Benchling cuts migration time by 75% with cloud agents

Watch on YouTube

Now Playing

Transcript

60 segments

0:03

Hi, I'm Emily. I am a software engineer

0:07

on the developer productivity team at

0:09

Benchling. We are really trying to

0:11

advance the pace of life sciences

0:14

development. We really have this culture

0:17

of experimentation, especially with AI.

0:19

Cursor really stood out right away

0:22

because of its customizability and

0:25

extensibility as compared to some of the

0:28

other tools. We have a very complex

0:30

developer environment. Think lots of

0:32

requirements, custom Docker files.

0:34

That's all running locally. But,

0:37

cloud agents [music] in particular, this

0:39

is what we're most excited about, we're

0:41

able to lift that entire local

0:43

development up into the cloud. We have

0:45

thousands of old enzyme tests in our

0:48

code base. We need to migrate them to

0:50

RTL, and it's blocking a major React

0:52

upgrade for us. It was moving very

0:54

slowly. We were quoting about a year to

0:57

get this entire [music] migration done.

0:59

But, what we've tried out in the past

1:01

quarter,

1:03

is we've rolled out cloud agents. Let's

1:05

orchestrate hundreds of agents that we

1:07

spin off and do migrations for us. Now,

1:11

the scale and the rate at which we're

1:13

converting our RTL tests

1:17

has [snorts] increased significantly

1:19

to the point we are quoting about 1

1:21

quarter to finish [music] this rather

1:23

than an entire year. We've really sort

1:25

of built up a whole developer experience

1:28

around cloud agents. 30% of our merged

1:32

PRs are actually from cloud agents now.

1:34

We're able to support that with skills

1:37

and context in order to match the

1:39

quality we still expect. Cursor is

1:41

really helping advance a lot of the

1:44

speed and the quality piece. And the

1:46

composer models have been very cost

1:49

efficient in that regard, and still

1:51

giving us a great output. [music] We

1:54

actually give developers the option to

1:56

choose out of

1:58

a set of AI tools and to choose out of

2:03

most available models. [music] We have

2:05

seen that people have gravitated towards

2:08

cursor because of its cloud agent

2:11

capability. Each quarter we run a

2:13

development survey. Last quarter we

2:16

rolled out cursor cloud agents. What we

2:18

saw from last quarter to this quarter

2:20

was a 20% jump [music] in AI developer

2:24

satisfaction. We also saw that 95% of

2:27

our developers are satisfied with our AI

2:29

tooling right now.

2:33

>> [music]

Interactive Summary

Ask follow-up questions or revisit key timestamps.

Emily, a software engineer at Benchling, explains how the team improved developer productivity and accelerated a massive code migration by implementing Cursor's cloud agents. By automating the migration of enzyme tests to RTL, they reduced the estimated project timeline from one year to just one quarter, while significantly increasing developer satisfaction.

Recently Distilled

Videos recently processed by our community