They can’t take it anymore
TPThe PrimeTime@ThePrimeTimeagen · 1.2M subscribers
- Views
- 1M
- 1,039,278
- Outlier
- ×2.9
- Median 364K
- Length
- 9:56
- Published
- Sep 23
- Wednesday
- Comments
- 2.5K
- 15.8K likes
Script
How it's written
- Words
- 2,134
- Words a minute
- 214
- Reading time
- 9 min
- Segments
- 7
The hook
0:00–0:26I am done with this It's over. The state of engineering right now is horrible. It's been a half a month since I started a new role at a big company. Nobody knows anything here. The spec, code, test, PRDs, tickets, resolution of those tickets, reports, etc. Everything is made by Claude code. Nobody on my team likes this. They're being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck,
Structure
I am done with this It's over. The state of engineering right now is horrible. It's been a half a month since I started a new role at a big company. Nobody knows anything here. The spec, code, test, PRDs, tickets, resolution of those tickets, reports, etc. Everything is made by Claude code. Nobody on my team likes this. They're being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow? People are working 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own. Everyone, literally everyone from L1 to L7 engineer here is doing the same thing. Talk to Claude. There is no sense of victory. Nobody is resolving bugs. In reality, nobody is even thinking anymore. Everything is done by LLMs. It is soul-sucking. I wouldn't mind it to be honest if we were at least given the time to check out the code and see what is going where. But no, the goal is to just ship, no matter what happens. Hey, how does this sound? Does this sound like what's currently going on with you? Is this the AI revolution that was supposed to be so fantastic? Is this the yes engineer that we are all told we will be over the next decade? Hey, you better press pause on those linear tickets because we got to do a little bit of yap, no, right? Hands off the keyboard, leave Claude alone because we just got to talk about what's going on. And we need to talk about how we can fix this. And does engineering even exist in 2026? Is it gone in 2027? I
Watch from 0:00guess more importantly, has this actually happened before? Yes. Yes, this has. But first, a quick thank you from the sponsor. Have you ever deployed to production and everything is green and awesome? But when you go to the website, oh my gosh, IT'S A 500! WELL, DON'T WORRY because Linear has your back. See, when that 500 hits, your error reporting can create a Linear ticket, which will kick off a Linear loop. So honestly, why spend 30, 40, 50 minutes of your time searching around with these production bugs when you can get automated pull requests that get you started with the correct fix. And just like that, a pull request issue is created and patch received, error fixed, and production is all good now. So, check out Linear today and actually save real time. The ad is the Linear engine. So, this kind of hits a bit personal to me because I over the last, you know, couple years I have I have struggled to integrate a lot of AI stuff and I've done more and more AI stuff. And I've also found just like a really kind of deep happy part of AI that I feel like I've changed over the last few months. And I I I feel pretty genuinely happy and I I kind of have a process how I think about it. And I do think it directly kind of reflects the opposite of the tweet we just read. So, let me tell you what I think is the actual solution and even more, why this has been going on for so long and what we're experiencing right now is technically nothing new. So, the person who said it the best is Mitchell Hashimoto. He calls it the whiteboard defense. I should be able to pull you aside at any moment and ask you to explain any customer-facing system that you've shipped. You should be able to clearly explain how it works and defend the decisions you've made.
Watch from 1:30This is my benchmark for responsible AI usage. Notice that this is like polar opposite of the tweet we just read. It literally is like, "No, there's actually still value in understanding what you've built, why you've built it. And if you're building something for a customer that you're ultimately responsible for, you should be able to slow down enough to at least know some of the nuts and bolts going on underneath the hood." And to me, this seems crazy because he's not even saying, "Hey, line by line familiarity." He doesn't even think He He doesn't even say you have to know exactly what function is being called. He cares about none of that detail. But he should be able to say, "Why did you use X instead of Y? What happens when an actor behaves maliciously? What data structure did you use here and why? Where does this fail?" Those are perfectly rational questions that if people could answer, you likely A will build a better product, but B I believe you will find yourself not having this soul-sucking experience where it's just all quad, all straight through is how he just he how he described it. From design docs to tickets to things being shipped, like nobody's reading anything. And this also this this type of approach where you don't actually have to know how things work, it doesn't work for all but the simplest of software. Eventually, that price does come due. Eventually, you picked several wrong abstractions.
Watch from 3:00It's going to be kind of hellish to undo that. And more importantly, the the reason why I say these things and the reason why I point to that out is because this has always been happening. So, in response to this, Uncle Bob says it best, which is what has been happening right now has always been happening. The irony here is that nothing has changed. Management is still asking why everything took so long. And this is so true. I still remember my very first kind of professional jobs in 2007, which was, "Hey, we need to build a feature, but at the same time, we need to know how many corners can we cut and how fast can we get this out into production?" See, back then, you had to kind of build something and you had to go, "Well, I want to pay myself a little bit into the future, so I want to fix a couple things, but man, it can't be perfect, so I'm going to slop a little bit over here. I'm going to fix a little bit of stuff over here because I believe this will be bigger impact fix. Hey, I'm going to get this to you in a week. I'm going to get it to you in 2 weeks." And the goal always was to try to make software a little bit better while trying to cut as many corners to be able to kind of play this balancing game of what corners are worth cutting and what ones aren't. That was in 2007. And from 2007 until 2024 when I quit after 10 years at Netflix, guess what? Same conversations every single time. We want to deliver a feature. How many corners are we cutting? What can we wait for a cycle after we get it out into production? And what do we need to take care of now?
Watch from 4:16Exact same questions that have always been asked. It's just now people feel a bit more disconnected from code, so the whole like I'm going to think into the future about the code base seems to kind of be slowly slipping from people's minds. It's not that the problems aren't already there. And anybody who takes any amount of time and reviews the Schrödinger code, the code works great, everything is fine, you look at it, you're horrified. How does this thing even work? So, I really do think that the fix to a lot of these problems is the ability to remember how to say no. Don't be just a yes engineer. Don't be someone that just shits out features because that's what's demanded of you. We had this power before AI to say, "Hey, things take a little bit of time." And now, if we just simply say no, it's infinitely cheap. We all know. Like anybody who's any good programmer can tell you, if you don't put up the proper guide rails, if you don't think about it, if you don't interact and ask the right questions, if you aren't actually actively involved in development, and you don't actually be able to answer these type of questions, you will just develop utter garbage. And it is really hard like it is soul-crushing. It's not fun to work in a shitty code base. It's not fun when the robot writes it, and it's not fun when you have to write it. Again, nothing's changed. I just kind of I I had to yap about this because I just feel like there's this like kind of like brain drain that's going on because what this is this is kind of like my behind-the-scenes theory, which is that we live in a world that's kind of bifurcated. There is a huge group of people who just work on like a website
Watch from 5:34hooked to one server that has all their data on remote database, log in with remote logging in system. And so, there's this kind of feel where you can just slop together crap, and it just works, right? It just seems to magically work because well, the problem space is rather straightforward. It's rather obvious how you're going to do stuff. You got server, you got client. And this from the outside appears to be a very kind of solved area of coding, if you will. And so, I think that there's a lot of people who exist in this world. And so, when they use AI, they just think that all things are solved. But then there's people that have to exist in the other world, where maybe there's a couple servers you have have think about. Maybe there's a few more interactions that aren't as simple as just thing executing. And the moment you do that, you find that there's so much complexity and so much like insanity that you can't just like let go of the wheel and say Dario, take the wheel for me. Just drive me into the the product success. All right, I just wanted to talk about this because I mean this this stuff worries me so much because I just hate the state of software right now. But me, I think we should be living in a world in which we can build really robust software so much better than like so much faster than we do like just 5 years ago. We're kind of living in this really cool time period, but instead everyone is miserable. And the thing that I most worry about even more is that juniors. Like how do juniors grow up in this world? Like how do you learn to develop curiosity and deeply how things work when everything is written for you? I would I don't know if I could have learned in this environment. Like I I just very
Watch from 7:05honestly, I don't know if I could have made it. I don't know. Not going to make it for me. I needed the difficult environment for me to be able to truly have to study these things. The creator of Clay, which is an immediate mode GUI, he I think he said it best. The catch-22 here is, seniors can explain the system because they've implemented them by hand before, but juniors are expected to generate code instead of writing it by hand. I I I don't honestly, I don't know how people learn what good systems look like without really learning them, without actually sitting down and building it yourself. And so someone who already happens to have the years of experience, this the whiteboard defense feels a lot easier. For those that have never really studied anything and don't really know how to program, I really don't know how you do this other than sit down, open up a text editor, learn some Vim motions, and program. By the way, I do actually take my own advice. I actually sit down and I open up my text editor and I program and I program things by hand to learn kind of specific libraries or large architectural decisions because I need to know how they feel. I need to understand that they're actually good. Maybe I'm just too stupid. Maybe I just have very small brain, little pea-sized brain, small IQ. I can't just simply read docs and go, "I understand the complete division." Even I need to sit down and write things sometimes. The name is the Primagen.
Watch from 8:37Full transcript
- 0:00
I am done with this It's over. The state of engineering right now is horrible. It's been a half a month since I started a new role at a big company. Nobody knows anything here. The spec, code, test, PRDs, tickets, resolution of those tickets, reports, etc. Everything is made by Claude code. Nobody on my team likes this. They're being forced to
- 0:21
ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow? People are working 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own. Everyone, literally everyone from L1 to L7 engineer here is doing the
- 0:42
same thing. Talk to Claude. There is no sense of victory. Nobody is resolving bugs. In reality, nobody is even thinking anymore. Everything is done by LLMs. It is soul-sucking. I wouldn't mind it to be honest if we were at least given the time to check out the code and see what is going where. But no, the goal is to just ship, no matter what
- 1:02
happens. Hey, how does this sound? Does this sound like what's currently going on with you? Is this the AI revolution that was supposed to be so fantastic? Is this the yes engineer that we are all told we will be over the next decade? Hey, you better press pause on those linear tickets because we got to do a little bit of yap, no, right? Hands off the keyboard, leave Claude alone because we just got to talk about what's going
- 1:23
on. And we need to talk about how we can fix this. And does engineering even exist in 2026? Is it gone in 2027? I guess more importantly, has this actually happened before? Yes. Yes, this has. But first, a quick thank you from the sponsor. Have you ever deployed to production and everything is green and awesome? But when you go to the website, oh my gosh, IT'S A 500! WELL, DON'T
- 1:45
WORRY because Linear has your back. See, when that 500 hits, your error reporting can create a Linear ticket, which will kick off a Linear loop. So honestly, why spend 30, 40, 50 minutes of your time searching around with these production bugs when you can get automated pull requests that get you started with the correct fix. And just like that, a pull request issue is created and patch
- 2:06
received, error fixed, and production is all good now. So, check out Linear today and actually save real time. The ad is the Linear engine. So, this kind of hits a bit personal to me because I over the last, you know, couple years I have I have struggled to integrate a lot of AI stuff and I've done more and more AI stuff. And I've also found just like a really kind of
- 2:26
deep happy part of AI that I feel like I've changed over the last few months. And I I I feel pretty genuinely happy and I I kind of have a process how I think about it. And I do think it directly kind of reflects the opposite of the tweet we just read. So, let me tell you what I think is the actual solution and even more, why this has been going on for so long and what we're experiencing right now is technically
- 2:46
nothing new. So, the person who said it the best is Mitchell Hashimoto. He calls it the whiteboard defense. I should be able to pull you aside at any moment and ask you to explain any customer-facing system that you've shipped. You should be able to clearly explain how it works and defend the decisions you've made. This is my benchmark for responsible AI usage. Notice that this is like polar opposite of the tweet we just read. It
- 3:08
literally is like, "No, there's actually still value in understanding what you've built, why you've built it. And if you're building something for a customer that you're ultimately responsible for, you should be able to slow down enough to at least know some of the nuts and bolts going on underneath the hood." And to me, this seems crazy because he's not even saying, "Hey, line by line familiarity." He doesn't even think He
- 3:28
He doesn't even say you have to know exactly what function is being called. He cares about none of that detail. But he should be able to say, "Why did you use X instead of Y? What happens when an actor behaves maliciously? What data structure did you use here and why? Where does this fail?" Those are perfectly rational questions that if people could answer, you likely A will build a better product, but B I believe
- 3:49
you will find yourself not having this soul-sucking experience where it's just all quad, all straight through is how he just he how he described it. From design docs to tickets to things being shipped, like nobody's reading anything. And this also this this type of approach where you don't actually have to know how things work, it doesn't work for all but the simplest of software. Eventually,
- 4:11
that price does come due. Eventually, you picked several wrong abstractions. It's going to be kind of hellish to undo that. And more importantly, the the reason why I say these things and the reason why I point to that out is because this has always been happening. So, in response to this, Uncle Bob says it best, which is what has been happening right now has always been happening. The irony here is that
- 4:32
nothing has changed. Management is still asking why everything took so long. And this is so true. I still remember my very first kind of professional jobs in 2007, which was, "Hey, we need to build a feature, but at the same time, we need to know how many corners can we cut and how fast can we get this out into production?" See, back then, you had to kind of build something and you had to
- 4:53
go, "Well, I want to pay myself a little bit into the future, so I want to fix a couple things, but man, it can't be perfect, so I'm going to slop a little bit over here. I'm going to fix a little bit of stuff over here because I believe this will be bigger impact fix. Hey, I'm going to get this to you in a week. I'm going to get it to you in 2 weeks." And the goal always was to try to make software a little bit better while trying to cut as many corners to be able to kind of
- 5:14
play this balancing game of what corners are worth cutting and what ones aren't. That was in 2007. And from 2007 until 2024 when I quit after 10 years at Netflix, guess what? Same conversations every single time. We want to deliver a feature. How many corners are we cutting? What can we wait for a cycle after we get it out into production? And what do we need to take care of now?
- 5:34
Exact same questions that have always been asked. It's just now people feel a bit more disconnected from code, so the whole like I'm going to think into the future about the code base seems to kind of be slowly slipping from people's minds. It's not that the problems aren't already there. And anybody who takes any amount of time and reviews the Schrödinger code, the code works great,
- 5:55
everything is fine, you look at it, you're horrified. How does this thing even work? So, I really do think that the fix to a lot of these problems is the ability to remember how to say no. Don't be just a yes engineer. Don't be someone that just shits out features because that's what's demanded of you. We had this power before AI to say, "Hey, things take a little bit of time."
- 6:15
And now, if we just simply say no, it's infinitely cheap. We all know. Like anybody who's any good programmer can tell you, if you don't put up the proper guide rails, if you don't think about it, if you don't interact and ask the right questions, if you aren't actually actively involved in development, and you don't actually be able to answer these type of questions, you will just
- 6:36
develop utter garbage. And it is really hard like it is soul-crushing. It's not fun to work in a shitty code base. It's not fun when the robot writes it, and it's not fun when you have to write it. Again, nothing's changed. I just kind of I I had to yap about this because I just feel like there's this like kind of like brain drain that's going on because what
- 6:56
this is this is kind of like my behind-the-scenes theory, which is that we live in a world that's kind of bifurcated. There is a huge group of people who just work on like a website hooked to one server that has all their data on remote database, log in with remote logging in system. And so, there's this kind of feel where you can just slop together crap, and it just works, right? It just seems to magically
- 7:17
work because well, the problem space is rather straightforward. It's rather obvious how you're going to do stuff. You got server, you got client. And this from the outside appears to be a very kind of solved area of coding, if you will. And so, I think that there's a lot of people who exist in this world. And so, when they use AI, they just think that all things are solved. But then there's people that have to exist in the other world, where maybe there's a
- 7:38
couple servers you have have think about. Maybe there's a few more interactions that aren't as simple as just thing executing. And the moment you do that, you find that there's so much complexity and so much like insanity that you can't just like let go of the wheel and say Dario, take the wheel for me. Just drive me into the the product
- 7:59
success. All right, I just wanted to talk about this because I mean this this stuff worries me so much because I just hate the state of software right now. But me, I think we should be living in a world in which we can build really robust software so much better than like so much faster than we do like just 5 years ago. We're kind of living in this really cool time period,
- 8:20
but instead everyone is miserable. And the thing that I most worry about even more is that juniors. Like how do juniors grow up in this world? Like how do you learn to develop curiosity and deeply how things work when everything is written for you? I would I don't know if I could have learned in this environment. Like I I just very honestly, I don't know if I could have made it. I don't know. Not going to make it for
- 8:41
me. I needed the difficult environment for me to be able to truly have to study these things. The creator of Clay, which is an immediate mode GUI, he I think he said it best. The catch-22 here is, seniors can explain the system because they've implemented them by hand before, but juniors are expected to generate code instead of writing it by hand. I I I don't honestly, I don't know how
- 9:03
people learn what good systems look like without really learning them, without actually sitting down and building it yourself. And so someone who already happens to have the years of experience, this the whiteboard defense feels a lot easier. For those that have never really studied anything and don't really know how to program, I really don't know how you do this other than sit down,
- 9:24
open up a text editor, learn some Vim motions, and program. By the way, I do actually take my own advice. I actually sit down and I open up my text editor and I program and I program things by hand to learn kind of specific libraries or large architectural decisions because I need to know how they feel. I need to understand that they're actually good. Maybe I'm just too stupid. Maybe I just have very small brain, little pea-sized
- 9:45
brain, small IQ. I can't just simply read docs and go, "I understand the complete division." Even I need to sit down and write things sometimes. The name is the Primagen.
Description
Linear helps teams track issues, manage projects, collaborate on docs, and ship products - and the crazy thing is their software is actually fast and a joy to use. Learn more: https://trm.sh/linear
Sources:
https://x.com/v0xium/status/210152610...
https://x.com/JonathanRoss321/status/...
https://x.com/mitchellh/status/210024...
https://x.com/nicbarkeragain/status/2...
/ theprimeagen - I Stream on Twitch
Great News? Want me to research and create video????: / theprimeagen
#theprimeagen #programming #computerscience #ai #chatgippty #terminaldotshop
