less credit. The things we get out the box are lowkey cracked- facial recognition, motor movement, self-training etc. Sure an AI can consume hundreds of years of data and become a glorified search engine and there is more potential output but unless it can utilize an incredibly accurate simulation and discover new things, it is still bound by human time in it's discoveries through good ol' trial and error.
Most people don't really realize how technically strong these guys are and the potential they can build at. Maybe they don't know how to code in react or whatever , but the point is that their fundamentals are so strong, especially the concrete mathematics they work with, its really the core of all computer science and they nail it. All this stuff we talk about in the industry are really just high level abstractions originally built by people who were really solid at math. So its guys like these, the so called nerdy guys as he says, that have the potential to create better solutions for the world. Its just that we need to push competitive programmers in a direction where they can apply their skills in the real world.
they are not programming in assembly mate
Sorry but no. I have enormous respect for John Von Nuemann who solved sorting in the most mathematically optimal way in 1945 (and other things). But he wasn't an engineer. He was a mathematician. Computer science is a field of mathematics. You don't see mathematicians building bridges. In fact they are notoriously bad at engineering. The same applies here. The manhattan project was made up of a few flashy physicists and thousands of engineers.
They could very well be good software engineers, but your conclusion is based on nothing but conjecture. And there are a lot of people that will tell you that Leetcode does not make you a good engineer. The fundamentals will help you a LOT. They will make you a BETTER engineer. But they are one piece of a very large puzzle of skills.
Especially for a senior software engineer in a FAANG company (for example). Coding ability isn't the factor that gets you to that level. They have those tough interviews because coding can't be what you struggle at. The design, management, how to sell, work with enormous ambiguity, get people to work well together and so on. Those are what get you beyond base level in those companies. The coding was the most basic requirement like being read/write is to be considered literate.
It's not that they don't know React, they did not have good error handling lmao💀, also this is just a chatgpt wrapper. A small company cannot even afford to make their own llm that is more powerful than tech giants. It was fine, they used tech giant product, but guess what they are lying too lmao. Doesn't really give off a good software engineer vibe, when all they are doing is scamming and lying for investor money
@@jvandervyvercorrect. they have potential but that doesn’t mean anything until they make a good product. CS, let alone math, and software engineering are different
Engineering is not a sub discipline of math, which I believe is what you're implying. You can see this most clearly in a bicycle. We just maybe a few years ago began to understand and model the way a bike works mathematically, but bikes are old as hell. Bill Hammock's excellent book The Things We Make elaborates on this point.
4:16 is simple when u know the trick to too. He does the operation quickly because he trained well and did his reps. (255-245)(255+245). He immediately got the ten, so his attention was at the addition right away since he would just need to put the 0 on the end once he got that. With enough training in addition, he recognized perfect additions to 100 very easily. 63 and 27, 55 and 45, etc. so he saw it was 500 quick too. Add the zero, 5000
11:15 also easy. Division rules: 2 check last digit, 4 check last 2 digits, 8 check the last 3 digits. Since 8 letters long 10/8 remainder is 2, so A.
wow..... thanks for the breakdown
Those problems are easy once you understand them. Still impressive for a kid his age
@@darekmistrz4364 most people can become smart if they are in the right environment. Maybe some point in the future, what we see as impressive will become the standard once education improves enough
You ignore the 2s when you do the addition and subtraction. So it really is quick
Could you explain what do you mean by "2 check last digit, 4 check last 2 digits, 8 check last 3 digits"? What does 2,4,8 and " checking the digits" refer to?
I feel bad for the girl who was competing against him in that mathcounts competition lol. Of course that for reaching that point she is insanely smart, but she got completely overshadowed by scott
She just studied different things; and wasn’t questioned on them.
did you get to watch the entire video?
@@mohammadhassan1649 sorry this isn't true. the question itself is a pretty generic math competition question. all it required was the identity x^2-y^2=(x+y)(x-y). anyone who did competition math enough to get to that point in mathcounts would know that identity like the back of their hand. what made him answering these questions impressive was the speed he had. i have 0 doubt the girl could also solve them in a few seconds, but not nearly as fast.
@@mohammadhassan1649 Just like how I failed every single Leetcode interview question
@@zhuhw absolutely love
It is a well-known fact though that interview performance (i.e. LC solving) does not correlate with job performance. There is a huge difference between solving 10-30 minute puzzles and working in codebases with 10s of millions of lines of code with decades of unwritten assumptions in underspecified problems, emergent behaviors, etc.
It is an industry-changing tech for sure, but not yet worried about myself as it is currently making me more effective.
(Btw. Yes on math importance being underestimated, speaking as a CS/math double major with two masters and a PhD)
Or working with people which matters when you don't have requirements spoon fed to you.
6:29 I’m not too sure about they explanation, but I think of it as 5 choose 2 (5C2) that represents the placement of 1 and 2, since they will always be in one order from left to right, and fill in the rest of the empty spaces with 3! permutations of numbers. It comes out to be 10 times 6 which is 60.
While he may be good at mental sports, I’m not convinced the team is made of genius programmers. Currently devin is just GPT wrapped in a sandbox. Nothing technically groundbreaking. This is all superficial marketing, and the substance has yet to be proven
Cope
@@ayushshshsh how?
you're right, the only innovation they have shown so far is a nice UI, they're very late to the party for their project, nothing unique or special yet.
@@kguyrampage95 agreed
@@ayushshshsh Why would you troll him when he is in fact stating facts? Mathematicians and good competitive programmers who are computer scientists have an ego problem, which is exacerbated by the size of the field (everyone wants to be in Tech and AI). Just because you can solve olympiad-level problems does not mean you can build AGI. It takes more than that, a variety of people from multiple disciplines and backgrounds.
I take 6 min to solve this problem in coding. But this guy is mind blowing.
var a= "MATHLETEMATHLETEMATHLETEMATHLETEMATHLETEMATHLETEMATHLETE";
var position = 2010;
const b = (Math.ceil(position/a.length) - 1) * a.length;
console.log(a.charAt((position-b)-1));
there are 8 letters in that sequence so you just divide 2010 by that. Since there is a remainder 2, that'll be the letter where it stops; which is the 2nd letter, "A".
Its a simple modulo arithmetic problem. It can be solved in 30 seconds by thinking it through.
i figured it was MATHLETE[2010%8] pretty fast :)
Scott Wu is a legendary grandmaster on Codeforces. The difference between him and Neetcode (who is probably an Expert/Candidate Master on Codeforces) is like the difference between Neetcode and someone who's only solved 10 LC problems 😅
I've been programming competitively and studying DSA for the better part of three years now, and I'm just breaking Specialist on Codeforces.
They really are that cracked. Plus, they went to Harvard lol.
Not saying this to discourage anyone though. In the real world, you aren't competing with Scott Wu or his brother. If he wanted to work at FAANG he'd have been hired ages ago 😅
Keep pushing guys 🎉
@@eti-iniERThese people spent their ENTIRE childhoods on this where they basically had unlimited time. There are plenty of people competing with scott wu and his brother. You are being too delusional over their ability to complete one genre of cognitive tasks. Where is this same energy for the putnam fellows who are honestly probably "smarter" than these kids.
side not but I hate people like you lmfao: any competitive domain has ppl like this guy that glaze the fuck out of the top 1% by stating the skill differential as if its some kind of insurmountable gap, discouraging people that haven't hit their peak from trying by saying "oh btw not saying this to discourage anyone tho guys !!" just because you took 3 years to hit specialist doesn't mean everyone has to be measured by your (frankly below avg) progress. Stop discouraging people
@@jepper6140 Competitive Programming is mostly practice, literally no one ever solved a few problems and got to Candidate Master. Most people who start Competitive Programming at a young age and keep at it with IQ > 90, will become red eventually.
@@jepper6140Putnam fellows and IMO medalists are so much smarter it’s not even a question.
Guy was answering the questions before I could even finish reading.
bro has math privilege.
For those intimidated by these test geniuses - mastering a boxing bag and mastering boxing in a ring with a real opponent are two very different things. These tests are the former. Yes, the bag's useful, you should probably spend time getting competent with it, but it's only one piece of the puzzle and is no substitute for experience in the ring. And sometimes the skills that count in the actual fight are quite tangential to the ones that applied with the bag.
where can i watch your stream??
14:00 actually Sam Altman said the same thing. There's gonna be a point in this AI bubble where we are gonna be forced to find better algorithms. Or try to find better ways to make smaller more accurate models, rather than general LLMs.
One more thing I'd like to add is what Primeagen said "AI is good with generating media and language, since you do not have to be 100% accurate. Math and Coding needs the 100% accuracy"
"AI is good with generating media and language, since you do not have to be 100% accurate. Math and Coding needs the 100% accuracy" -- Wow, pretty thought-provoking
I'm glad I could answer the problems, but that response time lmao gg. more grinding let's go
I couldn't even read the question in time, and he already answered it 🤣
The first question is a difference of squares, in case you thought he was a computer(still is but not a savant just very smart dude).
X^2-y^2=(x-y)(x+y) = (255-245)(255+245)=(10)(500)=5000
Most of these Olympiad question have tricks that these kids memorize and yes their math foundation is very good to be able to recall difference of squares and do the mental math in their head lol
You can be slightly more rigorous with the math problem. A 5 digit number has 5 slots. First realize there are exactly 5 choose 2 pairs of slots. For each pair of slots it’s possible for the numbers 1 and 2 to be arranged in 2 ways. In only half of them 1 is to the left of 2. So there is 5 choose 2 valid ways for the numbers 1 and 2 to be arranged. For the rest of the 3 numbers, there are 3 factorial ways for them to be arranged given each 5 choose 2 arrangement of 1 and 2.
Therefore 3 factorial times 5 choose 2 is the answer. (Simplifies to 60)
How did you find for half of them 1 is to the left of 2?
"This, is the crackhead who I'm talking about" 😂
Until we get true AGI, which is a long way off, we will need more SWE every year to deal with the increased system complexity in society.
I am COOKED
No you are not.
@@brainitesThe duality of man
Me too😂
Am I the only one who doesn’t get the “just divide by 2” solution given for the permutation problem? I came up with a solution that is 3! * (1+2+3+4) which is indeed 60, and you can generalize this up to ten digits and even then it’s still just half of the total permutations. What is the intuition for the solution he presented?
Think of it this way :
You can separate the 120 combinations into two groups : one where 1 is before 2, and the other is where 2 comes before 1
And then ask yourself: is there a reason why one group should be bigger than the other ? NO, because here the digits 1 and 2 play the EXACT same role.
Thats why the 2 groups are the same size, so each of them is size 120/2 = 60
I think I get it now, there are two groups which are exhaustive, 1 coming before 2 and 2 coming before 1. Since they are essentially saying the same thing, they must be equal and therefore it must be 2*x = 120 and solving for x is 60. That makes a lot of sense now, thanks for the respose!@@zakariaelghazi
There's so much more into developing software it's why someone like Steve Jobs can be so successful. This is more of a relief than a concern for competition. Sounds like a lot of IQ but you need high EQ to lead. That's what ppl like Steve Jobs have, a deep understanding of human beings and your customers. Time will tell
i understood the permutaion question but the fact that he did it so fast at that age is scary
if you learned it at a younger age and were trained to do those you would be just as good
@@sajeucettefoistunevaspasme
FACTS!
@@sajeucettefoistunevaspasme But there still is a barrier to what you can teach a 1 month old newborn.
@@darekmistrz4364 he wasn't one month old in the clip where he used permutation
i'm not saying that you can teach general relativity to a one month old but there are certainly ways of educating children that dramatically change their behaviour which will affect how "smart" he is
@@sajeucettefoistunevaspasme cope
you were very sympathetic throughout this video. even if I disagreed with you on some your hypotheses, I still have liked your reasoning, ideas and character. a primagen-lvl streaming commentery and interaction. stay safe, keep on improving, we are in great times
What did you disagree with?
Yeah we can see now how fake it was
How fake please explain
Wait till they hear about Gennady Korotkevich
I heard about the wonder kid years ago, similar background to Larry Page, both Korotevich's parents are Computer Science professors at Francysk Skaryna Homiel State University, and both of Page's parents were professors in Computer Science at Michigan State University.
he is friend of scott wu. All are friends. Gennady also tweeted about DEVIN ai
They're quite close to his level, actually.
Would love to see some “Neetcode Math” videos here or on your site. I’d curious to see which topics you’d highlight.
there are axioms in math as well though they seem very basic but really important assumptions
I think I can't agree with the last bit of the video. Machine learning system indeed do the system-2 thinking. They apply all the logics formulatically based on the given data. at least 100X faster than human. If you are given a choice between A and B, you would make some arguments and then choose 1. Machine learning does the same thing except it can do 1 million arguments in a second and choose A or B.
LLM's looks like they are on steriod, I get it. But that is because human perception is shallow. You can't really comprehend the fact that every data point is going to a very sophisticated machine learning model (those transformer architecture) and a giant book of logic was made when the LLM is trained with massive data.
It's just the retreiving process is so fast that it looks like it is on steriod but what it really does is, if you give choice A and B, LLM looks at the giant logic book and start at page 1. it make some decision and that page take it too say page 245. it looks up there and make some decision, and then that page hints it to page 812 and it make some more decision. And just like that it travels the whole logic book to finalize the prediction and gives it to you. Now again this analogy is really high level and what happens inside is not known to any human and also this is not comprehendable by humans either.
Quick observation: Scott Wu has a brother (Neal Wu). Both had impressive results in competitive programming. The video presents some information about Neal as being about Scott (e.g. the Leetcode #1 ranking and the Twitter profile).
12:52 is where it was information dense , after that there wasn't much content
0:18 1. Introduction to Devon AI Team
Overview of the talent-dense team behind Devon AI
0:53 2. Competitive Programmers' Fitness
Surprising revelation about competitive programmers' physical fitness
1:16 3. Insights into Competitive Programming
Exploring the skills and approach of competitive programmers
2:15 4. Founder's Exceptional Skills
Highlighting the exceptional skills of Scott Wu, the founder
3:14 5. Importance of Problem-Solving Skills
Discussion on the significance of problem-solving skills in software engineering
4:03 6. Math Prodigy behind Devon AI
Unveiling the math prodigy driving the innovation of Devon AI
5:06 7. Coding Olympics and Talent
Exploring the competitive coding landscape and early coding skills
5:21 8. Permutations Problem Explanation
Brief explanation of a permutations problem with five digits
5:43 9. Permutations and Symmetry
Calculating permutations and understanding symmetry in numbers.
6:35 10. Early Exposure to Math
Importance of early math education and competitions.
7:18 11. Launch of Coignition Labs
Introduction of Coignition Labs and the creation of Devin AI.
8:16 12. Investing in Disruptive Technology
Discussion on investing in disruptive technologies and innovation.
8:20 13. Debate on AI vs. Programmers
Exploring the potential of AI replacing programmers.
8:31 14. Cognition Lab Team
Overview of the team behind Cognition Lab and Devin AI.
9:03 15. Tech Innovation by Young Minds
Achievements of young developers in tech innovation.
9:51 16. Utilizing LULMs in AI
Discussion on the use of existing LULMs in AI development.
9:59 17. Skills in the Age of AI
Importance of fundamental skills in the era of AI.
10:15 18. Deciphering the 2010th Letter
Solving a coding problem through mathematical reasoning.
11:14 19. Competing with AI in Coding
Exploring the advantage of using AI in competitive programming.
12:03 20. The Mission Beyond Software Engineering
Discussing the broader goal of solving global challenges with AI.
12:50 21. Importance of Reasoning in Robotics
Highlighting the role of reasoning in mathematics and science.
13:39 22. Challenges in AI Reasoning
Addressing the complexity of reasoning abilities in AI development.
14:35 23. Efficiency vs. Replacement in AI
Debating the impact of AI on software engineering efficiency.
14:58 24. The Impact of AI on Engineering
Discussing the challenges and benefits AI brings to engineering and problem-solving.
15:21 25. AI Interacting with the Physical World
Exploring the need for AI systems to interact physically, such as through robots.
15:52 26. Data Companies vs. Self-Driving Cars
Analyzing the shift from data companies to the challenges of achieving fully self-driving cars.
16:42 27. Embodiment in Intelligent Agents
Understanding the concept of creating intelligent agents capable of navigating the real world.
17:15 28. Human vs. Computer Learning
Comparing the fundamental differences between human and computer learning processes.
18:15 29. System 1 and System 2 Thinking
Explaining the two systems of thinking in the brain and their functions.
18:35 30. Large Language Models and Original Thinking
Examining the limitations of large language models in terms of original thinking compared to human brains.
19:08 31. Evolutionary Differences in Reasoning
Discussing the evolutionary advantages of human brains in reasoning abilities.
19:38 32. System 1 Coding Mastery
Exploring the realm of fast and efficient coding.
19:43 33. System 2 Coding Complexity
Delving into the intricacies of problem-solving and human ingenuity.
19:52 34. The Billion-Dollar Question
Contemplating the future of coding and AI development.
19:57 35. Uncertainty and Speculation
Examining the unpredictability of AI advancements and expert opinions.
20:06 36. Elon Musk's Predictions
Analyzing Elon Musk's forecasts on AI surpassing human intelligence.
Generated with Tubelator AI Chrome Extension!
one of my friends - he dropped out of college and was also a USACO plat - its crazy seeing his intellect.
Reactions from Neel everytime the Olympiad clip appears...are hilarious lollll
That math question was pretty easy if you're good in school level algebra
255^2-245^2
= (255+245)×(255-245)
=500×10
=5000
If you're focused then you can solve it within 3 to 5 seconds.
You: That ma…
Scott Wu: 5000
@@sporefergieboy10 yes it only takes 3 sec
@@miteshranjanpanda1776 you would get utterly obliterated in a math competition, shut up you absolute noob lol
maybe i'm insane and coping but the fact that devin has "competitive developers" is like suuuper indicative that devin is meant to capture VC funding. like what kind of VC could resist a team of geniuses putting their brains together
Remember how everyone was singing praises for Sam Bankman-Fried, calling him a genius and all? And look how that turned out. Not saying this is the same deal, but it does make you think twice. Everything seems a bit too perfect-like, the team, what they're selling, their track record. Kinda makes you wonder, doesn't it? Always good to keep your eyes peeled for stuff like this.
SBF is fluff, but these ppl have the educational background, pure talent. Big diff. 😊
I am just enjoying the scam in the making unfold and see investors get their monies fried.
@@sayyidiskandarkhan3064Seriously, you think SBF didn't have the educational background? He is very talented.
@@brainitesthat is just a single cherry picked example
SBF? Never showed any skill just a scammer. Of course devin is overhyped for max funding but that doesn't mean they are scamming people. There are already quite some people have early access and no it of course it is still limited. It also still depend on the LLM under the hood. Complex "wrapper" will be more common soon.
I’d argue that the quickness of his answer of 60 was simply due to him seeing this question before. Not saying he isn’t extremely intelligent but it seemed a little too fast for him to read and compute that in a matter of seconds.
He had seen this TYPE of question before, but he computed given parameters: 5 numbers times 1/2 of the permutations = 60. Easy math.
@@darekmistrz4364 my argument isn’t the complexity of the question, but the rate of how quickly he answered it.
I think they can read the question on the ipad as well. The guy is there just to read it to the audience
If you've thought of giving up, please do give up, straight up.
People who love/like doing what they're doing never think of giving up what they're doing, ever.
People who love/like what they're doing will conjure endless reasons why they should continue. People who aren't sure, or don't really like what they're doing will dance around excuses to stop - this activity will drain and harm you.
So, give up. There's no shame in it. It just means it's not for you, it just means you won't be happy with it and it will burn you out. Find your own happiness elsewhere. To each their own.
Cracking a job interview using AI or chatgpt is prohibited for humans but when Devin AI clears the software Engineer interview , it was considered good....what an irony a machine is replacing human but human with machine is not allowed !!!
Except Devin is a joke. It does not work. You can be a math genius (Although I would not say so just because he is quick. Being quick is different from solving difficult problems. There are people who won the Field's medal who would not win that competition.), and still do dumb things.
Have you ever seen a codeforces question? Especially at the higher levels, there's no way the average development "difficult" problem is anything compared to competitive programming at that level.
And it doesn't work yet. It'll get better. (And actually it does work now with o1)
@@takeuchi5760 it was literally exposed as a fake piece of software, creating its own issues and trying to fix its own issues. Its not legit
2:50 - 4:02 harsh but truth spoken
People in the Western world grow up sheltered from the truth. Some still believe that with enough hard work anyone can be Wu😊
Yo neetcode/anyone else, do you have any recommendations of "math" activities for gains/fun?
Do project euler
@@astroflexx82project euler imo gets too esoteric for the average non-math major. standard competition math is definitely more accessible
This one didn't age like fine wine
What happened?
@@deepjyotideb1173 From what I've heard, DevinAI turned out to be a scam (same as most "AI products" coming these days)
@@deepjyotideb1173
He still thinks what he said
13:16 have you ever heard of godel's incompleteness theorem
I really don't think constantly making comparison to others is healthy as a Software Engineer. What these young people is doing is impressive but how does it translate to real world software? There are so many other attributes you need as a SWE, speed is the death of quality.
Agree with your point about comparisons, but What do you mean "how does it translate to real world software?"
They literally bootstrapped and shipped a functioning product in 5 months that allegedly has a decent improvement over baseline metrics of other top research labs, and they raised several million for it. If that doesn't quantify what "good software" is, then I dont know what to say.
And before you say "Yea but they shipped a subpar product after speedrunning it in 5 months!" Yes, but your "real" software engineers at google were twiddling their thumbs for 6 years sitting on the transformer paper and still messed up the release of their premier generative AI bot. These people are not just speed coding one tricks. They also know how to run a business and they know how go to have a good go to market and product strategy. Thats honestly better than most SWEs I know who barely have any product sense.
Good software is something that has a product market fit, generates traction and revenue, and solves a problem. They've arguably hit all those criteria.
@@ashwin3073 Wow, you're gonna come in all my comments and gas on, dude, get a life, jesus.
You also have like 10 other comments here, could say the same thing about you 🤷
@@ashwin3073 Sounds like a rat race
@@sachins5784 I dont think you know what a rat race is, because I don't see how it relates at all to what me or this other guy said.
there was a research paper from google or someone big that compute increase is proportional to solving more complex problem. they might've proven that the plateau only comes when you can't exceed compute power (I might be 100% misremembering the conclusion). anyone remembers that paper?
Where is devin? Is it solving leet code questions.
Hard skills: top of leetcode
Soft skills: Convince top of leetcode to work for you.
I mean, i always try to get better, better people serve as inspiration.
But i started to think that the real skills are not hard...........
5:15 his face was absolutely - what the help fkin sorcery going on here 🤣🤣🤣🤣🤣🤣🤣🤣
And this is why I decided to go into med school instead of software engineering lol
So, you'll be brushing dust off robot doctors while they're performing surgeries, get it.
@@SandraWantsCoke Software engineering coping really hard rn lol :(, sad your get rich quick easy scheme didn't work?
@@CoconingaBut can we be serious here? Very soon, robots will be able to replace most doctors, they will perform all known surgeries. It will no longer be one doctor performing one surgery and then maybe three other surgeries that day, it will be 30 robots doing it simultaneously on 30 patients. People who make and maintain these robots are the ones who will be making money.
I think its a bit overexageratted that some smart guy created the 1 millionth AI co pilot software.
his past is not representative of his present
Compute does matter a lot, as seen by what SORA does from 1-16x compute. But the underlying algorithm also needs to be correct and efficient for "AGI" level of reasoning to be feasible, imo.
Also the bit about humans, being able to do "original" thinking, is contentious for me. Lot's of original ideas come from the intersection of ideas from other sectors, or a combination of what others have thought about before. Like how piano playing led to the idea of frequency hopping for wifi networks. LLMs ability to have access to so much data, and the possibility of it linking ideas and data across different fields can lead to original ideas, that humans would take decades to learn and finally put together, in an eureka moment. LLMs also have a disadvantage, as the have no knowledge of the real world, only what is available to it through text.
I just started solving 3 leetcode questions a day for the past week, mostly mediums with some easy and one hard and I thought I was the shi😢
your are NOT him
Keep going! You will only get better over time the above comment is stupid because who needs to be him when you can be you and if you are superior to him then therefore become HIM
@@adamS-zw9rk real
@@stoppls1709 its a joke ,I know that 3 leetcode questions is pretty booboo lol. Ive done more in the past in a single day but this time I was being consistent 3 per day. "your are"
@@I61void my bad, you might just be him
The AI revolution is akin to the agricultural revolution, where technology reduced manual labor but created specialized jobs like harvester drivers, repairmen, and engineers, signifying not job loss but human development and occupational evolution.
even this comment I regenerated with GPT for you fellas, to make it shorter and save you time, cheers😉
That's Mark Cuban former owner of the Dallas Mavericks.
Still ended up building a useless gpt wrapper lol
I really don't know how I watch Neetcode's long videos and never feel it!
The problem with LLM is even its inventors don't even know how it works. LOL
I'd rather be a Carmack or a Tim Sweeney than whatever these "competitive programmers" are. I'm sure there's value in getting some degree of aptitude with those kinds of tests, but past about 90% I'd expect severely diminishing returns, at which point you're doing it to what... impress other nerds?
Yeah, honestly I wonder if the competitive guys actually end up producing things of note. I mean just looking at Devin, it just isn't impressive to me in any way. It's a wrapper over ChatGPT and that's basically it.
People don't understand, your environment and childhood matters a lot. Chances are he'd be average kid if born into a family that didn't encourage him academically. Don't feed bad, if you weren't pushed to learn,
True. Most people don't realize it's not about the talent but the environment and the upbringing these kids got that made them who they are today.
For me, I didn't had the best environment nor the best support from my family to do better in academics. I was emotionally neglected in my family. Constant quarrels betweens parents everyday have traumatized me to this day.
People keep saying we still don't have self driving cars. They have existed for a bit, Waymo, Cruise.
they want non-geofenced self driving better than or equal to human driving
@@erkinalp fair
Think you're good at something? Remember, there is always an asian who can do it better.
Here is a question i asked gemini, "hey i earned 150000 last year and spent some money on rent - 2000$, grocery 350$ and phone bills 5000$. How much money i have left?" a human would assume 150000 was probably yearly or ask a clarifying question. whereas gemini gave me this answer. "Great! It sounds like you are financially responsible by keeping track of your income and expenses. After accounting for rent ($2,000), groceries ($350), and phone bills ($5,000), you have $142,650 left." if i have to give correct input and context there is no difference between database and AI. there is a long way to go... We don't know when AI will take over but we know for sure that there is plenty of hype going around that is going to make many people rich.
On the other hand, GPT understands me just fine when I misspell words or even accidentally use wrong words. It understands what I mean.
Mayor of Yapperville right here
Thank you, bias af because he grinded a skill that will be made irrelevant in the next decade and now he equates software engineering to generating piss code on a timer.
@@cachestache2485 yeah. he's not convicing anyone but himself
@@cachestache2485
competitive programming is supposed to be a fun mind sports. Calling it generating "piss code" is a good way to call yourself stupid in public tho.
@@rohakdebnath8985 It is, what use is it to anyone? It's like solving a crossword puzzel, if you don't get that you clearly have no idea what you're talking about. To double check what have you built? Where have you worked? What interesting and unique problems have you solved. I'll tell you, nothing. You have less depth then the joke you're defending. Open your mouth just like him with nothing behind it. You grind leetcode let the professionals actually make things sheep.
@@cachestache2485
Its called a mind sport, calling it piss code is stupid and you are probably one. I cant waste my time arguing what the purpose of a mind sport it.
I was thinking the math questions were straightforward then I realized this kid is 8 years old, thought I was a nerd
Maybe I should quit university (I will finish bachelor's degree at SE in 2027) and go learn plumbing instead because I has do plumbing for my family quite often earlier...
No. It will be robots doing the plumbing. It will be hundreds of robots performing surgeries instead of one doctor.
@@SandraWantsCoke it was just joke but if it no then what job to choose if robots do all?
@@SandraWantsCokewhat?? You sound a little over the top, the cost to develop technologies to make energy efficient, 3d coordination, movement and price won't be worth for a very long time
@@Sg6CrossOver Interesting points, but I don't agree, since self driving vehicles already exist, and modern factories are a bunch of robots doing similar things to surgeries.
@@SandraWantsCoke They've should be autonamus 10 years ago, geforce was trying to do it nearly 20 years ago, all of that just to put on perspective how long it takes to advance those technologies. We're having advancements based on how much computer power and data in the internet increased in the last 10 years. It will take longer to achieve things that had been idealized 40 years ago.
With how bad some AI data sets can be, I wouldn't be surprised if their market advantage was to just write their own (albeit world class) coding data sets for 5 months straight.
I also feel like a plateau is coming. I can't explain why my gut tells me this, but I think we're going to stall a bit with transformers/LLMs/current gen. They will keep improving, but the "exponential" progress everyone things is/should be/will continue happening will turn out to be linear by next year. Just like we were many years between "ML" and "transformers," I feel there will be several years before transformers give way to the next thing. The hype cycle will probably fade before the next "AI Winter."
LLMs are really, really great at a bunch of things. Applying them to problems they are good it will be worth a lot of money to the likes of NVidia and ServiceNow and domain-specific automation. But most of the "moonshot" projects will probably fail to significantly improve on what we already have in the next couple of years ... and people will feel disillusioned at that point.
low iq take
not that my opinion should mean anything, but 10/10 video, dude
informative, hilarious, and super entertaining
I had watched Fireship’s video about Devin a while ago, but Neetcode really had the scoop on this - I totally thought some random 28 year old in SF made it instead of some brilliant world-class competitive programmer
bro eats sleeps and breathes leetcode
Man, you have every idea and reasoning just like me.. critical thinking, math logic and everything..
LOL I love how you avoid those things at the last moment:
10:12 cause I'm to l.... busy (lazy)
13:31 here's my f... five hundred page proof
Man, Devin better be a fucking antichrist-level event the way people are fawning over these math medals. If they didn't create their own AI, then they're bound to whichever product they are building on. I think that Devin all on its own is a great idea and really cool, just like AutoGPT was, but for some reason people seem to now think that the missing piece was math medals?
4:45 you're killing it🤣
Bro started spitting bullshit at the end😂
My imposter syndrome had reached new heights after watching this.
😂💯
Neal Wu actually puts up videos in here, youtube. I've been wondering where he works since I found out his leetcode score (oh, my god!). Now I know. It makes perfect sense, doesn't it?
yeah, every time I see his videos, I am getting sad, because the way he is submitting codes while I am finishing second sentence is crazy
thank you
Noice
You're welcome
end of the day you do what gives you happiness
The math Olympiad questions, just reminds me of further mathematics. Factorials and all its applications.
How does math relate to swe
For 99% of roles it doesn't
@@cachestache2485 Can Devin put the right coversheets on its TPS reports?
I rather don't believe in this project, but actually, I am trying to build something similar. I think that guys from the company build some pipelines manually to use output from some LLMs... The best solutions for coding will come from using the specialized neural network in the most direct way of doing it. I build something similar to them, but it's amateur-like if-ology.
The problem with AI right now is that it's based on data that people give them rather than a sentient consciousness.
Found you when the primagen reacted to your video a while ago - Been really loving the content lately thanks for all the work you put into it!
I thought I was pretty good ,but now I actually done.
Gotta learn more about human spirit and the fight against odds... Stats means a lot in a paper, but the human history is filled with people with no stats being extremely important and defeating giants of their fields. Dont let that worldly lie get into you.
Are you sure about that? The history is filled with smart people only. Every chip in your microwave and phone, the materials in your car, are all made by the smartest this world has to offer. Do you think it's some idiots who keep improving LCD panels? It's the best of the best.
Hey man, i really like the way you think. It's the first video i watched and i really enjoyed it.
My dad always told me when I was in school it was in the 2000s and 2010s, that there is a math and others in subjects.
This is discouraging no mattter what. Doesn't mean it should or will stop anything but it will still hit a spot
If you are founder and are building a product, remember, being amazing at programming is cool, but people buy your product not because it is a great code, but because you are solving their problem. Things like design, marketing and sales matter just as much if not more when it comes to building a great product. So don't compare yourself, we all have our advantages and disadvantages :)
The answer is in your rambling…. One might be smart to demonstrate their reasoning but it doesn’t necessarily mean the reason is perceived by someone else the some way. The reason people like me shut down when encountering leet coders