AI completion in it's current incarnation reminds me of a joke "banging your head over the concrete wall actually hardens your skull". they could have started with advanced code snippet engine with AI search (where snippet packs are made by humans) or something like that. it wouldn't have had the same hype, but it would probably be much more reliable and reproducible, hence, much more useful.
Yes, please, can we also replace all the parliaments and presidents and CEOs? Even if they might have been useful in the past (which i doubt), they are clearly a liability now. And they are sooo predictable in their rhetoric and decisions, i bet, you can simulate Russian, Chinese and US parliaments and high leaders altogether on a single modern GPU. Ah, also, we could optimize Wall street and traders in general.
I'd be worse if it was AI. AI would only focus on maximum profits. You can argue that the current CEO's the same, sure, but AI will just take it ot hte next level, finding worse exploits. And then, they'll just say "we didn't want this, but the AI calculated it's the best decision" and just use that an excuse to do anything.
This is outsourcing to India all over again. The sky is falling, you are getting replaced, all the jobs are going away! No, I'm going to make fat stacks off of fixing bad code bases again.
India does hurt the job market. But mostly for entry level. Every company thinks they can just hire India or China or whatever for cheap. But you're hiring people who are totally willing to lie to escape a rural village. I had a client where most of their Indian coworkers literally had chickens and cows audible in the background because they were rural farmers by day and this was their second job. It's a whole other world. And the diversity team definitely doesn't understand or care about any issues facing rural farmers who have to speak quietly on a call because they don't want to wake their family who are sleeping.
@@KevinJDildonik- I've heard animal sounds too a couple of times while talking to a "tech-support" in India, chickens and cows mooing in the background. lol Those entry level Indian jobs will be easily taken back by AI with no more farm animals sounds. lol
i've already encountered my first project where the majority of the code is ai generated, that pile of shit was dropped and being rewritten with proper programmers involved because it caused the company a huge production incident and reputational damage... It generates shit, if you dont refine that shit into normal quality code you will end up with unmaintainable shit that gets thrown away.
The fault of this lies with (precisely) the lies being told to people about AI's capabilities. It's like everyone hyping up a videogame you know is going to be shit. Any similitude to Tesla's FSD is pure coincidence
@@Titere05 correct, i believe people drastically overestimate what AI can do, i was surprised and impressed when i was told about the fact that the 80+% of the code was AI generated. Then as i had to work with the code it all made sense... you have to go out of your way to produce such garbage. AI will get there, but right now its very big risk to jump into it just on some fake demos out there.
Today, it produces shit. 5, 10, 20 years from now it won't. AI is going to force us to start talking about UBI and how citizens will exist in the new AI economy.
My experience with AI generated code: 1) AI generated code that didn't work at all. 2) When I provided a working code for the AI to add functionality, the resulting code didn't work at all. 3) Provided a working code to the AI, asked about how the code would behave - AI was wrong. Every AI presentation is carefully crafted to show how awesome it is, so managers (who have little to no tech knowledge) would buy in. It seems like a bubble. My biggest concern is that incompetent people will just copy-paste AI generated code without checking/testing it and it will cost lives. And from what I heard, some companies want to fire junior developers and replace them with AI...
So glad the algorithm threw you in front of me this morning. I’ve been in tech for 27 years. Started as design, then ended up in code because I have a knack for it. I kept thinking ‘aw another 5 years, the kids coming up now will be so much smarter than us, so go hard while it lasts’. Still good advice. The most useful thing I’ve learned in over 2 1/2 decades doing this, is not a particular language, or framework, or method. It’s learning. If you can keep learning, you’ll go so far in tech. Learn your core foundations then figure out how to incorporate the new into it. Rinse, repeat. Loved this video, man, exactly what I needed to see this morning. Before COFFEE no less.
@@NicodemusT even on a non AI worlds this already exists. How many companies have a functional team just to translate the client requirements to the devs. This is the X Y problem that prime talks about, the client think that he needs X but in reality he needs Y.
Some do, some dont. Life is nuance. Yall know there are literally TONS of stakeholders who have requirements AND know enough technical and business logic to be able to tell an LLM exactly what their firm needs and get an AI to do 95% of it? TODAY! Maybe AI wont replace ALL devs but at least 90% will be useless in 5 years time. Prove me wrong.
@@concernedcitizen6572 Doubt that will happen. AI isn't good enough to replace a single real developer today, yet you think they can replace 90% in 5 years. That's wild how you think they will improve like that. AI can't replace me until we reach AGI. And at that point, if we can achieve it, we have bigger issues. I still don't seem the path where AI becomes more than a tool existing developers use to be more productive.
99% of employers don’t know how to implement an AI dev into their dev teams. Being able to solve leet code or code a snake game does not equate to a full stack dev who is required to create and manage database scripts and publish releases and review design specs and code them and fix bugs and create PRs and write documentation and upgrade a library while fixing the breaking changes. We’re at least 10-15 years away from junior devs having to start to sweat.
this is the part no one seems to grasp. Implementation at scale is what separates theory from industry. AI automation need to be borderline unfuck-up-able in order replace teams of devs on massive projects. Cars cant even reliably distinguish the horizon from the ass end of an semi enough to convince us we are safe letting the onboard software take the wheel for good, yet the notion is devin with chat gpt is going to autonomously do CRUD for all fortune 500 companies databases, and sensitive financial data. Yeah show me someone on wall street who says let AI cook up algorithms on its own, no human oversight, implement itself into our trading tools, and lets go home and hope billions of dollors dont vanish. when that happens then maybe we should look into getting a cdl
@@ApexSportsbooks With respect to AI in the workplace, there is more to consider from a political and legal standpoint than the mere capability of AI to do the job. Can AI do specific directed tasks and possibly some of the tasks a developer can do? Yes, probably. Has HR streamlined itself with the ability to adopt policies regarding AI in the workplace and the relationship between AI devs and human devs? Are security protocols in place in order to ensure that these AI tools are secure enough to handle intellectual ideas and sensitive information with regards to proprietary information, PCI data, and HIPAA laws? Have insurance companies rewritten their policies and gained corporate sign off in the event a breach occurs due to AI involvement? Are there laws in place in the event personal data is compromised and used for nefarious activity due to AI involvement (who is responsible)? Are politicians prepared to take a major hit to their campaigns by promising job growth to their constituents while lobbying for the replacement of their jobs by AI at the same time? The list goes on, and the answer to all of these questions is a resounding "No". If anything, as it pertains to developers, I would prefer to see AI integrated into the tools we use, maybe even a "super Clippy" that assists you throughout your workday. This would allow people to do their jobs faster and with an increase to productivity. What used to be done in 80 hours can now easily be done in 40 hours. A direct replacement of humans by AI would erase the last 40 years of advancement in the development space, all the problems we have faced and found solutions for. AI would only introduce an entirely NEW set of problems we are not able to find solutions for so easily. In fact, the risk of AI and the lack of our ability to control it would be a much greater liability than humans could ever be. That is grounds for pause and there are those who have interests in seeing AI not advance as we would see it. It could become something else entirely before we even realize what to do with it on such a small scale, at our level. It would be a waste and somewhat naive to think that the purpose of AI is to develop CRUD applications for corporate advancement. AI is already being developed in ways that may transcend our own dimension. So no, I don't think junior devs have anything to be worried about. It would be like saying a bulldozer might end up replacing my garden shovel.
Not even close to having solid self driving cars, but people are convinced we are going to have complex software built only by AI? I think I will let those people be the first on a Boeing aircraft with purely AI generate software...
1. You dont need 100% AI generated code to have a terrible impact on the developer labor market lol even an AI that replaces simply 20-30% of the current developer labor market would be devastating for millions worldwide. 2. Why would computer-generated code be so scary? Guarantee the vast majority of software currently used in production airline environments was copied from somewhere. Who cares if it was copied by an LLM or a human? Now dont get me wrong likely there will still be human testers needed for the last bit but the vast majority of current developers work will be able to be completed by AIs for next to nothing in 3 years time. How that affects the labor market for devs is anybodys guess.
Same here. I started my first serious job in my last year of college, and I didn't know what to do with the money, I felt like I was swimming in it. Despite my salary being pretty small at the time
Having dependants really flips the script, now every extra dollar earned is by default consumed without a lot of overwatch and being a hardass. Saving money becomes arduous
The saltwater analogy is so perfect. That's exactly right. I think what will happen over long term AI dependency is that your "progress" becomes asymptotic - more and more expenses for smaller and smaller improvements.
Unfortunately I regularly see ads like "Senior full stack developer $60k salary" EDIT: for everyone saying "that's huge in my country" I am referring to the ads that explicitly state "US-residents only"
@@lplsI’ve heard in other countries the environment is still like 2015 . I’ve heard in Chile developers only make 20-30k but due to cost of living it’s a comfortable life . Anyone can get a job knowing html and css . In Norwegian countries it’s the same as well . Same in Germany and Romania . The wages aren’t as high but easily employable with little knowledge. It’s really only America and India where the skill gap is huge
That's about my current rate in Poland exactly as a senior full stack (.net + react)... maybe tiny bit less. And Poland is a pretty popular destination for US startups. But there are things that kind of work in your favor: the timezone and the attitude of US employers. Not sure if I'd ever agree to work in US TZ (would definitely need 6 figures) since I do have a family and cult-like behaviors of various Google-wannabes are a straight up red flag.
Many of the takes not worrying about the industry seem to come from skilled coders in jobs. New coders don’t have the opportunity to get those jobs and become skilled. I live in a fairly decent-sized metro, and there are maybe 4 coding jobs total aren’t for senior roles. And those still want 2-3 years exp. I still apply for them but it’s unlikely I’ll ever get called back. Further, back to the cab driver take, no self-driving and Uber didn’t totally kill the industry, but I work on in loan servicing currently, we had taxi medallion loans. It was one of the hardest things we’ve ever done. It was so bad, we transferred all of those loans within a year, because we were quite literally talking cab drivers off the ledge. They lost millions. Each one of those drivers is a small business, even if they drive for a larger company. Uber and Lyft absolutely destroyed their lives and their futures. Prime is right about the skilled vs unskilled devs, but for those of us farther down the ladder, it really seems like we may not even get a shot to try for a while, and if you’re older, like I am, that’s scary because you don’t have the time to wait.
I have been in the industry for 30 years. AI is a disruptive technology on tier with the web. Back in mid 90's Visual Basic was all the rage then all of sudden the web took off and changed everything. We are in the early days of AI Revolution. Things will change and new opportunities will emerge. Instead of doom and gloom take this time to learn solid engineering skills which don't change because of the hot new technology or JavaScript framework. If you skill set now is focus on Front End Development its time to become a full stack developer. A good book to read from the 90's is the Pragmatic Programmer.
Being a junior dev I really appreciate these kinds of insights coming from a senior . Thank you. Apart from that book, how can one learn those solid engineering skills you mentioned?
@@smnomad9276 Practice. Making errors and learning from them. Asking questions and looking for answers every time you make something. And dont trust blindly everything you see or read, sometimes it takes a lot of time to realize that mantras and gospel might sometimes be utter bullshit.
@@smnomad9276 The best way is to get a good mentor. If that is not possible youtube is full of great videos or design patterns and SOLID Principles. Then review open source code identify patterns other developers are using. Never stop learning.
@@Salantor Explain? How is full stack declining? Every application has 3 tiers at minimum Front End, API and database. As a developer we need to keep learning and become good at building applications. Don't let a framework or library define your skillset.
Devin is vaporware. There said it. High water mark for manual programmers will come a couple years after big non-tech companies have adopted AI to build applications and there is no one who knows how any of them work and no new features can be added. You'll have companies pop up that build the AI tools to figure out your AI codebase and there will be companies that say screw AI and hire developers to rebuild everything.
I just want to throw this out there: automating "other people's jobs" isn't necessarily the downer that people make it out to be. Several years ago, I automated phone-number transfer monitoring because two employees at our telecommunications company were spending their *entire* time doing that -- in no small part because the query websites were *slow* -- and by automating this, they were able to go *back* to their customer support/sales work! Automating jobs means that (sometimes, at least) people are free to do something else that they would rather do instead.
Unskilled is better than 60% of the graph. Juniors around me can't solve problems I was solving in high school. I have to hold hands all day, every day.
There are people who have been on my project for 8 years, who still don't know the absolute basics of the language we are working in. It's absolutely insane.
On second thought, I would say take it easy on them. Some are extremely talented than others. Some of us didn't need any directions right from the beginning of our career and were on par with seniors. I have to often active my empathy when dealing with juniors I have to babysit.
Human developers ask questions, and remodel their solutions on the fly. Robots make assumptions, and charge right off the edge of the nearest cliff (Note - some mf’s are actually robots too)
One could take some small comfort in the idea that they thought the machinegun would be the end of the riffleman because it was 400 times more powerful. Then ww1 came and they just killed 400 times more soldiers and not a single rifleman was replaced.
I mean for me as a young person, I don't care about whether or not I'm making 60k or 100k as a software developer. I just care that I get a job after getting through school. I try to push myself and try new things in the development world and I try to develop very deep knowledge, but at the end of the day that won't matter if I can't get hired in the first place and end up working a completely different job.
Yeah, there are lots of senior level roles out there. Some are not at senior level pay. Very few junior roles and some mid level roles. I've had multiple recruiters tell me they get 100s of applications for each job posting.
That's kinda the problem, everyone wants seniors because they created this cycle of hiring new people, not raising their salary in step with their performance and abilities, people jumping ship, rinse and repeat. Now they don't want to invest into new and "junior" people as much and just want the people who already know what they're doing, except those people are jaded, are shrinking by the day and subsequently there won't be a new generation on that level to replace them at this rate. There's plenty of work out there, just not for the demographic of developers that are most affected by this shift in the IT world, not to mention that even seniors are getting shafted by companies even when they meet and exceed qualifications.
That's because you cant get senior engineers if you never hire junior engineers. If you really think the job market right now is good, youre smoking some really good stuff.
I have zero worry that developers are going to be replaced by LLM's. I tried using one of these LLM's to handle coding; and frankly the think is clueless about how to properly layout a large program. Its good at following a pattern though. That's about where it ends. Put it on a medium to large size complex task and it falls apart.
Yep good with small snippets but anything above that forget it. And dealing with large code bases with complex interconnections seems so far away as nearly impossible at least for another 10 years, maybe more.
Absolutely nailed it. It takes an absurdly long time to become "just" well rounded in software development. Not to mention the constant learning one needs to do. Very few careers consist of that level of "staying on it".
As someone who built a fairly popular jQuery carousel in 2009 (coda slider / liquid slider) I feel we'll just level up with all the upcoming pressure, so long as we just keep at it.
The question “when will we have enough programs written?” is like asking “when will we be done creating laws?” Or “when will be done writing stories?” I think software development is more like this than farming. 🤷♀️
Good video. Optimism is always best - as long as it's not paired with ignoring events as they happen. But basing life decisions on hype trajectories is not wise. Solid developer skills will translate into whatever the future brings. Deep technical specialization is always risky. But being a developer who can turn requirements into actual solutions is a more fundamental skill simply coding - even if coding is for many the main tool used to accomplish the transformation. The same goes for troubleshooting, debugging, enterprise IT, keep critical live systems running, etc. If we get to a point where AI can take over all of that work without any need for being directed by or pairing up with humans (and be cost-effective compared to humans) - then we're heading for a utopia because that translates all the way down - and enables frictionless automation of pretty much anything at a trivial cost. So don't worry about that. As for the more realistic trajectory where AI becomes able to churn out a lot of code, but doesn't have agentic independence at cost-efficiency - I am not so sure the majority of work will be in maintaining the terrible code bases which will result from enthusiastic and/or reckless adoption of such AI. Sure, there will be some clean-up jobs and some horrible systems to keep afloat until they can get ditched for something more sane - but I think it's much more likely developers will be needed to pair up with AI. So, I do think developers might need to get used to the job changing. Get comfortable working with AI as they are now - their limitations and their strength. Because I don't there will keep being a choice. And they will get more agentic in nature. We might have AI 'colleagues' that queue up requests for assistance, work to be done, questions, etc. We might have AI watching as we work - a virtual pair programmer. For some this will just be a natural extension of how they already work - for others it might be a huge negative, to no longer be "alone" for the majority work. Someone who wants to become a dev, but hates working with AI and hates working with other people - might find it the coming years tough. That kind of person could find a great niche until now. But I think there is hope even for the most anti-social - as AI can be tailored to the individual.
No not at all LMAO! Some jobs will be fairly safe from this. 99% of people will want a human therapist. No one that makes decisions cares if when their value-added product goes down its fixed by an LLM or a human 😂
im a software enginner student, my father is an accomplished programmer, and the way i see it rn, mind you i havnet done any interniship yet, is that it seems to be a "easy" to get in area but hard to stick around, due to the constant change in tecnology and the insane skill level required, thats why i always try to not use libraries and build my own stuff whenver possible, and then theres the ML market right now, which i relly dont wanna go to, but different from web development i feel like i might have to T^T.
In a job, they won't let you write your own libraries if an open source option exists. Because it is seen as a waste of time. Learning about API of some of these libraries is exhausting when sometimes I feel like I can make that library or feature on my own. But soon you will understand that the amount of development that went into that library is so much that they have had better programmers than you and me fix many edge cases that you don't even realise exist at this point in your own implementation. That's the reason you're not allowed to spend a lot of time writing your own packages in production environment. Learn to pick the right package for the job and make a habit of reading docs.
@@Adam-nw1vy yes, but I don't have much experience with templates. So can't say. If we are talking about building UI yourself, then there are a lot of UI libraries like chakra ui, material etc. lot of things are doable. If something isn't, then you would have to make it custom. I recently had to make a multi month calendar with various accompanying tool components this way. Because the most similar one I could find that could do calendar lacked responsiveness.
@@Steel0079reading the docs is a given whatever, and i understand that places have production standarts, makes sense streamlining production, but i still think that for practice and learning its good to know HOW something would work from the ground up.
Nobody knows the future but the fact that so many people are having so much anxiety in the tech industry right now is clearly a symptom that we’re on the precipice of a big change that nobody knows the implications of yet.
@@Cat-vs7rc maybe but even at my job, I’ve seen a few people mostly juniors and some seniors with at least a decade of experience who don’t know how things work, they know how to write code but don’t know why they’ll use some functions instead of others or sometimes the code is very low quality or the algorithms are not good. I’m very confused by it sometimes. It feels like they just memorized things and repeat it everyday not willing to dig deeper to understand how things work. And we work with Indians too who are based in India for everything related to infrastructure, I’m just gonna say if you have a problem it’s better to find the solution yourself
@@imhassaneOne way you could classify professionals is by the ratio of interest/disinterest in the nuances of their jobs. Surprise, many devs simply don't care that much. They just want to do their hours and go home, doing the minimum required. And this applies to most jobs.
there are multiple factors in a corporate work. one might say "its not what you know, its who you know" a nepotism and other factors as well such as communication skills. and interviews are a flawed by either way its the way it is
I will give a little advice here... Careful with any forecast. War is not about the best weapons, business is not about having the best product. Coding will not be about AI
Here is what I think is gonna happen: Industry hits itself in the face because of AI. AI first tries to code and push work on github and other repositories, making the processed work unsuitable to use as training dataset with a tinge of error/hallucination when asked something complex or improper prompts. More AI models propagate this hallunication bit by bit (if content is not vetted properly) and industry comes to a screeching halt after this recession because now you need to hire back the devs you fired as AI just doesn't cut it now. Let's say this doesn't happen at all. Then when a super tiny bugs comes in, your systems burns because AI cannot find it. Now what? Will company have 5 AI models and a more than a few senior or principal engineers that debug the hot garbage AI wrote? most of them will keep jumping ship until they are paid top dollar to debug something like that. Think people.
What bug would AI not be able solve that only a human could and why? Tell me ill wait. They both literally have access to the same information, the same docs, the same exception lists to process. One is expontentially faster and cheaper than the other. And it obviously doesnt care about the consequences of that truth.
@@concernedcitizen6572I used chatgpt and it could not solve a bug in GDScript so yea buddy it can create made up functions libraries. I think devs will be paid way less and our job will be to fix those bugs if we accept it. Same thing happen when I tried to get the AI to write assembly. The AI does not think its trained on DataSets. If it runs into a unknown problem it does the best it can. Humans can stop, actually think and solve the issue. Once the issue is solved we can feed that solution into the dataset. Ai improves based on human expriences, the program which is really a Large Languange Model still does not think. It is no different then haviny a hugr dictionary with all the problems solved at your finger tips. Imagine if a person made a program with all known solutions and solved programming patterens and gave it to a Software Engineer, this solution would actually be better then this so called AI. Instead they gatekeep the industry with Leetcode type question and brag that an AI with all the current solution feed to the dataset passed but an entry level dev without google to be his dictionary failed. If you want to make it fair, give that entry level dev the same access to look things up. In reality they are both similiar but the AI will save the company money unless its makes up functions.
@@concernedcitizen6572 yes obviously it doesn’t care, because AI doesn’t think. It’s not true intelligence. AI just very complex auto complete. It will do what it can with the data you give it, but if the data is bad, it won’t know the difference. Worse, AI has a major problem with hallucinating information.
@@concernedcitizen6572bro the human mind is complicated. We can remember shit that happened years ago. Ai can't do that. Code base can get huge. And ai is very limited compared to the human brain. In software development everything is tied together and when you make a change to one thing it can affect hundreds of other things. If you have hundreds of thousands of lines of codes the ai won't be able to keep all of that in its context as it is very limited. When it comes time to solve complicated issues like that only humans can do that. Ai is good at generating small snippets of codes and that's it. Unless they can completely simulate an entire human brain. I don't see ai replacing humans.
The graph in the end: I highly doubt AI gonna try to automate the less demanded easy tasks first. They gonna do the most demanded easy task. Then less demanded ones, then maybe they try to go higher in difficulty. But I agree, we're still so far from automating programmer job, not even near to that, Devin isn't enough.
it's impossible for you to say that. It seems a lot of these comments are saying what they want to hear. If you think that AI isnt going to wipe out at least 80% of developer jobs you are crazy. AI is not perfect now but can do a lot right now. Imagine AI in 5 years time. It will be able to do the job of 500 developers in a day. A lot of developer jobs are going and we need to accept it.
LLMs can do 70%+ of the tasks that developers currently do faster than the vast majority of devs. TODAY. I am not sure if this will lead to layoffs but within 1-3 years time there will be AIs that can do almost all of the tasks that we currently associate with development work. Maybe what we think of as the typical tasks of "developers" will simply change to accomodate AI. But the truth is that slowly but surely orgs will realize there is next to no value paying people $50-100 an hour to stare at a screen and try to figure out problems AI can figure out for next to nothing much quicker.
You talking about how hard carrousels used to be reminded me of the time I built a treeview taking the results of an Oracle Hierarchical Query when all we had was Vanilla Javascript and early JQuery
There's way WAY too many insanely overpaid developers who get paid to sit on their butt and Google stuff. This will be the huge market correction that fixes that issue. Skilled labor will have its comeback, as Prime said, and I'm all here for it.
@@YTDeletes90PercentOfMyCommentsThis is so inaccurate LOL. i can assure you only good devs are good at googling. I'd say only 10% of corporate programmers are professional googlers, the rest try to get by, take weeks to find the answer, or ask the good dev for help (not hating on the asking for help, mind you)
As a senior web dev what I see right now is that as a frontend dev you get more and as a backend dev your salary goes down (at least here in stockholm for senior roles)
There are programmers who only copy-paste simple javascript/python code, have no idea how computers actually work and don't know what a compiler is. Such programmers rarely even have to think. Most if not ALL of the thinking has already been done for them and is available on StackOverflow. Then there are programmers whose field of expertise simply REQUIRES that they know how computers work. Like operating systems developers, compiler developers, embedded developers, optimized game engine developers, among other specialties. Only the former group that I mentioned will be replaced by this trash-generating "AI". Because the code and systems the human programmers in that group generate IS ALREADY trash.
Second group is at risk too. Who can understand hlw computers work better than computers themselves. (Note here I dont think computers is exactly the right terminology here just using your term)
@@concernedcitizen6572 This comment perfectly encapsulates how little you truly understand. No, a model that does next best token prediction for a given natural language context window (aka an LLM) does not understand how a computer works "because it is a computer". That you think an LLM is performing any reasoning whatsoever displays more about your knowledge than anything else.
@@krityaan lol I know AI does not "reason" calm down lmao. OS devs, embedded systems devs most devs in general do tasks to add value to a business. ANY type of dev that is writing code and implementing development concelts that alreasy exist (aka 99% of devs) are at risk since AI can and soon will be able to build and troubleshoot and add value to businesses quicker than the vast majority of devs. Aint nobody special.
Fuck tht dude who said those things on twitter. I have learned alot from u over past 6 months. And your yelling and jokes only makes things easier to remember. Your a teacher we all needed. Thts so wrong tht people bring the meth thing up to put you down. The life experince you have shared are much more valuable thn a non meth user. Infact sir, how u fought of your addiction is much more inspiring. Thank you for being so true, honest and more power to u
The issue is that a lot of programmers today are block-box programmers; they have no idea what their code is doing... since it's not really their code. And since their code is largely what feeds (re)Generative AI, the situation will only worsen over time.
I hate to be that linguistics guy but profane’s generally understood meaning today is as a synonym to vulgar (see: profanity). Coming from the Latin “pro”, meaning “before” or “outside” as understood by the Romans, and “fanum”, a temple or sanctuary. Profanus literally meant “unholy”, but later was used colloquially to describe things outside of a metaphorical ideology rather than the literal temples, morphing profane into its current meaning as “disrespectful”. The idea of it “making common” is actually its newer meaning, and continues to be its less-used definition
I like the emphasis on how difficult it is to become a really effective programmer. You can learn how to crank out some baseline code in a matter of months - in some ways, coding is just taking the thought process in your brain and giving those instructions to a computer. But to do it in a way that the code exists at the right level of complexity and readability and maintainability is much, much harder and reliant on hard-earned experience.
The frontend market situation sounds bizzare. On the backend, companies literally stopped hiring experience people and they just hire the top 0.1% from the "I need a job, or I will starve" crowd.
What a weird way for these employers to gate keep. Can't there be a set of standards that isn't just "muh arbitrary interview guidelines that include asking someone how you get out of a blender"?
Human laziness is infinite, so the need for automation is infinite. The only thing that changes is the current Zeitgeist of what people pay for getting automated. And as a programmer, you sit right in the middle of making automation happen. The more AI is rolled out, the cheaper automation becomes, the higher demand becomes for automation experts because more and more automation needs to be managed. This is the rebound effect and it is a well researched economic effect. So, don't worry. Yes, FAANG massively over hired, but everyone else is madly scrambling to keep up with automating their businesses. Don't be complacent, learn to adapt, and don't worry. This is just the beginning of our golden age as programmers.
This statement is NOT true - "And right now, except for a tiny group of gilded AI wizards, most programmers have either seen their prospects stalled or become more precarious." Actually, there are many "non-AI specialist" programmers who are still extremely sought and have never even started worrying about all this new trash-generating "AI" stuff. Of course I'm talking about low-level developers who ACTUALLY know how real computers work. OS developers, compiler developers, embedded developers, game engine developers, etc.
this.. I'm 30+ years in the trade and have never been out of a job for more than 3 weeks. I get offers weekly on Linkedin. There are jobs outside of FAANG which everybody seems to jizz about for some reason
In my country there's too few React devs, so they're getting paid crazy amounts of money. Imagine that. Now try to imagine an AI writing maintainable React code. Now laugh.
Look Devin makes English a very, very high level programming language. Natural programming languages can get you of to a good start but they're shit for engineering because they are not explicitly accurate - you spend more time writing a desciption/instruction than you do just writing the code. It has been tried before - the issue wasn't successful natural language implementation but the verbose and imprecise nature. Its like 3D movies all over again.
Started off as a pabx technician in 1991, I learned turbo Pascal, then turbo C then turbo C++. I stuck with Borland and now embarcadero products. I don't do Microsoft products other than the operating system
Ten years ago it was autonomous vehicle will take over and make drivers obsolete. Five years later it was the blockchain will take over every transaction on the internet, making banks, currencies and payment systems obsolete. Now it’s A.I. will make knowledge/creative workers obsolete. Give me a break. I do see value in Chat GPT, and I understand how it will get smarter over time. But I see it as a companion to workers, not a replacement. Just how our “autonomous” systems became autopilot/super cruise. The day I see a client understanding their requirements and explicitly stating what they want from a software product, that’s the day I’ll worry.
@@trappedcat3615Yes exactly just like no organizations entrust their sensitive info and business logic with external tooling like github and copilot. ;)
WRONG. from 1990 to present pay went up 50% while cost of living in most cases double to tripled. If you need I can send you the data from cost of average home then and now or you can look it up the same as I did. I looked up pay from then to till now, cost of food, housing, rent, school, ... Average income went from 50K to 75K prospectively where homes went from 140K to 300K, college doubled.average gal gas 1990 1.15, average rent $447 today $1700.
What's going to happen is there will be very few high level developers that will manage the AI systems. Basically, the developers will clean up or adjust certain areas that AI hasn't got exactly right. It's time to accept that AI will wipe out 80% of developer jobs in around 5 years. I know everyone is saying stuff to comfort themselves into thinking AI will not affect them, but the reality is AI is going to replace a lot of developers. Anyone who thinks AI ain't coming for their jobs is living in dream world. Tech lead warned us about this, but many laughed and called him a hater, and someone who is working for the companies to discourage everyone. So many companies are now using AI and turning their backs on human developers. Tech lead was right. Development and programming will be completely dead in 10 years for 90% of developers.
I do believe that AI will take major of dev jobs at some point, but I think that before this happens there are other types of the job in queue like a manager.
@@vViktor1 Manager will be safer. Managers TYPICALLY know business needs better than their developers do. Once managers have a tool that they can communicate with effectively and it can ship their value-added features and fix bugs faster than their average developer then developer job becomes less and less needed. Managers and Product Owners still needed to come up with and prioritize business requirements
well, in 2008 we already had JavaScripts libraries for carousel, like dojo toolkit, older than jquery for 1 year. And yes, dojo toolkit had a carousel ready to use, and other libraries that have have fade into obscurity, before the age of famous libraries in js, there was a lot of websites dedicated to dhtml with ready available widgets to copy and use. It wasn't that dry dude. In 2008 I was working in a bank using Ext.js a UI library for the corporate world, all components with a default windows looking style with blue tones, they even had a WYSIWYG UI editor, I don't know if they still have it.
this will be a good thing for programmers all your side projects and ideas can now become a real thing with help of ai starting a startup would be easier now
9:00 is the key. When I graduated everyone wanted to be on finance or business administration, engineering was just an excuse to prove you were worth hiring. 5-10y later there was a reduced pool of real talent, and you could tell just by looking at your yoy salary growth. Over the last 2-3 years there’s been a LOT of people trying to become devs (particularly web), with vague promises of getting good salaries with 6 month courses. Now the pool is overflowing with people with poor skills, so if you can’t prove your talent quickly, you won’t get hired. Is that simple.
If the rate of progress continues accelerating as it has in the past few years, you are 100% wrong. If it slows down, you are 100% right. The first signs of computers understanding sentiment was 2016 and in 2022 chat GPT came out - now AI is ubiquitous. We are now in an early stage of the curve (AI sophistication/time) developing - we dont know what kind of curve it is... It could be a logistic/sigmoid curve or an exponential curve - If the rate of improvement of AI models continues on the exponential, we are fucked. I would say its a 50/50, the only thing i can say for sure is that nobody knows the the future and therefore nobody knows the answer, so we are stuck guessing...
My observation is that the curve is asymptotic. 100% coverage becomes infinitely harder as the limit is approached. The limit being full human capability.
@@vripiatbuzoi9188 Yes, it is asymptotic to any finite set of goals - by definition, you cannot really improve once the problems in the set are "solved". But when Im refering to "sophistication", I mean something like "the sum of results across all conceivable problems" - which is unfortunately not a quantifiable quantity. Also, I dont see a reason to a-priory assume that human capability is the limit - which is also a very vage limit since we dont really know what the limits of humans are. There are many examples of narrow NN-based AI models surpassing humans - there might be a way to do similar thing to general AI (Probably not for LLMs of today, but it might be possible for future architectures).
For income, we "feel" the changes up or down, but we also "feel" differences between us and others around us. So if my pay is good but I can see lots of people who earn 50% more, I'm gonna feel poor. This can cause *serious* social unrest.
I am a 10x engineer it takes 10 engineers to come in and correct all the mistakes and production outages I cause
We should work together. Together, we could 10x that to 100! With a few more of us, we could save ALL the jobs!
Sounds like a team where I can finally fit in 😅
Guess i am not the only one 😂
Are you my alter ego?
Isn't that -10x instead?
Guaranteed in maybe 2 or 5 years we'll see articles about "Why we stopped using AI to develop our codebases"
wanna bet?
2years yes, but by 5 it’ll have cycled again.
@@Kane0123🤣🤣🤣🤣🤣🤣
Or even “Why we never ended up using AI to develop our codebases”
AI completion in it's current incarnation reminds me of a joke "banging your head over the concrete wall actually hardens your skull".
they could have started with advanced code snippet engine with AI search (where snippet packs are made by humans) or something like that. it wouldn't have had the same hype, but it would probably be much more reliable and reproducible, hence, much more useful.
Developers are edging🗣
Improves coding abilities
Ha any really 10xer is a student of Royce Du Pont , and most likely is already watching porn on the job while edging 😮
Hell yeah 😂
😫😩💦
Group edging starts in T-5 minutes
I am waiting to see Nvidia CEO replaced by AI.
Yes, please, can we also replace all the parliaments and presidents and CEOs? Even if they might have been useful in the past (which i doubt), they are clearly a liability now. And they are sooo predictable in their rhetoric and decisions, i bet, you can simulate Russian, Chinese and US parliaments and high leaders altogether on a single modern GPU. Ah, also, we could optimize Wall street and traders in general.
We're just waiting for more trainin data 😜
I got my mojo crunching the numbers 🙃
I suspect that using AI to replace the C suite and marketing makes more sense than using it to replace labor.
I'd be worse if it was AI. AI would only focus on maximum profits. You can argue that the current CEO's the same, sure, but AI will just take it ot hte next level, finding worse exploits. And then, they'll just say "we didn't want this, but the AI calculated it's the best decision" and just use that an excuse to do anything.
This is outsourcing to India all over again. The sky is falling, you are getting replaced, all the jobs are going away! No, I'm going to make fat stacks off of fixing bad code bases again.
The real answer right here.
AI will probably take a lot of the mundane programming jobs from India this time.... think about that.
India does hurt the job market. But mostly for entry level. Every company thinks they can just hire India or China or whatever for cheap. But you're hiring people who are totally willing to lie to escape a rural village. I had a client where most of their Indian coworkers literally had chickens and cows audible in the background because they were rural farmers by day and this was their second job. It's a whole other world. And the diversity team definitely doesn't understand or care about any issues facing rural farmers who have to speak quietly on a call because they don't want to wake their family who are sleeping.
@@KevinJDildonik- I've heard animal sounds too a couple of times while talking to a "tech-support" in India, chickens and cows mooing in the background. lol
Those entry level Indian jobs will be easily taken back by AI with no more farm animals sounds. lol
i've already encountered my first project where the majority of the code is ai generated,
that pile of shit was dropped and being rewritten with proper programmers involved because it caused the company a huge production incident and reputational damage...
It generates shit, if you dont refine that shit into normal quality code you will end up with unmaintainable shit that gets thrown away.
Thank you!
I feel like the only people who claim that AI is ready to replace devs have never done any software project in their lives.
The fault of this lies with (precisely) the lies being told to people about AI's capabilities. It's like everyone hyping up a videogame you know is going to be shit. Any similitude to Tesla's FSD is pure coincidence
@@Titere05 correct, i believe people drastically overestimate what AI can do, i was surprised and impressed when i was told about the fact that the 80+% of the code was AI generated. Then as i had to work with the code it all made sense... you have to go out of your way to produce such garbage.
AI will get there, but right now its very big risk to jump into it just on some fake demos out there.
Today, it produces shit. 5, 10, 20 years from now it won't. AI is going to force us to start talking about UBI and how citizens will exist in the new AI economy.
My experience with AI generated code:
1) AI generated code that didn't work at all.
2) When I provided a working code for the AI to add functionality, the resulting code didn't work at all.
3) Provided a working code to the AI, asked about how the code would behave - AI was wrong.
Every AI presentation is carefully crafted to show how awesome it is, so managers (who have little to no tech knowledge) would buy in. It seems like a bubble.
My biggest concern is that incompetent people will just copy-paste AI generated code without checking/testing it and it will cost lives. And from what I heard, some companies want to fire junior developers and replace them with AI...
Developers aren't just edging, they're also gooning
What do all those words even mean lol
if you aren't gooning you consider ropemaxxing @@liquidsnake6879
@@liquidsnake6879 its all alpha generation speak (basically kids). Its sexual practices
Developers are gooning like it's NNN all year around baby!!!
@@TheManinBlack9054 No no, sigma gen, please.
So glad the algorithm threw you in front of me this morning.
I’ve been in tech for 27 years. Started as design, then ended up in code because I have a knack for it.
I kept thinking ‘aw another 5 years, the kids coming up now will be so much smarter than us, so go hard while it lasts’.
Still good advice.
The most useful thing I’ve learned in over 2 1/2 decades doing this, is not a particular language, or framework, or method.
It’s learning.
If you can keep learning, you’ll go so far in tech.
Learn your core foundations then figure out how to incorporate the new into it. Rinse, repeat.
Loved this video, man, exactly what I needed to see this morning.
Before COFFEE no less.
We live in the golden era. It's not too late yet to make your own JavaScript framework.
jajajjaja
Don’t encourage them lol
Devin will only succeed when clients are able to precisely describe the requirements of theirs project. They don't have a clue about what they want.
For now.
@@NicodemusT even on a non AI worlds this already exists. How many companies have a functional team just to translate the client requirements to the devs. This is the X Y problem that prime talks about, the client think that he needs X but in reality he needs Y.
Some do, some dont. Life is nuance. Yall know there are literally TONS of stakeholders who have requirements AND know enough technical and business logic to be able to tell an LLM exactly what their firm needs and get an AI to do 95% of it? TODAY! Maybe AI wont replace ALL devs but at least 90% will be useless in 5 years time. Prove me wrong.
@@concernedcitizen6572 >at least 90% will be useless in 5 years time. Prove me wrong.
I will go back to that comment in 2029
@@concernedcitizen6572 Doubt that will happen. AI isn't good enough to replace a single real developer today, yet you think they can replace 90% in 5 years. That's wild how you think they will improve like that.
AI can't replace me until we reach AGI. And at that point, if we can achieve it, we have bigger issues.
I still don't seem the path where AI becomes more than a tool existing developers use to be more productive.
99% of employers don’t know how to implement an AI dev into their dev teams. Being able to solve leet code or code a snake game does not equate to a full stack dev who is required to create and manage database scripts and publish releases and review design specs and code them and fix bugs and create PRs and write documentation and upgrade a library while fixing the breaking changes. We’re at least 10-15 years away from junior devs having to start to sweat.
this is the part no one seems to grasp. Implementation at scale is what separates theory from industry. AI automation need to be borderline unfuck-up-able in order replace teams of devs on massive projects. Cars cant even reliably distinguish the horizon from the ass end of an semi enough to convince us we are safe letting the onboard software take the wheel for good, yet the notion is devin with chat gpt is going to autonomously do CRUD for all fortune 500 companies databases, and sensitive financial data. Yeah show me someone on wall street who says let AI cook up algorithms on its own, no human oversight, implement itself into our trading tools, and lets go home and hope billions of dollors dont vanish. when that happens then maybe we should look into getting a cdl
I think 10-15 years is immense over estimation
Maybe 2-3 years till Junior debs are replaced
@@ApexSportsbooks With respect to AI in the workplace, there is more to consider from a political and legal standpoint than the mere capability of AI to do the job. Can AI do specific directed tasks and possibly some of the tasks a developer can do? Yes, probably. Has HR streamlined itself with the ability to adopt policies regarding AI in the workplace and the relationship between AI devs and human devs? Are security protocols in place in order to ensure that these AI tools are secure enough to handle intellectual ideas and sensitive information with regards to proprietary information, PCI data, and HIPAA laws? Have insurance companies rewritten their policies and gained corporate sign off in the event a breach occurs due to AI involvement? Are there laws in place in the event personal data is compromised and used for nefarious activity due to AI involvement (who is responsible)? Are politicians prepared to take a major hit to their campaigns by promising job growth to their constituents while lobbying for the replacement of their jobs by AI at the same time? The list goes on, and the answer to all of these questions is a resounding "No".
If anything, as it pertains to developers, I would prefer to see AI integrated into the tools we use, maybe even a "super Clippy" that assists you throughout your workday. This would allow people to do their jobs faster and with an increase to productivity. What used to be done in 80 hours can now easily be done in 40 hours. A direct replacement of humans by AI would erase the last 40 years of advancement in the development space, all the problems we have faced and found solutions for. AI would only introduce an entirely NEW set of problems we are not able to find solutions for so easily. In fact, the risk of AI and the lack of our ability to control it would be a much greater liability than humans could ever be. That is grounds for pause and there are those who have interests in seeing AI not advance as we would see it. It could become something else entirely before we even realize what to do with it on such a small scale, at our level. It would be a waste and somewhat naive to think that the purpose of AI is to develop CRUD applications for corporate advancement. AI is already being developed in ways that may transcend our own dimension.
So no, I don't think junior devs have anything to be worried about. It would be like saying a bulldozer might end up replacing my garden shovel.
@@ApexSportsbooksNo. 10-15 years is the absolute worst case. I predict that demand for software experts will take off exponentially.
Not even close to having solid self driving cars, but people are convinced we are going to have complex software built only by AI?
I think I will let those people be the first on a Boeing aircraft with purely AI generate software...
Yes it's ridiculous to even think that.
1. You dont need 100% AI generated code to have a terrible impact on the developer labor market lol even an AI that replaces simply 20-30% of the current developer labor market would be devastating for millions worldwide.
2. Why would computer-generated code be so scary? Guarantee the vast majority of software currently used in production airline environments was copied from somewhere. Who cares if it was copied by an LLM or a human? Now dont get me wrong likely there will still be human testers needed for the last bit but the vast majority of current developers work will be able to be completed by AIs for next to nothing in 3 years time. How that affects the labor market for devs is anybodys guess.
You should check out some videos of the FSD12 beta 😃
Bro get up to speed. The coding AI will make good AI for cars
@@christophkogler6220You mean official propaganda from Tesla? I'm sure it's super accurate and realistic, not at all concerned with stakeholders
Richest I ever felt was first job out of college in 2011. Still on the college budget with an actual salary was amazing
Same here. I started my first serious job in my last year of college, and I didn't know what to do with the money, I felt like I was swimming in it. Despite my salary being pretty small at the time
Having dependants really flips the script, now every extra dollar earned is by default consumed without a lot of overwatch and being a hardass. Saving money becomes arduous
And then you got a wahman and now you do not have money left for anything. A (dumb) story as old as time. :)
Same
Went from $15 an hour to $100k a year with signing bonuses and profit shares. Wild pay straight out of school.
At least it used to be.
The saltwater analogy is so perfect. That's exactly right. I think what will happen over long term AI dependency is that your "progress" becomes asymptotic - more and more expenses for smaller and smaller improvements.
Just look at any futurist predictions from 50-100 years ago, and you'll see a lot of technologies that plateau'd hard.
Unfortunately I regularly see ads like "Senior full stack developer $60k salary"
EDIT: for everyone saying "that's huge in my country" I am referring to the ads that explicitly state "US-residents only"
😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆😆
If it's remote, it's because 60k/yr on some parts of the world is a heck of a salary.
@@lpls It's lifetime savings in many places.
@@lplsI’ve heard in other countries the environment is still like 2015 . I’ve heard in Chile developers only make 20-30k but due to cost of living it’s a comfortable life . Anyone can get a job knowing html and css . In Norwegian countries it’s the same as well . Same in Germany and Romania . The wages aren’t as high but easily employable with little knowledge. It’s really only America and India where the skill gap is huge
That's about my current rate in Poland exactly as a senior full stack (.net + react)... maybe tiny bit less.
And Poland is a pretty popular destination for US startups.
But there are things that kind of work in your favor: the timezone and the attitude of US employers.
Not sure if I'd ever agree to work in US TZ (would definitely need 6 figures) since I do have a family and cult-like behaviors of various Google-wannabes are a straight up red flag.
I'm on firefox
A browser that can't Edge® 😢
And?
@@plaintext7288isnt that a good thing? lol
They are a developer that's not on edge...
Underrated by the TH-cam algo
Many of the takes not worrying about the industry seem to come from skilled coders in jobs. New coders don’t have the opportunity to get those jobs and become skilled. I live in a fairly decent-sized metro, and there are maybe 4 coding jobs total aren’t for senior roles. And those still want 2-3 years exp. I still apply for them but it’s unlikely I’ll ever get called back.
Further, back to the cab driver take, no self-driving and Uber didn’t totally kill the industry, but I work on in loan servicing currently, we had taxi medallion loans. It was one of the hardest things we’ve ever done. It was so bad, we transferred all of those loans within a year, because we were quite literally talking cab drivers off the ledge. They lost millions. Each one of those drivers is a small business, even if they drive for a larger company. Uber and Lyft absolutely destroyed their lives and their futures.
Prime is right about the skilled vs unskilled devs, but for those of us farther down the ladder, it really seems like we may not even get a shot to try for a while, and if you’re older, like I am, that’s scary because you don’t have the time to wait.
I have been in the industry for 30 years. AI is a disruptive technology on tier with the web. Back in mid 90's Visual Basic was all the rage then all of sudden the web took off and changed everything. We are in the early days of AI Revolution. Things will change and new opportunities will emerge. Instead of doom and gloom take this time to learn solid engineering skills which don't change because of the hot new technology or JavaScript framework. If you skill set now is focus on Front End Development its time to become a full stack developer. A good book to read from the 90's is the Pragmatic Programmer.
Dunno about that, full stack seems to be on the decline.
Being a junior dev I really appreciate these kinds of insights coming from a senior . Thank you. Apart from that book, how can one learn those solid engineering skills you mentioned?
@@smnomad9276 Practice. Making errors and learning from them. Asking questions and looking for answers every time you make something. And dont trust blindly everything you see or read, sometimes it takes a lot of time to realize that mantras and gospel might sometimes be utter bullshit.
@@smnomad9276 The best way is to get a good mentor. If that is not possible youtube is full of great videos or design patterns and SOLID Principles. Then review open source code identify patterns other developers are using. Never stop learning.
@@Salantor Explain? How is full stack declining? Every application has 3 tiers at minimum Front End, API and database. As a developer we need to keep learning and become good at building applications. Don't let a framework or library define your skillset.
Devin is vaporware. There said it.
High water mark for manual programmers will come a couple years after big non-tech companies have adopted AI to build applications and there is no one who knows how any of them work and no new features can be added. You'll have companies pop up that build the AI tools to figure out your AI codebase and there will be companies that say screw AI and hire developers to rebuild everything.
I just want to throw this out there: automating "other people's jobs" isn't necessarily the downer that people make it out to be. Several years ago, I automated phone-number transfer monitoring because two employees at our telecommunications company were spending their *entire* time doing that -- in no small part because the query websites were *slow* -- and by automating this, they were able to go *back* to their customer support/sales work!
Automating jobs means that (sometimes, at least) people are free to do something else that they would rather do instead.
Unskilled is better than 60% of the graph. Juniors around me can't solve problems I was solving in high school. I have to hold hands all day, every day.
There are people who have been on my project for 8 years, who still don't know the absolute basics of the language we are working in. It's absolutely insane.
Wow! You are obviously babysitting them.
On second thought, I would say take it easy on them. Some are extremely talented than others. Some of us didn't need any directions right from the beginning of our career and were on par with seniors. I have to often active my empathy when dealing with juniors I have to babysit.
@@JaekSean No way same lmao
Can you describe what problems juniors around you aren't able to solve.
Human developers ask questions, and remodel their solutions on the fly.
Robots make assumptions, and charge right off the edge of the nearest cliff
(Note - some mf’s are actually robots too)
Beep beep. I'm here to make my next bug.
@@Steel0079You will develop ze bugs and be happy about it
Developers are living on the edge of edging, oh yes.
For the lack of a better term. A coomder.
@@FlockofSmeagles Yea, pretty much. We only live once, why not explore our potential. 😹
One could take some small comfort in the idea that they thought the machinegun would be the end of the riffleman because it was 400 times more powerful. Then ww1 came and they just killed 400 times more soldiers and not a single rifleman was replaced.
And the number of bullets used per kill went up dramatically- even compared to the muzzle loaded musket era
ya because that was the birth of suppressive fire.@@steveoc64
I mean for me as a young person, I don't care about whether or not I'm making 60k or 100k as a software developer. I just care that I get a job after getting through school. I try to push myself and try new things in the development world and I try to develop very deep knowledge, but at the end of the day that won't matter if I can't get hired in the first place and end up working a completely different job.
Completely agree on you on that one. The real threat that everyone is putting under the rug is offshoring of jobs.
I love how you corrected a demand/difficulty graph by drawing an explanation on how to read that same graph :D
"Layoffs are dragging on" as LinkedIn tells me that hiring for senior engineers is up 7%
Yeah, there are lots of senior level roles out there. Some are not at senior level pay. Very few junior roles and some mid level roles. I've had multiple recruiters tell me they get 100s of applications for each job posting.
Yes, because now they can hire senior at the cost of juniors. Not sure what your point is here, the market is horrible, specially at Linkedin.
That's kinda the problem, everyone wants seniors because they created this cycle of hiring new people, not raising their salary in step with their performance and abilities, people jumping ship, rinse and repeat.
Now they don't want to invest into new and "junior" people as much and just want the people who already know what they're doing, except those people are jaded, are shrinking by the day and subsequently there won't be a new generation on that level to replace them at this rate. There's plenty of work out there, just not for the demographic of developers that are most affected by this shift in the IT world, not to mention that even seniors are getting shafted by companies even when they meet and exceed qualifications.
That's because you cant get senior engineers if you never hire junior engineers. If you really think the job market right now is good, youre smoking some really good stuff.
I used to get constantly spammed by recruiters, to the point of suspending my Linkedin. I don't get that anymore.
i was not prepared for those feet
I have zero worry that developers are going to be replaced by LLM's.
I tried using one of these LLM's to handle coding; and frankly the think is clueless about how to properly layout a large program. Its good at following a pattern though. That's about where it ends. Put it on a medium to large size complex task and it falls apart.
Yep good with small snippets but anything above that forget it. And dealing with large code bases with complex interconnections seems so far away as nearly impossible at least for another 10 years, maybe more.
@henriquemarques6196Ok, ill be sure to make a dare out of that and talk then.
Absolutely nailed it. It takes an absurdly long time to become "just" well rounded in software development. Not to mention the constant learning one needs to do. Very few careers consist of that level of "staying on it".
ThePrimeagen: Do you trust me?
Me: With every cell in my body.
As someone who built a fairly popular jQuery carousel in 2009 (coda slider / liquid slider) I feel we'll just level up with all the upcoming pressure, so long as we just keep at it.
I used to work at Ram Ranch, those weren't good time :(
I'm sorry for all of the friends you lost when you were raided by the US Marines.
The question “when will we have enough programs written?” is like asking “when will we be done creating laws?” Or “when will be done writing stories?” I think software development is more like this than farming. 🤷♀️
Good video. Optimism is always best - as long as it's not paired with ignoring events as they happen. But basing life decisions on hype trajectories is not wise. Solid developer skills will translate into whatever the future brings. Deep technical specialization is always risky. But being a developer who can turn requirements into actual solutions is a more fundamental skill simply coding - even if coding is for many the main tool used to accomplish the transformation. The same goes for troubleshooting, debugging, enterprise IT, keep critical live systems running, etc. If we get to a point where AI can take over all of that work without any need for being directed by or pairing up with humans (and be cost-effective compared to humans) - then we're heading for a utopia because that translates all the way down - and enables frictionless automation of pretty much anything at a trivial cost. So don't worry about that.
As for the more realistic trajectory where AI becomes able to churn out a lot of code, but doesn't have agentic independence at cost-efficiency - I am not so sure the majority of work will be in maintaining the terrible code bases which will result from enthusiastic and/or reckless adoption of such AI. Sure, there will be some clean-up jobs and some horrible systems to keep afloat until they can get ditched for something more sane - but I think it's much more likely developers will be needed to pair up with AI.
So, I do think developers might need to get used to the job changing. Get comfortable working with AI as they are now - their limitations and their strength. Because I don't there will keep being a choice. And they will get more agentic in nature. We might have AI 'colleagues' that queue up requests for assistance, work to be done, questions, etc. We might have AI watching as we work - a virtual pair programmer. For some this will just be a natural extension of how they already work - for others it might be a huge negative, to no longer be "alone" for the majority work. Someone who wants to become a dev, but hates working with AI and hates working with other people - might find it the coming years tough. That kind of person could find a great niche until now. But I think there is hope even for the most anti-social - as AI can be tailored to the individual.
I thought developers are now using Microsoft Edge
The day AI is able to really actually replace programmers is going to be able to replace anyone. So yes, it's useless to worry
I can't wait for that day to come. I'm excited for the future! Bring on the AI!
That's the day we get 100% free time and universal income and let the machines do the work. Love it.
@@JamesMowerywaiting for the promissed AGI so we can all become hippies
No not at all LMAO! Some jobs will be fairly safe from this. 99% of people will want a human therapist. No one that makes decisions cares if when their value-added product goes down its fixed by an LLM or a human 😂
@@JamesMoweryWill we get free food then?
Senior dev here, I've been using copilot and I have not been worried at all about "AI" until Devon. I'm nervous now.
No need, it's all hype and cherry picked scenarios.
im a software enginner student, my father is an accomplished programmer, and the way i see it rn, mind you i havnet done any interniship yet, is that it seems to be a "easy" to get in area but hard to stick around, due to the constant change in tecnology and the insane skill level required, thats why i always try to not use libraries and build my own stuff whenver possible, and then theres the ML market right now, which i relly dont wanna go to, but different from web development i feel like i might have to T^T.
In a job, they won't let you write your own libraries if an open source option exists. Because it is seen as a waste of time. Learning about API of some of these libraries is exhausting when sometimes I feel like I can make that library or feature on my own. But soon you will understand that the amount of development that went into that library is so much that they have had better programmers than you and me fix many edge cases that you don't even realise exist at this point in your own implementation. That's the reason you're not allowed to spend a lot of time writing your own packages in production environment. Learn to pick the right package for the job and make a habit of reading docs.
@@Steel0079 Would you say the same thing applies to things like templates? Often clients want to use a template.
@@Adam-nw1vy yes, but I don't have much experience with templates. So can't say. If we are talking about building UI yourself, then there are a lot of UI libraries like chakra ui, material etc. lot of things are doable. If something isn't, then you would have to make it custom. I recently had to make a multi month calendar with various accompanying tool components this way. Because the most similar one I could find that could do calendar lacked responsiveness.
@@Steel0079reading the docs is a given whatever, and i understand that places have production standarts, makes sense streamlining production, but i still think that for practice and learning its good to know HOW something would work from the ground up.
Nobody knows the future but the fact that so many people are having so much anxiety in the tech industry right now is clearly a symptom that we’re on the precipice of a big change that nobody knows the implications of yet.
I honestly don’t understand how some people get jobs in tech. A lot of people don’t know basic things
Social Networking skills. Why do you think most tech CEO's are Indians but very few founders are Indians?
@@Cat-vs7rc maybe but even at my job, I’ve seen a few people mostly juniors and some seniors with at least a decade of experience who don’t know how things work, they know how to write code but don’t know why they’ll use some functions instead of others or sometimes the code is very low quality or the algorithms are not good. I’m very confused by it sometimes. It feels like they just memorized things and repeat it everyday not willing to dig deeper to understand how things work. And we work with Indians too who are based in India for everything related to infrastructure, I’m just gonna say if you have a problem it’s better to find the solution yourself
They’re prob just BS’d the interview and then made the right friends
@@imhassaneOne way you could classify professionals is by the ratio of interest/disinterest in the nuances of their jobs. Surprise, many devs simply don't care that much. They just want to do their hours and go home, doing the minimum required. And this applies to most jobs.
there are multiple factors in a corporate work. one might say "its not what you know, its who you know" a nepotism and other factors as well such as communication skills. and interviews are a flawed by either way its the way it is
I am a skilled Flash ActionScripting man, all these new soy devs can't compete with me or my Arch install speed runs.
I will give a little advice here... Careful with any forecast. War is not about the best weapons, business is not about having the best product. Coding will not be about AI
We love you, you are our No. 1 Financial advice chanel
The wrod you were looking for is not profane but pedestrian.
Or something involving "vulgar".
I mean, I wouldn’t say it’s my primary browser, but it ain’t that bad
Thank alphabet
So everyone is going to ignore how prime decided to randomly assign a new meaning to the word "profane"? Couldn't find his definition of it anywhere.
Here is what I think is gonna happen:
Industry hits itself in the face because of AI. AI first tries to code and push work on github and other repositories, making the processed work unsuitable to use as training dataset with a tinge of error/hallucination when asked something complex or improper prompts.
More AI models propagate this hallunication bit by bit (if content is not vetted properly) and industry comes to a screeching halt after this recession because now you need to hire back the devs you fired as AI just doesn't cut it now.
Let's say this doesn't happen at all. Then when a super tiny bugs comes in, your systems burns because AI cannot find it. Now what? Will company have 5 AI models and a more than a few senior or principal engineers that debug the hot garbage AI wrote? most of them will keep jumping ship until they are paid top dollar to debug something like that.
Think people.
What bug would AI not be able solve that only a human could and why? Tell me ill wait.
They both literally have access to the same information, the same docs, the same exception lists to process. One is expontentially faster and cheaper than the other. And it obviously doesnt care about the consequences of that truth.
@@concernedcitizen6572I used chatgpt and it could not solve a bug in GDScript so yea buddy it can create made up functions libraries. I think devs will be paid way less and our job will be to fix those bugs if we accept it.
Same thing happen when I tried to get the AI to write assembly. The AI does not think its trained on DataSets. If it runs into a unknown problem it does the best it can. Humans can stop, actually think and solve the issue. Once the issue is solved we can feed that solution into the dataset.
Ai improves based on human expriences, the program which is really a Large Languange Model still does not think. It is no different then haviny a hugr dictionary with all the problems solved at your finger tips.
Imagine if a person made a program with all known solutions and solved programming patterens and gave it to a Software Engineer, this solution would actually be better then this so called AI. Instead they gatekeep the industry with Leetcode type question and brag that an AI with all the current solution feed to the dataset passed but an entry level dev without google to be his dictionary failed. If you want to make it fair, give that entry level dev the same access to look things up. In reality they are both similiar but the AI will save the company money unless its makes up functions.
@@concernedcitizen6572 yes obviously it doesn’t care, because AI doesn’t think. It’s not true intelligence. AI just very complex auto complete. It will do what it can with the data you give it, but if the data is bad, it won’t know the difference. Worse, AI has a major problem with hallucinating information.
Crickets to the above comment. So much coping with no facts to back it up lol
@@concernedcitizen6572bro the human mind is complicated. We can remember shit that happened years ago. Ai can't do that. Code base can get huge. And ai is very limited compared to the human brain. In software development everything is tied together and when you make a change to one thing it can affect hundreds of other things. If you have hundreds of thousands of lines of codes the ai won't be able to keep all of that in its context as it is very limited. When it comes time to solve complicated issues like that only humans can do that. Ai is good at generating small snippets of codes and that's it. Unless they can completely simulate an entire human brain. I don't see ai replacing humans.
The graph in the end: I highly doubt AI gonna try to automate the less demanded easy tasks first. They gonna do the most demanded easy task. Then less demanded ones, then maybe they try to go higher in difficulty. But I agree, we're still so far from automating programmer job, not even near to that, Devin isn't enough.
I really don't understand why people fear devin, it's just never going to succeed at doing what they are advertising it to be.
it's impossible for you to say that. It seems a lot of these comments are saying what they want to hear.
If you think that AI isnt going to wipe out at least 80% of developer jobs you are crazy.
AI is not perfect now but can do a lot right now. Imagine AI in 5 years time. It will be able to do the job of 500 developers in a day.
A lot of developer jobs are going and we need to accept it.
LLMs can do 70%+ of the tasks that developers currently do faster than the vast majority of devs. TODAY. I am not sure if this will lead to layoffs but within 1-3 years time there will be AIs that can do almost all of the tasks that we currently associate with development work. Maybe what we think of as the typical tasks of "developers" will simply change to accomodate AI. But the truth is that slowly but surely orgs will realize there is next to no value paying people $50-100 an hour to stare at a screen and try to figure out problems AI can figure out for next to nothing much quicker.
@@jackblack1801 how will you adjust? Career change?
@@jackblack1801 then pls leave the industry don't stay ...🙏🙏.
@@concernedcitizen6572 People doing html , css in companies. What you expect. They are not real engineers.
You talking about how hard carrousels used to be reminded me of the time I built a treeview taking the results of an Oracle Hierarchical Query when all we had was Vanilla Javascript and early JQuery
it's crazy to me that people saw a well refined probabilistic model and thought "shit, we are now at a post computers era!!"
Because 90% of corporate programmers are professional googlers. AI is simply letting management google for them
There's way WAY too many insanely overpaid developers who get paid to sit on their butt and Google stuff. This will be the huge market correction that fixes that issue. Skilled labor will have its comeback, as Prime said, and I'm all here for it.
@@YTDeletes90PercentOfMyCommentsThis is so inaccurate LOL. i can assure you only good devs are good at googling. I'd say only 10% of corporate programmers are professional googlers, the rest try to get by, take weeks to find the answer, or ask the good dev for help (not hating on the asking for help, mind you)
@@JamesMoweryExactly. But of course you never google ever do you? So you will for sure be safe. 😜
As a senior web dev what I see right now is that as a frontend dev you get more and as a backend dev your salary goes down (at least here in stockholm for senior roles)
Yea that's because LLM is better at language than images right now. front-end is next
There are programmers who only copy-paste simple javascript/python code, have no idea how computers actually work and don't know what a compiler is. Such programmers rarely even have to think. Most if not ALL of the thinking has already been done for them and is available on StackOverflow. Then there are programmers whose field of expertise simply REQUIRES that they know how computers work. Like operating systems developers, compiler developers, embedded developers, optimized game engine developers, among other specialties. Only the former group that I mentioned will be replaced by this trash-generating "AI". Because the code and systems the human programmers in that group generate IS ALREADY trash.
Second group is at risk too. Who can understand hlw computers work better than computers themselves. (Note here I dont think computers is exactly the right terminology here just using your term)
@@concernedcitizen6572 This comment perfectly encapsulates how little you truly understand.
No, a model that does next best token prediction for a given natural language context window (aka an LLM) does not understand how a computer works "because it is a computer". That you think an LLM is performing any reasoning whatsoever displays more about your knowledge than anything else.
@@krityaan lol I know AI does not "reason" calm down lmao.
OS devs, embedded systems devs most devs in general do tasks to add value to a business. ANY type of dev that is writing code and implementing development concelts that alreasy exist (aka 99% of devs) are at risk since AI can and soon will be able to build and troubleshoot and add value to businesses quicker than the vast majority of devs. Aint nobody special.
@@concernedcitizen6572 LLMs understand shit. They approximate the mean result of "if someone says a, then most likely b will follow".
"I YELL! THINGS! ON THE INTERNET!" We all do, Prime.
Fuck tht dude who said those things on twitter. I have learned alot from u over past 6 months. And your yelling and jokes only makes things easier to remember. Your a teacher we all needed.
Thts so wrong tht people bring the meth thing up to put you down. The life experince you have shared are much more valuable thn a non meth user. Infact sir, how u fought of your addiction is much more inspiring. Thank you for being so true, honest and more power to u
I am gonna be Senior Plumber Engineer, fu Devin
Good luck. Takes years to get certified and get an actual job as a plumber.
look out for optimus and figure 01, they are built to take those jobs.
The issue is that a lot of programmers today are block-box programmers; they have no idea what their code is doing... since it's not really their code. And since their code is largely what feeds (re)Generative AI, the situation will only worsen over time.
So, have we asked Devin's developers when they expect Devin to replace them?
they don't care, they plan to make bank and retire using their equity
I hate to be that linguistics guy but profane’s generally understood meaning today is as a synonym to vulgar (see: profanity). Coming from the Latin “pro”, meaning “before” or “outside” as understood by the Romans, and “fanum”, a temple or sanctuary. Profanus literally meant “unholy”, but later was used colloquially to describe things outside of a metaphorical ideology rather than the literal temples, morphing profane into its current meaning as “disrespectful”. The idea of it “making common” is actually its newer meaning, and continues to be its less-used definition
Threat of AI? *laughs in C*
i love this channel man, it gives me life!
ive been edging for a while
I like the emphasis on how difficult it is to become a really effective programmer. You can learn how to crank out some baseline code in a matter of months - in some ways, coding is just taking the thought process in your brain and giving those instructions to a computer. But to do it in a way that the code exists at the right level of complexity and readability and maintainability is much, much harder and reliant on hard-earned experience.
Electrical engineer turned sw dev here ✋
Consider computer engineering. We need better, cheaper hardware.
Should move back to EE. Im considering it. EE much harder for an AI to replace
Me too!
i tried to use chatgippity for EE and i am left almost always disappointed
No one is talking about Prime saying software dev is harder than trades. Anyone who has actually done both want to comment? Seemed wild to me.
That escalated quickly in the end with feet and gaps...
1:25, hm... washing dishes is paying better! (comparing the stresses)
Yeah come back once you do that for a full time job lol
@@bigPauliee Certainly it's better than coding. Then again, what job isn't better than coding?
Do dishwashing teams spend half the day in a “daily standup” and other planning meetings?
Or can they just get on with washing the dishes ?
He didn't manage to convince me with words. Guy's financial advice still did seem good.
The foot thing, though. This was the moment I've understood.
Hold up.. the developers are doing WHAT now?
The frontend market situation sounds bizzare. On the backend, companies literally stopped hiring experience people and they just hire the top 0.1% from the "I need a job, or I will starve" crowd.
Livin on the edge!
your really having fun with this :)
I am trying
What a weird way for these employers to gate keep.
Can't there be a set of standards that isn't just "muh arbitrary interview guidelines that include asking someone how you get out of a blender"?
Get out of Blender? Simple chief, click the X button on the top right corner
@@Titere05 It opened the windows startup menu, what do I do now ?
Human laziness is infinite, so the need for automation is infinite. The only thing that changes is the current Zeitgeist of what people pay for getting automated. And as a programmer, you sit right in the middle of making automation happen.
The more AI is rolled out, the cheaper automation becomes, the higher demand becomes for automation experts because more and more automation needs to be managed. This is the rebound effect and it is a well researched economic effect.
So, don't worry. Yes, FAANG massively over hired, but everyone else is madly scrambling to keep up with automating their businesses.
Don't be complacent, learn to adapt, and don't worry. This is just the beginning of our golden age as programmers.
Junior devs are trying so hard to look professional and passionate about programming.
Meanwhile... 23:24 Senior devs ladies and gentleman.
Least flexible senior dev
"Flying high on tailwinds" - I see what he did there.
This statement is NOT true - "And right now, except for a tiny group of gilded AI wizards, most programmers have either seen their prospects stalled or become more precarious." Actually, there are many "non-AI specialist" programmers who are still extremely sought and have never even started worrying about all this new trash-generating "AI" stuff. Of course I'm talking about low-level developers who ACTUALLY know how real computers work. OS developers, compiler developers, embedded developers, game engine developers, etc.
this.. I'm 30+ years in the trade and have never been out of a job for more than 3 weeks. I get offers weekly on Linkedin. There are jobs outside of FAANG which everybody seems to jizz about for some reason
I have been thinking about embedded and hardware work to help protect myself.
You are probably in webdev?@@concernedcitizen6572
In my country there's too few React devs, so they're getting paid crazy amounts of money. Imagine that. Now try to imagine an AI writing maintainable React code. Now laugh.
@@Titere05 can't be real, which country? I'll fly there
Look Devin makes English a very, very high level programming language. Natural programming languages can get you of to a good start but they're shit for engineering because they are not explicitly accurate - you spend more time writing a desciption/instruction than you do just writing the code. It has been tried before - the issue wasn't successful natural language implementation but the verbose and imprecise nature. Its like 3D movies all over again.
Mom I’ve made you proud 😭
Started off as a pabx technician in 1991, I learned turbo Pascal, then turbo C then turbo C++. I stuck with Borland and now embarcadero products. I don't do Microsoft products other than the operating system
Never goon.
Ten years ago it was autonomous vehicle will take over and make drivers obsolete. Five years later it was the blockchain will take over every transaction on the internet, making banks, currencies and payment systems obsolete. Now it’s A.I. will make knowledge/creative workers obsolete. Give me a break. I do see value in Chat GPT, and I understand how it will get smarter over time. But I see it as a companion to workers, not a replacement. Just how our “autonomous” systems became autopilot/super cruise. The day I see a client understanding their requirements and explicitly stating what they want from a software product, that’s the day I’ll worry.
am I on edge?
14:13 - Don't worry, Prime - I'm listening... :)
Your wizzdom and shower thoughts are very inspiring!
I develop IoT on the Edge. Do I count?
The electricians that need to take 10 years to get to the "Just OK" level are probably filtered out due to survivorship bias.
You need 20 years of sitting in front of the computer to maybe not be replaced by AI in the next 5 years.
For micro services, yes. Sensitive operations will never rely on AI.
@@trappedcat3615Yes exactly just like no organizations entrust their sensitive info and business logic with external tooling like github and copilot. ;)
WRONG. from 1990 to present pay went up 50% while cost of living in most cases double to tripled. If you need I can send you the data from cost of average home then and now or you can look it up the same as I did. I looked up pay from then to till now, cost of food, housing, rent, school, ... Average income went from 50K to 75K prospectively where homes went from 140K to 300K, college doubled.average gal gas 1990 1.15, average rent $447 today $1700.
What's going to happen is there will be very few high level developers that will manage the AI systems.
Basically, the developers will clean up or adjust certain areas that AI hasn't got exactly right.
It's time to accept that AI will wipe out 80% of developer jobs in around 5 years.
I know everyone is saying stuff to comfort themselves into thinking AI will not affect them, but the reality is AI is going to replace a lot of developers.
Anyone who thinks AI ain't coming for their jobs is living in dream world.
Tech lead warned us about this, but many laughed and called him a hater, and someone who is working for the companies to discourage everyone.
So many companies are now using AI and turning their backs on human developers.
Tech lead was right. Development and programming will be completely dead in 10 years for 90% of developers.
I do believe that AI will take major of dev jobs at some point, but I think that before this happens there are other types of the job in queue like a manager.
@@vViktor1 Manager will be safer. Managers TYPICALLY know business needs better than their developers do. Once managers have a tool that they can communicate with effectively and it can ship their value-added features and fix bugs faster than their average developer then developer job becomes less and less needed. Managers and Product Owners still needed to come up with and prioritize business requirements
@@concernedcitizen6572great points and believe the problematic areas you mentioned, they are currently working on them.
It's a wrap for developers
@@vViktor1i think the managers will be the last to go simply because many have connections higher up.
well, in 2008 we already had JavaScripts libraries for carousel, like dojo toolkit, older than jquery for 1 year. And yes, dojo toolkit had a carousel ready to use, and other libraries that have have fade into obscurity, before the age of famous libraries in js, there was a lot of websites dedicated to dhtml with ready available widgets to copy and use. It wasn't that dry dude. In 2008 I was working in a bank using Ext.js a UI library for the corporate world, all components with a default windows looking style with blue tones, they even had a WYSIWYG UI editor, I don't know if they still have it.
this will be a good thing for programmers all your side projects and ideas can now become a real thing with help of ai starting a startup would be easier now
9:00 is the key. When I graduated everyone wanted to be on finance or business administration, engineering was just an excuse to prove you were worth hiring.
5-10y later there was a reduced pool of real talent, and you could tell just by looking at your yoy salary growth.
Over the last 2-3 years there’s been a LOT of people trying to become devs (particularly web), with vague promises of getting good salaries with 6 month courses.
Now the pool is overflowing with people with poor skills, so if you can’t prove your talent quickly, you won’t get hired. Is that simple.
I will cherish the foot. Thank you. You made me smile
Thank you for the financial advice, I just spent my life savings and invested my pension into Nvidia stock. New sub!
cost of living, for the average person, went up far higher than salary. For you it might not seem like it, but they have.
At this point could you even imagine what "Prime gone wild" would actually have to mean? :D
If the rate of progress continues accelerating as it has in the past few years, you are 100% wrong. If it slows down, you are 100% right.
The first signs of computers understanding sentiment was 2016 and in 2022 chat GPT came out - now AI is ubiquitous.
We are now in an early stage of the curve (AI sophistication/time) developing - we dont know what kind of curve it is...
It could be a logistic/sigmoid curve or an exponential curve - If the rate of improvement of AI models continues on the exponential, we are fucked.
I would say its a 50/50, the only thing i can say for sure is that nobody knows the the future and therefore nobody knows the answer, so we are stuck guessing...
My observation is that the curve is asymptotic. 100% coverage becomes infinitely harder as the limit is approached. The limit being full human capability.
@@vripiatbuzoi9188 Yes, it is asymptotic to any finite set of goals - by definition, you cannot really improve once the problems in the set are "solved". But when Im refering to "sophistication", I mean something like "the sum of results across all conceivable problems" - which is unfortunately not a quantifiable quantity.
Also, I dont see a reason to a-priory assume that human capability is the limit - which is also a very vage limit since we dont really know what the limits of humans are.
There are many examples of narrow NN-based AI models surpassing humans - there might be a way to do similar thing to general AI (Probably not for LLMs of today, but it might be possible for future architectures).
For income, we "feel" the changes up or down, but we also "feel" differences between us and others around us. So if my pay is good but I can see lots of people who earn 50% more, I'm gonna feel poor. This can cause *serious* social unrest.
We're living right on the pops of CloudFlare s edge
Zoned out with video in background, came back to Prime farming subs by showing his feet. At least, that's what I assume was happening.