文本记录English

Head of Claude Code: What happens after coding is solved | Boris Cherny

1h 27m 45s20,046 字数2,765 segmentsEnglish

完整文本记录

0:00

100% of my code is written by quad code.

0:02

I have not edited a single line by hand

0:05

since November. Every day I ship 10, 20,

0:07

30 p requests. So at the moment I have

0:09

like five [music] agents running while

0:10

we're recording this.

0:11

>> Yeah. Yeah. Do you miss writing code?

0:13

>> I have never enjoyed coding as much as I

0:15

do today because I don't have to deal

0:17

with all the minutia. Productivity per

0:19

engineer has increased 200%.

0:21

>> There's always this question, should I

0:22

learn to code? In a year or two, it's

0:23

not [music] going to matter. Coding is

0:24

largely solved. I imagine a world where

0:26

everyone is able to program. Anyone can

0:28

just build software anytime. What's the

0:30

next big shift [music] to how software

0:31

is written?

0:32

>> Quad is starting to come up with ideas.

0:33

It's looking through feedback. It's

0:35

looking at bug reports. It's looking at

0:36

telemetry for bug fixes and things to

0:38

ship a little more like a co-orker or

0:40

something like that.

0:41

>> A lot of people listening to this are

0:42

product managers and they're probably

0:44

sweating. I think by the end of the

0:45

year, everyone's going to be a product

0:46

manager and everyone codes. The title

0:48

software engineer is going to start

0:49

[music] to go away. It's just going to

0:50

be replaced by builder and it's going to

0:52

be painful for a lot of people.

0:56

Today my guest is Boris Churnney, head

0:58

of Claude Code at Anthropic. It is hard

1:01

to describe the impact that Claude Code

1:03

has had on the [music] world. Around the

1:05

time this episode comes out will be the

1:07

one-year anniversary of Claude Code. And

1:09

in that short time, it has completely

1:11

transformed the job of a software

1:13

engineer and it is now starting to

1:15

transform the jobs of many other

1:17

functions in tech which we talk about.

1:19

Cloud code itself is also a massive

1:22

driver of anthropic overall growth over

1:24

the past year. They just raised a round

1:26

at over [music] $350 billion. And as

1:29

Boris mentions, the growth of Claude

1:31

Code itself is still accelerating. Just

1:33

in the past month, their daily active

1:35

users has doubled. Boris is also just a

1:37

really interesting, thoughtful,

1:39

deepinking human. And during this

1:41

conversation, we discover we were born

1:43

in the same city in Ukraine. That is so

1:46

funny. I had no idea. A huge thank you

1:48

to Ben Man, Jenny Wen, and Mike Griger

1:50

for suggesting topics for this

1:52

conversation. Don't forget to check out

1:53

lennisprodpass.com

1:55

for an incredible set of deals available

1:57

exclusively to Lenny's newsletter

1:58

subscribers. Let's get into it after a

2:01

short word from our wonderful sponsors.

2:04

Today's episode is brought to you by DX,

2:06

the developer intelligence platform

2:08

designed by leading researchers. To

2:10

thrive in the AI era, organizations need

2:12

to adapt quickly. But many organization

2:15

leaders struggle to answer pressing

2:16

questions like [music] which tools are

2:18

working? How are they being used? What's

2:21

actually driving value? DX provides the

2:23

data and insights that leaders need to

2:25

navigate this shift. With DX, companies

2:28

like Dropbox, Booking.com, Adion, and

2:30

Intercom get a deep understanding of how

2:33

AI is providing value to their

2:34

developers and what impact AI is having

2:37

on engineering productivity. To learn

2:39

more, visit DX's website at

2:41

getdx.com/lenny.

2:44

That's getdx.com/lenny.

2:48

Applications break in all kinds of ways.

2:50

Crashes, slowdowns, regressions, and the

2:53

stuff that you only see once real users

2:55

show up. Sentry catches it all. See what

2:58

happened where, and why, [music] down to

3:00

the commit that introduced the error,

3:01

the developer who shipped it, and the

3:04

exact line [music] of code all in one

3:06

connected view. I've definitely tried

3:08

the five tabs [music] and Slack thread

3:10

approach to debugging. This is better.

3:13

Sentry shows you how the request moved,

3:15

what ran, what slowed down, and what

3:18

users saw. Seir, Sentry's AI debugging

3:21

agent, takes it from there. It uses all

3:23

of that Sentry context to tell you the

3:25

root cause, suggest a fix, and even

3:27

opens a PR for you. It also reviews your

3:29

PR and [music] flags any breaking

3:31

changes with fixes ready to go. Try

3:33

Sentry and SER [music] for free at

3:36

centry.io/lenny

3:38

and use code Lenny for $100 in Sentry

3:40

credits. That's s n t r y.io/lenny.

3:49

Boris, thank you so much for being here

3:52

and welcome to the podcast. [music]

3:54

>> Yeah, thanks for having me on.

3:55

>> I want to start with a a spicy question.

3:58

About 6 months ago, I don't know if

3:59

people even remember this, you actually

4:01

left Anthropic. You joined Curser and

4:05

then two weeks later, you went back to

4:07

Anthropic. What happened there? I don't

4:09

think I've ever heard the actual story.

4:11

>> [laughter]

4:12

>> It's the fastest job change that I've

4:13

ever had.

4:17

Um, I joined Cursor because I'm a big

4:19

fan of the product and honestly I met

4:22

the team and I was just really

4:23

impressed. Uh, they're an awesome team.

4:25

Uh, I still I still think they're

4:27

awesome and they're just building really

4:28

cool stuff and kind of they they saw

4:30

where AI coding was going I think before

4:31

a lot of people did. So the idea of

4:34

building good product was just very

4:35

exciting for me. I think as soon as I

4:38

got there, what I started to realize is

4:40

what I really missed about Ant was the

4:42

mission. And that's actually what

4:44

originally drove me to Ant also cuz uh

4:48

but before I joined Anthropic, I was,

4:49

you know, I was working in big tech and

4:50

then I was at some point I wanted to

4:52

work at a at a lab to just help shape

4:55

the future of this crazy thing that that

4:58

we're building in some way. And the

5:00

thing that drew me to anthropic was the

5:01

mission. And it was, you know, it's all

5:02

about safety. And when you talk to

5:04

people at Enthropic, just like find

5:06

someone in the hallway, if you ask them

5:08

why they're here, the answer is always

5:10

going to be safety. Um, and so this kind

5:13

of like missiondrivenness just really

5:14

really resonated with me. And I just

5:16

know personally it's something I need in

5:18

order to be happy. Um, and I that's just

5:22

a thing that I really missed. And I

5:23

found that, you know, whatever the work

5:25

might be, no matter how exciting, even

5:27

if it's building a really cool product,

5:28

it's just not really a substitute for

5:29

that. Um, so for me it was actually u it

5:33

was pretty obvious that that I was

5:34

missing that pretty quick.

5:35

>> Okay. So let me follow the thread of

5:37

just coming back to anthropic and the

5:39

work you've done there. This podcast is

5:41

going to come out around the year

5:42

anniversary of launching cloud code. So

5:45

I'm going to spend a little time just

5:46

reflecting on the impact that you've

5:49

had. There's um this report that

5:51

recently came out that I'm sure you saw

5:53

by semi analysis that showed that 4% of

5:55

all GitHub commits are authored by cloud

5:58

code now. and they predicted it'll be a

6:00

fifth of all code commits on GitHub by

6:03

the end of the year. The way they put it

6:05

is while we blinked, AI consumed all

6:07

software development.

6:10

The day that we're recording this,

6:11

Spotify just put out this uh headline

6:13

that their best developers haven't

6:14

written a line of code since December

6:17

thanks to AI. More and more of the most

6:20

advanced senior engineers, including

6:22

you, are sharing the fact that you don't

6:24

write code anymore, that it's all AI

6:26

generated. and many aren't even looking

6:28

at code anymore is how far we've gotten

6:31

in large part thanks to this little

6:33

project that you started and that your

6:34

team has scaled over the past year. I'm

6:37

curious just to hear your reflections on

6:39

on this past year and the impact that

6:41

your work has had. These numbers are

6:43

just totally crazy, right? Like four 4%

6:45

of all commits in the world is just way

6:48

more than I imagined and like like you

6:49

said, it still feels like the starting

6:51

point. Um these are also just public

6:53

commits. So we actually think if you

解锁更多

免费注册以访问高级功能

互动查看器

观看带有同步字幕、可调节叠加层和完整播放控制的视频。

免费注册以解锁

AI 摘要

获取由 AI 立即生成的视频内容摘要、要点和结论。

免费注册以解锁

翻译

一键将字幕翻译成 100 多种语言。以任何格式下载。

免费注册以解锁

思维导图

将字幕可视化为交互式思维导图。一目了然地了解结构。

免费注册以解锁

与字幕聊天

提出关于视频内容的问题。直接从字幕中获取由 AI 驱动的答案。

免费注册以解锁

从您的字幕中获得更多

免费注册并解锁交互式查看器、AI 摘要、翻译、思维导图等。无需信用卡。