HomeVideos

Github might be in trouble

Now Playing

Github might be in trouble

Transcript

252 segments

0:00

So according to the information, Open AI

0:03

is developing an internal alternative to

0:06

Microsoft GitHub. Now, I personally have

0:10

just absolutely no idea why OpenAI would

0:14

ever want to replace GitHub. I mean, I

0:16

thought they were best friends with

0:18

Microsoft. I thought they loved

0:20

Microsoft products. I would have just

0:23

never guessed there to ever be a reason

0:25

ever to switch off of GitHub.

0:29

Not a lot of green in that status graph.

0:31

The reality is I do not think the uptime

0:33

of GitHub really has any impact on

0:36

OpenAI's decision to be making a source

0:38

control internally if the information is

0:41

to be correct. Now, I'm going to do

0:43

something that I don't often do on this

0:45

channel. Okay, I am going to make a fool

0:48

of myself by making a prediction into

0:51

the future about technology. And it's

0:54

pretty obvious why you should pretty

0:55

much never do that is because technology

0:57

moves so fast. How could you ever make

0:59

any predictions? Like I would have never

1:02

personally guessed that a front-end

1:04

library known as React would eventually

1:06

lead to remote code execution and

1:08

several of these CVEes in 2025. The last

1:12

12 months cinema. Now before I make that

1:15

prediction, I'm going to get that bag.

1:16

Let me tell you a little bit about

1:18

hiring. All right. Hey, hiring engineers

1:20

is broken right now. AI resumes, fake

1:23

profiles, and senior devs who don't even

1:24

use Vim. G2I fixes that, not the Vim

1:27

part, the hiring part, because they have

1:29

prevetted 8,000 plus engineers through

1:32

real technical interviews. So, you can

1:34

review quality candidates in days, not

1:37

months. And I've talked about G2I before

1:40

for backend and front-end roles, but if

1:41

you're also interested in AI roles, G2I

1:44

needs to be the first place you go and

1:45

check out. Get a 7-day trial plus $1,500

1:48

off using my code. Visit g2i.co/prime.

1:53

But hold on, there's more. You know, I

1:55

love React Miami, right? Well, now

1:57

there's another conference called AI

1:58

Engineer that's going to take place also

2:00

in Miami, right next to React Miami. So,

2:02

if you don't want to have skill issues

2:04

like I have with AI, you need to go to

2:06

the conference. Use code prime 50 off

2:09

for 50 off and I'll see you in Miami.

2:14

All right, the prediction. I think

2:15

version control is going to change a

2:18

huge amount over the course of the next

2:20

few years and I think it's going to be a

2:22

bigger and faster change than it was

2:24

from tortois svn to git. Now a lot of

2:27

you may not know this but git hasn't

2:29

been around that long as far as the

2:31

history of computing goes. Git is 30

2:33

years younger than the oldest version

2:35

control system and it's one year younger

2:38

than mean girls which every time I think

2:40

about that it just I don't know. I don't

2:43

know how I feel. It doesn't make me feel

2:45

that good. Git is a version control

2:47

that's strongly tied to the file system

2:50

and it really encapsulates the ideas of

2:52

that day and of course it's creator

2:55

Linus Torva. So with my prediction of

2:57

everybody moving away, you can obviously

2:59

tell I'm panicking because one of my

3:01

courses on boot.dev is to learn Git. I

3:04

mean boot.dev/prime 25% off. Am I am I

3:08

right, boys? Now I know there's a lot of

3:10

you that are probably fairly skeptical

3:12

of what I'm saying right now. And I get

3:14

that. Hey, I would be pretty skeptical

3:16

if I heard people just spouting such

3:18

nonsense because I am kind of like

3:20

yapping. I'm blathering at this point

3:22

about this entire topic. Just hear me

3:24

out. Follow my reasoning. May I first

3:27

present exhibit A. This is pretty much

3:30

before any of the AI revolution. Okay,

3:32

this is two years ago back when AI was

3:34

real real bad at everything. This is

3:36

just a meme. We This We love this. Okay,

3:38

this is one of my favorite things of all

3:41

time ever posted on the internet. And

3:43

now I'd like to present you exhibit B,

3:46

OpenClaw. It has more stars than React

3:51

and Linux on GitHub. Single-handedly one

3:54

of the most popular repos of all time.

3:57

AI is just at the very beginning and the

4:00

storm of people coming in and realizing

4:02

that they can now do things to computers

4:04

that they've never been able to do

4:06

because they lacked the either the

4:07

motivation or the mental capacity to be

4:10

able to become a programmer. Now they

4:12

can just be like, "Yo, yo, OpenClot,

4:14

make all this stuff for me." And of

4:16

course, what happens when people start

4:18

building software? They quickly realize

4:20

the value of being able to keep track of

4:23

everything that they've ever made.

4:24

underneath the hood, an agent could, you

4:27

know, make a commit every single time it

4:29

makes a change and then push that up to

4:31

some sort of GitHub, but that person

4:32

would have to go and they'd have to make

4:34

a GitHub and then after making a GitHub,

4:36

they'd have to like get a nit or get

4:38

remote, you know, like there's just so

4:40

many things that they're unlikely to do,

4:42

at least in my personal opinion, because

4:44

it is just a lot of computer work. And I

4:46

think people are going to keep on

4:47

getting more and more abstracted away

4:48

from computers. And so to me it seems to

4:51

make a lot more sense that these

4:53

companies that are developing these

4:54

tools such as codeex with open AI or

4:57

really cursor with what is anyphere? Any

5:01

sphere? That's right. Any sphere with

5:02

cursor to me it's like they're the ones

5:04

that are truly positioned to just create

5:07

version control systems that are

5:08

effectively transparent to agent

5:11

changes. And so agents are just going to

5:12

be making these big set of changes and

5:14

making these little commits or these

5:16

little, you know, change logs as they

5:18

go. and the person that is developing a

5:20

lot of scare quotes on the word

5:22

developing these pieces of software will

5:24

be able to kind of walk through those

5:25

changes go okay my app broke and I was

5:27

unable to fix it so let's walk back four

5:29

places let's start again from right here

5:31

for those that kind of know version

5:33

control systems this sounds very

5:35

familiar doesn't it yes JJ jiu-jitsu

5:37

this is pretty much the premise of JJ

5:40

which seems to be an excellent alignment

5:41

with how these agents are operating also

5:44

the diffs they look really good in JJ

5:47

you know I've never used JJ, but the

5:49

more I look into JJ, I'm like, "Oh my

5:51

gosh, this is what I want." Like, this

5:54

diff is absolutely beautiful. Now,

5:56

technically with Git, you can also use

5:58

colored words, which will look like

6:00

this. But man, JJ's that that diff and

6:02

JJ was

6:05

that was looking good. And I say all

6:07

this because I don't think GitHub has

6:09

some sort of strangle hold on

6:11

developers. I don't think it really

6:13

matters to this next generation of

6:15

people coming in who are trying to

6:17

develop software. I don't think they

6:19

care that it's hosted on GitHub or

6:22

GitLab or wherever it's at. They just

6:24

want things that don't disappear. And

6:26

honestly, I genuinely thought this was

6:29

going to be the case when Cursor bought

6:32

graphite. graphite was the ability to

6:34

kind of have the material approach to

6:37

stacked diffs and all of that and be

6:39

able to move changes more easily with a

6:41

team and kind of get rid of some of the

6:43

annoying parts of working with git. So I

6:45

just naturally thought like this was

6:47

already going to happen but it hasn't

6:49

happened yet. So it looks like Open AAI

6:51

might just be developing an alternative

6:53

to GitHub internally. They probably also

6:56

have achieved AGI internally. Sam Alman

6:59

is probably wearing multiple polos at

7:02

this moment. It's a weird time kind of

7:03

to be in this whole software world

7:04

because giants in which I never thought

7:07

would change such as GitHub I now

7:10

foresee as one of the first things to

7:12

actually get radically reduced in this

7:15

AI revolution. I I just figured they

7:17

would be on top coming out on top of

7:18

this because hey, more code, more things

7:19

to store. But now that I'm kind of

7:21

looking at it, I'm thinking, no,

7:23

actually, they're going to be the first

7:24

ones gone. And any of these AI

7:26

companies, whether it is going to be

7:27

cursor or open AI that eventually

7:29

creates an entirely closed ecosystem,

7:31

it's going to create just more and more

7:33

training data. they can watch and have

7:35

full control over the prompts that were

7:38

used, the code that was generated, the

7:40

things that were released to production,

7:41

the things that were marked good. The

7:43

amount of just like data and operations

7:46

that's going to come from whatever

7:47

company is the first one to really jump

7:49

on this is going to be incredible. And

7:51

honestly, I'm just sh I shocked again. I

7:54

just can't believe Cursor hasn't already

7:56

done this. All right, that's it. That's

7:57

it. That's all I just want to yap about

7:59

it a little bit because it's just a

8:00

strange time. I I'm just feeling weird

8:02

about this whole world that we live in.

8:04

Anyhow, the name is I'm probably I'm I

8:07

mean, personally probably going to

8:09

continue to use GitHub for a long time.

8:11

Like I don't see any reason not to learn

8:13

Git right now. I think there's just

8:14

going to be an extended period of time

8:16

of weird tumultuous change that's going

8:18

to exist for a while and then there's

8:19

probably going to be two islands of

8:20

people. Honestly, the people that just

8:22

truly have no idea what they're doing

8:24

and they're going to be completely

8:26

enclosed in a wall garden like cursor

8:28

where they develop all their code in it.

8:29

They're able to have all the browser

8:31

testing in it. They have all the things

8:32

being committed transparently. They

8:34

don't even realize it's being pushed to

8:36

some sort of remote origin. They just

8:37

have no idea how anything works. They're

8:39

just like that that guy on the Truman

8:41

show that's like, "Yes, now make that

8:43

theme more purple." Like that's it.

8:45

That's all they really have inside of

8:46

their head. Then there's going to be

8:47

this other side, the people that artisal

8:49

people, the people that want to just

8:50

like maybe they use half vibing, half

8:52

just hand coding, whatever they do, but

8:54

they're the ones kind of in control of

8:56

everything and they understand how

8:57

computers work. It's just going to be a

8:58

weird world. And I am, you know, I was

9:01

born in a different time, so I'm just

9:03

naturally in the other side again.

Interactive Summary

The speaker discusses the surprising news that OpenAI is reportedly developing an internal alternative to Microsoft GitHub. He predicts a significant and rapid change in version control systems, suggesting this shift will be even more profound than the transition from Tortoise SVN to Git. This change is driven by the rise of AI agents, which are enabling a new generation of users to build software without traditional programming knowledge or the motivation for complex version control tasks. The speaker envisions future version control systems being transparent to these agent-driven changes, with companies like OpenAI or Cursor potentially creating closed ecosystems that provide vast training data and control, ultimately diminishing GitHub's current dominance. He concludes by foreseeing two distinct groups of developers: those operating entirely within AI-driven "walled gardens" without understanding the underlying mechanics, and "artisanal" developers who maintain a deep understanding and control over their systems.

Suggested questions

6 ready-made prompts