HomeVideos

"I suck" -ThePrimeagen

Now Playing

"I suck" -ThePrimeagen

Transcript

492 segments

0:00

Before I show you my talk on what is my

0:02

value that I gave at Omicron, I wanted

0:04

to give like a little bit of a

0:05

precursor, a little bit of kind of

0:07

setting up here so you understand what's

0:09

happening. So, Omicron just got done

0:11

taking place in New York. The conference

0:12

was awesome. The talks were awesome.

0:14

They're all going to be on YouTube. I

0:15

highly recommend you checking them all

0:17

out. It's about kind of the celebration

0:19

of computers, about our interest in it,

0:21

and how they're actually really fun and

0:23

like our current future is actually a

0:25

pretty cool future where you can build

0:27

anything and things are more accessible

0:29

than they've ever been. I'd really like

0:30

to thank Dell for sponsoring Omicron and

0:33

making it possible for me and the team

0:35

to come out there and really make this

0:36

event awesome. This computer is

0:38

absolutely fantastic. This is the best

0:40

laptop I've ever used a Linux on ever.

0:43

The battery life is incredible. It's

0:44

been on all morning and I'm still at

0:46

90%. And holy cow, look at this OLED

0:49

screen. It is so good looking. You see

0:52

Chrome running and I'm not out of

0:53

battery? The magical world we live in.

0:55

Thanks, Dell. There were a couple few

0:56

audio problems, so if the audio does go

0:58

in and out, don't worry, it gets fixed

1:00

or some of it might be a little bit

1:01

missing. So, it's really just like a

1:03

peek behind the curtains into what I am

1:05

thinking. So, love to hear your

1:06

thoughts, your comments, if you're

1:07

feeling the same way. Am I just full of

1:09

crap? Let me know in the comments. Hope

1:11

you enjoy it. I get a lot of messages

1:13

and the messages go something along the

1:14

lines of, "Hey,

1:16

you know, I've been working for 1 year,

1:18

2 years, 5 years, 10 years. Hey, I'm in

1:20

college right now.

1:22

What do I do?

1:24

Like, what's the changing environment,

1:26

how things are moving? Like, I am

1:28

genuinely freaked out. I don't know

1:30

where things are going. Is everything

1:32

I've done for the last 1, 5, 10, 15, 20

1:35

years

1:36

is it worthless? Am I worth less now?

1:40

And this is a question I just see like

1:41

constantly to the point where

1:43

at least multiple times a month

1:45

someone's like saying they're going to

1:46

kill themselves over this. Like, it's

1:47

like insane messages going out there

1:49

because people are, you know, I think a

1:51

lot of people feel this kind of feeling.

1:53

And so

1:55

I guess in some sense I sympathize with

1:56

that because how could you not, right?

1:58

Like, I don't know if you've been on

1:59

Twitter in the last couple days, but

2:01

there's a very aggressive strain of AGI

2:03

going on right now. And we're achieving

2:05

it daily at this point. So, I

2:07

I'd highly recommend a good social

2:10

distancing for 24 to 48 hours to help

2:12

slow the spread because it's going to be

2:13

achieved several more days in a row if

2:15

we do not stop.

2:17

But for me, I actually feel it quite a

2:19

bit in a sense that I feel this exact

2:21

same sense of loss and I feel this

2:23

confusion and I

2:24

things and for the last 6 months I've

2:27

kind of been in a

2:29

you know, a bit of a funk. Shall I say

2:30

if TJ I don't know where TJ went, but

2:32

TJ, several phone calls to him being

2:34

like, "TJ,

2:36

I'm confused right now. Feeling a little

2:38

control. And uh and the reason being is

2:40

for the last 20 years

2:42

I'm not There's a handful of days, but

2:44

pretty much I've done 6,000 days of

2:46

programming in the last 20 years. I

2:48

don't know how many hours that equates

2:49

to, but it's a lot of hours. I'm not

2:51

that quick with math.

2:52

Uh

2:54

Of those 20 years, 14 of them have been

2:56

spent doing Vim motions of some form,

2:59

whether it's an intelligent J or a God

3:01

bless my soul, I did VS Code for 6

3:02

months. Sorry, Maria. Uh

3:05

I did a lot You know, I did a lot of it,

3:06

right? And then I eventually landed home

3:08

in Neovim. It was a great time. But and

3:09

so I look at that

3:11

and I realized something kind of

3:13

strange, I guess.

3:14

See, you guys were out smoking with

3:17

cigarettes, partying,

3:20

drinking grape juice, and I was out

3:23

sharpening my blade, honing my skills,

3:26

speed running Vim tutor, becoming the

3:28

best you can possibly be at that

3:30

program.

3:32

You guys were probably having premarital

3:34

sex. I was learning programming

3:36

languages, okay? I knew Go, JavaScript,

3:38

C, Rust, Zig, even some Jai at one

3:41

point. Uh I've done I've done them all.

3:43

And obviously this is an Arch Linux

3:45

conference. I can tell a lot of chads

3:47

out there. And so you guys were out

3:49

sculpting your bodies, lifting heavy

3:50

weights. No. I was becoming the ultimate

3:53

generalist. I could build anything. I I

3:56

at one point I had to build a Pac

3:57

Perturber, sending something that looked

3:59

like G streamer so we could test the

4:01

latency from JavaScript to when audio

4:03

hits HDMI and making sure that, hey, if

4:06

we're going to be doing live video

4:07

games, are we going to hear pops, right?

4:09

Like, I have just been doing this thing

4:10

over and over again, building developer

4:12

tools, taking some idea and how do you

4:14

make it into something people can look

4:16

at and feel and test. And

4:20

well,

4:21

have I invested poorly? It's a question

4:23

I keep asking myself over and over

4:25

again. Have I invested poorly in my

4:28

future?

4:29

I guess in some sense it feels like

4:31

that, right? I think that a lot of

4:33

people can look at that and say, "Well,

4:35

no one's

4:37

So, I actually don't know what I just

4:38

said there and the rest of the audio

4:41

blips, which are another six of them

4:44

throughout the next 10 minutes, I can

4:46

generally guess what I'm about to say.

4:48

From here on out you will see one of the

4:49

you know, the old Jackie Chan movies,

4:51

how the lips didn't quite match up to

4:53

the dubs. Yeah, you There will be dubs

4:55

going on. That's not worth as much

4:57

anymore. And so I kind of feel a little

4:58

bit like Jim Cramer. You know, like

5:00

maybe there should be like an inverse

5:01

prime index fund out there because it's

5:03

way better.

5:04

And so you know, perhaps this is just

5:06

like part of being on X. You know, it's

5:09

not necessarily the most healthy

5:10

platform to be able to have any sort of

5:12

correct perspective. It is just a

5:14

platform to have a perspective.

5:18

Speaking of perspective,

5:19

have you guys heard of introspection?

5:24

Figured at least one person would like

5:25

that. Marc Andreessen out there, I'm so

5:27

sorry if you ever hear

5:28

I introspected by accident. I won't do

5:30

it again, I swear. I'm not giving in to

5:33

communism, don't worry. Uh so

5:35

last 6 months I have been doing this. I

5:37

have been introspecting and I have I

5:39

keep asking myself the same question,

5:40

which is

5:41

do I have value? What is my value?

5:44

Is my work history

5:47

less?

5:48

Is is like my time now worth less

5:51

because of what the changes and all

5:53

this? And I I know that a lot of people

5:54

are feeling the exact same thing. So, I

5:57

just have to answer that question. What

5:59

is my value?

6:01

And for a long time I guess I kind of

6:03

gave into this notion or I started

6:04

thinking about this notion or testing

6:06

this notion like, is my value just

6:08

having taste? This is that fancy word

6:10

VCs keep using. I don't really know what

6:12

it means, but apparently what that means

6:14

is that your value as an engineer is how

6:17

nice you can make a website look or

6:19

maybe how nice your command line flags

6:21

can be eaten by Claude Bot because

6:23

that's what's going to happen, right? Or

6:24

sorry, not Claude Bot. Sorry for the

6:25

trademark infringement, Dario. Please

6:26

don't come after me. Open Claw. Uh and

6:29

so

6:30

is that it? Is that Is that what we are

6:32

as engineers now? Just taste?

6:34

Just making sure things look nice? Make

6:36

circle not square?

6:38

Uh

6:38

Maybe. No, I mean, I hope not, but

6:40

that's that's what it kind of seems if

6:42

you read a lot of the writings right

6:43

now.

6:44

I guess another thing I thought about is

6:45

like, okay, is it lines of code? Because

6:47

during my peak, during my my heyday,

6:49

non-Adderall supported amount of lines

6:52

of code could be like 15,000 in a week.

6:54

I could I could type really fast. I can

6:56

Vim hard, man.

6:58

But then

6:59

Gary Tan does like 37,000 in a day.

7:03

And he goes like 7 days a week.

7:06

Like, there's no way I can compete on

7:08

lines of code. So, my value it can't it

7:10

can't be in that either because that I

7:12

mean, I'm getting mugged left and right

7:13

by a guy who dressed in a lobster

7:15

outfit. Like, that's not it's not going

7:16

to work out for me.

7:18

Um

7:19

so this is kind of the difficult part of

7:21

the talk.

7:22

I didn't really know how to do this. Um

7:25

mostly because the transition is really

7:27

bad.

7:29

Trans

7:31

we're going to do We're going to talk

7:32

about a time when I was much, much

7:34

younger.

7:35

It was changing a job. I just got done

7:36

by purchasing a house in Bozeman,

7:38

Montana. Okay, if you don't know

7:40

anything about Bozeman, Montana, it's

7:42

now like

7:43

Beverly Hills of Montana. Very

7:46

expensive. It's ridiculous. But I bought

7:47

mine on a handshake deal for 205. Look

7:50

at it in the books, that's it's crazy

7:52

it's crazy different now. Uh and

7:54

part of that is that I lived in this,

7:56

you know, nice little house outside of

7:57

town and I had a nice job. My wife

7:59

actually worked at the company called

8:00

Zoot where she learned how to use Vim

8:02

before me. Huh? Hubba hubba. That was

8:05

crazy. That was very very attractive. Um

8:10

I I hope she liked that. She might not

8:12

like what I just said there. Okay,

8:13

anyways.

8:15

So,

8:16

I got an offer to go work at a different

8:18

company. And this company ultimately, I

8:20

would say, became the company that

8:21

shaped who I am as an engineer. I think

8:22

a lot of people can kind of remember

8:24

that one company that maybe pushed them

8:25

a lot harder, did all that, you know,

8:28

however that looks. And for me, I had

8:30

that company. It's called WebFilings at

8:31

the time, now known as Workiva.

8:34

But I didn't know that at the time,

8:36

right? When you take a job, you don't

8:38

know how a job's going to be. And I I

8:39

was sitting there thinking like, I By

8:41

the way, I'm really bad with debt. I

8:42

hate debt. And so just like sitting

8:44

there, I was just freaking out. I just

8:46

bought my first house. Now I'm going to

8:48

a job. Am I going to be good enough for

8:49

this job? I was just like so in the

8:52

zone, focused on just these things that

8:54

could possibly happen going to a new

8:56

job. And I remember, I was I was my wife

8:59

was right here and I was just walking

9:00

We're walking to our neighbors and I was

9:02

just so focused on it. And for the first

9:05

time in my lifetime

9:06

life hit me in the face.

9:09

And by life I mean a 2 by 4 that

9:11

happened to be sticking out of a truck.

9:14

And due to the fact that I was so in my

9:16

head and so upset, I walked directly

9:19

into it and I was laying on the ground

9:20

and I was shocked. It was a very

9:22

shocking experience.

9:24

And that was actually the thing that

9:25

snapped me out of it. Surprisingly

9:27

enough, when you're super pissed getting

9:28

hit in the head with a 2 2 by 4.

9:30

Slightly recommend it, but it has to be

9:32

on accident.

9:34

But that moment I realized like I was so

9:36

focused on a future that I was crafting

9:39

out of a narrative that I don't even

9:40

know

9:41

that I didn't see the actual real and

9:44

obvious danger directly in front of me.

9:46

I made up dragons that didn't even exist

9:48

until I was laying on the ground. And

9:50

kind of

9:52

that that same thing happened to me

9:54

again just recently.

9:56

Sorry.

9:59

Which was

10:00

I had the same kind of two-by-four

10:01

moment. And so, for the last 6 months, I

10:03

genuinely have been very, very worried.

10:06

Uh I've asked the same question, what is

10:08

my value? What am I doing here?

10:10

Um I for the last 2 years have had a

10:12

couple medical issues. It's just like

10:14

everything doesn't seem to be lining up

10:16

like it normally has for some portion of

10:18

my life. And so, it's been it's been

10:20

difficult. And my second two-by-four

10:23

moment, if you will, was I saw this

10:25

tweet.

10:27

And I'm not going to name names, and I

10:28

don't want to make fun of anyone for

10:29

this.

10:30

But, the tweet looks something like

10:32

this.

10:34

Man, I'm trying to solve this really

10:35

trivial issue. Has anyone figured this

10:37

out yet?

10:39

And then there's a screenshot. And the

10:40

screenshot was Claude or some one of the

10:43

one of the many robots having its

10:45

internal monologue.

10:46

And inside the internal monologue, it

10:48

said,

10:50

"Hey, forking Chromium is a really

10:53

interesting question.

10:55

We should pursue dot dot dot dot dot."

10:59

And I thought,

11:01

"How do you build a web project

11:04

in which the result of whatever you've

11:07

made choices of as engineers led you to

11:10

forking Chromium? This is the worst

11:12

decision I've ever seen. There's nothing

11:14

trivial about it. Modern 2023 to 2025

11:17

hardware takes 6 hours to compile the

11:19

project. Like, this is not a How did you

11:22

get here? Like, this is insanity." And

11:24

then it just kind of dawned on me there

11:27

that

11:28

all the decisions I've made and learned

11:30

and earned along the way,

11:33

those only get, you know, if AI is to be

11:36

a true multiplier, then every one of

11:39

those little decisions actually do

11:41

matter. Because if you don't, you're

11:42

forking Chromium. Again, it is a

11:44

terrible idea. Do not fork Chromium for

11:47

any reason. Like, there's very few

11:48

reasons you should do that. And I just

11:50

sat I just sat there, and I was like,

11:52

"You can't explain to a date like you're

11:54

not going to just magically end up with

11:56

a Boyce-Codd normalized database. Yeah,

11:58

shout out Boyce-Codd. Where's my

12:00

Boyce-Codd boys at? Anyways, but you're

12:02

not going to just

12:03

end there unless if you know what it is.

12:05

You're not going to know why you should

12:06

break that. You're not going to know why

12:07

you should use standard in and standard

12:09

out versus a web server. You're not

12:10

going to know why you should make all

12:11

these individual choices and these these

12:14

decisions, they compound over and over

12:16

again. It's not some sort of like, 'Oh,

12:18

you have a five end problem.' No, you

12:20

have a two to the end problem. And every

12:22

one of these just get more and more

12:23

weighty. And so, yeah. Maybe you use

12:26

Neovim and open code.

12:28

>> Trad coder, I don't know if you know

12:29

that, but I'm a big fan of it. But, I

12:30

could see why generation of code is

12:32

super fantastic. Because

12:34

if it works like with TJ, you could just

12:35

have magical next to Machi theme just

12:37

come popping up, and that's like super

12:39

fantastic. And there's a lot of problems

12:41

that are just simply really, really well

12:43

understood and documented 5,000 times.

12:46

built another parser. I don't think

12:47

anybody needs to build another parser

12:49

except in bash. Am I right? Uh

12:52

Sorry, this is

12:53

very stupid. Don't don't worry about

12:54

that. Don't Ignore that.

12:56

Um

12:58

I forgot where I was. It was very funny

12:59

though. I um

13:02

And so, with that, it's like I guess

13:04

the the big two takeaways is what do you

13:06

do if you have all these skills? Well,

13:07

even if you don't somehow we do get to

13:09

this future where you're never typing

13:11

another character. It is actually so

13:13

valuable to have experience. If

13:15

anything, if the cost of a line of code

13:17

has dramatically dropped, then the cost

13:20

of the right line of code has

13:22

dramatically increased. It has to.

13:24

Because that means everything is free.

13:25

And when you have a bunch of options,

13:27

like, who here's made their own pizza?

13:29

Your first time you make your own pizza,

13:30

it's terrible. You're like, "Dude, I

13:32

love onions. Oh my gosh, SAUSAGE IS

13:34

GREAT. OH, I LOVE PINEAPPLE. HEY, I

13:35

love" And then all of a sudden you just

13:37

have every ingredient underneath the sun

13:38

on your pizza, and you're like, "This is

13:39

the worst pizza I've ever made in my

13:40

entire lifetime." Cuz when you have

13:42

every option, it actually becomes

13:44

dramatically harder to pick the right

13:46

option.

13:47

And so, for those that don't have skill,

13:48

I'm throwing around a hot new term.

13:51

Probably get this on on some of those

13:52

highfalutin podcasts out there, but

13:54

toxic productivity.

13:56

You don't have to be productive at all

13:57

times. Like, earning experience is more

14:00

valuable than completing something

14:01

instantaneously. Like, I think there's

14:03

this mindset or this idea that if you're

14:05

not continuously showing having some

14:08

sort of end result that you can click or

14:10

do something, then you haven't been

14:12

productive. But, I can tell you the

14:14

biggest and most influential moments in

14:16

my life have not been from making

14:18

something that's clickable, but from

14:20

failing over and over again to

14:22

understand why we do something a certain

14:25

way. And sure, I could have read the

14:27

friendly manual. But, I decided to save

14:29

5 minutes of reading the friendly manual

14:30

with 6 hours of debugging, and I learned

14:32

a lot from that. And so, please, if

14:35

you're new, like, there's still so much

14:37

hope. There's still so much cool things.

14:38

We live in a world where you can

14:39

actually be like, "Write me an OCaml

14:41

parser." And like, that actually can

14:43

happen. You can just do that now. That's

14:45

neat. Even though you wouldn't want to

14:46

maintain it, you can still do that.

14:48

That's super cool. And so, I don't want

14:50

people to get lost in this weird kind of

14:52

world we have crafted where everything

14:54

happens to be the end of the world.

14:56

AGI's being achieved. Dario, once again

14:58

I said, "Coding will be gone now in 12

15:00

months." It happened again. Can you

15:02

believe that? It's just like, "Well,

15:04

maybe.

15:05

But, good decision-making can't be gone.

15:07

Like, I don't see how you can get from A

15:09

to B without someone with good

15:11

decision-making. And I don't mean taste,

15:13

I mean actual engineering

15:15

decision-making that goes into this. Why

15:16

should you pick this data format versus

15:18

that data format? Why should you

15:19

serialize? Why should you do There's I

15:21

mean, The decisions are endless. And so,

15:25

that's all I really wanted to say.

15:26

Because that's how I got over my last 6

15:28

months of swirling around. And I know

15:30

that a lot of people are swirling

15:31

around. And I hope you realize the fun

15:33

have

15:34

and the goodness. And finally, I'll I'll

15:36

I'll end with a DHH quote. It's fun to

15:39

be competent.

15:43

>> [applause]

Interactive Summary

The speaker addresses the widespread anxiety among engineers about the future of their careers due to AI advancements, sharing his personal struggle with feeling his two decades of programming experience might be losing value. He recounts two "two-by-four moments" – a past physical accident and a recent realization sparked by a disastrous technical decision (forking Chromium) – that helped him understand true engineering value. He concludes that while AI can generate code, the ability to make good decisions, borne from experience and understanding *why* certain choices are made, becomes exponentially more valuable. The speaker emphasizes that competence, learning from failures, and focusing on sound decision-making, rather than just raw productivity or superficial "taste," are crucial for engineers in this evolving landscape.

Suggested questions

5 ready-made prompts