Why Casey Muratori avoids AI
- Views
- 66.8K
- 66,798
- Outlier
- ×7.3
- Median 9.2K
- Length
- 6:41
- Published
- May 15
- Friday
- Comments
- 259
- 1.8K likes
Script
How it's written
- Words
- 1,214
- Words a minute
- 181
- Reading time
- 6 min
- Segments
- 5
The hook
0:00–0:28You're one of the few people who actually Well, I I guess a lot of people don't like AI programming, but you're one of the few people who from what I know, like very much abstains from it. And so, I I I want to get your kind of opinions on that. Uh my perspective on AI programming has basically been the same since I was first asked about it, which was, you know, uh I guess a year ago now at this point. Uh kind of before people started to think it got good. Like mo- most I mean, there
Structure
You're one of the few people who actually Well, I I guess a lot of people don't like AI programming, but you're one of the few people who from what I know, like very much abstains from it. And so, I I I want to get your kind of opinions on that. Uh my perspective on AI programming has basically been the same since I was first asked about it, which was, you know, uh I guess a year ago now at this point. Uh kind of before people started to think it got good. Like mo- most I mean, there are obviously people who were claiming it was amazing several years ago. Uh but most like serious programmers I know um nobody really agreed with that for quite some time. Um and most people said like, you know, it started getting useful, you know, sometime around December, you know, with with a particular Anthropic release or whatever. This This, you know, this December. So, I've been saying the same thing since well before that. Uh which is that from an objective standpoint, AI programming to me is more like Well, I have some concerns about it. Like, I'm concerned that people might adopt it too early. Um I'm concerned they don't understand the ramifications of the code they're producing. Um you know, I'm I'm worried the impact it'll have on security, on performance, uh these sorts of things on reliability.
Watch from 0:00Uh I'm concerned about all those things, but those are just like concerns about people using the technology for more than what it currently can handle. And that's That will just be a concern I I have the same concerns about human Right, humans, too. Right? Like like I don't want somebody who doesn't really know how to write uh um a cryptography library writing a cryptography library, right? So, so I have those concerns about it, but those are just general concerns. Um that's not why I don't personally use it, because I think as long as you understand those things are and are making good assessments. Like, if you're a programmer who's capable capable of making rational assessments about whether the AI is producing code that's good enough for you to use, whether it's improving whether it's actually improving your productivity not just making you feel like it did. If you're capable of making those assessments well, then I don't see the problem with it after the copyright issue like there's the legal issues and and is this ethical? That's a separate thing. But in terms of the technical perspective as long as you're able to sort of correctly assess that, then I don't think AI is particularly a problem one way or the other. If you think it's helping you in your programming pipeline, great.
Watch from 1:17For me personally, I don't use it for the same reasons that I don't do a lot of things that a lot of programmers do. And that's that many programmers come from a perspective that is very different from mine in terms of what they are doing. Many programmers come at programming from a much more business-oriented perspective. Their goal is to make a specific thing and sell it or make a particular thing and release it, you know, as as a free piece of free software or whatever it is. And the goal is just to have the software. They don't really care how they got there, right? So, if you're coming at it from that perspective, whether you programmed it yourself line by line or whether you use AI doesn't really make a difference, right? Um because the end result as long as the end as long as you can get the AI to do as good of a job as you did or maybe in the future hopefully better than you did even, right? Uh then it's a a win because you were basically using some new tooling effectively uh to accomplish a goal that you wanted to accomplish.
Watch from 2:33But for me, I just like the act of programming. That's always kind of been why I did it and it's just kind of like a happy accident that I happen to live in an era where you got paid well to do that. But that wasn't why I did it. Like I didn't go into it as a job. I went into it as a as a hobby. Um and so I just like doing that stuff. I mean, I literally spent all of yesterday programming in assembly language. So I wasn't even using a compiler, right? And there was no particular reason for it other than I wanted to investigate some behavior of a thing that was best investigated by just playing with the assembly directly. That's just what it was, right? And I don't pretend that those are smart business decisions. I wouldn't go tell somebody at a company that they should start programming everything in assembly language cuz this compiler is, you know, doesn't work very well or something. Compilers work very well. Uh it's just that's what I enjoy doing. I like kind of digging in. I like editing the files myself. I like being the the more control over I have about what comes out, the better. And even compilers sometimes bother me cuz I don't have as direct control over the hardware as I do with something where I can specifically say like what the instructions are and what order uh they go in or whatever else, right?
Watch from 3:49Uh the ordering kind of now the CPU likes to do some some reordering, but they still are there still are ramifications for the order of instructions, believe it or not, even on out of order CPUs. So anyway, uh so hopefully most people who know me wouldn't find this strange. If you're the kind of person who would get excited about like let's take a look at how texture mapping worked on the PlayStation 1, right? Then it makes sense why you wouldn't care about it cuz it's just like look, you're already you already just like this stuff from a purely like experiential standpoint. Nobody Nobody should ever texture map the way the PlayStation 1 texture map. It's texture mapping was awful, right? So like it's not an argument about quality, it's not an argument about product, it's nothing like that. It's just it's about a fundamental like um love of the act of programming and the curiosity about how it works. And to me, AI, people will say, "Well, it's just another tool like a compiler." It's not. You're typing in English, it's very imprecise and difficult to control. It's not the same experience. It really isn't. And so, while I don't have any problem with someone else saying that's the tool that they would like to use, it's just it isn't a good match for someone who want, you know, who wants to do the poking directly, right? It's it's just extending, it's making an even less direct connection between you and the program, and that's exactly the opposite of what I want. So, hopefully that makes some sense. It's not an economic argument, it's just a it's just a what do I enjoy?
Watch from 5:06Full transcript
- 0:00
You're one of the few people who actually Well, I I guess a lot of people don't like AI programming, but you're one of the few people who from what I know, like very much abstains from it. And so, I I I want to get your kind of opinions on that. Uh my perspective on AI programming has basically been the same since I was first asked about it, which was, you know, uh
- 0:20
I guess a year ago now at this point. Uh kind of before people started to think it got good. Like mo- most I mean, there are obviously people who were claiming it was amazing several years ago. Uh but most like serious programmers I know um nobody really agreed with that for quite some time. Um and most people said
- 0:42
like, you know, it started getting useful, you know, sometime around December, you know, with with a particular Anthropic release or whatever. This This, you know, this December. So, I've been saying the same thing since well before that. Uh which is that from an objective standpoint, AI programming to me is more like Well, I have some concerns about it. Like, I'm
- 1:02
concerned that people might adopt it too early. Um I'm concerned they don't understand the ramifications of the code they're producing. Um you know, I'm I'm worried the impact it'll have on security, on performance, uh these sorts of things on reliability. Uh I'm concerned about all those things, but those are just like concerns about people using the technology for more than what
- 1:24
it currently can handle. And that's That will just be a concern I I have the same concerns about human Right, humans, too. Right? Like like I don't want somebody who doesn't really know how to write uh um a cryptography library writing a cryptography library, right? So, so I have those concerns about it, but those are just general concerns.
- 1:45
Um that's not why I don't personally use it, because I think as long as you understand those things are and are making good assessments. Like, if you're a programmer who's capable capable of making rational assessments about whether the AI is producing code that's good enough for you to use, whether it's improving whether it's actually
- 2:05
improving your productivity not just making you feel like it did. If you're capable of making those assessments well, then I don't see the problem with it after the copyright issue like there's the legal issues and and is this ethical? That's a separate thing. But in terms of the technical perspective as long as you're able to sort of correctly assess that, then I don't
- 2:25
think AI is particularly a problem one way or the other. If you think it's helping you in your programming pipeline, great. For me personally, I don't use it for the same reasons that I don't do a lot of things that a lot of programmers do. And that's that many programmers come from a perspective
- 2:46
that is very different from mine in terms of what they are doing. Many programmers come at programming from a much more business-oriented perspective. Their goal is to make a specific thing and sell it or make a particular thing and release it, you know, as as a free piece of free software or
- 3:07
whatever it is. And the goal is just to have the software. They don't really care how they got there, right? So, if you're coming at it from that perspective, whether you programmed it yourself line by line or whether you use AI doesn't really
- 3:27
make a difference, right? Um because the end result as long as the end as long as you can get the AI to do as good of a job as you did or maybe in the future hopefully better than you did even, right? Uh then it's a a win because you were basically using some new tooling effectively uh to accomplish a goal that you wanted to accomplish.
- 3:49
But for me, I just like the act of programming. That's always kind of been why I did it and it's just kind of like a happy accident that I happen to live in an era where you got paid well to do that. But that wasn't why I did it. Like I didn't go into it as a job. I went into it as a as a hobby. Um and so I just like doing that stuff. I mean, I
- 4:11
literally spent all of yesterday programming in assembly language. So I wasn't even using a compiler, right? And there was no particular reason for it other than I wanted to investigate some behavior of a thing that was best investigated by just playing with the assembly directly. That's just what it was, right? And I don't pretend that those are smart
- 4:32
business decisions. I wouldn't go tell somebody at a company that they should start programming everything in assembly language cuz this compiler is, you know, doesn't work very well or something. Compilers work very well. Uh it's just that's what I enjoy doing. I like kind of digging in. I like editing the files myself. I like being the the more control over I have about what comes out, the better.
- 4:54
And even compilers sometimes bother me cuz I don't have as direct control over the hardware as I do with something where I can specifically say like what the instructions are and what order uh they go in or whatever else, right? Uh the ordering kind of now the CPU likes to do some some reordering, but they still are there still are ramifications for the order of instructions, believe it or not, even on out of order CPUs.
- 5:15
So anyway, uh so hopefully most people who know me wouldn't find this strange. If you're the kind of person who would get excited about like let's take a look at how texture mapping worked on the PlayStation 1, right? Then it makes sense why you wouldn't care about it cuz it's just like look, you're already you already just like this stuff from a
- 5:35
purely like experiential standpoint. Nobody Nobody should ever texture map the way the PlayStation 1 texture map. It's texture mapping was awful, right? So like it's not an argument about quality, it's not an argument about product, it's nothing like that. It's just it's about a fundamental like um love of the act of programming and the
- 5:57
curiosity about how it works. And to me, AI, people will say, "Well, it's just another tool like a compiler." It's not. You're typing in English, it's very imprecise and difficult to control. It's not the same experience. It really isn't. And so, while I don't have any problem with someone else saying that's the tool that
- 6:18
they would like to use, it's just it isn't a good match for someone who want, you know, who wants to do the poking directly, right? It's it's just extending, it's making an even less direct connection between you and the program, and that's exactly the opposite of what I want. So, hopefully that makes some sense. It's not an economic argument, it's just a it's just a what
- 6:39
do I enjoy?
Description
The Full Episode:
Youtube - • Why Programmers adopt Bad Ideas - Casey Mu...
Spotify - https://open.spotify.com/episode/4OeS...
Apple - https://podcasts.apple.com/us/podcast...
🚀 https://neetcode.io/yt - A better way to prepare for Coding Interviews
#neetcode #python




