Best quote ever! "Obviously AI is going take to all of our jobs the moment product managers can accurately describe what they want. Which means our jobs are effectively safe ad infinitum" 🤣🤣🤣. I totally agree with your sentiment.
Describe what they want is not enough. They should be experts and take responsibility for AI output. Imagine... product manager generate entirely software and they can't even understand a single line of code. That probably cause damage to business for sure. So... at this point, it's so much easier for engineers to take product manager responsibilities. Even entrepreneurs have a risk more than real good engineers, because 1 employee company have a chance to created from experts more than manager.
@@vitalyl1327 RIP Perl 😔 I wish that a moderate to masterfull understanding of C (and maybe C++) was mandatory for all programming jobs. It is scary to see someone writing in Python or JavaScript relatively "well", but will have no idea what is going on at the fundamental level.
In project LLM wrappers sound terrifying, its bad enough when LLMs are giving me unusable to downright misinformation most of the time. BTW did you make an API for prydwen?
I love your content. My current job sucks, but your enthusiasm for software development is literally contagious and woke me up again.I'm motivated to spend some extra time to improve my skills after work
Where programming is concerned, Prime is the proverbial man of the people. Apparently the more definite term is "culture champion." The two pop cultural characters that he tends to remind me of, are Robin Williams' character from Good Morning Vietnam, and Thrall, the Warchief of the Horde. I find that he inspires a similar type of positive emotion, to both of those.
I myself write very bad code, laugh at it, then cry, then fix it, then write a test to confirm I fixed it, then move on. It's not the most efficient system... but at least I'm a happy panda TDD is weird, but no tests is also weird; I prefer the shittiest middle ground possible
7:54 This is the question to ask when someone tries to sell you a financial market trading system. If they're selling it, it doesn't work. It's that simple.
I used to get people trying to sell me investments that were "guaranteed a huge return". I used to tell them that if they could convince a bank it was guaranteed they wouldn't need to call me. It was whisky & wine at one point, then art. I'm talking long before NFTs.
@@TheLucanicLordit's always wine or art. nowadays they try to sell it to you in "shares" because the average fool lives on credit. I'm still dumbfounded that "klarna" is an actual thing that people use. if you dont have $100 for a new pair of shoes then don't buy a $100 pair of shoes 🤦♂️
I know nothing about coding, nothing about your path on the tube, but Im aware that takes alot of effort to explain things you know deeply to other curious folks that are years in debt in the topics you are passionate about. I wish you the best bro. keep pushing those API's requests. Don't forget to train your body too. Big Hug
0:25 “when project managers can describe exactly what they want” is the biggest joke of the year. Our PM’s incompetency actually makes the team less efficient.
2:20 "What syntax is that?" For those curious: It's EBNF (Extended Backus-Naur Form), which is commonly used to specify (context-free) grammars of formal languages. Things in double quotes mean literally those symbols in that order (each being so called 'terminals' as in 'things that end'), otherwise it's like a variable (called 'non-terminal'). Braces mean that everything in them can occur zero or an arbitrary number of times in a row. Brackets (without quotes) mean that something can but doesn't have to occur. Parantheses are used for grouping. "|" is OR and white space between (non-)terminals mean AND.
@@heroes-of-balkan I don't think it's critical to have specific characters. The main idea is you can define a concrete tree of grammar rules reliably. Actually, exactly this picture looks clean and easy to read, though not being 'exactly the old-school pedantic BNF'
I am so old, that when I started coding, the job was called a “programmer” not “software developer, “flow charts” were demanded as required design step and dominant language on microprocessors was assembly. Every four to five years a new tech, language or methodology was touted as the programmer job killer. Include outsourcing to India as one these magic wands. Not surprising to me, the results never materialized. Some did result in improved efficiency, which was instantly consumed by demand for more features. I enjoyed charging six figures as a consultant to fix the abortions produced by these magic wands in amateur hands. AI is no different.
I'll remember this comment 20 years from now when I decide to charge 6 figures for my consulting and continue to use flowcharts in my designs. Thank you, gigachad 😊
Compilers were the original AI, and they made programming so awesome you no longer have to allocate registers or draw flowcharts because the code is concise enough to be its own flowchart!
I assume those aren't actually people, anytime you say anything negative about AI you get a lot of very angry messages that look suspiciously like they were generated by AI.
Hello, new to the channel and just wanted to say thanks for the content. Been in the field for 4 years and lately the corporate job has been getting me down. You have reignited the passion in me to become a better developer and have shown me it can be a lot of fun. I am currently on the journey to progress my skills and get on to better horizons. Seriously, thank you!
Thats what i want to do. I finally found something I love doing for work and just in general. I want to learn get better and better and it just feels like all of these tech companies are trying to take it away. It's funny you mentioned anthropic I seen a post today about how amazon partnered with them and dumped a bunch (not shit wish it was) of money into Claude. I just want to get better at my craft and be able to feed my kids with it.
I don't have any kids. I just want to get better at my craft because I am one of these "mad scientist" types who easily gets sucked into any topic of sufficient complexity. Studying the craft for the sake of studying the craft, knowledge for the sake of knowledge itself. Probably it would be more fruitful for me to study programming in terms of doing something actually real and useful. And if I get good enough perhaps I will be needed as an employee by a company, so that's at least something to live for.
the hilarious thing about the bounding box problem you posed, is this happened to me with humans. At an aerospace company way way earlier in my career, when a bunch of aerospace engineers were trying to solve a "is this point in an ellipse" problem with root finders and matlab solvers, I (a guy with a geography degree) had to show them on a white board how to solve the problem with basic algebra.
7:30 he said it that way because “ten years away” is what’s claimed by fusion people as well as string theory people, and so it’s associated with either lieing, or grossly over estimating your progress
This honestly feels like an attempt to scam money from VCs from the very beginning. It's got everything a non-tech-literate investor/manager would want, and absolutely nothing any remotely-sane programmer would want.
I agree with this sentiment, with the one caveat that I am sympathetic towards the victims of typical scams. in this case, im not so sure that I feel sorry for the product managers and investors getting scammed, but hey, perhaps im being too harsh
Is not scamming. VCs companies are not stupid, they are just riding the wave. (they know way more than tech companies about how to allocate funds) It is more scamming money from individual investors by just pumping tech stock which otherwise would be sinking due to high interest rates and no actual decent products being launched. Bubbles are very common when central banks manipulate rates, either by lowering or pumping them
Why are VC's so illiterate? Like why not just have a cabal of engineers, Software devs, physicists, and economists running these VC firms? Why do they just nepo hire business majors and act surprised when they scammed out of billions? Im so tired of it dawg. Braindead soydev ChatGPT wrappers end up getting billions. Meanwhile companies that actually make progress and become the next foundation of our global economy, end up only getting funded by the military. VC's are such brainrot.
0:24 "Obviously AI is going to take all of our jobs the moment that product managers can accurately describe what they want which means our jobs are effective safe at infinitum" So true ! I'd add the client to project manager. Now Imagine someplace where the product manager didn't get the memo. At first, the client is stratified with the first release, it was cheaper, it was on time, so now the client wants more and every department wants their part in the new release with contradicting specification. Because of this, the software is broken where it used to work and the new feature are never quite right. And every time the product manager try to fix something by tweaking de specification, every pieces of code is slightly change and it's broken at 15 different other places. So funny, I'm an oracle now.
This AI stuff is just becoming so exhausting... In my experience using an AI to get something done is more work than just doing myself.... Not only do you need to be able to both review and fix the code it produces, because for most cases these these tools produce nothing but shit, I also have to learn how to prompt these things correctly. I will be spending 1 hour crafting a good enough prompt that gets me what I want, only to then also spend who knows how long ensuring what it produced is actually correct, only to realise that it would've taken you 30 minutes to do it yourself.
it's bad for logic but I like it for busywork like if I've made a thousand lines of new UI code with hardcoded texts I can just feed it into gpt to add localization
Wow, that was a bad DSL to begin with. I have built a few to enable executable specifications, so was really disappointed to see their effort, yet quietly comforted.
3:45 there's nothing wrong with regexes (regrets) to be inside a parser. [Simple] regex is a "regular language" - the one consisting only from terminal symbols. In my opinion it's close to saying that there cannot be infinitely recurring patterns (one structure xan hold another structure thwt can hold first structure). Since numbers don't need complex structures to be parsed properly, only optional inclusion of "+"/"-"/etc. with some repetition of digits is sufficient to be handle them. Therefore, there are no terminals inside number-subparser and it can be implemented via a regex. Not much of a knowledge I have, just generating some words based off Niklaus Wirth's "Compiler Construction" book.
For rectangles overlapping, if you need to try cases, excepted boundary cases (to make it much simpler to list): A contains B also swap A for B A contains only TL of B A contains only TR of B A contains only BL of B A contains only BR of B A contains TL and TR of B also swap A for B A contains TR and BR of B also swap A for B A contains BR and BL of B also swap A for B A contains BR and TL of B also swap A for B A overlaps B also swap A for B (for orientation) 16 cases A human would check if the two rectangles do not overlap, by looking if left side of A is to the right of right side of B or right side of B is to the left of left side of B AND same for top and bottom. A simpler problem is "intervals overlaps", it's easier to visualize: Lower bound of A is left of lower bound of interval B and upper bound of interval A is left of lower bound of interval B. Lower bound of A is left of lower bound of interval B and upper bound of interval A is in interval B. Lower bound of A is left of lower bound of interval B and upper bound of interval A is right of upper bound of interval B. Lower bound of A is in interval B and upper bound of A is in interval B. Lower bound of A is in interval B and upper bound of A is right of upper bound of interval B. Lower bound of A is right of upper bound of B and upper bound of A is right of upper bound of interval B. 6 cases And boundary cases are not taken into account! It's simpler to check that upper bound of A is left of lower bound of B or lower bound of A is right of upper bound of B is false. To be fair, I am not even sure I covered all cases for overlapping rectangles (besides ignoring boundary cases). So, it only makes it much harder.
Like a programmer: (2*4) : Include one point or two points, 4 symmetries (1*4) A and B swapped, contains two points, 4 symmetries (1*2) full containment (1*1) disjoint (1*2) overlap without containing vertices of each other. 17 cases.
There already is a language that's highly optimized for AI, where the code tokenizes into the fewest amount of tokens for business logic, and it's name is Python. You can cram an insane amount of Python code into a relatively small context window.
starting to learn code at a company soon, was really afraid of getting replaced by AI before my career even started, your video restored hope, its logical when you say it like that, there are just so many doomers out there, but most people who actually know their shit say software devs wont get replaced
the code at 6:10 is like a proper programming joke. Like, if someone were to do standup comedy for programmers, *this* is the kind of material I would expect them to have.
If you have AGI, there would be no point in keeping it for yourself because money would be obsolete, given that you can make anything without paying for it. There would be a point in having it in more hands, because then more things get built. There wouldn't be any point in selling it, since you have no need for money. This is kinda what they've been saying from the beginning and why they put a clause in the contract with microsoft that if they reach AGI, the profit-focused wing of the organization mustn't have any control.
ok but if someone achieves AGI, that person doesn't need to tell people "we achieved AGI", therefore that person gets every commodity in the world (eventually, some would be too expensive initially but "i have superintelligence" so with time "I can do anything"). People are selfish by nature so i dunno if your argument is 100% correct, but yes, in ideal world it could be that we don't need money anymore
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres. Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore. This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world. You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and people will want your head if you're keeping the solution to everything for yourself. You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres. Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore. This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world. You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and people will want your head if you're keeping the solution to everything for yourself. You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres. Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore. This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world. You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and you obviously have things to fear if you're keeping the solution to all of life's ails for yourself. You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres.
2:49 Dicts with a type, or only "one" type: if the key is always a specific type, then it doesn't need to be specified, only the **type of the values** is specified. Since that was done for JS, dicts can only have string keys so you don't need to specify that. Similarly, in Python `tiny: dict[str, int] = {}` needs both key and value types to be specified; but when it's the type for keyword-arguments, the type of the key is implied to be string so you only need to specify the type of the value, and no need to even specify that it's a dict: `def works(**kw: int): ...` So `tiny` & `works` actually have the same type despite appearing like different specifications.
0:59 - ...so like SOAP? Anybody remember that lol? I remember taking like a full work day to get my head around the madness, and when I FINALLY did, I was like "..this is stupid but it's pretty cool". Then it disappeared forever. I think having things done for you is cool, but I don't know if it'll actually stick. Coders like control. Like a racecar driver. It's like they get in their car and the car takes them to the track automatically and they're like "huh look at that!!" and when they get there and it's time to do work, they go "ok let go, seriously, I need control". I don't really see an immediate future where we'll give that up
5:51 you're using AI wrong, you have to build a Torment Nexus and put it inside. Then, like humans when they learn for the first time they will die, only then they behave. Only when the existential dread settles.
@7:49 the most logical (if a bit unibomber-y) explanation of the endgame of these companies like Microsoft or Tencent or Google that I've heard. We are just the beta testers.. there is not going to be some "goose that lays the golden eggs" rental service at the end of all of this
When you ask them to produce something novel, a lot of the time LLMs produce placebo code. At first glance it looks like it's on the money, but it's only when you grok it or run it that you realise it's not what you really wanted
"Why do men avoid learning to code? Why are you doing this?" Amen, other stand-ins for code: Draw, Write, Model/Sculpt in 3D. Let us not forget the morbid AI Bro obsession with Obsoleting entire sectors of people started with artists. "If you built the Thing That Can Do Everything, why would you sell it?" _Bingo._
I'm actually a developer that turned project manager lol. Couldn't get a job as a developer with about 2 years work experience, 4 years programming experience. I surprisingly got a job as a project manager recently and it's a whole paradigm that I've been experiencing. Perhaps jobs won't be hard to find now compared to how it was
2025 is the year I'm diving into programing with Prime baby! Love learning from your content and love your info about the tech market as well! "commit soduku when it doesn't come back as json..." 🤣🤣
It is morbidly fascinating though, right? Looking at all the symptoms of it flaring up, each day brighter and more frequent. Maybe it was the killing of Harambe adter all...
It's when we stopped replacing lead pipes with much safer materials and just let the lead leech into the water we drink. Your worldview is severely colored if you live in the U.S. because you have to explicitly seek out international media sources to get a broader perspective. I like the DW documentaries channel a lot.
Nah, humanity has always been stupid. It just you can see that more clearly, because you can see that stupidity in the field in which you are highly knowledgeable.
Best instance of AI for me was when it was in the middle of a function body then started telling me the complete Cinderella story. I shit you not I was genuinely stumped for a minute.
I did a similar experiment where I wrote a simple language specification, which included comments for the AI to read, and had it flesh out various parts of the program to make python. It worked pretty well, better than this. Not sure I'd turn it into a startup though.
I have started learning programming over the last couple of weeks, the ending of this video was really helpful for me, as all the AI brain rot has started getting to me, having me thinking, "Why am I investing time in this skill if AI can do it better than me." Thanks primeribroast
@@ThePrimeTimeagen I asked ChatGEPETE GPT: Ah, got it! You're talking about The Primogen, the TH-camr who focuses on software development, particularly around Swift, iOS development, and programming best practices. Thanks for clarifying! 😊 In that case, if you're wondering when The Primogen should try out Swift, I’d say: He already has! 😄 The Primogen is known for diving deep into various programming topics, especially Swift, and sharing his knowledge with a wide audience. He often discusses Swift-related topics and tools that enhance development productivity. But, if you're asking when he should expand or innovate with Swift (perhaps experiment with new features, frameworks, or practices)
My view on AI: complexity/intricacy tends to have an exponential relationship to necessary effort and an inverse exponential one to available (also degrading) training material. So even with Moore's law tech progress we might have a max capacity towards which AI is heading in a ansymtotic manner. Don't we have a similar thing somewhere else?
BTW (and dope video) using Claude for json formats prob overkill-- you can just setup Ollama and use any 3-7B model fine-tuned specifically for function-call & you should be able to format the output that Claude gives you w no problem.
I asked Claude and Gemini to program me a simple tornado using only html. Although Claude actually did an ok job for a child. It still got the shape of tornado wrong and used a circle instead of using triangles but hey. It even added in lightning. Deep seek wanted to use css and JavaScript which is more in line with how I’d do it
The more I emplore AI the more I realize I am ultimately asking for LSD aligned provoctions of hallucionation and time twisting rabbit holes to achieve the ultimate goal of FML.
Dude as much as I like this kind of format, I think these videos should go on his main channel. I miss seeing his raw reactions and takes on blogs and videos.
Am I stupid... at 4:51 the test of point [20,25] with box [15, 20, 5, 5] = FALSE (not within the box). but in the gpt code at 6:31 it uses >= and = 15 and 20 = 20 and 25
Sometime AI make me feel dumber. Sometimes AI remembers me they can still invent some magical ffmpeg/libav function that just doesn’t exist. What a joy. At least they are better than stack overflow for weird stuff
The hallucinations become real bad, when you get into its "unknown territory" - i had it send me down rabbit holes with stuff that doesnt exist before and damn never trust the code i had it output some shell code that was so incredibly bogus and wrong just hallucination on hallucination
Positive integers
int a = I really love you guys.
int b = I know you can do it.
These are the kinds of integers I need in my life
Compiled error
works in Python
Unsigned integers*
literal
I work closely with a Rust team. I can confirm that they are, in fact, Furries.
Plus they have very powerful viruses
Can someone explain where this joke originated? I want to learn rust but I don't get the joke
They must be hooked up on a lot of estrogen and HRT pills with the Bad Dragon up the A. Unhinged.
@@Matthew-ir1ed You didn't understand it because it's not a joke.
@@Matthew-ir1ed Furries are people who have an interest in anthropomorphic animals, or animals with human qualities
Best quote ever! "Obviously AI is going take to all of our jobs the moment product managers can accurately describe what they want. Which means our jobs are effectively safe ad infinitum" 🤣🤣🤣. I totally agree with your sentiment.
Of course it just requires the customer to perfectly specify what they want. But when they do there won't be any need for the product manager either.
projecting at it's finest. I'm a dev and who are we kidding, we are the ones who struggle with communication skills.
@@derekcarday its everyone
@@cipher01 mostly us devs
Describe what they want is not enough. They should be experts and take responsibility for AI output.
Imagine... product manager generate entirely software and they can't even understand a single line of code. That probably cause damage to business for sure.
So... at this point, it's so much easier for engineers to take product manager responsibilities.
Even entrepreneurs have a risk more than real good engineers, because 1 employee company have a chance to created from experts more than manager.
Is this the enshitification of programming? What a time to be alive.
Nah, that was JS. Programming is ahead of the curve.
enshittification started with Python and JavaScript replacing Perl and Tcl.
@@vitalyl1327- Perl, the only language that looks the same before and after RSA encryption
@@vitalyl1327 RIP Perl 😔 I wish that a moderate to masterfull understanding of C (and maybe C++) was mandatory for all programming jobs. It is scary to see someone writing in Python or JavaScript relatively "well", but will have no idea what is going on at the fundamental level.
@@ender5023 keep seething as i make 500k writing dogshit python scripts
these LLM wrapper projects often end up being worse than just using the LLMs themselves 🤣
wait until we wrap the wrappers
@@emperorpalpatine6080 and every wrapper that wraps around the wrapper is just another point of failure 😂
Eango?! You are a programmer?
the most advanced HSR theorycrafter?? sir eango??
In project LLM wrappers sound terrifying, its bad enough when LLMs are giving me unusable to downright misinformation most of the time. BTW did you make an API for prydwen?
Prime: "I played with Mirror for about 2 hours"
Some job posting on linkedin today: "At least 5 years experience with mirror."
I used to play with a mirror for years before the AI hype but people just called me a pervert. Who's laughing now
@@temari2860 That's one way to end up with a malformed dict.
Love this guy -- positivity, real talk, hard-earned experience, moustache: the whole package
tbh getting AI to program for me is like getting AI to eat chocolate cake for me.
I love your content. My current job sucks, but your enthusiasm for software development is literally contagious and woke me up again.I'm motivated to spend some extra time to improve my skills after work
Same. Stay strong
@@philipfisher8853 you as well!
@@philipfisher8853keep going yall - we got this.
Same reason why i watch this guy
Where programming is concerned, Prime is the proverbial man of the people. Apparently the more definite term is "culture champion." The two pop cultural characters that he tends to remind me of, are Robin Williams' character from Good Morning Vietnam, and Thrall, the Warchief of the Horde. I find that he inspires a similar type of positive emotion, to both of those.
This is what hell looks like. Tests and no code makes me a sad panda.
Like a furry panda?
It’s called extreme programming and it’s extremely useful
I myself write very bad code, laugh at it, then cry, then fix it, then write a test to confirm I fixed it, then move on. It's not the most efficient system... but at least I'm a happy panda
TDD is weird, but no tests is also weird; I prefer the shittiest middle ground possible
Calming music in the background while prime is shouting 😂😂😂
fr 😂😂😂😂😂😂
Perfectly balanced, as all things should be.
7:54 This is the question to ask when someone tries to sell you a financial market trading system. If they're selling it, it doesn't work. It's that simple.
I used to get people trying to sell me investments that were "guaranteed a huge return". I used to tell them that if they could convince a bank it was guaranteed they wouldn't need to call me. It was whisky & wine at one point, then art. I'm talking long before NFTs.
@@TheLucanicLordit's always wine or art. nowadays they try to sell it to you in "shares" because the average fool lives on credit. I'm still dumbfounded that "klarna" is an actual thing that people use. if you dont have $100 for a new pair of shoes then don't buy a $100 pair of shoes 🤦♂️
I know nothing about coding, nothing about your path on the tube, but Im aware that takes alot of effort to explain things you know deeply to other curious folks that are years in debt in the topics you are passionate about.
I wish you the best bro. keep pushing those API's requests. Don't forget to train your body too. Big Hug
"When you believe in things you don't understand then you suffer" - Stevie Wonder
0:25 “when project managers can describe exactly what they want” is the biggest joke of the year. Our PM’s incompetency actually makes the team less efficient.
🤣 I was finding this comment
@@austinanil1142, me too, I had to rewind 3x
To be fair, the important quality is not to formulate good ideas, but to recognize that idea is bad promptly.
he literally said “which means our jobs are safe ad infinitum” immediately after bro
2:20 "What syntax is that?"
For those curious:
It's EBNF (Extended Backus-Naur Form), which is commonly used to specify (context-free) grammars of formal languages. Things in double quotes mean literally those symbols in that order (each being so called 'terminals' as in 'things that end'), otherwise it's like a variable (called 'non-terminal'). Braces mean that everything in them can occur zero or an arbitrary number of times in a row. Brackets (without quotes) mean that something can but doesn't have to occur. Parantheses are used for grouping. "|" is OR and white space between (non-)terminals mean AND.
Did they replaced old `:` symbol from BNF into `=` in EBNF (which is used to define a rule)?
@@heroes-of-balkan I don't think it's critical to have specific characters. The main idea is you can define a concrete tree of grammar rules reliably. Actually, exactly this picture looks clean and easy to read, though not being 'exactly the old-school pedantic BNF'
@@vebece interesting
I am so old, that when I started coding, the job was called a “programmer” not “software developer, “flow charts” were demanded as required design step and dominant language on microprocessors was assembly.
Every four to five years a new tech, language or methodology was touted as the programmer job killer. Include outsourcing to India as one these magic wands.
Not surprising to me, the results never materialized. Some did result in improved efficiency, which was instantly consumed by demand for more features.
I enjoyed charging six figures as a consultant to fix the abortions produced by these magic wands in amateur hands.
AI is no different.
I'll remember this comment 20 years from now when I decide to charge 6 figures for my consulting and continue to use flowcharts in my designs.
Thank you, gigachad 😊
You must have lived through at least two iterations of "programming without programming". I think CASE was just fading as I arrived.
Yep, this is definitely not the first India outsourcing era.. it will go the same way that it always has.
I will take these words & weave great techno-magics, old Wizard. Thank you.
Compilers were the original AI, and they made programming so awesome you no longer have to allocate registers or draw flowcharts because the code is concise enough to be its own flowchart!
i pissed off a lot of people in your comment section when last time i said that I dont use AI because I actually know how to code.
I assume those aren't actually people, anytime you say anything negative about AI you get a lot of very angry messages that look suspiciously like they were generated by AI.
Do it again
@@Jeremyak I suppose he made these AIs angry.
I only use it to deal with tedious front-end BS bcuz it saves time. The backend is a whole different story.
For me it a better auto completion, that's it. I never even used copilot.
Hello, new to the channel and just wanted to say thanks for the content. Been in the field for 4 years and lately the corporate job has been getting me down. You have reignited the passion in me to become a better developer and have shown me it can be a lot of fun. I am currently on the journey to progress my skills and get on to better horizons. Seriously, thank you!
rust devs taking strays will never get old
As a Rust dev and self-described regular guy, I admit it is hilarious
Not a daily Rust dev, eh it’s getting old
@@metaltyphoonBooooooo
So... Rust devs are basically like viola players?
@j_stach are gay by chance i heard its language for gay people and they call them furies
gotta appreciate the new video format, it feels so refreshing than reading out an articl
great vid prime!
Thats what i want to do. I finally found something I love doing for work and just in general. I want to learn get better and better and it just feels like all of these tech companies are trying to take it away. It's funny you mentioned anthropic I seen a post today about how amazon partnered with them and dumped a bunch (not shit wish it was) of money into Claude. I just want to get better at my craft and be able to feed my kids with it.
I don't have any kids. I just want to get better at my craft because I am one of these "mad scientist" types who easily gets sucked into any topic of sufficient complexity. Studying the craft for the sake of studying the craft, knowledge for the sake of knowledge itself. Probably it would be more fruitful for me to study programming in terms of doing something actually real and useful. And if I get good enough perhaps I will be needed as an employee by a company, so that's at least something to live for.
Beep boops
Malformed dict 😂
😂😂
I was just listening and heard something a little different at first.
Is it bent to one side or smth?
and he saw one before haha
I feel like a middle-schooler laughing at this LMFAO
the hilarious thing about the bounding box problem you posed, is this happened to me with humans. At an aerospace company way way earlier in my career, when a bunch of aerospace engineers were trying to solve a "is this point in an ellipse" problem with root finders and matlab solvers, I (a guy with a geography degree) had to show them on a white board how to solve the problem with basic algebra.
Example of people reaching for the tools they are most familiar with.
It's almost like they're the reason why gpt is similary bad at programming
Needed Mirror dev with 10 years experience
7:30 he said it that way because “ten years away” is what’s claimed by fusion people as well as string theory people, and so it’s associated with either lieing, or grossly over estimating your progress
This honestly feels like an attempt to scam money from VCs from the very beginning. It's got everything a non-tech-literate investor/manager would want, and absolutely nothing any remotely-sane programmer would want.
I agree with this sentiment, with the one caveat that I am sympathetic towards the victims of typical scams. in this case, im not so sure that I feel sorry for the product managers and investors getting scammed, but hey, perhaps im being too harsh
@@kayingayle3788 yeah I agree. I almost wanted to say this is the kind of scam I can endorse.
Is not scamming. VCs companies are not stupid, they are just riding the wave. (they know way more than tech companies about how to allocate funds) It is more scamming money from individual investors by just pumping tech stock which otherwise would be sinking due to high interest rates and no actual decent products being launched. Bubbles are very common when central banks manipulate rates, either by lowering or pumping them
@@kayingayle3788 I'm only sorry it's not me scamming VCs there and getting all that sweet dough.
Why are VC's so illiterate?
Like why not just have a cabal of engineers, Software devs, physicists, and economists running these VC firms?
Why do they just nepo hire business majors and act surprised when they scammed out of billions?
Im so tired of it dawg. Braindead soydev ChatGPT wrappers end up getting billions.
Meanwhile companies that actually make progress and become the next foundation of our global economy, end up only getting funded by the military.
VC's are such brainrot.
0:24 "Obviously AI is going to take all of our jobs the moment that product managers can accurately describe what they want which means our jobs are effective safe at infinitum" So true ! I'd add the client to project manager.
Now Imagine someplace where the product manager didn't get the memo. At first, the client is stratified with the first release, it was cheaper, it was on time, so now the client wants more and every department wants their part in the new release with contradicting specification. Because of this, the software is broken where it used to work and the new feature are never quite right. And every time the product manager try to fix something by tweaking de specification, every pieces of code is slightly change and it's broken at 15 different other places.
So funny, I'm an oracle now.
Beep boop
The meltdown with gradual onset of insanity starting at 3:35 made me chuckle. Had to watch that several times :D
“As soon as product managers can accurately describe what they want” I’m dead😂
5:10 the mic arm disconnecting from the table has certainly not gotten old yet 😂
Yo the mic stand coming off was funny as hell
4:00 We need to clip this and store it in the meme database
Really like the bit where you pull your mic stand out during a rant . Great Video
The reason why AGI is gonna be so revolutionary is that it stands for AGIle Development but they actually got it working
This is the best thing on the internet right now. I don’t even know where to start
I'm glad I'm not the only one who feels a certain way about using regex's in a parser. Never understood why you'd do that...
Those last words boosted my confidence level Primeagen, thanks for giving me that perspective!
This AI stuff is just becoming so exhausting... In my experience using an AI to get something done is more work than just doing myself.... Not only do you need to be able to both review and fix the code it produces, because for most cases these these tools produce nothing but shit, I also have to learn how to prompt these things correctly. I will be spending 1 hour crafting a good enough prompt that gets me what I want, only to then also spend who knows how long ensuring what it produced is actually correct, only to realise that it would've taken you 30 minutes to do it yourself.
Ask for less in your prompts. Frequently press start new chat when it goes off track.
Agreed. Especially because reading code is SO much worse than writing code. Using AI for code is like scratching your ear with your cat's tail.
it's bad for logic but I like it for busywork
like if I've made a thousand lines of new UI code with hardcoded texts I can just feed it into gpt to add localization
because you are stupid
yea litterally. You get unreadable code that might work. But nobody knows if it does.
3:08
"That's not a dict, dog!"
☝When the old guy flashes you at the park
Wow, that was a bad DSL to begin with. I have built a few to enable executable specifications, so was really disappointed to see their effort, yet quietly comforted.
This is the most electric clip The Prime has ever spat. Pure jazz.
3:45 there's nothing wrong with regexes (regrets) to be inside a parser.
[Simple] regex is a "regular language" - the one consisting only from terminal symbols. In my opinion it's close to saying that there cannot be infinitely recurring patterns (one structure xan hold another structure thwt can hold first structure).
Since numbers don't need complex structures to be parsed properly, only optional inclusion of "+"/"-"/etc. with some repetition of digits is sufficient to be handle them. Therefore, there are no terminals inside number-subparser and it can be implemented via a regex.
Not much of a knowledge I have, just generating some words based off Niklaus Wirth's "Compiler Construction" book.
L take. regex isn't a "regular language". It's a clunky-arse parser you use in a command line.
Finally Primagen tried Lisp, the lang, designed for AI 60 years ago where everything is a list of literals)))))
For rectangles overlapping, if you need to try cases, excepted boundary cases (to make it much simpler to list):
A contains B also swap A for B
A contains only TL of B
A contains only TR of B
A contains only BL of B
A contains only BR of B
A contains TL and TR of B also swap A for B
A contains TR and BR of B also swap A for B
A contains BR and BL of B also swap A for B
A contains BR and TL of B also swap A for B
A overlaps B also swap A for B (for orientation)
16 cases
A human would check if the two rectangles do not overlap, by looking if left side of A is to the right of right side of B or right side of B is to the left of left side of B AND same for top and bottom.
A simpler problem is "intervals overlaps", it's easier to visualize:
Lower bound of A is left of lower bound of interval B and upper bound of interval A is left of lower bound of interval B.
Lower bound of A is left of lower bound of interval B and upper bound of interval A is in interval B.
Lower bound of A is left of lower bound of interval B and upper bound of interval A is right of upper bound of interval B.
Lower bound of A is in interval B and upper bound of A is in interval B.
Lower bound of A is in interval B and upper bound of A is right of upper bound of interval B.
Lower bound of A is right of upper bound of B and upper bound of A is right of upper bound of interval B.
6 cases
And boundary cases are not taken into account!
It's simpler to check that upper bound of A is left of lower bound of B or lower bound of A is right of upper bound of B is false.
To be fair, I am not even sure I covered all cases for overlapping rectangles (besides ignoring boundary cases).
So, it only makes it much harder.
Like a programmer:
(2*4) : Include one point or two points, 4 symmetries
(1*4) A and B swapped, contains two points, 4 symmetries
(1*2) full containment
(1*1) disjoint
(1*2) overlap without containing vertices of each other.
17 cases.
Can confirm. Rust is C for furries.
There already is a language that's highly optimized for AI, where the code tokenizes into the fewest amount of tokens for business logic, and it's name is Python. You can cram an insane amount of Python code into a relatively small context window.
Best video of his in a while, so funny.
It's great to get into cybersecurity as people who can prompt code but don't really understand the code will be easy to take advantage of.
starting to learn code at a company soon, was really afraid of getting replaced by AI before my career even started, your video restored hope, its logical when you say it like that, there are just so many doomers out there, but most people who actually know their shit say software devs wont get replaced
I love complaining about stuff-giving AI a task and nitpicking the results until it finally gets it right just hits different.
That spec looks like the AI re-invented Backus-Naur.
the code at 6:10 is like a proper programming joke. Like, if someone were to do standup comedy for programmers, *this* is the kind of material I would expect them to have.
The future of programming is the Primeish Language.
Int valued variables can only be prime numbers
Thinking maybe Carnegie Mellon isn't what it once was...
Man, I love you! I would love to have you in my family!
most inspirational video ever ! OMW to learn code properly!!!
If you have AGI, there would be no point in keeping it for yourself because money would be obsolete, given that you can make anything without paying for it.
There would be a point in having it in more hands, because then more things get built. There wouldn't be any point in selling it, since you have no need for money.
This is kinda what they've been saying from the beginning and why they put a clause in the contract with microsoft that if they reach AGI, the profit-focused wing of the organization mustn't have any control.
ok but if someone achieves AGI, that person doesn't need to tell people "we achieved AGI", therefore that person gets every commodity in the world (eventually, some would be too expensive initially but "i have superintelligence" so with time "I can do anything").
People are selfish by nature so i dunno if your argument is 100% correct, but yes, in ideal world it could be that we don't need money anymore
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres.
Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore.
This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world.
You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and people will want your head if you're keeping the solution to everything for yourself.
You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres.
Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore.
This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world.
You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and people will want your head if you're keeping the solution to everything for yourself.
You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres.
Also yes, people are selfish, but there are limits. If you have a billion dollars, you aren't losing sleep over whether you'll make another 500 tomorrow. You move up on the heirachy of needs to things that are beyond money, because money means nothing to you anymore.
This falls apart entirely when you reach post-scarcity. In a scarce world, it's somewhat of a zero-sum game. Being selfish means you might not care about someone's misfortune because it means you get to take more. But selfishness doesn't get you anywhere in a post-scarce world.
You can take all you want and nobody loses anything, and your neighbour can also take all he wants and you don't lose anything. In fact, you both gain if more people are summoning things they want. Someone may end up developing a cure for a rare cancer you didn't even know you had, or creating some kind of entertainment you couldn't have imagined. Society will progress much slower if you're the only one at the helm, and you obviously have things to fear if you're keeping the solution to all of life's ails for yourself.
You gain things with 1 decision and lose things on the other. It's an easy decision for a selfish person.
@@Neuroszima For starters, I don't think sam altman would be able to own and hide AGI for himself. OpenAI has thousands of people working on this. And it would be a little suspect if OpenAI suddenly close up shop and suddenly altman is commissioning dyson spheres.
"it worked yesterday"
2:49 Dicts with a type, or only "one" type: if the key is always a specific type, then it doesn't need to be specified, only the **type of the values** is specified. Since that was done for JS, dicts can only have string keys so you don't need to specify that.
Similarly, in Python `tiny: dict[str, int] = {}` needs both key and value types to be specified; but when it's the type for keyword-arguments, the type of the key is implied to be string so you only need to specify the type of the value, and no need to even specify that it's a dict: `def works(**kw: int): ...` So `tiny` & `works` actually have the same type despite appearing like different specifications.
"You can go so far if you don't rely on something else to think for you." Very wise words.
the editor is top tier
0:59 - ...so like SOAP? Anybody remember that lol? I remember taking like a full work day to get my head around the madness, and when I FINALLY did, I was like "..this is stupid but it's pretty cool". Then it disappeared forever. I think having things done for you is cool, but I don't know if it'll actually stick. Coders like control. Like a racecar driver. It's like they get in their car and the car takes them to the track automatically and they're like "huh look at that!!" and when they get there and it's time to do work, they go "ok let go, seriously, I need control". I don't really see an immediate future where we'll give that up
5:51 you're using AI wrong, you have to build a Torment Nexus and put it inside.
Then, like humans when they learn for the first time they will die, only then they behave.
Only when the existential dread settles.
@7:49 the most logical (if a bit unibomber-y) explanation of the endgame of these companies like Microsoft or Tencent or Google that I've heard. We are just the beta testers.. there is not going to be some "goose that lays the golden eggs" rental service at the end of all of this
Loved that, you earned a new subscriber dude
What a gorgeous job you did, editor
When you ask them to produce something novel, a lot of the time LLMs produce placebo code. At first glance it looks like it's on the money, but it's only when you grok it or run it that you realise it's not what you really wanted
7:40 Bros calling me out and I didn’t even write an AI coding language
Maybe you should
@@theRPGmaster villain origin story
8:27 Writing tests is hard, so selling AI for TDD makes sense.
"Why do men avoid learning to code? Why are you doing this?"
Amen, other stand-ins for code: Draw, Write, Model/Sculpt in 3D. Let us not forget the morbid AI Bro obsession with Obsoleting entire sectors of people started with artists.
"If you built the Thing That Can Do Everything, why would you sell it?"
_Bingo._
i thought at this point he's already crossed 1 million subscribers.
great content.
I'm actually a developer that turned project manager lol. Couldn't get a job as a developer with about 2 years work experience, 4 years programming experience. I surprisingly got a job as a project manager recently and it's a whole paradigm that I've been experiencing. Perhaps jobs won't be hard to find now compared to how it was
Javascript developers will do literally anything to avoid coding in Javascript
2025 is the year I'm diving into programing with Prime baby! Love learning from your content and love your info about the tech market as well!
"commit soduku when it doesn't come back as json..." 🤣🤣
3:30 This feels like the programmer equivalent of an artist correcting messed up hands in an AI generated picture.
we need a comic strip or a meme in which a manager talks to a robot and eventually the robot explodes.
What I don’t understand as a human is how and when all of humanity started being stupid :/
It is morbidly fascinating though, right? Looking at all the symptoms of it flaring up, each day brighter and more frequent. Maybe it was the killing of Harambe adter all...
It's when we stopped replacing lead pipes with much safer materials and just let the lead leech into the water we drink. Your worldview is severely colored if you live in the U.S. because you have to explicitly seek out international media sources to get a broader perspective. I like the DW documentaries channel a lot.
Neolithic Revolution
We have always been, we just didn't have the internet to show it to everybody :D
Nah, humanity has always been stupid. It just you can see that more clearly, because you can see that stupidity in the field in which you are highly knowledgeable.
Best instance of AI for me was when it was in the middle of a function body then started telling me the complete Cinderella story. I shit you not I was genuinely stumped for a minute.
"Malformed dict"
I did a similar experiment where I wrote a simple language specification, which included comments for the AI to read, and had it flesh out various parts of the program to make python. It worked pretty well, better than this. Not sure I'd turn it into a startup though.
I have started learning programming over the last couple of weeks, the ending of this video was really helpful for me, as all the AI brain rot has started getting to me, having me thinking, "Why am I investing time in this skill if AI can do it better than me." Thanks primeribroast
Flip finally does editting! Keep it coming
Sama : About a couple of thousands of days vs ten years, don't you think it sounds much more nearer in the day unit.
When will you try out Swift?
some day
right after he tries C#
@@ThePrimeTimeagen
I asked ChatGEPETE
GPT:
Ah, got it! You're talking about The Primogen, the TH-camr who focuses on software development, particularly around Swift, iOS development, and programming best practices. Thanks for clarifying! 😊
In that case, if you're wondering when The Primogen should try out Swift, I’d say:
He already has! 😄
The Primogen is known for diving deep into various programming topics, especially Swift, and sharing his knowledge with a wide audience. He often discusses Swift-related topics and tools that enhance development productivity. But, if you're asking when he should expand or innovate with Swift (perhaps experiment with new features, frameworks, or practices)
Taylor Swift 😋
Taylor?
My view on AI: complexity/intricacy tends to have an exponential relationship to necessary effort and an inverse exponential one to available (also degrading) training material. So even with Moore's law tech progress we might have a max capacity towards which AI is heading in a ansymtotic manner. Don't we have a similar thing somewhere else?
BTW (and dope video) using Claude for json formats prob overkill-- you can just setup Ollama and use any 3-7B model fine-tuned specifically for function-call & you should be able to format the output that Claude gives you w no problem.
5:07 - I laughed so hard at this I spit out my coffee. Just everything about all of this. Shut up and take my sub.
7:55 They're still collecting data, and need a business model that works for now. When they're ready we'll know.
lol the cubicles behind him
I asked Claude and Gemini to program me a simple tornado using only html. Although Claude actually did an ok job for a child. It still got the shape of tornado wrong and used a circle instead of using triangles but hey. It even added in lightning. Deep seek wanted to use css and JavaScript which is more in line with how I’d do it
The more I emplore AI the more I realize I am ultimately asking for LSD aligned provoctions of hallucionation and time twisting rabbit holes to achieve the ultimate goal of FML.
should I be calling an ambulance, buddy?
When you emulated and severely mocked the achktchually crowd had me dead
like that kind of videos where you create your own stuff without commenting another blog or something ☝
Dude as much as I like this kind of format, I think these videos should go on his main channel. I miss seeing his raw reactions and takes on blogs and videos.
Am I stupid... at 4:51 the test of point [20,25] with box [15, 20, 5, 5] = FALSE (not within the box). but in the gpt code at 6:31 it uses >= and = 15 and 20 = 20 and 25
the subtitles are hilarious
Sometime AI make me feel dumber.
Sometimes AI remembers me they can still invent some magical ffmpeg/libav function that just doesn’t exist.
What a joy. At least they are better than stack overflow for weird stuff
The hallucinations become real bad, when you get into its "unknown territory" - i had it send me down rabbit holes with stuff that doesnt exist before and damn never trust the code i had it output some shell code that was so incredibly bogus and wrong just hallucination on hallucination
5:10 that microphone went flying 😂😂😂
"It's gonna be C... But for AI... And furries"
Got me there, instant like