Latest / The Pragmatic Engineer / The present, past and future of GitHub

Transcript

The present, past and future of GitHub · The Pragmatic Engineer

0:01
SPEAKER_00: So GPT-3 came out right after the build conference in 2020.
0:04
Kevin Scott and Sam Altman did a session about transformers and
0:07
Dutch language models.
0:08
SPEAKER_00: And then after that,
0:08
GPT-3 got into the preview and we got access to that through
0:12
the OpenAI-Microsoft
0:13
SPEAKER_00: partnership.
0:14
And we realized together with OpenAI that it was able to write
0:17
decent code in
0:18
different programming languages and would not mix up the syntax
0:21
between Python,
0:21
SPEAKER_00: Ruby,
0:22
SPEAKER_00: and JavaScript.
0:23
And then OpenAI fine-tuned a model that was called Codex that
0:26
was specific for
0:27
SPEAKER_00: these coding scenarios.
0:28
SPEAKER_00: And so in 2020,
0:30
SPEAKER_00: And in August,
0:30
SPEAKER_00: we wrote a paper with three ideas we had,
0:32
SPEAKER_00: text-to-code,
0:33
SPEAKER_00: code-to-text as in describing code,
0:35
SPEAKER_00: and conversational coding is what we called it,
0:36
SPEAKER_00: which then today is known as chat.
0:38
SPEAKER_00: And those two later scenarios didn't work well enough.
0:40
SPEAKER_00: But text-to-code,
0:41
as in prompting the model within the editor and ultimately
0:44
building autocompletion,
0:45
worked so well that very quickly we saw our internal hubbers
0:49
adopting the tool
0:50
giving it really high scores saying this is great I want to
0:52
keep using this it's
0:53
not the typical management says you have to use it and you
0:56
don't want to but
0:57
ultimately writing in the early days 25% of the code in those
1:00
files it was enabled
1:01
and then shortly thereafter that number got into the 50% range
1:04
or 46% I think early
1:05
SPEAKER_00: 2023
1:07
SPEAKER_00: And so that was the early days of Copilot.
1:09
SPEAKER_00: And then June 2021, we went into the public preview.
1:11
SPEAKER_00: And within a few months, this had grown to a million users.
1:14
SPEAKER_00: And we saw more and more folks on social media saying,
1:17
SPEAKER_00: well,
1:17
SPEAKER_00: I was skeptical that this could ever work.
1:19
But it actually is good enough that I don't want to work
1:21
without it anymore.
1:22
GitHub recently turned 17 years old,
1:24
but how did it start and how did the platform evolve?
1:27
SPEAKER_01: I sat down with GitHub CEO Thomas Zumke,
1:30
SPEAKER_01: who has been a GitHub user for 16 years,
1:32
SPEAKER_01: has been working at GitHub for seven years,
1:34
SPEAKER_01: and has been the CEO of the company for the last four years since 2021.
1:38
SPEAKER_01: In today's episode,
1:39
we discuss GitHub's remote-first and async-first engineering
1:42
culture and why the
1:43
SPEAKER_01: company barely uses email,
1:45
SPEAKER_01: why GitHub seemingly stops shipping features between 2015 and 2020,
1:49
SPEAKER_01: and how they created GitHub Copilot in 2021.
1:52
GitHub's interesting internal tools in the early days,
1:55
like Haystack,
1:56
GitHub TV,
1:56
and HALP.
1:57
SPEAKER_01: How GitHub sees about 10 billion API requests per day,
2:01
SPEAKER_01: that's about 120,000 per second,
2:03
SPEAKER_01: and why Thomas doesn't believe that autonomous AI tools would solve
2:07
SPEAKER_01: hard-engineering challenges.
2:09
SPEAKER_01: and many more topics.
2:10
SPEAKER_01: If you're interested in the past,
2:11
SPEAKER_01: present,
2:12
SPEAKER_01: or future of a high-grilled company like GitHub,
2:14
SPEAKER_01: this episode is for you.
2:16
If you enjoy the show,
2:16
please subscribe to the podcast on any podcast platform and on
2:19
YouTube.
2:20
SPEAKER_01: So, Thomas, welcome to the podcast.
2:22
SPEAKER_01: Thank you so much for having me.
2:24
It's great to meet a person after exchanging emails and putting
2:28
a name to the face.
2:29
SPEAKER_00: And reading your newsletter every week.
2:31
SPEAKER_00: Twice a week, I guess.
2:33
SPEAKER_01: I appreciate that.
2:34
SPEAKER_01: That's where interaction started.
2:35
SPEAKER_01: Before we jump into the history,
2:37
I just wanted to go through a couple of just interesting things
2:40
that people might
2:41
SPEAKER_01: or might not know about GitHub.
2:43
SPEAKER_01: Number one is a tech stack.
2:44
It is pretty commonly known or it used to be that this was a
2:48
Ruby on Rails monolith.
2:49
SPEAKER_01: At some point,
2:49
it might have been the biggest in the world and then I think
2:52
Shopify might have
2:53
SPEAKER_01: come.
2:54
SPEAKER_01: But today, is it still a Ruby on Rails monolith or have things changed?
2:59
It's still Ruby on Rails,
3:00
and I think it's still one of the largest Ruby on Rails
3:03
applications.
3:03
SPEAKER_00: The other one you didn't mention was Twitter.
3:05
SPEAKER_00: It was in the early days also Ruby on Rails.
3:07
SPEAKER_00: In the early days, yeah.
3:08
I think all three companies in one form or another have
3:11
somewhat moved away from
3:13
SPEAKER_00: this being the only part of the stack.
3:14
So we still have a large monolith with about 700 engineers
3:19
within the company.
3:21
SPEAKER_00: contributing to it, you know, at different times of the year.
3:24
SPEAKER_00: Not all of them are working constantly in the monolith.
3:26
SPEAKER_00: We actually just passed 2 million Git commits into the monolith and,
3:31
SPEAKER_00: you know,
3:31
SPEAKER_00: tens of thousands of pull requests.
3:33
SPEAKER_00: But we also have moved beyond just that one architecture.
3:38
SPEAKER_00: So, for example, we're using more and more React on the front end.
3:42
SPEAKER_00: We have a number of, you know, services that are outside the monolith.
3:46
SPEAKER_00: For example, the Copilot API is written in Go,
3:50
SPEAKER_00: API calls for inference.
3:52
SPEAKER_00: The GitHub Actions sits on a completely different tech stack.
3:55
SPEAKER_00: GitHub Actions.
3:57
SPEAKER_01: I would imagine Rails might not be the best for a very heavy workflow.
4:00
And it has actually a history because a large part of the
4:05
action stack came from
4:06
SPEAKER_00: what was back then
4:08
Visual Studio Team Services and then became Azure DevOps and so
4:11
Azure Pipelines.
4:13
SPEAKER_00: And so there was a lot of .NET code in that code base.
4:16
SPEAKER_00: And over time,
4:17
we are now evolving that into a more modern architecture with
4:20
the goal of one of
4:22
SPEAKER_00: the other posts of getting to four nines and beyond.
4:25
SPEAKER_00: And so our tech stack is very diverse today.
4:28
SPEAKER_00: Obviously, we're doing also Swift for our iPhone app.
4:30
SPEAKER_00: We have Kotlin for our Android app.
4:32
SPEAKER_00: We're using all the clouds beyond also having our own
4:38
SPEAKER_00: metal in commercial data centers.
4:40
And we have like a modern software company with all the
4:44
complexity and all the
4:45
SPEAKER_00: challenges that everybody else has as well.
4:48
SPEAKER_01: And I guess all the not simple trade-offs,
4:50
SPEAKER_01: for example,
4:50
SPEAKER_01: it's easy to point,
4:51
SPEAKER_01: you know,
4:51
SPEAKER_01: the community loves to say Rails does not scale or why would you do it?
4:54
SPEAKER_01: But I mean,
4:55
SPEAKER_01: you are one example,
4:56
obviously there's other companies we mentioned that it feels
4:58
like it's less about
4:59
SPEAKER_01: the technology itself,
5:01
SPEAKER_01: but what you're building on it and,
5:02
SPEAKER_01: you know,
5:03
SPEAKER_01: your history of it and then
5:04
SPEAKER_00: I mean,
5:05
we can spend a whole hour probably just talking about the
5:07
benefits of having a
5:09
SPEAKER_00: monorepo with a monolith versus hundreds of microservices.
5:15
SPEAKER_00: When GitHub was,
5:16
SPEAKER_00: in 2007,
5:17
SPEAKER_00: was founded for,
5:18
SPEAKER_00: I think,
5:18
SPEAKER_00: more than a year,
5:19
SPEAKER_00: it was only the three original founders,
5:21
SPEAKER_00: Chris,
5:22
SPEAKER_00: PJ,
5:22
SPEAKER_00: and Tom,
5:23
SPEAKER_00: plus Scott,
5:25
SPEAKER_00: who's now the
5:27
SPEAKER_00: founder of Git Butler and Scott Charconnell,
5:29
who is kind of like the first employee that was often seen as
5:32
the fourth founder.
5:33
And so,
5:33
you know,
5:34
a team of four is way better off of working in a single code
5:37
base.
5:38
SPEAKER_00: You can move much faster.
5:39
I think that was always important for GitHub to be able to move
5:44
really fast and
5:45
SPEAKER_00: ship new features.
5:45
SPEAKER_00: And it's much easier to,
5:46
SPEAKER_00: you know,
5:46
SPEAKER_00: get to learn the code base and try to understand the dependencies.
5:51
SPEAKER_00: As the company scales and the product scales,
5:53
SPEAKER_00: it makes more sense to move away from that architecture.
5:56
SPEAKER_01: I'm so glad you're saying this.
5:57
The other day I was talking with an early Amazon employee and
6:00
he was saying the
6:00
SPEAKER_01: same thing,
6:00
SPEAKER_01: how early Amazon had made sense.
6:02
SPEAKER_01: And then as you grow, you know, you figure out your trade-offs.
6:05
SPEAKER_01: If you want to build a great product, you have to ship quickly.
6:08
SPEAKER_01: But how do you know what works?
6:10
SPEAKER_01: More importantly, how do you avoid shipping things that don't work?
6:14
SPEAKER_01: The answer, Statsig.
6:16
SPEAKER_01: Statsig is a unified platform for flags,
6:19
SPEAKER_01: analytics,
6:19
SPEAKER_01: experiments,
6:20
SPEAKER_01: and more,
6:21
combining five plus products into a single platform with a
6:24
unified set of data.
6:26
SPEAKER_01: Here's how it works.
6:27
SPEAKER_01: First, StatSick helps you ship a feature via FeatureFly or Config.
6:31
SPEAKER_01: Then it measures how it's working,
6:34
SPEAKER_01: from alerts and errors,
6:35
SPEAKER_01: to replays of people using that feature,
6:37
SPEAKER_01: to measurement of top-line impact.
6:39
SPEAKER_01: Then you get your analytics,
6:41
SPEAKER_01: user account metrics,
6:42
SPEAKER_01: and dashboards to track your progress over time,
6:44
SPEAKER_01: all linked to the stuff you ship.
6:46
SPEAKER_01: Even better,
6:47
SPEAKER_01: Statsik is incredibly affordable with a super generous free tier,
6:50
SPEAKER_01: a starter program with $50,000 of free credits,
6:53
SPEAKER_01: and custom plans to help you consolidate your existing spend on flags,
6:56
SPEAKER_01: analytics,
6:57
SPEAKER_01: or A-B testing tools.
6:58
SPEAKER_01: To get started, go to statsik.com slash pragmatic.
7:02
SPEAKER_01: That is S-T-A-T-S-I-G dot com slash pragmatic.
7:06
SPEAKER_01: Happy building.
7:07
SPEAKER_01: This episode was brought to you by Graphite,
7:09
SPEAKER_01: the developer productivity platform that helps developers create,
7:12
SPEAKER_01: review,
7:12
SPEAKER_01: and merge smaller code changes,
7:14
SPEAKER_01: stay unblocked,
7:15
SPEAKER_01: and ship faster.
7:17
SPEAKER_01: Code review is a huge time sink for engineering teams.
7:20
Most developers spend about a day per week or more reviewing
7:23
code or blocked
7:24
SPEAKER_01: waiting for a review.
7:26
SPEAKER_01: It doesn't have to be this way.
7:28
SPEAKER_01: Graphite brings stack pull requests,
7:30
the workflow at the heart of the best-in-class internal code
7:32
review tools at
7:33
SPEAKER_01: companies like Meta and Google,
7:34
SPEAKER_01: to every software company on GitHub.
7:37
SPEAKER_01: Graphite also leverages high-signal,
7:39
codebase-aware AI to give developers immediate actionable
7:42
feedback on their pull
7:43
SPEAKER_01: requests,
7:44
SPEAKER_01: allowing teams to cut down on review cycles.
7:47
SPEAKER_01: Tens of thousands of developers at top companies like Asana,
7:50
SPEAKER_01: Ramp,
7:50
SPEAKER_01: Tekton,
7:51
SPEAKER_01: and Vercel rely on Graphite every day.
7:53
Start stacking with Graphite today for free and reduce your
7:56
time to merge from days to hours.
7:58
SPEAKER_01: Get started at gt.dev slash pragmatic.
8:02
SPEAKER_01: That is g for graphite, t for technology.dev slash pragmatic.
8:06
SPEAKER_01: Now,
8:06
SPEAKER_01: you mentioned something really interesting,
8:08
SPEAKER_01: which I also heard,
8:09
that you run your own data centers or historically have run
8:11
your own data centers.
8:12
But I talked with some current GitHub folks,
8:14
and they said that you,
8:15
in some parts,
8:16
still do so.
8:17
SPEAKER_01: Why did you do this?
8:19
SPEAKER_01: How is it changing?
8:20
SPEAKER_01: And are you,
8:21
SPEAKER_01: obviously,
8:21
SPEAKER_01: since you're now part of Microsoft as well,
8:23
SPEAKER_01: does it make sense to use some parts of Azure?
8:25
SPEAKER_01: And if so, what parts are you using?
8:27
SPEAKER_01: And you also mentioned other clouds.
8:28
SPEAKER_01: So how are you thinking of own data center, which, again,
8:31
We used to say,
8:32
oh,
8:32
it's just moved to the cloud and there's now an opposite thing
8:35
as well.
8:36
SPEAKER_01: What's your take on this?
8:36
SPEAKER_00: GitHub started on the cloud.
8:40
SPEAKER_00: Really?
8:40
SPEAKER_00: In the early days,
8:41
SPEAKER_00: it was,
8:42
SPEAKER_00: to my memory,
8:43
it was engineered as a platform as a service provider and then
8:47
AWS behind the
8:49
SPEAKER_00: scenes.
8:50
SPEAKER_00: Oh, okay.
8:51
It's easy to forget in today's world where lots of startups
8:55
announce huge funding rounds.
8:56
SPEAKER_00: GitHub had no funding.
8:58
SPEAKER_00: until I think five years into its journey.
9:01
Back then,
9:03
unheard of 100 million came from around,
9:06
led by Andreessen Horowitz.
9:07
SPEAKER_00: But until that point in time, GitHub was bootstrapped.
9:10
SPEAKER_00: And it had very fast adoption.
9:13
SPEAKER_00: When it launched in April 2008,
9:15
SPEAKER_00: very quickly,
9:16
SPEAKER_00: the number of users,
9:17
SPEAKER_00: the number of both public and private repositories
9:20
SPEAKER_00: kept going.
9:21
You can still find people talking about that they're begging
9:24
the founders to
9:25
introduce a paid tier so they have the confidence that the
9:28
company can survive.
9:29
SPEAKER_00: Will be around, yes.
9:30
SPEAKER_00: Will be around.
9:31
SPEAKER_00: And so it was a natural business decision for the founders to say,
9:35
how can we host this with a more optimized cost given that we
9:39
are bootstrapping
9:40
SPEAKER_00: this and we're having to pay for the service?
9:42
SPEAKER_00: from our own revenue stream.
9:45
And they couldn't be burning money because they had no money
9:49
from venture capitalists.
9:50
SPEAKER_00: And so I think that was part of starting the decision.
9:52
SPEAKER_00: The other one is storing Git data.
9:55
What we call today our Git systems team in file storage in
10:00
itself is probably not
10:01
SPEAKER_00: the best use case for the cloud that existed.
10:05
SPEAKER_00: you know,
10:06
SPEAKER_00: in 2008,
10:07
SPEAKER_00: 2009,
10:07
SPEAKER_00: where you had limited access to that layout and where,
10:11
SPEAKER_00: you know,
10:11
especially the networking part of having multiple nodes
10:15
communicating with each
10:16
SPEAKER_00: other was much more complex than it is today.
10:18
SPEAKER_00: And so the decision was made to move to
10:20
SPEAKER_00: servers in commercial data centers.
10:22
SPEAKER_00: So we own the servers.
10:23
SPEAKER_00: We still own the servers and the storage and all that.
10:26
SPEAKER_00: But the data center is a commercial data center.
10:28
SPEAKER_00: So we don't own the data center.
10:29
SPEAKER_00: So it's owned in the sense of, I'm making air quotes for listeners.
10:32
SPEAKER_00: It's owned in the sense of, we managed the servers in those data centers.
10:36
SPEAKER_00: Now, today's GitHub still is a lot of that.
10:38
SPEAKER_00: The core is in these own data centers,
10:41
SPEAKER_00: but things like GitHub Actions,
10:43
SPEAKER_00: GitHub Codespaces,
10:44
GitHub Copilot run on Azure in the cloud and leverage the scale
10:49
that you naturally
10:49
SPEAKER_00: lead,
10:50
SPEAKER_00: whether it's CPUs and actions or GPUs in Copilot.
10:54
SPEAKER_00: Of course,
10:55
SPEAKER_00: Copilot,
10:55
given its multi-model choice and has a phobic models and OpenAI
10:59
models and Gemini
11:00
SPEAKER_00: models,
11:00
SPEAKER_00: not all of these models are on Azure.
11:02
SPEAKER_00: So naturally, we're also working with other providers.
11:05
SPEAKER_00: often with both the first-party API,
11:08
SPEAKER_00: aka the API that OpenAI or Anthropic provide,
11:11
SPEAKER_00: and other clouds that have vertex bedrock and so on.
11:15
And we,
11:15
in fact,
11:16
now have GitHub fully hosted on Azure for data residencies.
11:19
If you,
11:20
as an enterprise customer in Europe,
11:22
want your data all stored in Europe,
11:24
SPEAKER_00: That's kind of a big deal.
11:25
You're getting a GitHub organization on a stamp that fully runs
11:29
on Azure in the
11:30
SPEAKER_00: Netherlands and in Sweden.
11:32
SPEAKER_00: And we have the same for Australia.
11:33
SPEAKER_00: We've just announced a US stamp that will be FedRAM compliant.
11:37
SPEAKER_00: And we're going to expand into more regions in the years to come.
11:41
SPEAKER_01: Yeah,
11:41
SPEAKER_01: I feel these are the really kind of boring things,
11:44
but they start to become really important when you are at a
11:47
company that is now
11:48
SPEAKER_01: starting to care about these things.
11:49
SPEAKER_00: Well, they're not boring from an interesting perspective.
11:51
SPEAKER_00: Well,
11:52
SPEAKER_01: not boring,
11:53
SPEAKER_01: but it's a thing that I think from outside,
11:56
SPEAKER_01: you know,
11:57
SPEAKER_01: it's not the shiny features,
11:58
SPEAKER_01: but I agree.
11:58
SPEAKER_01: An interesting perspective, that actually sounds pretty challenging.
12:02
SPEAKER_00: The challenge,
12:02
SPEAKER_00: you know,
12:03
SPEAKER_00: if you go back to the Rubio and Rails conversation is,
12:06
SPEAKER_00: and if you know a little bit about Rails and how,
12:08
SPEAKER_00: and we do not only run on Rails,
12:10
SPEAKER_00: we also run on MySQL.
12:11
SPEAKER_00: And so whenever you change,
12:12
SPEAKER_00: you know,
12:13
SPEAKER_00: the schema of a database table,
12:14
SPEAKER_00: a so-called migration is run.
12:16
SPEAKER_01: Yep.
12:17
SPEAKER_00: How do you do that if you not only have one stem in one region,
12:20
SPEAKER_00: but you actually have five or 10 stems?
12:23
SPEAKER_00: And we deploy hundreds of times per day.
12:26
SPEAKER_00: In fact,
12:26
SPEAKER_00: I think this year,
12:27
we're forecasting 12 million deploys across about 1,000 or so
12:32
IC engineers that we
12:34
SPEAKER_00: have in the company.
12:34
SPEAKER_00: And so how do you do that?
12:35
SPEAKER_00: How do you deploy,
12:36
SPEAKER_00: how do you go from one,
12:37
SPEAKER_00: you know,
12:37
SPEAKER_00: a central stamp on our US data center cloud,
12:41
SPEAKER_00: if you will,
12:42
SPEAKER_00: to having,
12:43
SPEAKER_00: you know,
12:43
deployment that goes to all these different stamps and run the
12:46
migrations and if
12:47
SPEAKER_00: you need to,
12:48
SPEAKER_00: are able to hold back.
12:49
SPEAKER_00: So that was a lot of the engineering work that went into that.
12:52
SPEAKER_00: Less,
12:53
SPEAKER_00: so I'd say,
12:54
the actual work of moving this to Azure was less work than
12:57
getting the engineering
12:58
SPEAKER_00: system to DX,
12:59
SPEAKER_00: the developer experience,
13:00
SPEAKER_00: actually aligned with this.
13:01
SPEAKER_00: Now we are running in multiple regions.
13:04
Yeah one interesting thing that I always surprise people who
13:08
don't know GitHub all
13:09
that well is people think well it's now Microsoft right so it
13:12
must work the same
13:13
way as Microsoft and then you know one like fact that shows
13:16
this is not the case
13:17
SPEAKER_01: GitHub is still remote first it started as remote first
13:20
SPEAKER_01: initially,
13:20
SPEAKER_01: and it was actually one of the few companies that even before COVID,
13:24
SPEAKER_01: truly remote first.
13:26
SPEAKER_01: And as I understand, even today, it's remote first.
13:29
SPEAKER_01: Tell me about this on how you managed to keep this,
13:33
SPEAKER_01: how it's working inside a larger company,
13:35
SPEAKER_01: which Microsoft is not remote first.
13:37
SPEAKER_01: There are,
13:37
SPEAKER_01: I understand,
13:38
SPEAKER_01: parts of it that are more supportive of it,
13:39
SPEAKER_01: but generally,
13:40
SPEAKER_01: people are often in the office.
13:42
SPEAKER_00: So the three founders, they actually met in San Francisco.
13:44
So you could argue that very,
13:45
very early phase was kind of like three people in the same
13:48
place.
13:49
SPEAKER_00: And that's often how that goes.
13:52
SPEAKER_00: But yeah,
13:52
SPEAKER_00: very quickly,
13:53
SPEAKER_00: GitHub started hiring super fans,
13:56
SPEAKER_00: people that were interested in GitHub,
13:58
SPEAKER_00: promoting GitHub,
13:59
SPEAKER_00: helped drive in the communities.
14:00
SPEAKER_00: Those people got hired.
14:02
SPEAKER_00: into GitHub and they were naturally everywhere in the world.
14:05
SPEAKER_00: And so very quickly,
14:06
GitHub moved from a culture that was headquarter centric to a
14:10
remote first culture.
14:12
SPEAKER_00: And that gave us an advantage when the pandemic came, COVID.
14:17
SPEAKER_00: Because we were already used on being on video calls, on using Slack.
14:22
SPEAKER_00: In fact,
14:22
SPEAKER_00: you know,
14:23
I think GitHub is very different than Microsoft when it comes
14:26
to how the company
14:27
SPEAKER_00: communicates.
14:28
SPEAKER_00: My joke always is when I wake up in the morning,
14:29
SPEAKER_00: because I'm obviously both,
14:30
SPEAKER_00: I'm a Microsoft executive and I'm a GitHub employee.
14:34
SPEAKER_00: When I wake up in the morning on the GitHub side,
14:35
SPEAKER_00: I have lots of Slack messages,
14:37
SPEAKER_00: DMs,
14:38
SPEAKER_00: channels to look into.
14:40
SPEAKER_00: And maybe,
14:40
SPEAKER_00: you know,
14:41
a handful of customer emails or emails from you that can
14:44
communicate with me in
14:44
SPEAKER_00: Slack.
14:45
SPEAKER_00: On the Microsoft side, it's the other way around.
14:46
I have dozens of emails to pay attention to and maybe a few
14:51
Teams messages.
14:52
And so I think that shows the async part of our culture that
14:56
has grown over the last 17 years.
14:59
SPEAKER_00: And it is so important at GitHub.
15:01
SPEAKER_00: We are using GitHub for everything.
15:04
SPEAKER_00: And so all employees across all the functions,
15:06
SPEAKER_00: not just engineering and product,
15:07
SPEAKER_00: but HR and comms and finance,
15:10
SPEAKER_00: all the functions work on GitHub.
15:12
SPEAKER_00: And so they have repositories to describe their team.
15:14
You know,
15:15
they're using pull requests to make changes,
15:17
for example,
15:17
to our terms of service.
15:20
SPEAKER_00: You know,
15:20
every company-wide announcement is a pull request against the
15:23
repository that we
15:24
call the hub that gets published as a pages page that is only
15:29
internally accessible
15:30
SPEAKER_00: and that goes to our identity provider.
15:32
SPEAKER_00: And so even there, we don't do company-wide emails, almost never.
15:36
We do an announcement on the hub and there's a link that is
15:39
posted in a Slack
15:40
SPEAKER_00: channel that everybody can see and then we...
15:43
SPEAKER_00: take the conversation there.
15:44
SPEAKER_00: So yeah,
15:44
today we are very much a remote-first company with people all
15:48
around the world in
15:49
SPEAKER_00: different time zones.
15:51
SPEAKER_00: Almost everything is async.
15:52
Obviously,
15:53
there's meetings,
15:54
and our town hall meeting is called the get-together.
15:56
SPEAKER_00: And so that has to be in some time zone.
16:00
SPEAKER_00: And the challenge there is that if you do it,
16:04
SPEAKER_00: adjusting it for the folks in the U.S., North America, and Europe.
16:08
We are missing out those in Australia and India because that's
16:11
just impossible to meet.
16:12
SPEAKER_00: So we try,
16:13
SPEAKER_00: you know,
16:14
different times of a year to have one that is in the
16:16
APEC-friendly time zone and,
16:18
SPEAKER_00: you know,
16:18
SPEAKER_00: a few that are European-friendly.
16:20
Often,
16:20
you know,
16:20
it's not the way when you do like one early in the morning on
16:23
the West Coast.
16:24
That's actually bad time for the Europeans because that's kind
16:26
of like when you
16:27
SPEAKER_00: pick up your kids from school and go and have family dinner.
16:30
SPEAKER_00: So many of them actually prefer a later
16:33
point on the day on the west coast so it is later in the
16:36
evening when maybe
16:37
everybody's watching tv or the kids are already in bed so yeah
16:39
we're trying to
16:40
really keep that spirit and it enables us to hire talent um you
16:44
know in different
16:45
places um in different uh phases of their life and i think that
16:49
helps us as a
16:50
SPEAKER_00: company you know to be as agile as possible
16:53
SPEAKER_01: Yeah,
16:53
SPEAKER_01: it is pretty cool to hear because GitHub,
16:55
SPEAKER_01: I mean,
16:56
SPEAKER_01: you can use GitHub in many ways,
16:57
SPEAKER_01: but like one popular way is to use it at this async way,
17:00
SPEAKER_01: right?
17:00
So it's kind of cool that you're using it the same way as a lot
17:04
of the startups are.
17:04
SPEAKER_00: I think one of the core pieces there is not only as the startups are,
17:08
SPEAKER_00: but as the open source ecosystem is using it,
17:10
SPEAKER_00: right?
17:10
SPEAKER_00: But the open source ecosystem by design is a
17:13
SPEAKER_00: Async, yes.
17:14
SPEAKER_00: And they're not all in the same place.
17:16
That would never work if an open source project could only be a
17:20
contributor if
17:21
SPEAKER_00: you're in the same place and all the same time zone.
17:24
And so given that we're catering to both commercial customers,
17:28
startups,
17:28
enterprises,
17:30
those that are paying us money for GitHub Enterprise and
17:32
corporate business,
17:33
SPEAKER_00: but we're also catering,
17:34
SPEAKER_00: serving the open source economy,
17:37
SPEAKER_00: the open source ecosystem,
17:38
SPEAKER_00: and as such as a company,
17:39
SPEAKER_00: even though the GitHub core itself was never open source,
17:43
SPEAKER_00: and we're operating pretty much like an open source project.
17:46
One fun fact I've heard is you have some really kind of fun and
17:50
interesting internal tools.
17:52
SPEAKER_01: Some are still here, some are not.
17:54
SPEAKER_01: One tool that I understand is no longer there was called Haystack,
17:57
SPEAKER_01: which was an internal exception tracking app.
18:00
SPEAKER_01: And someone told me it was the best piece of software they've seen.
18:02
Can you share a little bit about what past fun internal tools
18:06
you have and what
18:07
SPEAKER_01: kind of internal tools do you still have today,
18:09
SPEAKER_01: especially for what engineers use?
18:11
SPEAKER_00: If you look back into GitHub's history,
18:13
SPEAKER_00: and we don't have the time to cover it all today,
18:16
SPEAKER_00: I can recommend if I'm allowed to another podcast,
18:20
SPEAKER_00: the Acquired podcast.
18:21
SPEAKER_00: They had an episode on June 5th,
18:24
SPEAKER_00: 2018,
18:24
so the day after the acquisition announcement that covers a lot
18:27
of the history of
18:28
SPEAKER_00: GitHub in the early days,
18:30
SPEAKER_00: and in general,
18:31
SPEAKER_00: a really good podcast.
18:34
SPEAKER_00: And they actually mentioned this,
18:35
SPEAKER_00: that GitHub,
18:37
SPEAKER_00: when it started,
18:38
SPEAKER_00: had a very non-traditional way of operating.
18:40
SPEAKER_00: There were no managers.
18:42
SPEAKER_00: Everybody was encouraged to work on their passion projects.
18:46
SPEAKER_00: And as such, GitHub, for many years, didn't have a traditional way
18:50
SPEAKER_00: IT infrastructure company.
18:53
Everybody was built in-house because there was the belief we as
18:58
GitHub can do it
18:58
SPEAKER_00: better than a tool that we can buy off the shelf.
19:01
SPEAKER_00: Maybe we come back to this later when we talk about AI.
19:04
SPEAKER_00: That future might actually come back to some startups.
19:07
SPEAKER_00: So we had a tool with our own internal video streaming platform.
19:12
SPEAKER_00: It was called githubber.tv.
19:14
SPEAKER_00: And so all the town halls,
19:16
SPEAKER_00: And so we have, you know, streamed on our own platform.
19:19
SPEAKER_00: And of course, that no longer exists.
19:21
SPEAKER_00: And we're using Loom for that.
19:24
SPEAKER_00: Haystack, you already mentioned that was exception tracking.
19:27
SPEAKER_00: And now keep in mind,
19:28
SPEAKER_00: 15 years ago,
19:29
SPEAKER_00: there wasn't like established players in that market that you could use,
19:33
SPEAKER_00: especially not in the Ruby Rails work.
19:35
SPEAKER_00: So Haystack was my such tool today.
19:37
SPEAKER_00: It's Sentry that we're using for that.
19:39
SPEAKER_00: Another one was HALP,
19:41
SPEAKER_00: H-A-L-P,
19:42
which was our internal support system where tickets got routed
19:46
for folks to work on
19:47
SPEAKER_00: the ticket,
19:47
SPEAKER_00: have internal communication on the ticket.
19:49
SPEAKER_00: Yeah.
19:50
SPEAKER_00: And now this is all in Zendesk and HALP has been retired.
19:53
SPEAKER_00: But, you know, some of these ideas have led to new startups.
19:58
SPEAKER_00: Hubbers is what we call GitHub employees,
19:59
where Hubbers have taken the idea and maybe they're a bit
20:01
frustrated that we're
20:02
SPEAKER_00: shutting down their internal passion project,
20:05
SPEAKER_00: have taken that and created the startup of it.
20:08
SPEAKER_00: I think,
20:09
SPEAKER_00: you know,
20:09
SPEAKER_00: culturally,
20:10
we are still encouraging folks pretty much to experiment and to
20:13
incubate new ideas,
20:14
SPEAKER_00: but we're trying to make those external-facing products.
20:17
SPEAKER_00: And there's a number of examples from GitHub's history as well there.
20:21
SPEAKER_00: Electron comes to mind, which came out of Atom, which was our editor.
20:26
SPEAKER_00: Oh, yes.
20:26
SPEAKER_00: It's hugely influential.
20:27
SPEAKER_00: Our desktop app or comment line interfaces are all open-source projects,
20:31
SPEAKER_00: so everybody can see what we're doing there and can take the ideas.
20:34
SPEAKER_00: And, you know, we mentioned tech stack earlier.
20:36
SPEAKER_00: The CLI, for example, is with Mongo.
20:38
Whoever wants to learn how to build a CI can use the GitHub CI
20:42
as an example.
20:43
One thing that keeps surprising me about GitHub is every now
20:47
and then there's a
20:48
SPEAKER_01: security incident that is in the press that is a big deal.
20:51
SPEAKER_01: Not always, but sometimes it starts with company X.
20:55
SPEAKER_01: had a security issue.
20:56
And it starts like,
20:57
oh,
20:57
they detected it because GitHub alerted it for them.
21:00
SPEAKER_01: And a very famous one was in 2022, Heroku had a security incident.
21:06
SPEAKER_01: And it started by not Heroku noticing that something leaked,
21:10
SPEAKER_01: but GitHub itself noticed some strange patterns.
21:13
SPEAKER_01: They contacted the Heroku security team who then confirmed it.
21:16
SPEAKER_01: And this just struck me a little bit striking.
21:18
SPEAKER_01: Like, hold on.
21:19
SPEAKER_01: Like, how is...
21:21
SPEAKER_01: GitHub having better or very strong security,
21:24
SPEAKER_01: I'm not going to say better,
21:25
SPEAKER_01: but in some ways in this case,
21:27
SPEAKER_01: clearly they got earlier.
21:29
SPEAKER_01: What is your approach to security?
21:30
SPEAKER_01: And this was, again, this was not even right now, this was years back.
21:34
SPEAKER_01: It just suggests to me,
21:35
SPEAKER_01: like,
21:36
you probably think about security a little bit differently than
21:39
maybe some other
21:40
SPEAKER_00: companies.
21:40
SPEAKER_00: Security is part of our culture.
21:41
SPEAKER_00: We have the saying, again, security is priority zero for everybody.
21:46
SPEAKER_00: Okay.
21:47
SPEAKER_00: And in fact, our previous CISO, Mike Hanley, introduced this concept.
21:51
SPEAKER_00: He came from Duo, the two-factor authentication company.
21:55
He introduced this concept that when the question is asked who
21:57
he is on the
21:58
SPEAKER_00: security team,
21:59
SPEAKER_00: every hubber raises their hand.
22:01
SPEAKER_00: And so we institutionalize that security is part of what we do.
22:05
SPEAKER_00: Of course,
22:05
SPEAKER_00: we want to drive innovation and make customers happy,
22:08
SPEAKER_00: but we cannot be in the market if we lose the trust.
22:10
SPEAKER_00: And of course, the worst thing
22:12
SPEAKER_00: you know,
22:13
SPEAKER_00: in my life would be that I get a call and say,
22:15
SPEAKER_00: I lost,
22:15
SPEAKER_00: you know,
22:16
SPEAKER_00: the private repositories.
22:18
SPEAKER_00: We lost, you know, as a platform.
22:19
SPEAKER_00: It's so important to companies, your code.
22:21
SPEAKER_00: To an attacker and now have to do the damage control.
22:25
And so it's really important for us,
22:26
the CSO,
22:27
the chief security officer reports directly to me.
22:30
And we have a team of about 150 or so employees that take care
22:35
of security from multiple angles.
22:37
SPEAKER_00: One is of course, quite intelligence.
22:40
SPEAKER_00: working closely with the Microsoft security team,
22:43
SPEAKER_00: the so-called Cyber Defense Operations Center,
22:45
SPEAKER_00: and those folks that look into finding security vulnerabilities,
22:50
SPEAKER_00: that teaming,
22:51
SPEAKER_00: not only for the traditional software building,
22:54
SPEAKER_00: but also for the AI pieces.
22:56
And you can imagine at the scale of GitHub,
22:59
but even more so at the scale of Microsoft,
23:01
SPEAKER_00: There's a lot of threat intelligence.
23:02
SPEAKER_00: When you combine all that intelligence, you get a lot of signal.
23:06
And one such example is a very trivial example is that if you
23:10
have a person,
23:11
SPEAKER_00: if you,
23:12
SPEAKER_00: for example,
23:12
SPEAKER_00: your user ID accesses not only the pragmatic engineer repositories,
23:17
SPEAKER_00: but also those of,
23:18
SPEAKER_00: let's say,
23:18
SPEAKER_00: BMW at Mercedes,
23:20
SPEAKER_00: that person is very unlikely to exist in the industry.
23:25
SPEAKER_00: Even if you're working for a systems integrator,
23:29
SPEAKER_00: then you typically don't work on a BMW and a Mercedes project together.
23:32
So there's a lot of intelligence that we can collect by having
23:35
the holistic view,
23:36
SPEAKER_00: even more so when you combine it with all the other Microsoft systems.
23:39
And we also have a security lab where we have a team of
23:42
researchers that are
23:44
hunting for security vulnerabilities using our own security
23:47
products like CodeQL,
23:49
which is query language that you can use to find variations of
23:54
existing code flaws,
23:56
SPEAKER_00: code vulnerabilities.
23:58
tries to find those in open source projects and then reports
24:01
them to the open
24:02
SPEAKER_00: source maintainer in a responsible way,
24:04
SPEAKER_00: helping them to fix those.
24:05
SPEAKER_00: And then in time of disclosure, hopefully, the bug is already fixed.
24:09
SPEAKER_00: So we are doing a lot of that.
24:10
SPEAKER_00: We are investing into a lot of that.
24:12
SPEAKER_00: But of course, nobody is perfect.
24:14
And so it's really crucial for us as a company to have this
24:17
ingrained into our
24:17
SPEAKER_00: culture,
24:18
SPEAKER_00: to not have secrets and source code.
24:19
SPEAKER_00: In fact, we're blocking all the pushes.
24:21
SPEAKER_00: Exactly.
24:23
SPEAKER_01: I guess open source might help here, your big focus on open source.
24:26
SPEAKER_01: But you mentioned 150 people working on security.
24:28
As context,
24:29
how many people or engineers are in GitHub roughly these days?
24:33
SPEAKER_00: Yeah.
24:33
SPEAKER_00: So GitHub all up is about 3,000 employees.
24:37
About half or so are in what I would call engineering product
24:40
design or EPD.
24:43
SPEAKER_01: So that's about 10% if I just look at the security.
24:45
SPEAKER_00: Roughly 10% of their engineering function.
24:48
SPEAKER_00: Wow.
24:50
Engineers,
24:51
I see engineers,
24:52
it's a little bit under a thousand that work and build
24:55
software.
24:55
SPEAKER_00: And then it's,
24:55
SPEAKER_00: of course,
24:56
SPEAKER_00: managers and product managers and technical program managers,
24:58
SPEAKER_00: designers and all of that.
24:59
One interesting thing I've heard from a current GitHub employee
25:02
is you are starting
25:03
SPEAKER_01: to look into hiring junior developers,
25:06
SPEAKER_01: developers with less experience,
25:07
SPEAKER_01: which is a little bit different because most of the industry,
25:10
SPEAKER_01: we don't hear this.
25:11
SPEAKER_01: Yeah.
25:12
SPEAKER_01: why are you deciding on this?
25:14
SPEAKER_01: This is just a really positive thing to hear.
25:16
SPEAKER_01: I'd love to hear more about your thinking.
25:17
SPEAKER_00: We're really excited about working with young people.
25:19
SPEAKER_00: You saw some of those in the second day keynote yesterday as well.
25:24
SPEAKER_00: And in fact, just on Monday, our intern class for the summer started.
25:29
We have three groups that start on different days and go
25:31
through the internship.
25:32
SPEAKER_00: So those are the
25:33
very early in career folks that often,
25:35
or not often,
25:36
that still go to college or university.
25:38
And this is a program that we institutionalized after the
25:43
pandemic was over,
25:44
SPEAKER_00: and you could bring people back in person,
25:46
SPEAKER_00: because obviously,
25:47
SPEAKER_00: or half in person,
25:49
SPEAKER_00: where they at least have,
25:49
SPEAKER_00: you know,
25:50
SPEAKER_00: available.
25:51
to meet with folks at our San Francisco headquarters or here in
25:56
the Pacific
25:57
SPEAKER_00: Northwest or elsewhere in the world.
25:58
SPEAKER_00: And so we have an intern program,
26:00
and often these interns get a full-time offer at the end of the
26:04
internship,
26:04
SPEAKER_00: or maybe they come back for a second year.
26:07
SPEAKER_00: And so it's A,
26:07
SPEAKER_00: lovely to see that those folks that bring fresh ideas,
26:12
SPEAKER_00: a great amount of energy,
26:14
SPEAKER_00: the latest learnings from college and university,
26:18
SPEAKER_00: and often,
26:18
SPEAKER_00: you know,
26:19
SPEAKER_00: a different,
26:19
SPEAKER_00: you know,
26:20
SPEAKER_00: diverse background into the company and,
26:23
SPEAKER_00: you know,
26:24
SPEAKER_00: come in and often,
26:25
SPEAKER_00: you know,
26:26
SPEAKER_00: the folks,
26:27
SPEAKER_00: you know,
26:27
SPEAKER_00: that are younger in Korea bring a new perspective to the team and say,
26:32
SPEAKER_00: here,
26:33
SPEAKER_00: this is,
26:33
SPEAKER_00: you know,
26:34
SPEAKER_00: why don't we try this or I want to incubate this idea.
26:36
SPEAKER_00: And so we are excited about having a
26:40
SPEAKER_00: this kind of like both junior and senior population in the company.
26:43
SPEAKER_00: And it's not only true in engineering.
26:44
SPEAKER_00: You know, it's the same if you look into sales or into marketing.
26:48
SPEAKER_00: Obviously,
26:48
SPEAKER_00: the way,
26:49
SPEAKER_00: you know,
26:49
SPEAKER_00: marketing works today is very different than it was,
26:53
SPEAKER_00: you know,
26:54
SPEAKER_00: five,
26:54
SPEAKER_00: 10 years ago when the press,
26:56
SPEAKER_00: you know,
26:56
SPEAKER_00: traditional media played a much bigger role.
26:59
Today,
27:00
you've got to be,
27:01
you know,
27:01
in a podcast and you have to be on YouTube and TikTok.
27:05
SPEAKER_00: Like today.
27:06
SPEAKER_00: Just like today, but you see more and more of that happening.
27:10
And of course,
27:10
when you hire people earlier in Korea,
27:14
they give you those ideas.
27:15
SPEAKER_00: They give you that input.
27:17
SPEAKER_00: And you know how it is.
27:18
SPEAKER_00: You talk to a lot of people in companies.
27:20
SPEAKER_00: There's always a debate.
27:22
There's always a debate between those that are defending how
27:25
we've always done it
27:27
SPEAKER_00: in a 50-year-old company like Microsoft that obviously exists.
27:30
SPEAKER_00: And those that come in and say, hey, nobody in my peer group
27:35
SPEAKER_00: watches TV anymore or CNBC or whatever.
27:37
SPEAKER_00: And they're all watching just three-minute videos on TikTok.
27:42
SPEAKER_00: And so I think that's really important.
27:44
And we're trying to have a nice balance between early career
27:48
engineers,
27:49
senior engineers.
27:50
We have a number of distinguished engineers,
27:52
and they all contribute back to our planet.
27:54
SPEAKER_01: No,
27:54
I love it because when I worked both as an engineering manager
27:58
or as an engineer,
27:59
SPEAKER_01: having an intern,
28:00
SPEAKER_01: it's just like boosted morale.
28:02
SPEAKER_01: We were so much more enthusiastic.
28:04
SPEAKER_01: I'm not going to say that we necessarily did more work.
28:06
SPEAKER_01: I think we might have, but we got fresh ideas.
28:09
SPEAKER_01: And it's just so rewarding looking back at those interns.
28:12
SPEAKER_01: I now see them many, many years later.
28:14
SPEAKER_01: Some are now principal engineers, some are CTOs.
28:17
But the one question that a lot of people who are maybe less
28:20
technical than you and
28:21
SPEAKER_01: who have not seen the benefit of internships,
28:23
SPEAKER_01: they're thinking like,
28:23
SPEAKER_01: oh,
28:24
SPEAKER_01: I'm hearing all these AI things,
28:25
SPEAKER_01: including Microsoft is saying it's so efficient.
28:28
SPEAKER_01: They're now talking about AI colleagues on the keynote.
28:31
SPEAKER_01: Why?
28:32
SPEAKER_01: And now this is not Microsoft people,
28:35
SPEAKER_01: but we're hearing people at Google say like,
28:36
SPEAKER_01: oh,
28:37
SPEAKER_01: it's now at a level of a junior engineer.
28:40
SPEAKER_01: And they're thinking,
28:40
SPEAKER_01: hmm,
28:41
SPEAKER_01: maybe let's hold off on hiring new grads because they're telling us,
28:45
SPEAKER_01: you know,
28:45
some of these so-called experts that it's now at the level of a
28:47
junior engineer,
28:48
SPEAKER_01: so maybe we don't need junior engineers.
28:50
What would you suggest to people who are in this thinking and
28:53
maybe not as informed?
28:54
SPEAKER_01: Because clearly you've decided it makes sense and,
28:57
SPEAKER_01: you know,
28:57
SPEAKER_01: this thing doesn't apply,
28:59
SPEAKER_01: but their thinking is real.
29:01
SPEAKER_00: it's backwards in many ways.
29:02
I think actually folks that go to high school now or to college
29:06
or even kids
29:08
SPEAKER_00: earlier in their education,
29:10
SPEAKER_00: they get to use AI much faster.
29:12
SPEAKER_00: They get it because they are taking this with an open mind.
29:16
SPEAKER_00: They don't have to
29:17
SPEAKER_00: This is how we always done it.
29:19
SPEAKER_00: Don't touch, never touch a running system.
29:20
They haven't been in an experience where some change that was
29:24
applied from upper
29:25
management or from the engineers themselves has led to a big
29:29
outage and things like
29:30
SPEAKER_00: that.
29:30
SPEAKER_00: So they're more open-minded.
29:31
SPEAKER_00: You kind of see that when you look at media, as we already talked about,
29:36
kids are much more open to just scoring through charts and
29:41
consuming the content
29:42
SPEAKER_00: they'd like to consume while,
29:43
SPEAKER_00: you know,
29:43
SPEAKER_00: when you and I were a child,
29:44
we had linear TV and you got to watch on the living room TV
29:48
what your mom and dad
29:49
SPEAKER_00: decided is on tonight,
29:51
SPEAKER_00: right?
29:51
SPEAKER_00: And so they have much more freedom.
29:52
And I think this will lead to ultimately junior engineers
29:57
coming or interns coming
29:58
SPEAKER_00: into the companies and thinking in like prompting skills,
30:01
SPEAKER_00: you know,
30:02
SPEAKER_00: like experience with different models.
30:04
SPEAKER_00: Like,
30:05
SPEAKER_00: Intern doesn't mean I haven't worked for another company before.
30:07
SPEAKER_00: It might be my second or third internship,
30:10
or I might have already been coding with a group of students on
30:14
my own project,
30:15
SPEAKER_00: on my own app.
30:16
So you get a lot of good new ideas and outside perspective that
30:19
is ultimately crucial
30:21
SPEAKER_00: to compete in the market.
30:22
SPEAKER_00: And I think vice versa,
30:23
these interns then go back into college and they have worked on
30:26
something real.
30:27
SPEAKER_00: I remember my time at university,
30:29
I always had the feeling I got to get out in the industry and
30:32
learn how it's
30:32
SPEAKER_00: actually done.
30:33
Because while you learn coding in university,
30:37
but you don't really have an engineering system.
30:39
SPEAKER_00: You're not having...
30:40
SPEAKER_00: tech debt.
30:41
SPEAKER_00: You're not working on a legacy code base.
30:43
Becoming an engineer in a company means you're working on
30:46
somebody else's code,
30:47
SPEAKER_00: often now GitHub 17 years old code,
30:51
SPEAKER_00: and you have to follow up on decisions other people made.
30:54
SPEAKER_00: And that's not how you learn coding in a university.
30:57
And so I think it's both for us important to have those fresh
31:01
ideas within the
31:01
SPEAKER_00: company,
31:01
but also for us important to give back to those that in the
31:06
future become GitHub
31:07
SPEAKER_00: customers.
31:08
SPEAKER_01: I love it, and I feel it's also just the cheapest way to do it.
31:11
SPEAKER_01: You're not going to have the highest salary for the new people.
31:15
SPEAKER_01: This is a silly way to think about it.
31:16
I talked with the product manager recently,
31:18
and I just had a deja vu when you said it.
31:21
SPEAKER_01: This product manager was saying,
31:23
this is while we were doing the Gen Z research about the latest
31:25
generation.
31:26
SPEAKER_01: She said,
31:27
SPEAKER_01: I love Gen Z.
31:28
SPEAKER_01: She's like,
31:29
SPEAKER_01: for years as a product manager,
31:30
SPEAKER_01: I'm sitting in front of engineers telling them,
31:33
SPEAKER_01: oh,
31:33
SPEAKER_01: get product thinking,
31:34
SPEAKER_01: think about the customer.
31:35
SPEAKER_01: And they're like, just tell us what we need to do.
31:37
SPEAKER_01: And she's like, the new generation comes in and they have opinions.
31:40
SPEAKER_01: They're like, I think our customers would like this.
31:42
SPEAKER_01: I have tested our competitor.
31:44
SPEAKER_01: Here's what they're doing.
31:45
SPEAKER_01: And she's like, I love it.
31:47
SPEAKER_01: So I wonder if by hiring these people and actually just ignoring,
31:52
SPEAKER_01: as you say,
31:52
SPEAKER_01: the backward thinking,
31:53
SPEAKER_01: you might get the next generation of software engineers.
31:55
SPEAKER_00: Gen Z is already too old, I think, to some degree.
31:58
Because Gen B,
31:59
I think,
31:59
starts 1980 or something like that,
32:01
the millennials.
32:02
SPEAKER_00: Gen Alpha, right?
32:03
SPEAKER_00: Like, they're 60 years old, are coding today.
32:05
SPEAKER_00: And so that's Gen Alpha or,
32:07
SPEAKER_00: you know,
32:07
SPEAKER_00: soon the,
32:07
SPEAKER_00: whatever,
32:08
SPEAKER_00: the Gen Beta,
32:08
SPEAKER_00: I guess,
32:09
SPEAKER_00: is the next one.
32:10
SPEAKER_00: And so,
32:10
SPEAKER_00: but you're right,
32:11
SPEAKER_00: you know,
32:11
you're getting folks that didn't learn to code on a Commodore
32:16
64 on a PC that,
32:17
SPEAKER_00: you know,
32:17
SPEAKER_00: have,
32:18
SPEAKER_00: have,
32:19
SPEAKER_00: A lot of low-level thinking,
32:20
SPEAKER_00: but also learned in a different environment than those that grew up with
32:25
SPEAKER_00: smartphones,
32:26
SPEAKER_00: that grew up with being always connected.
32:28
SPEAKER_00: They just have,
32:29
SPEAKER_00: just from an expectation perspective,
32:31
they're looking at this often systems problem they're trying to
32:34
solve with a
32:35
SPEAKER_00: different perspective.
32:36
SPEAKER_01: Yeah, I wonder just like a silly thing.
32:39
Would you rather hire an engineer with the same years of
32:42
experience of like someone
32:43
SPEAKER_01: who has,
32:43
SPEAKER_01: you know,
32:43
like has traditional coding and learned in university or the
32:47
same years old.
32:48
You learned in university and did some university projects or
32:52
this engineer or this
32:54
SPEAKER_01: fresh grad who did a university,
32:55
but they actually have before that five years of building
32:59
programs in Roblox and
33:00
SPEAKER_01: they actually built this like hundred person thing.
33:03
SPEAKER_01: It's like the skills that that person brings and the mindset, you know,
33:06
SPEAKER_00: And there's a third group,
33:08
and there's lots of folks at Microsoft and GitHub that don't
33:11
have a formal
33:12
SPEAKER_00: university education and either dropped out or never went.
33:15
That goes back to the earlier discussion about being a
33:19
remote-first company.
33:20
SPEAKER_00: You hire people because they're passionate about the product.
33:23
You hire people because they have a green contribution graph on
33:26
their GitHub profile.
33:27
because they have shown they contributed back to the Ruby on
33:30
Rails code base,
33:31
SPEAKER_00: to the Git open source code base,
33:33
where at GitHub we traditionally have employed folks that are
33:37
spending a lot of
33:38
SPEAKER_00: time on maintaining Git together with the open source community.
33:41
So I think a lot of what actually plays a role to us is that
33:44
folks have shown the
33:45
SPEAKER_00: right mindset to join GitHub.
33:48
SPEAKER_00: If you don't have a green contribution graph on your GitHub profile,
33:52
SPEAKER_00: That,
33:52
SPEAKER_00: I think,
33:52
matters more to us than whether you have five years at one
33:55
company and five years
33:56
SPEAKER_00: at another company.
33:57
SPEAKER_00: Of course,
33:57
SPEAKER_00: we take people to an interview loop,
33:59
and I think increasingly we're thinking about how do we
34:02
leverage AI within the
34:03
SPEAKER_00: interview loop.
34:05
SPEAKER_00: There's nothing wrong about that from my perspective.
34:07
SPEAKER_00: In fact, I would say if you want to...
34:10
SPEAKER_00: get a job in a tech company very soon,
34:12
SPEAKER_00: you're going to be asked to show your prompting skills,
34:15
SPEAKER_00: your co-pilot skills,
34:16
SPEAKER_00: if you will,
34:17
and how you use something like agent mode or the coding agent
34:20
that we introduced
34:21
SPEAKER_00: this week to get the exercise done.
34:25
Because the goal of the future engineer is no longer to run it
34:28
all from scratch.
34:29
And the goal is to combine their prompting skills and agent
34:33
open source libraries
34:34
into getting that problem solved much faster than they could
34:37
have done three,
34:37
SPEAKER_00: three years ago.
34:38
So with that,
34:39
these were some of the,
34:40
like,
34:40
a bunch of interesting facts.
34:42
I wanted to talk about the past and then later the present and
34:46
the future of GitHub
34:47
SPEAKER_01: starting with the past.
34:48
SPEAKER_01: So Git,
34:50
SPEAKER_01: I did a podcast episode with Linux kernel maintainer,
34:53
SPEAKER_01: Greg K.H.,
34:55
SPEAKER_01: and he told a story about how Git was created in 2005.
34:58
SPEAKER_01: In fact,
34:58
SPEAKER_01: recently there's a podcast with GitHub,
35:00
SPEAKER_01: a staff engineer at GitHub,
35:02
SPEAKER_01: interviewed Linux Torwards.
35:03
SPEAKER_01: In 2005, Linux Torwards,
35:06
SPEAKER_01: got frustrated with source control for the Linux project specifically.
35:09
SPEAKER_01: There were some commercial things, we can't see the history.
35:12
SPEAKER_01: He sat down in 10 days, he wrote Git, which was obviously open source.
35:15
It was meant for Linux only,
35:17
but they released it and turns out it's pretty good.
35:19
SPEAKER_01: GitHub was founded in, I believe, 2008.
35:22
SPEAKER_01: Can you tell me about these early days?
35:25
SPEAKER_01: What happened then?
35:26
SPEAKER_01: What do you understand from talking with the early founders?
35:28
The first commit was October 2007,
35:30
so it was about two years after Linux had done Git,
35:34
right?
35:34
SPEAKER_00: And then October 2007 to,
35:36
SPEAKER_00: I believe,
35:37
February 2008 is when they started beta testing it with folks
35:42
in the Ruby Orells
35:43
SPEAKER_00: community.
35:44
SPEAKER_00: And then...
35:45
SPEAKER_00: April is when the public version of GitHub launched.
35:48
SPEAKER_00: That was really fast as well.
35:50
SPEAKER_00: You know, October to April, five, six months.
35:52
SPEAKER_00: You know,
35:53
SPEAKER_00: if you go back,
35:54
SPEAKER_00: you know,
35:54
SPEAKER_00: to the history books,
35:56
SPEAKER_00: it's Chris and Tom,
35:58
SPEAKER_00: Tom Preston-Borner and Chris Warnstrad meeting in a bar,
36:02
SPEAKER_00: as I remembered,
36:02
SPEAKER_00: and having seen Git and figuring out that
36:07
There isn't a place where you can boost your open source
36:11
repositories and your
36:12
SPEAKER_00: private repositories together.
36:13
SPEAKER_00: There was obviously things like SourceForge,
36:16
SPEAKER_00: which was the most popular open source platform back then,
36:20
SPEAKER_00: but also painful to use.
36:23
SPEAKER_00: I think that's fair to say.
36:25
SPEAKER_00: I had a couple of projects on SourceForge as well.
36:27
It was kind of designed for a different era of the web and
36:31
didn't make the jump in
36:32
SPEAKER_00: what ultimately became known as web developers.
36:36
SPEAKER_00: as Web 2.0 or Cloud Native.
36:39
SPEAKER_00: And so I think the idea was, OK, Git is here.
36:41
SPEAKER_00: Git is amazing.
36:43
But I want to host it without setting up a server and
36:46
installing the back end and
36:48
SPEAKER_00: managing SSH access.
36:50
SPEAKER_00: And so they started building around that idea.
36:53
SPEAKER_00: What always made GitHub GitHub is that they put the developers first.
36:57
If you go on the Wayback Machine and look at the first GitHub
37:00
homepage,
37:01
it isn't very fancy.
37:03
SPEAKER_00: It looks like a commit log.
37:04
SPEAKER_00: It's a very short announcement of,
37:06
SPEAKER_00: hey,
37:06
SPEAKER_00: we shipped this new feature,
37:07
SPEAKER_00: try it out,
37:08
SPEAKER_00: and let us know what you think.
37:09
SPEAKER_00: The story about our logo, what is now known as the Octocad,
37:14
SPEAKER_00: You know, it's a funny one.
37:16
And I learned that actually from the Acquired podcast because
37:18
they learned from the
37:20
Twitter founders how they got their logo and they bought it on
37:22
some stock graphic
37:24
SPEAKER_00: page from a British designer,
37:26
SPEAKER_00: I believe.
37:27
SPEAKER_00: And so they looked,
37:28
SPEAKER_00: you know,
37:28
SPEAKER_00: for other graphics that the same designer had done and bought,
37:32
SPEAKER_00: you know,
37:33
SPEAKER_00: this hybrid of an octopus and a cat and made that the logo...
37:40
SPEAKER_00: So there was a lot of,
37:41
SPEAKER_00: as a bootstrap startup,
37:42
SPEAKER_00: that's scrappy,
37:44
that has a cool technology that might become the future of
37:48
version control.
37:50
SPEAKER_00: And then they built what is now known as a cloud-native company.
37:55
SPEAKER_00: And yeah, it grew fast.
38:00
Back in the day,
38:01
probably the product was the fastest product market fit.
38:05
SPEAKER_00: Nowadays, there's other companies that have similar.
38:08
SPEAKER_00: It's a different time.
38:09
It's a different time now,
38:10
and I think it was unheard of how quickly it spread like
38:14
wildfire.
38:15
SPEAKER_00: I was in 2008 still working in the automotive industry.
38:18
SPEAKER_00: I was at Bosch, so I didn't use GitHub in the first year.
38:21
Then I quit Bosch at the end of 2008,
38:23
not because of GitHub,
38:24
but because of the iPhone SDK.
38:26
SPEAKER_00: And it became an ISV building apps for German companies.
38:31
SPEAKER_00: And so I created, I think, my GitHub account in early 2009.
38:35
SPEAKER_00: We can look it up, I think, March or April 2009.
38:37
And then I was at RailsConf,
38:39
the Ruby on Rails conference in Las Vegas in 2009 and saw Chris
38:43
one stress giving a talk about github and then what they are up
38:46
to and that's i
38:47
would say when i became a github fan and and when i started
38:50
using github in my
38:51
projects um but until that point you would use some help self
38:54
was a tool you know
38:55
install it on some wood server or maybe you already had the
38:58
cloud and and maintain
38:59
all that yourself and ever since people just you know log in
39:02
with github and create
39:03
SPEAKER_00: repo
39:04
SPEAKER_01: Yeah,
39:04
SPEAKER_01: when I looked through the history,
39:05
SPEAKER_01: just the things that I could find,
39:07
SPEAKER_01: what struck me as like,
39:08
SPEAKER_01: so like as the first couple of the 2007 and 2008,
39:11
SPEAKER_01: the service was launched,
39:12
SPEAKER_01: I think you said March or something like that,
39:14
SPEAKER_01: April,
39:15
SPEAKER_01: sorry.
39:16
And already in 2008,
39:17
the Reddit,
39:19
Yahoo,
39:20
Twitter,
39:20
and Facebook already onboarded
39:22
SPEAKER_01: Facebook was surprising to me because I knew they didn't use Mercurial,
39:25
SPEAKER_01: but apparently they used it for their open source things.
39:28
SPEAKER_01: These were very hip and very trendy companies at the time.
39:31
Obviously,
39:32
from there on,
39:32
as you said,
39:33
in 2009,
39:33
this was the obvious place to find.
39:36
SPEAKER_01: Why do you think it was so popular?
39:38
SPEAKER_01: What was the reason?
39:39
SPEAKER_01: Was it open source?
39:41
SPEAKER_01: Because Git wasn't even like back, we're talking 2008, 2009.
39:44
SPEAKER_01: Like now we know Git is great and it works.
39:47
SPEAKER_01: But back then this was not clear.
39:48
SPEAKER_01: Everyone was using SVN, I remember at the time.
39:51
And Microsoft had their own really heavily invested,
39:54
called TFS,
39:55
Team Foundation Server.
39:56
SPEAKER_01: Why do you think this product market fit just hit so bad?
39:59
SPEAKER_01: And devs were like, yeah, GitHub is a place to be.
40:02
SPEAKER_00: Git, you know...
40:02
solve the pain point that we had with all these other version
40:07
control systems.
40:08
SPEAKER_00: You know,
40:08
SPEAKER_00: subversion,
40:10
SPEAKER_00: TFS,
40:10
SPEAKER_00: you know,
40:11
SPEAKER_00: had its own version control until much data when they added Git support.
40:14
SPEAKER_00: Yes.
40:15
SPEAKER_00: They're painful to use and they're painful to use,
40:18
especially when you had multiple teams working together within
40:21
the same code base.
40:23
SPEAKER_00: Git bought this concept of you have the full repository on your machine,
40:27
SPEAKER_00: so you can switch from one branch to another
40:29
SPEAKER_00: even without having an internet connection.
40:32
SPEAKER_00: And you get always everything on your computer,
40:34
so you could make changes to multiple branches and then push
40:38
upstream or share it
40:39
SPEAKER_00: with your coworkers.
40:40
SPEAKER_00: And in some ways,
40:41
SPEAKER_00: GitHub,
40:42
SPEAKER_00: I think,
40:42
is the irony of software development tools because you have a
40:47
decentralized system
40:48
SPEAKER_00: like Git,
40:48
SPEAKER_00: and then you have a central hub.
40:51
SPEAKER_00: that the world depends on today and where everybody pushes to.
40:55
So this notion of decentralized systems is still something that
41:00
sounds great when
41:00
SPEAKER_00: developers are chatting about technology,
41:03
SPEAKER_00: but I think humans are longing for home.
41:07
SPEAKER_00: And today at GitHub,
41:08
SPEAKER_00: we have the saying,
41:09
SPEAKER_00: we're the home of software developers,
41:11
SPEAKER_00: where they collaborate human to human and soon human to agent.
41:15
SPEAKER_00: So I think that played a big role.
41:17
You had a new version control system that people were
41:20
intrinsically motivated to
41:21
SPEAKER_00: try out because they were frustrated with using some versions.
41:24
SPEAKER_00: Or even CPS, you know, and something like that.
41:27
SPEAKER_00: And just wanted something new to solve that problem.
41:30
SPEAKER_00: And then it was just easy to shoot an account and try it out.
41:36
SPEAKER_00: Even user interfaces like browsing the repo,
41:39
SPEAKER_00: clicking through the file tree,
41:41
GitHub innovated a lot in that compared to what was there
41:44
before and making that
41:45
Ajax request so they would load really fast and you wouldn't
41:48
wait for the full page
41:49
SPEAKER_00: to reload.
41:50
GitHub invented the pull request,
41:52
this notion of that you can fork a repo.
41:55
SPEAKER_00: make your modifications.
41:56
SPEAKER_00: If you'd like to keep your fork, you can just do that.
41:58
SPEAKER_00: That's totally cool.
41:59
SPEAKER_00: Or you can send a pull request back.
42:00
So we're asking the other maintainer to merge in your changes
42:04
before that diff
42:05
SPEAKER_00: files for exchange.
42:07
SPEAKER_01: This episode is brought to you by Augment Code.
42:09
SPEAKER_01: You're a professional software engineer, vibes will not cut it.
42:13
SPEAKER_01: Augment Code is the AI assistant built for real engineering teams.
42:17
SPEAKER_01: It ingests your entire repo,
42:18
SPEAKER_01: millions of lines,
42:19
SPEAKER_01: tens of thousands of files,
42:21
SPEAKER_01: so every suggestion lands in context and keeps you in flow.
42:24
SPEAKER_01: With Augment's new remote agent,
42:26
SPEAKER_01: queue apparel tasks like bug fixes,
42:28
SPEAKER_01: features,
42:29
SPEAKER_01: and refactors,
42:30
SPEAKER_01: close your laptop,
42:31
SPEAKER_01: and return to ready-for-review poll requests.
42:33
SPEAKER_01: Where other tools stall, Augment Code sprints.
42:37
AugmentCode never trains or sells your code,
42:39
so your team's intellectual property stays yours.
42:42
SPEAKER_01: And you don't have to switch tooling.
42:43
SPEAKER_01: Keep using VS Code, JetBrains, Android Studio, or even Vim.
42:47
SPEAKER_01: Don't hire an AI for vibes.
42:49
SPEAKER_01: Get the agent that knows you and your code base best.
42:52
SPEAKER_01: Start your 14-day free trial at augmentcode.com slash pragmatic.
42:57
SPEAKER_01: Well,
42:57
SPEAKER_01: I think this is one of the biggest thing that people miss,
42:59
SPEAKER_01: as I understood,
43:00
SPEAKER_01: like understanding.
43:00
So like Linus Toro's design Git,
43:02
and a lot of GitHub or the Git functionality comes from there.
43:06
SPEAKER_01: But the Linux workflow was email-based.
43:08
SPEAKER_01: So actually in Git,
43:09
SPEAKER_01: there is a command,
43:10
SPEAKER_01: I don't remember the exact command,
43:12
SPEAKER_01: where I think it's something email,
43:13
SPEAKER_01: where you send an email with your changes bundled,
43:16
SPEAKER_01: with your Git changes bundled.
43:18
SPEAKER_01: And that's how Linux works.
43:19
SPEAKER_01: They just use email.
43:21
SPEAKER_01: And GitHub does not do this.
43:22
SPEAKER_01: Indeed, GitHub invented the pull request, which is...
43:26
SPEAKER_01: a big difference between even today how Linux uses Git and GitHub does.
43:29
SPEAKER_01: And yeah, it was a smash.
43:31
SPEAKER_01: When was the pull request invented?
43:32
SPEAKER_01: Do you know?
43:33
SPEAKER_01: Was it early on?
43:33
SPEAKER_00: I'm getting the date from it.
43:34
SPEAKER_00: I think it was a year or so into GitHub's journey.
43:37
SPEAKER_00: Pretty early.
43:37
SPEAKER_00: It definitely wasn't there on day one.
43:40
SPEAKER_00: It came sometime later.
43:41
SPEAKER_00: We can look it up in the show notes.
43:44
SPEAKER_00: But it was a new thing at the time.
43:47
SPEAKER_00: And I think it
43:48
It's ultimately liberating for software developers where you
43:52
send a pull request
43:53
SPEAKER_00: and if the maintainer doesn't like it,
43:55
SPEAKER_00: they can just close the pull request or leave it open.
43:57
SPEAKER_00: Others can see it and if they want to pick it up,
44:02
SPEAKER_00: they can just jump over to your fork and keep using it there.
44:04
And so you see a lot of that also happening that people just
44:08
keep their forks independent.
44:11
because they don't want to have the obligation to engage with
44:14
the main project.
44:15
I have a bunch of forks where I made some small modification
44:19
for my own hobby project.
44:21
SPEAKER_00: I don't see a big reason that that needs to go back into the main branch.
44:27
SPEAKER_00: Some of that might be just a bit hackery as well.
44:30
SPEAKER_00: And so it makes it much easier for developers that want to collaborate,
44:33
SPEAKER_00: to collaborate with each other.
44:34
SPEAKER_00: And for the others,
44:35
SPEAKER_00: it democratizes how they engage with open source,
44:38
SPEAKER_00: how they learn from it,
44:39
SPEAKER_00: how they mix and match.
44:40
SPEAKER_00: and so on.
44:41
So what kind of a gathering if I'm putting it together like
44:44
open source was clearly
44:45
one thing this new collaborative workflow especially with open
44:48
source but also and
44:49
I think one thing you didn't mention but I wonder if you're if
44:52
it is a part is how
44:54
even if you use GitHub you can always move to your own git
44:56
server it's very easy to
44:57
migrate it also works offline so even if GitHub is offline like
45:02
yes it's a central
45:03
hub but if there's a GitHub outage I mean we're not happy about
45:06
it and people
45:06
SPEAKER_01: complain these days and there's not many of them to be fair
45:09
But even when that happens,
45:10
I mean,
45:11
oh,
45:11
I guess I can keep working locally.
45:14
SPEAKER_01: I can keep pushing.
45:15
SPEAKER_01: So you might not,
45:16
SPEAKER_01: if it's just,
45:17
SPEAKER_01: again,
45:17
SPEAKER_01: if it's just like a 30-second outage or a one-minute one,
45:20
SPEAKER_01: again,
45:21
SPEAKER_01: not advocating for it,
45:22
SPEAKER_01: but you might not even notice because you're just doing your local thing.
45:26
So I think as a developer,
45:28
I think we appreciate when we know we are not locked into it.
45:31
SPEAKER_01: Oh, by the way, it's free.
45:32
Like a lot of the,
45:34
sorry,
45:34
not GitHub per se,
45:35
but for open source,
45:36
it was always free,
45:37
right?
45:37
SPEAKER_01: Yeah.
45:37
SPEAKER_00: It was always free for open source.
45:39
And you might say GitHub invented a version of the freemium
45:44
model that is not
45:45
SPEAKER_00: ad-based,
45:45
SPEAKER_00: where traditionally freemium is ad-based.
45:47
SPEAKER_00: The free users get something for free with advertisement,
45:52
and then the paid users get rid of the advertisement and get
45:55
more features.
45:56
GitHub originally had the separation between everything you do
45:58
in public as free
46:00
SPEAKER_00: because it helps the community,
46:01
SPEAKER_00: obviously helps the social network,
46:03
SPEAKER_00: quote unquote.
46:03
SPEAKER_00: Yes.
46:05
SPEAKER_00: GitHub growing.
46:06
SPEAKER_00: And then when you wanted private repositories, you had to pay for them.
46:10
SPEAKER_00: Shortly after the acquisition, early 2019, we changed that.
46:14
SPEAKER_00: And instead of paying for private repositories,
46:17
SPEAKER_00: we actually made private repositories free as well.
46:19
SPEAKER_00: We did, yeah.
46:19
SPEAKER_00: Because we had a lot of individual developers giving us feedback.
46:23
SPEAKER_00: especially when you travel around the world.
46:26
SPEAKER_00: For an American,
46:26
SPEAKER_00: $4 a month might not be a lot,
46:29
SPEAKER_00: but for somebody in a different country,
46:30
SPEAKER_00: that is much more money relative to their income.
46:34
SPEAKER_00: And so private repository is free,
46:36
SPEAKER_00: and today we charge for what you could call enterprise features,
46:39
SPEAKER_00: like single sign-on or branch protection,
46:42
SPEAKER_00: things that a student or a hobbyist probably doesn't need.
46:46
If I'm honest,
46:47
when I work on my hobby projects,
46:49
I barely write any tests.
46:50
SPEAKER_00: I certainly don't send myself a pull request.
46:53
SPEAKER_00: And now I have to review my own code.
46:54
SPEAKER_00: I just merge, push into main branch, and that's cool.
46:58
SPEAKER_00: But yeah, GitHub innovated in the business model space as well.
47:01
And you mentioned the scenarios where it's helpful to use
47:06
GitHub or Git on your local machine.
47:10
SPEAKER_00: In 2007, 2008, flights didn't have Wi-Fi, at least in my memory.
47:14
SPEAKER_00: No, no.
47:15
SPEAKER_00: Wi-Fi wasn't as prevalent.
47:17
SPEAKER_00: That's very recent.
47:18
As far as I remember,
47:19
I don't know if Starbucks,
47:20
maybe Starbucks already had it.
47:22
But there was much more places where you would work just local
47:25
on your machine for
47:27
SPEAKER_00: quite a while and being able to commit into the repository.
47:31
SPEAKER_00: There wasn't, you know, subversion didn't have a push at it.
47:33
SPEAKER_00: You would always commit into the upstream,
47:37
SPEAKER_00: change branches,
47:38
SPEAKER_00: roll back all these operations that are now natural with Git.
47:42
SPEAKER_00: They weren't natural in 2007, 2008.
47:47
So GitHub started off,
47:48
it started to get real big traction,
47:50
2008,
47:50
2009,
47:50
2010.
47:51
The next big kind of milestone I remember from a developer's
47:55
perspective is,
47:56
SPEAKER_01: until then,
47:57
SPEAKER_01: to use GitHub,
47:58
you could use the web version and you use the local command
48:01
line interface,
48:02
SPEAKER_01: you just learned all the commands,
48:03
SPEAKER_01: you know,
48:03
SPEAKER_01: there's books about,
48:04
SPEAKER_01: like,
48:05
SPEAKER_01: the easy ones are easy and then like cherry picking and all that,
48:07
SPEAKER_01: I was always confused by that.
48:09
And then in around 2014 was the first time that GitHub actually
48:14
released desktop
48:15
SPEAKER_01: applications,
48:16
SPEAKER_01: but not only did it release desktop applications,
48:18
SPEAKER_01: it actually released the Atom text editor,
48:20
SPEAKER_01: which I loved.
48:21
SPEAKER_01: I used it for so long.
48:22
SPEAKER_01: It was a lightweight editor.
48:24
It released the Electron framework,
48:26
which powered Atom,
48:27
and then GitHub desktop in 2014,
48:29
2015.
48:29
SPEAKER_01: Why do you think GitHub
48:33
SPEAKER_01: waited that long to have these desktop applications?
48:36
SPEAKER_01: What was the goal with Atom and how did this help?
48:38
Did it actually help get more people who just liked a nice user
48:42
interface and were
48:42
SPEAKER_01: not the command types?
48:44
SPEAKER_00: I don't know the history of all these apps.
48:46
SPEAKER_00: My suspicion is that when you say waited this long,
48:50
I'd make the counter argument in how long is long for a startup
48:54
to run the core
48:55
product to having multiple other products that may or may not
48:58
become a distraction.
49:00
SPEAKER_00: I think each of those solved a specific problem that the team saw.
49:05
SPEAKER_00: Atom came out, you know, I think a year or so before VS Code launched.
49:11
SPEAKER_00: And so clearly the time was ripe, you know, for a new generation of IEs.
49:17
SPEAKER_00: Both of them were open source.
49:18
SPEAKER_00: You know,
49:18
SPEAKER_00: for the acquisition,
49:19
they both ended in the same company and then already made the
49:22
decision to continue
49:24
SPEAKER_00: our journey on VS Code and retire Atom.
49:26
SPEAKER_00: Yeah.
49:26
SPEAKER_00: It goes back to this developer-first culture.
49:29
It goes back to GitHub's origins where you would join as an
49:32
employee and you
49:33
SPEAKER_00: wouldn't have a manager and you wouldn't have a backlog.
49:35
SPEAKER_00: You would have the... So back then, really, there were no managers.
49:39
SPEAKER_00: Well, there was obviously a CEO and a little bit, but that was very flat.
49:43
SPEAKER_00: And people, when they... Onboarding was always at headquarters.
49:47
SPEAKER_00: So even though it's a remote-first company,
49:48
you would spend a week or so in our headquarters in San
49:51
Francisco and to learn how
49:52
SPEAKER_00: to make an Espresso.
49:55
SPEAKER_00: And that was literally one of the onboarding videos.
49:57
SPEAKER_00: And, you know, they got to have some whimsical future, whimsical culture.
50:04
SPEAKER_00: And you got told, you know, pick something that you're interested in.
50:08
SPEAKER_00: And that naturally leads,
50:10
SPEAKER_00: you know,
50:10
SPEAKER_00: to developers picking things that they want to solve for themselves,
50:14
SPEAKER_00: that they care about.
50:15
SPEAKER_00: And so Adam was actually, you know, Chris Wonstrad, the CEO at the time.
50:21
SPEAKER_00: And then for most of GitHub's journey until the acquisition,
50:24
SPEAKER_00: was one of the core maintainers of the Allen project.
50:28
SPEAKER_00: You don't hear that often, CEO being a core maintainer.
50:33
SPEAKER_00: A lot of that was about the passion of the people working there.
50:36
SPEAKER_00: I don't know exactly the history of desktop,
50:38
but my assumption there would be that they saw a need to expand
50:42
the people that can
50:43
SPEAKER_00: use GitHub without knowing all the intricacies of Git.
50:48
In many ways,
50:49
the GitHub user interface on GitHub.com also serves that
50:52
purpose.
50:52
SPEAKER_00: If you think,
50:53
SPEAKER_00: for example,
50:54
SPEAKER_00: about blame views,
50:56
blame on GitHub.com is just picking a button and see who
50:59
modified which file,
51:00
SPEAKER_00: which line in the file.
51:02
SPEAKER_00: While if you do get blame on the comment line,
51:04
SPEAKER_00: it's much,
51:05
SPEAKER_00: much harder to figure out what's what,
51:08
SPEAKER_00: especially if you don't know a lot about software development.
51:10
So it's all about growing the user base and expanding the
51:14
people that get utility automatically.
51:18
SPEAKER_01: I'll say that I use GitHub Desktop a lot.
51:20
SPEAKER_01: I mean, everything is different.
51:21
SPEAKER_01: And I know how to use command line.
51:22
SPEAKER_01: But I just like that I don't have to think about it again.
51:26
SPEAKER_01: So I appreciated what it came out.
51:28
SPEAKER_00: I barely use it.
51:29
SPEAKER_00: But I use the GitHub integration in VS Code a lot for the same purpose.
51:34
SPEAKER_00: And so I think even that,
51:37
SPEAKER_00: Git as a protocol,
51:40
SPEAKER_00: SSH access and HTTP access,
51:42
SPEAKER_00: which also,
51:42
SPEAKER_00: by the way,
51:43
SPEAKER_00: wasn't something that was very prevalent in the early days of GitHub.
51:48
And it's now much more used because it's easier to access
51:51
through HTTP than through SSH.
51:54
I think all of these pieces are ultimately helping to create
51:58
the developer
51:59
SPEAKER_00: experience that we now consider normal,
52:01
SPEAKER_00: but by no means was normal 20 years ago.
52:05
We're getting to the Microsoft acquisition,
52:07
so we've now passed 2015,
52:09
it's now 2016 or so.
52:11
SPEAKER_01: You were in the inside of Microsoft, you were already working there.
52:14
SPEAKER_01: How did this come about?
52:16
SPEAKER_01: What was Microsoft thinking?
52:18
SPEAKER_01: Why was even Microsoft interested in GitHub?
52:21
SPEAKER_01: And then how did the purchase go from your perspective?
52:24
SPEAKER_00: It was 2018,
52:25
SPEAKER_00: early 2018,
52:25
SPEAKER_00: that back then Nat Friedman,
52:30
my boss within Microsoft CDP in the developer division doing
52:34
mobile developer
52:35
SPEAKER_00: tools,
52:35
SPEAKER_00: Nat came from Xamarin,
52:37
SPEAKER_00: so had joined Microsoft himself for an acquisition.
52:40
SPEAKER_00: I had joined for the Hockey App acquisition.
52:42
SPEAKER_00: So we were like a bit of a...
52:44
SPEAKER_00: outside-in team within Microsoft with lots of people coming from acquired
52:48
SPEAKER_00: companies,
52:48
SPEAKER_00: lots of cloud-native experience,
52:51
SPEAKER_00: a non-traditional stack for Microsoft.
52:55
SPEAKER_00: At the same time, also, Microsoft had changed.
52:57
SPEAKER_00: It was now about four years since Satya Nadada had become the CEO.
53:02
SPEAKER_00: Satya had very early on embraced open source,
53:05
SPEAKER_00: then VS Code was launched,
53:09
SPEAKER_00: .NET became open source.
53:10
So a number of steps had happened where Microsoft changed its
53:15
fundamental position
53:16
SPEAKER_00: on how it views open source,
53:18
SPEAKER_00: how it views the stack and where the future is going.
53:22
You may remember this part of the strategy back then was cloud
53:26
first,
53:26
mobile first.
53:29
SPEAKER_00: In many ways, that is still true today.
53:31
SPEAKER_00: We're no longer saying those two things.
53:35
SPEAKER_00: If you think about how you interact with Microsoft products today,
53:39
SPEAKER_00: they are on every device,
53:41
SPEAKER_00: regardless of who makes the device and what upbring system.
53:44
SPEAKER_01: This was not obvious then because it was still Windows first.
53:47
SPEAKER_01: It was still a feeling that Microsoft is only Windows.
53:49
I heard this was said everywhere and I think outside of
53:53
Microsoft,
53:53
we were skeptical,
53:54
honestly.
53:55
SPEAKER_00: But,
53:55
SPEAKER_00: you know,
53:56
SPEAKER_00: and obviously the cloud stack,
53:57
SPEAKER_00: the compute stack that Microsoft offers today,
54:00
SPEAKER_00: not only to its own products,
54:01
SPEAKER_00: but to any company in the industry,
54:04
SPEAKER_00: really.
54:05
One thing I learned early in my journey at Microsoft is that
54:08
when you're a small
54:09
SPEAKER_00: startup,
54:10
SPEAKER_00: you consider,
54:11
SPEAKER_00: you know,
54:11
SPEAKER_00: other companies in the space competitors and they,
54:14
SPEAKER_00: you know,
54:14
SPEAKER_00: quote unquote,
54:15
SPEAKER_00: the enemy.
54:16
At Microsoft,
54:17
competitors are at the same time also customers and often
54:20
partners.
54:22
SPEAKER_00: We are providing,
54:22
SPEAKER_00: you know,
54:23
SPEAKER_00: if you take,
54:23
SPEAKER_00: for example,
54:23
SPEAKER_00: Azure AI Foundry,
54:24
SPEAKER_00: our inference stack for larger English models,
54:27
SPEAKER_00: we're providing those not only to our own co-pilots,
54:30
we're providing those to many other companies that may directly
54:34
or indirectly
54:34
SPEAKER_00: compete with GitHub or with other Microsoft products,
54:38
SPEAKER_00: right?
54:38
SPEAKER_00: And so 2018, the worldview in Microsoft had changed.
54:42
SPEAKER_00: I think the reputation started to change.
54:45
We at Microsoft started to believe that we could pull off an
54:48
acquisition like GitHub,
54:49
where
54:50
you know make github part of microsoft and and and not lose all
54:54
the open source
54:55
maintainers because they like microsoft you know it is not
54:58
going to be with steward
54:59
um of github and we said very early on clear principles of what
55:04
that looks like and
55:05
2018 was about two years after the linkedin acquisition a bit
55:08
longer since the
55:09
minecraft acquisitions we had kind of like two examples of
55:12
acquisitions where we
55:13
SPEAKER_00: kept the brand and the platform
55:16
uh independent um even even today when you go to to the
55:19
linkedin home page you
55:21
don't see a lot of microsoft there um seven years after the gap
55:25
acquisition uh
55:26
there's there's no microsoft uh on the github home page um
55:29
unless you're using uh
55:30
entra id um you know as your as your single sign-on provider or
55:34
of course we have
55:35
blog posts where we talk about um you know copilot and and the
55:38
microsoft file model
55:39
SPEAKER_00: and things like that
55:41
SPEAKER_00: And Minecraft,
55:43
SPEAKER_00: many kids,
55:43
SPEAKER_00: I think,
55:44
realized much later in their childhood that one of their
55:47
favorite games is actually
55:49
the same company that also produces the operating system and
55:52
the professional
55:53
SPEAKER_00: network and GitHub,
55:55
SPEAKER_00: right?
55:55
SPEAKER_00: But when you look at those two acquisitions and then you see GitHub,
56:00
SPEAKER_00: it feels like a natural consequence,
56:02
SPEAKER_00: like the natural next step for Microsoft to go down that path.
56:06
SPEAKER_00: Microsoft always has been a developer-first company.
56:09
SPEAKER_00: You know, the first product was basic for the Altair.
56:12
SPEAKER_00: And,
56:12
SPEAKER_00: you know,
56:12
SPEAKER_00: here at the conference center,
56:13
I don't know if you had a chance to see the 1975 experience
56:16
where you work on a
56:17
SPEAKER_00: fake Altair and pass some coding tests.
56:21
You know,
56:21
then that was basic for the Apple II and obviously Visual
56:24
Studio in the 1990s.
56:27
SPEAKER_00: I think my first version of Visual Studio was Visual C.
56:30
1.0 came with a book and then had Visual J,
56:33
which was like Microsoft's Java clone.
56:35
SPEAKER_00: And then I bought like a full-scale
56:37
SPEAKER_00: Visual Studio 7 for way too much money back then,
56:42
SPEAKER_00: at least for where I was as a high school student.
56:45
SPEAKER_00: Obviously,
56:46
SPEAKER_00: in today's world,
56:48
selling an IDE at a one-time price would be something that has
56:51
no market fit
56:52
SPEAKER_00: anymore.
56:53
SPEAKER_00: Microsoft lacked access to the open source ecosystem.
56:58
SPEAKER_00: It didn't have the same attraction to cloud-native developers.
57:04
SPEAKER_00: It was just being not trendy at all.
57:07
SPEAKER_00: And it was important from a business perspective to have that ecosystem.
57:11
SPEAKER_00: It was important for us to put developers first again.
57:15
SPEAKER_00: And so that was actually the first principle of three principles.
57:19
And we believed we can re-accelerate GitHub because GitHub in
57:23
2018 had a bit of the
57:24
SPEAKER_00: reputation of not being able to innovate anymore for various reasons.
57:29
SPEAKER_00: And we can talk about that as more in detail.
57:32
SPEAKER_00: But it was really important to think about the acquisition,
57:34
SPEAKER_00: not a Microsoft gets a lot out of GitHub,
57:37
SPEAKER_00: but Microsoft is willing to invest into GitHub.
57:40
SPEAKER_00: You know,
57:40
GitHub in 2018 at the time to start the conversation for them
57:43
was about 700
57:43
SPEAKER_00: employees all up.
57:45
SPEAKER_00: And you mentioned earlier, it's now over 3,000 employees.
57:48
SPEAKER_00: So that investment surely has happened.
57:51
SPEAKER_00: And we have leveraged a lot of the Microsoft stack.
57:54
And Copilot comes to mind where we're using Azure AI Foundry
57:57
and the relationship
57:57
SPEAKER_00: with OpenAI to build our AI cogeneration product.
58:02
And then the last one was,
58:03
of course,
58:04
also Microsoft wanted something back.
58:05
SPEAKER_00: And so GitHub Accelerate Microsoft.
58:07
SPEAKER_00: And the easy answer there is, how did we achieve that?
58:10
SPEAKER_00: By growing revenue.
58:12
SPEAKER_00: And we announced last July,
58:15
So about a year ago now that revenue has grown to over $2
58:18
billion annual recurring
58:20
SPEAKER_00: revenue run rate.
58:21
And that keeps growing and we're very happy with these business
58:24
results and
58:25
SPEAKER_00: contributing back to Microsoft's earnings.
58:27
SPEAKER_00: But in that order,
58:29
SPEAKER_00: developers first making great developer products,
58:31
SPEAKER_00: accelerate GitHub,
58:32
and then eventually GitHub accelerating Microsoft and giving
58:36
back both on the
58:36
SPEAKER_00: product side with Copilot or on the revenue side with our results.
58:40
But I think this last thing is just worth emphasizing in the
58:43
sense that I think as
58:44
SPEAKER_01: a developer,
58:45
SPEAKER_01: I really like that.
58:47
SPEAKER_01: I like using products where I know they'll be around.
58:49
So you're saying that GitHub's revenue has been consistently
58:52
going up more
58:52
SPEAKER_01: enterprises,
58:53
SPEAKER_01: more companies,
58:54
SPEAKER_01: more teams paying for professional use cases to help their work.
58:58
SPEAKER_00: If I remember correctly,
58:59
2017 was the last official revenue announcement that GitHub
59:03
made before the
59:04
SPEAKER_00: acquisition.
59:05
SPEAKER_00: And they were at about $200 million run rate.
59:07
SPEAKER_00: And they announced at that time that they're looking for a new CEO.
59:10
SPEAKER_00: So Chris was ready to step down.
59:13
And that ultimately resulted in the acquisition then in June
59:17
2018 through an
59:18
SPEAKER_00: acquisition and a new CEO,
59:20
SPEAKER_00: Ned Friedman.
59:21
But like 2017,
59:21
$200 million to 2024,
59:22
in seven years,
59:23
revenue went up a factor of 10.
59:29
SPEAKER_01: Yeah.
59:29
SPEAKER_01: Well, it's kind of reassuring to hear that because there's always a risk.
59:34
SPEAKER_01: We know Microsoft is a for-profit company.
59:37
SPEAKER_01: It's such an adult report to shareholders.
59:39
SPEAKER_01: It's good to know that those things are going well as well.
59:42
SPEAKER_00: Well,
59:42
SPEAKER_00: and I think for a long time,
59:45
SPEAKER_00: developer tools were not seen as a revenue driver in the industry.
59:51
SPEAKER_00: I think there were a few examples like
59:53
SPEAKER_00: companies in JetBrains comes to mind or JFrog,
59:56
SPEAKER_00: there's a bunch of companies that,
59:58
SPEAKER_00: you know,
59:58
SPEAKER_00: we're focusing on developer tools.
1:00:00
SPEAKER_00: But I think,
1:00:01
SPEAKER_00: you know,
1:00:02
SPEAKER_00: if you turn to back five years or before that,
1:00:05
SPEAKER_00: because let's say seven and a half years,
1:00:07
SPEAKER_00: and we talk with venture capitalists about an idea that we have,
1:00:09
SPEAKER_00: you and I,
1:00:10
SPEAKER_00: about a developer tool startup,
1:00:11
SPEAKER_00: they're going to tell you,
1:00:12
SPEAKER_00: you need something else beyond just developer tools,
1:00:15
SPEAKER_00: because that's where the big money is.
1:00:17
SPEAKER_00: And today, this world obviously has dramatically changed.
1:00:19
SPEAKER_01: Absolutely.
1:00:20
SPEAKER_01: But as I was talking with Scott Guthrie,
1:00:21
the interesting thing was that the last company that was able
1:00:24
to make developers
1:00:25
SPEAKER_01: pay for developer tools a lot of money,
1:00:27
SPEAKER_01: we're talking thousands of dollars,
1:00:29
SPEAKER_01: was Microsoft in the 2000s,
1:00:31
SPEAKER_01: where this was pre when internet was small.
1:00:33
SPEAKER_01: And developers didn't just pay for developer tools,
1:00:35
SPEAKER_01: but they pay for MSD and documentation for access to the latest software.
1:00:39
SPEAKER_01: And that time has passed.
1:00:41
SPEAKER_01: And we were talking about Scott on some of the learnings.
1:00:43
SPEAKER_01: But it's interesting how Microsoft figured that out.
1:00:46
SPEAKER_01: I feel Microsoft is very good at figuring out this great mix where,
1:00:50
SPEAKER_01: Back then,
1:00:51
SPEAKER_01: I worked at a Hungarian company,
1:00:52
we paid about $1,000 per developer in Hungary because it made
1:00:56
us so much more
1:00:57
SPEAKER_01: productive.
1:00:59
SPEAKER_01: Today I part on this one.
1:01:00
SPEAKER_01: But you mentioned something interesting, which I also noticed.
1:01:04
Looking back,
1:01:04
and I remember back when GitHub Desktop came out,
1:01:07
it was great.
1:01:08
SPEAKER_01: I think I started to use GitHub more.
1:01:10
SPEAKER_01: But then from 2015, nothing really happened.
1:01:13
SPEAKER_01: Microsoft bought GitHub, great.
1:01:15
SPEAKER_01: And I remember on online forums as well,
1:01:17
SPEAKER_01: people were saying,
1:01:18
SPEAKER_01: like,
1:01:18
SPEAKER_01: GitHub is not doing anything.
1:01:19
SPEAKER_01: And then the next thing,
1:01:20
SPEAKER_01: I looked at the change log,
1:01:21
SPEAKER_01: and the only notable thing I've seen shipped,
1:01:24
SPEAKER_01: it started in 2021.
1:01:25
Then GitHub sponsors came out,
1:01:27
the package registry,
1:01:28
co-pilot,
1:01:29
code spaces.
1:01:30
SPEAKER_01: But what happened there from 2015 to 2020?
1:01:31
SPEAKER_01: It almost felt like from...
1:01:34
SPEAKER_01: I'm sure people were working,
1:01:35
SPEAKER_01: but from the outside,
1:01:36
SPEAKER_01: it just looked like nothing,
1:01:37
SPEAKER_01: like barren land and more people being hired.
1:01:39
SPEAKER_01: And you alluded to this.
1:01:41
SPEAKER_01: What was going on behind the scenes?
1:01:42
SPEAKER_00: I mean,
1:01:42
SPEAKER_00: obviously,
1:01:43
SPEAKER_00: the time before Microsoft,
1:01:45
SPEAKER_00: 2015 to 2018,
1:01:45
SPEAKER_00: I only have hearsay and a little bit of insight.
1:01:49
SPEAKER_00: I think a big problem that GitHub had,
1:01:53
SPEAKER_00: especially before the Microsoft acquisition and for some time after that,
1:01:57
SPEAKER_00: was that...
1:01:58
The platform had grown so much and so many people were relying
1:02:02
on it being
1:02:03
SPEAKER_00: available and not broken.
1:02:05
Plus,
1:02:06
many developers did the big bend and still are very much in
1:02:11
love with the brand.
1:02:13
SPEAKER_00: The Octocad, you know, internally we call it Mona.
1:02:16
SPEAKER_00: How do you call it?
1:02:16
SPEAKER_00: Mona.
1:02:17
SPEAKER_00: Mona.
1:02:18
SPEAKER_00: Mona Lisa, you know, Mona.
1:02:20
SPEAKER_00: It's something that is recognized.
1:02:23
You get asked about it on the street or in a coffee shop or in
1:02:26
the Lego store
1:02:26
because obviously the employees there are often also college
1:02:29
students and have
1:02:30
SPEAKER_00: been.
1:02:30
I mean,
1:02:31
I've been,
1:02:31
you know,
1:02:32
I had on an airplane the crew member asking about GitHub.
1:02:36
SPEAKER_00: I'm like, how do you know GitHub?
1:02:37
SPEAKER_00: And she's like,
1:02:37
She was like,
1:02:38
well,
1:02:38
I used that in college for some computer science course that I
1:02:43
took.
1:02:43
SPEAKER_00: And I think if you have something that is so beloved by the space,
1:02:48
SPEAKER_00: that also means the expectations are really high.
1:02:50
It's often that the disappointment is even bigger if you
1:02:53
actually care about something.
1:02:55
SPEAKER_00: But if you don't give a damn about this thing,
1:02:58
SPEAKER_00: then whether it's up or down or whether it's broken,
1:03:00
SPEAKER_00: it doesn't really matter to you because you hate it anyway.
1:03:03
And so I think those two things combined led to people really
1:03:06
being worried about
1:03:07
shipping stuff because you would change something and the
1:03:10
reaction from developers
1:03:12
SPEAKER_00: when you change something,
1:03:13
whether it's in the user interface or things work or hate
1:03:16
limits and things like
1:03:16
SPEAKER_00: that,
1:03:17
SPEAKER_00: is often there's a loud minority yelling on the internet.
1:03:23
And there's a silent majority that are actually fine with the
1:03:25
change,
1:03:26
but don't say anything.
1:03:26
SPEAKER_00: And a small change might lead to an outage,
1:03:29
and then you're the one that made three lenders of change in
1:03:32
your first week,
1:03:32
SPEAKER_00: and you brought GitHub down.
1:03:33
SPEAKER_00: That's obviously not encouraging in your journey.
1:03:36
So I think that created an organization that actually still
1:03:41
shipped a lot of
1:03:42
SPEAKER_00: things,
1:03:42
SPEAKER_00: and what we call staff ship,
1:03:44
SPEAKER_00: what Microsoft called stock coding,
1:03:45
SPEAKER_00: so features that landed
1:03:47
SPEAKER_00: within the GitHub organization that Hubbers would see when they use.
1:03:50
SPEAKER_00: One funny thing is we have,
1:03:53
SPEAKER_00: even today,
1:03:53
SPEAKER_00: as we're using GitHub every day,
1:03:55
SPEAKER_00: I often actually don't know if that feature is available to you,
1:04:00
SPEAKER_00: because I only see GitHub how a Hubber would see it.
1:04:03
SPEAKER_00: And there's a way to disable all these features, but how often
1:04:07
well,
1:04:07
no,
1:04:07
I can hide the staff bar and then it disables the features that
1:04:11
you don't see.
1:04:12
SPEAKER_00: But how often do you really do that?
1:04:14
And do you then remember that in a conversation that Gergain
1:04:16
doesn't have the
1:04:17
SPEAKER_00: feature,
1:04:17
SPEAKER_00: but I have it.
1:04:18
And I thought it already shipped,
1:04:19
but it is actually slated for next week.
1:04:21
SPEAKER_00: So I think that created a culture where things were internally shipped,
1:04:26
SPEAKER_00: but never made it to the public because everybody's worried,
1:04:31
SPEAKER_00: is that good enough to ship it?
1:04:32
SPEAKER_00: You know, CEO change was in flight.
1:04:35
SPEAKER_00: There were a number of cultural issues.
1:04:36
And then as we closed the deal and we came in,
1:04:40
we made private repos public relatively fast.
1:04:43
SPEAKER_01: That was the fast one.
1:04:43
SPEAKER_00: Yeah, sorry.
1:04:44
SPEAKER_00: That's true.
1:04:44
At that universe in 2018,
1:04:46
the first version of GitHub Actions was announced.
1:04:49
SPEAKER_00: That was even before the deal closed.
1:04:51
But it was,
1:04:52
you know,
1:04:53
hopefully the folks building it,
1:04:54
forgive me,
1:04:55
it was a bit of a hack.
1:04:56
SPEAKER_00: built on top of containers in Google Cloud.
1:05:00
And we quickly saw from the data that the product wasn't
1:05:05
actually meeting the
1:05:06
SPEAKER_00: requirements from developers.
1:05:07
SPEAKER_00: Many developers wanted full CI-CD,
1:05:10
SPEAKER_00: while GitHub Actions originally was designed more as workflow automation.
1:05:14
And we kept that work for automation part,
1:05:15
but evolved it into what it is now.
1:05:17
SPEAKER_00: And over the course of 2019,
1:05:20
2020 sponsors was announced in mid-2019 at the last ever
1:05:24
satellite conference in
1:05:25
SPEAKER_00: Berlin.
1:05:26
SPEAKER_00: And then COVID came and that ruined the plans that we had.
1:05:30
SPEAKER_00: As we came back from it,
1:05:31
SPEAKER_00: we were very careful with bringing back Universe in a smaller scale,
1:05:35
SPEAKER_00: given that there was still lots of travel restrictions.
1:05:37
SPEAKER_00: But it took us a while to get the organization back into a state
1:05:42
where we could both innovate really fast and keep the
1:05:47
fundamentals in place,
1:05:48
SPEAKER_00: keep security,
1:05:49
SPEAKER_00: availability,
1:05:52
SPEAKER_00: other things like accessibility,
1:05:54
SPEAKER_00: things that you might not actually see have changed.
1:05:57
SPEAKER_00: Because if everything is working, you don't notice it.
1:06:00
SPEAKER_00: But it created a lot of investment behind the scenes.
1:06:03
SPEAKER_00: So we are feeling good about the pace of innovation that we have today.
1:06:07
SPEAKER_00: But yeah, it took us a while to turn the ship around.
1:06:11
SPEAKER_00: And even in a team of 700 people, that takes a while.
1:06:14
And certainly with 3,000 people,
1:06:16
it is something that requires a lot of energy and focus.
1:06:19
SPEAKER_00: Yeah.
1:06:20
I think it's just a good reminder from the outside,
1:06:22
a company that is doing great.
1:06:24
SPEAKER_01: Because GitHub,
1:06:25
SPEAKER_01: I think all developers can agree,
1:06:26
SPEAKER_01: from the outside it looked like it's doing really good.
1:06:29
Like,
1:06:29
yeah,
1:06:30
you can have some things that you need to work through,
1:06:31
right?
1:06:32
SPEAKER_01: And growing too fast is amazing.
1:06:34
SPEAKER_01: I think the best problem for any team to have.
1:06:37
But it's still a problem,
1:06:38
right,
1:06:38
that you need to solve and you get through it.
1:06:41
Co-pilot was the biggest flash that I can remember from the
1:06:45
past few years,
1:06:46
SPEAKER_01: clearly,
1:06:46
SPEAKER_01: especially because when it came out,
1:06:48
SPEAKER_01: it just felt so early.
1:06:49
SPEAKER_01: Did the beta come out in 2022?
1:06:50
SPEAKER_01: 21?
1:06:50
SPEAKER_01: 21, yes.
1:06:50
The beta came out in 2021 because it was built on an earlier
1:06:54
model before ChatGPT launched.
1:07:02
SPEAKER_01: Was it on GPT-2?
1:07:03
SPEAKER_00: No, it was GPT-3.
1:07:06
SPEAKER_00: So GPT-3 came out right after the build conference in 2020,
1:07:10
SPEAKER_00: which was obviously,
1:07:11
SPEAKER_00: given the pandemic,
1:07:13
SPEAKER_00: fully remote.
1:07:14
But Kevin Scott and Sam Altman did a session about transformers
1:07:18
and Dutch language models.
1:07:18
SPEAKER_00: And then after that, GPT-3
1:07:21
SPEAKER_00: got into the preview,
1:07:23
SPEAKER_00: and we got access to that through the OpenAI-Microsoft partnership,
1:07:26
and we realized together with OpenAI that it was able to write
1:07:31
decent code in
1:07:32
different programming languages and would not mix up the syntax
1:07:35
between Python,
1:07:37
SPEAKER_00: Ruby,
1:07:37
SPEAKER_00: and JavaScript,
1:07:39
SPEAKER_00: Somebody asked me that earlier this week.
1:07:41
It was my biggest surprise that it can do that even though it
1:07:44
doesn't have a compiler built in.
1:07:46
SPEAKER_00: It's not able to validate the code it generates.
1:07:48
SPEAKER_00: It doesn't have the syntax.
1:07:50
SPEAKER_00: It just knows it from the training set.
1:07:52
Then OpenAI fine-tuned a model that was called Codex that was
1:07:58
specific for these
1:08:00
SPEAKER_00: coding scenarios.
1:08:02
SPEAKER_00: So in 2020,
1:08:03
SPEAKER_00: in August,
1:08:04
SPEAKER_00: we wrote a paper with three ideas we had,
1:08:06
SPEAKER_00: text-to-code,
1:08:07
SPEAKER_00: code-to-text as in describing code,
1:08:09
SPEAKER_00: and conversational coding is what we called it,
1:08:10
SPEAKER_00: which then today is known as chat.
1:08:12
SPEAKER_00: And those two letter scenarios didn't work well enough.
1:08:16
SPEAKER_00: We had a describe code,
1:08:17
SPEAKER_00: and sometimes it worked,
1:08:18
SPEAKER_00: and oftentimes it was wrong,
1:08:19
SPEAKER_00: and you know how it is.
1:08:20
SPEAKER_00: When developers see that,
1:08:21
SPEAKER_00: that the description is not actually what the code does,
1:08:24
SPEAKER_00: you quickly detract,
1:08:25
and you get a low net promoter score or whatever you're
1:08:29
measuring satisfaction
1:08:30
SPEAKER_00: with.
1:08:31
SPEAKER_00: But text-to-code,
1:08:32
SPEAKER_00: as in,
1:08:33
SPEAKER_00: you know,
1:08:33
prompting the model within the editor and ultimately building
1:08:37
autocompletion,
1:08:37
SPEAKER_00: that worked so well that very quickly we saw our internal,
1:08:42
SPEAKER_00: you know,
1:08:43
SPEAKER_00: Hubbers adopting the tool,
1:08:45
SPEAKER_00: giving it really high scores,
1:08:47
SPEAKER_00: saying,
1:08:47
SPEAKER_00: you know,
1:08:47
SPEAKER_00: this is great.
1:08:48
SPEAKER_00: I want to keep using this.
1:08:49
It's not the typical management says you have to use it and you
1:08:53
don't want to.
1:08:54
SPEAKER_00: But ultimately writing,
1:08:55
SPEAKER_00: you know,
1:08:56
SPEAKER_00: in the early days,
1:08:57
SPEAKER_00: 25% of the code in those files,
1:08:59
SPEAKER_00: it was enabled.
1:09:00
And then shortly thereafter,
1:09:01
that number got into the 50% range or 46%,
1:09:04
I think,
1:09:05
early 2023.
1:09:06
SPEAKER_00: And so that was the early days of Copilot.
1:09:09
SPEAKER_00: And then June 2021, we went into the public preview.
1:09:12
And within,
1:09:12
you know,
1:09:13
a few months,
1:09:14
the debate just had grown to a million users.
1:09:16
SPEAKER_00: And you saw more and more folks on social media saying,
1:09:19
SPEAKER_00: well,
1:09:19
SPEAKER_00: I was skeptical that this could ever work.
1:09:22
but it actually is good enough that I don't want to work
1:09:24
without it anymore.
1:09:25
SPEAKER_01: Yeah,
1:09:25
because the surprising thing for me is most of us remember the
1:09:29
launch of ChatGPT in
1:09:31
SPEAKER_01: November 2022,
1:09:31
SPEAKER_01: and everything exploded from there.
1:09:34
SPEAKER_01: It was the product that made people go,
1:09:36
SPEAKER_01: wow,
1:09:37
SPEAKER_01: inside tech,
1:09:37
SPEAKER_01: outside tech,
1:09:38
SPEAKER_01: and a lot of AI startups started up,
1:09:39
SPEAKER_01: and a lot of actually coding startups started after that,
1:09:42
SPEAKER_01: so like in early 2023.
1:09:44
SPEAKER_01: However, by that time, GitHub Copilot was already, I think it
1:09:49
SPEAKER_01: It worked with beta in 2021, right?
1:09:50
SPEAKER_01: But in 2022, I think it became public.
1:09:53
SPEAKER_00: So it was GA already?
1:09:54
SPEAKER_00: It was generally available, yeah.
1:09:56
SPEAKER_00: It was generally available in June 2022,
1:09:58
SPEAKER_00: and we started charging for it for individuals in August 2022.
1:10:02
SPEAKER_00: And then ChatGPT actually had two moments, but one of them was...
1:10:07
SPEAKER_00: Not noticed by many.
1:10:08
SPEAKER_00: So OpenAI showed at Microsoft Ignite conference,
1:10:12
SPEAKER_00: which is the fall event,
1:10:13
SPEAKER_00: showed actually ChatGPT in the keynote.
1:10:16
SPEAKER_00: It wasn't called ChatGPT,
1:10:18
but they showed a similar scenario that is what you see now the
1:10:22
agent doing,
1:10:23
SPEAKER_00: where you ask it to write some code and make modifications to that code,
1:10:26
SPEAKER_00: and then it launched.
1:10:28
SPEAKER_00: in late November, and the world changed for us dramatically.
1:10:32
SPEAKER_00: For once,
1:10:32
SPEAKER_00: because we had a product and market that was really good for developers,
1:10:37
and so our customer conversations shifted from us having to
1:10:40
convince customers that
1:10:41
AI is actually something they should look into to customers
1:10:45
asking us to come
1:10:47
SPEAKER_00: talk about how we build Copilot, how it can help their developers.
1:10:50
SPEAKER_00: The first, you know, user studies come in.
1:10:52
SPEAKER_00: We were the first one,
1:10:54
the first company comparing a group without Copilot with a
1:10:57
group with Copilot.
1:10:58
And the 5% number is obviously,
1:11:00
you know,
1:11:02
both well known and often used today as a...
1:11:06
SPEAKER_00: productivity improvement, which was never intended to be.
1:11:09
SPEAKER_00: It's a case study in that month scenario.
1:11:11
SPEAKER_00: It made the developers 55% faster, right?
1:11:13
SPEAKER_00: There's lots of other things that developers do day in,
1:11:15
SPEAKER_00: day out where you don't get the co-pilot to make you that much faster.
1:11:20
SPEAKER_00: But it was a very exciting time.
1:11:21
And then in 2023,
1:11:22
we launched a chat,
1:11:24
GPT-4,
1:11:25
brought it into the CLI,
1:11:26
voice came out,
1:11:27
right?
1:11:27
SPEAKER_00: It's
1:11:28
SPEAKER_00: A lot of these things feel like they have been around forever,
1:11:31
SPEAKER_00: but it's only been a little bit over two years.
1:11:33
SPEAKER_00: It's only been.
1:11:34
SPEAKER_00: And then 2023,
1:11:36
SPEAKER_01: on a post from GitHub,
1:11:39
SPEAKER_01: I'm not sure if it was you,
1:11:40
SPEAKER_01: so I don't want to misinterpret you,
1:11:41
SPEAKER_01: but it was written that just as GitHub was founded on Git,
1:11:44
SPEAKER_01: today we are refounded.
1:11:45
SPEAKER_00: Sorry,
1:11:45
SPEAKER_01: so it was you writing it that just as GitHub was founded on Git,
1:11:48
SPEAKER_01: today we are refounded on Copilot.
1:11:50
SPEAKER_01: This was in 2023, and Copilot was clearly very strong.
1:11:53
SPEAKER_01: What has this change meant since and now it's been two years?
1:11:57
SPEAKER_00: This was at GitHub Universe 2023.
1:12:00
SPEAKER_00: So I think in that year, the conference was in early November.
1:12:03
SPEAKER_00: It is our biggest event of the year,
1:12:05
and it was one of the lines in the keynote and I think in the
1:12:07
press release or blog
1:12:08
SPEAKER_00: post.
1:12:08
SPEAKER_00: And we talked a lot about the early days of GitHub, right?
1:12:11
SPEAKER_00: The GitHub founders did not invent Git.
1:12:14
SPEAKER_00: Linux did.
1:12:16
But they realized there's a new technology that changes how
1:12:19
developers work,
1:12:20
SPEAKER_00: not only to version the code,
1:12:22
SPEAKER_00: but how they collaborate, right?
1:12:24
GitHub,
1:12:25
if we think about it,
1:12:26
GitHub is designed for asynchronous collaboration of
1:12:30
developers.
1:12:30
SPEAKER_00: That's what GitHub is really all about.
1:12:33
SPEAKER_00: And with Copilot repeated history,
1:12:35
SPEAKER_00: because we didn't invent transformers,
1:12:37
SPEAKER_00: we didn't invent large language models,
1:12:39
SPEAKER_00: we didn't train GPT-2 or 3 or codecs.
1:12:42
SPEAKER_00: But we saw something that changes how developers work.
1:12:47
And,
1:12:47
you know,
1:12:47
if you're honest,
1:12:48
we saw an opportunity for ourselves to accelerate.
1:12:51
SPEAKER_00: You know, my backlog is endless.
1:12:54
There's lots of GitHub issues in repositories that were filed
1:12:57
where the issue was
1:12:58
SPEAKER_00: filed in 2014 with customer feedback and,
1:13:01
SPEAKER_00: you know,
1:13:02
Every enterprise seller since then posted on it,
1:13:05
I have the same problem.
1:13:06
SPEAKER_00: And it's not like we don't want to get to that issue,
1:13:08
but it's like there's also a thousand other things that are
1:13:11
also a high priority.
1:13:13
And we mentioned security issues and availability and all these
1:13:16
things are also high priority.
1:13:17
SPEAKER_00: When we saw GP3 and then Codex,
1:13:20
we saw an opportunity to bring the effort down for our own
1:13:23
developers and,
1:13:24
SPEAKER_00: of course,
1:13:25
SPEAKER_00: every developer on the planet to get more done
1:13:29
SPEAKER_00: Not that they actually get to an empty backlog,
1:13:31
SPEAKER_00: because I don't believe that,
1:13:32
SPEAKER_00: because Javon's paradox means more ideas will get stacked.
1:13:35
SPEAKER_00: In fact, now we have all the AI ideas.
1:13:38
SPEAKER_00: If you will,
1:13:39
now GitHub and Microsoft are closer together from the sense
1:13:43
that Microsoft's
1:13:43
SPEAKER_00: developer division
1:13:45
you know,
1:13:45
has for over 25 years produced IDEs and Copilot started as an
1:13:49
IDE feature.
1:13:50
SPEAKER_00: And while GitHub was the developer platform,
1:13:52
SPEAKER_00: now those are,
1:13:53
SPEAKER_00: you know,
1:13:53
SPEAKER_00: the Venn diagram of GitHub and VS Code and VS is much more overlapping.
1:13:58
SPEAKER_00: And we see that here at the conference and we see that in our day-to-day
1:14:01
SPEAKER_00: interaction between GitHub and DevDiv.
1:14:03
SPEAKER_01: And just on build, two big announcements related to GitHub.
1:14:08
SPEAKER_01: One is open sourcing copilot extension.
1:14:11
SPEAKER_01: Why did you decide to do that?
1:14:14
SPEAKER_01: So what's the idea here?
1:14:17
SPEAKER_01: Because we have seen open sourcing with VS Code.
1:14:19
SPEAKER_01: In fact,
1:14:20
open sourcing VS Code gave way to some of the competing editors
1:14:25
that are also
1:14:26
SPEAKER_01: adopted,
1:14:27
SPEAKER_01: Windsurf,
1:14:27
SPEAKER_01: Cursor.
1:14:28
SPEAKER_01: I think there's smaller ones as well.
1:14:31
SPEAKER_01: What is your hope with this move?
1:14:33
SPEAKER_01: And why?
1:14:34
It feels counterintuitive because Copilot has kind of been an
1:14:38
edge for Microsoft,
1:14:39
right?
1:14:40
SPEAKER_00: Isn't it funny that it's counterintuitive?
1:14:42
We live in a world now where Microsoft is embracing open source
1:14:46
and is open
1:14:47
SPEAKER_00: sourcing its plant while startups are taking open source,
1:14:50
SPEAKER_00: forking it,
1:14:50
SPEAKER_00: and making it closed source.
1:14:52
SPEAKER_00: This is like if you had told me that 10 years ago,
1:14:54
SPEAKER_00: you would have said you'll have it backwards,
1:14:57
SPEAKER_00: buddy.
1:14:59
SPEAKER_00: And look, we published VS Code in 2015 under the MIT license.
1:15:06
SPEAKER_00: So the license allows it.
1:15:07
SPEAKER_00: And of course, people can do with the VS Code code base what they want.
1:15:11
SPEAKER_00: That's fair game.
1:15:12
And so we're stepping into our own footsteps of VS Code over
1:15:18
the last 10 years.
1:15:20
And we felt the time is right to bring the co-pilot extension
1:15:24
into the main
1:15:25
SPEAKER_00: editor code base.
1:15:27
And that means it needs to be open source because we made a
1:15:31
promise to the
1:15:32
SPEAKER_00: community when we published VS Code that we will keep it open source.
1:15:35
SPEAKER_00: You mentioned we copiloted in 2021, 2022.
1:15:39
It didn't take long until people reverse engineered the VS Code
1:15:42
extension at the
1:15:44
SPEAKER_00: end of the day JavaScript.
1:15:45
We were writing it in TypeScript,
1:15:46
but it's delivered as obfuscated JavaScript.
1:15:49
And so you still can find those blog posts of how we compose
1:15:53
the prompt,
1:15:54
how long the prompt,
1:15:55
SPEAKER_00: The context window is, you know, that we're using adjacent tabs.
1:15:57
SPEAKER_00: All that information is already out there,
1:15:59
SPEAKER_00: but you couldn't take that information and contribute back or,
1:16:04
SPEAKER_00: you know,
1:16:04
SPEAKER_00: look at the code base,
1:16:05
SPEAKER_00: learn from it,
1:16:05
SPEAKER_00: or take,
1:16:07
SPEAKER_00: you know,
1:16:07
SPEAKER_00: Copilot and its APIs and its system prompt and put that into
1:16:11
let's say code.org or the interview tool you're building
1:16:14
because you want your
1:16:15
applicants to use a copilot and maybe even show how they can
1:16:18
sign in with GitHub at
1:16:19
SPEAKER_00: the same time,
1:16:20
SPEAKER_00: not given the full VS Code experience,
1:16:22
SPEAKER_00: you want that scope experience for an interview loop.
1:16:24
SPEAKER_00: So I think from our perspective,
1:16:26
SPEAKER_00: it follows exactly the idea of open source,
1:16:28
SPEAKER_00: which is you can learn from it,
1:16:29
SPEAKER_00: you can contribute to it,
1:16:30
SPEAKER_00: you can fork it,
1:16:31
SPEAKER_00: you can do whatever you want with it.
1:16:33
It ultimately helps the developer community to have that piece
1:16:37
of client code available.
1:16:39
SPEAKER_00: And
1:16:40
SPEAKER_00: you know,
1:16:40
SPEAKER_00: now you're going to ask,
1:16:42
SPEAKER_00: or I'm going to ask myself the question,
1:16:43
SPEAKER_00: so where's the value generated and aren't we losing all the revenue?
1:16:47
And we believe that the ultimate value generation for these AI
1:16:51
tools is in the
1:16:52
SPEAKER_00: platform,
1:16:52
SPEAKER_00: whether it's the computer inference layer,
1:16:55
SPEAKER_00: whether it's the security or control layer,
1:16:57
SPEAKER_00: right?
1:16:58
SPEAKER_00: Enterprise is one control of what model is enabled,
1:17:00
SPEAKER_00: who can use AI,
1:17:03
SPEAKER_00: in which project you might use AI,
1:17:05
given that you might have signed legal terms with whoever
1:17:08
provided your source code
1:17:09
SPEAKER_00: that you aren't using certain technology,
1:17:12
SPEAKER_00: and inference layer,
1:17:15
SPEAKER_00: the security layer,
1:17:16
SPEAKER_00: and at the collaboration layer where people collaborate,
1:17:19
SPEAKER_00: and you see that with human-to-human collaboration,
1:17:23
And we were going to see that with human-to-agent collaboration
1:17:25
with our new coding agent.
1:17:26
We're bringing that agent into the Gear platform and we're
1:17:29
giving it certain permissions.
1:17:31
SPEAKER_00: And we are protecting the agent from doing things that you might,
1:17:34
SPEAKER_00: you know,
1:17:35
SPEAKER_00: allow for a human,
1:17:36
SPEAKER_00: but you want control over that when the agent does it.
1:17:39
SPEAKER_01: Looking back at the history of Microsoft,
1:17:40
I do appreciate that Microsoft has been open always to the idea
1:17:44
of being a
1:17:45
SPEAKER_01: platform,
1:17:45
SPEAKER_01: but a good platform.
1:17:46
SPEAKER_01: For example,
1:17:47
SPEAKER_01: JetBrains,
1:17:47
SPEAKER_01: they might disagree with this,
1:17:49
but I remember the first time I bought a JetBrains project for
1:17:52
$200 a year was this
1:17:54
SPEAKER_01: extension called ReSharper.
1:17:56
So Microsoft opened up their IDE to allow extension points and
1:18:00
JetBrains honestly
1:18:01
SPEAKER_01: rolled a way better autocomplete than Microsoft themselves had.
1:18:03
SPEAKER_01: And they made a bunch of money off of it.
1:18:05
SPEAKER_01: And I think this helped them spin out these things.
1:18:07
SPEAKER_01: But when I look at,
1:18:08
SPEAKER_01: let's say,
1:18:08
SPEAKER_01: Apple's Xcode,
1:18:09
SPEAKER_01: the reason I don't like it and I'm very vocal about it,
1:18:11
SPEAKER_01: there is no extensions.
1:18:12
SPEAKER_01: I'm stuck with what it is.
1:18:14
SPEAKER_01: I would sometimes pay money for someone I cannot.
1:18:17
SPEAKER_01: So I just really like how...
1:18:20
SPEAKER_01: Microsoft seems to be keeping this idea of,
1:18:22
SPEAKER_01: like,
1:18:22
SPEAKER_01: sometimes opening up the business so others can succeed,
1:18:25
SPEAKER_01: startups can do it.
1:18:26
SPEAKER_01: It's better for everyone, especially as for developers.
1:18:29
SPEAKER_01: We just want better tools.
1:18:30
SPEAKER_00: As you mentioned, Xcode, we actually have a Copilot extension for Xcode.
1:18:33
SPEAKER_00: It's a bit of, you know, hackery given the extensions.
1:18:36
SPEAKER_00: Given the limitations, yeah.
1:18:38
SPEAKER_00: But it does have auto-completion,
1:18:39
SPEAKER_00: it has chat,
1:18:40
and just this week also we ship agent mode as a preview into
1:18:43
the Xcode extension.
1:18:45
We have it in the JetBrains extension as well,
1:18:47
across all the different JetBrains ADEs.
1:18:50
We even bring it to Eclipse,
1:18:52
and that is in preview now,
1:18:54
including agent mode.
1:18:54
SPEAKER_00: Because if you look into larger companies,
1:18:58
They have developers on all of these and probably another dozen
1:19:03
of other editors
1:19:03
that developers prefer and where the company doesn't want to
1:19:06
prescribe which IDE to
1:19:08
SPEAKER_00: use.
1:19:08
But it is important for companies to standardize their
1:19:11
engineering system stack and
1:19:12
SPEAKER_00: to have certain approved tools.
1:19:15
SPEAKER_00: Maybe it's multiple of those.
1:19:16
SPEAKER_00: Obviously,
1:19:17
SPEAKER_00: if you think about,
1:19:18
SPEAKER_00: for example,
1:19:18
SPEAKER_00: Artifactory or JFrog,
1:19:20
SPEAKER_00: often companies have that in the stack together with GitHub.
1:19:23
or hashi cop terraform and of course you have a cloud and a key
1:19:26
key store a secret
1:19:28
store and a key value store and whatnot but it is uh important
1:19:31
to have uh copilot
1:19:32
available across all these ies and our xcode extension is
1:19:36
already open source and
1:19:37
was open source before this week because an open source
1:19:40
maintainer avoided and then
1:19:42
SPEAKER_00: we did a commercial agreement with that open source maintainer
1:19:45
SPEAKER_00: To take it over, right?
1:19:46
SPEAKER_00: So we, you know, the maintainer benefited from it.
1:19:49
SPEAKER_00: We benefited from it.
1:19:50
SPEAKER_00: But so that means the system prompt,
1:19:51
SPEAKER_00: the APIs,
1:19:52
SPEAKER_00: all that,
1:19:53
SPEAKER_00: you can already find in Xcode extension.
1:19:55
SPEAKER_00: And so now we're continuing that journey with VS Code and others.
1:19:59
SPEAKER_01: The last question I wanted to touch on, this is,
1:20:02
Agent mode or GitHub agents was also an announcement and the
1:20:07
agents are getting
1:20:08
SPEAKER_01: more and more autonomous now.
1:20:10
SPEAKER_01: As developers, there is a bit of this fear.
1:20:12
SPEAKER_01: There's some,
1:20:14
SPEAKER_01: not necessarily Microsoft,
1:20:14
SPEAKER_01: but there's this vision of in the future,
1:20:16
SPEAKER_01: a software engineer,
1:20:17
you will manage all these agents and you'll be more of a
1:20:19
manager than a software
1:20:20
SPEAKER_01: engineer.
1:20:21
SPEAKER_01: And there's two things here.
1:20:22
A,
1:20:22
a fear of like,
1:20:23
are they taking away the parts of the work that I like?
1:20:26
SPEAKER_01: Yeah.
1:20:27
Even if not my job,
1:20:28
I'll have a job,
1:20:28
but it just seems like a boring job to be a manager.
1:20:31
And then the other thing is,
1:20:34
what could happen with software engineering jobs?
1:20:36
SPEAKER_01: And then you said earlier that you believe this will increase,
1:20:39
SPEAKER_01: actually,
1:20:39
SPEAKER_01: just recently in public.
1:20:41
SPEAKER_01: So what is your view on agents,
1:20:43
and what advice would you give to software engineers who want
1:20:46
to say software
1:20:47
SPEAKER_01: engineers,
1:20:47
SPEAKER_01: but obviously want to up-level?
1:20:50
SPEAKER_01: How should we look at this thing?
1:20:51
SPEAKER_01: Because it feels like a frigging big change.
1:20:52
SPEAKER_00: I don't like the term autonomous because I think autonomous,
1:20:55
SPEAKER_00: at least in my understanding,
1:20:56
SPEAKER_00: means the thing figures out itself what it should work on.
1:21:00
SPEAKER_00: That's not how these things work.
1:21:02
They have a level of autonomy when you assign a task to them of
1:21:05
how they're
1:21:06
SPEAKER_00: implementing the task.
1:21:07
SPEAKER_00: But even that autonomy is restricted by the MCP service,
1:21:12
a model context protocol that you configure for it or the tools
1:21:14
you allow it to use
1:21:15
SPEAKER_00: or the repository you give it access to because obviously as an engineer,
1:21:20
SPEAKER_00: it's a human engineer,
1:21:22
I can look beyond the one depository boundary and see what has
1:21:26
happened in other
1:21:27
SPEAKER_00: depositories while the agent needs to get permission to do that.
1:21:33
SPEAKER_00: We are definitely far away from full self-driving.
1:21:37
SPEAKER_00: If you take cars as a metaphor here,
1:21:39
SPEAKER_00: we're certainly getting into more driver assistance systems.
1:21:43
SPEAKER_00: As developers, we have always automated whatever we can automate.
1:21:46
SPEAKER_00: A compiler is a tool that helps me to not write assembler.
1:21:52
SPEAKER_00: as a language.
1:21:52
SPEAKER_00: Thank God.
1:21:54
But I'm sure when compilers and languages like BASIC and others
1:21:58
were introduced,
1:21:59
SPEAKER_00: there were also skeptics saying,
1:22:01
I'm losing control over the instruction set and the
1:22:05
optimizations that I can do,
1:22:06
SPEAKER_00: and my app is so much faster.
1:22:07
And I certainly remember my Commodore 64 days when there was a
1:22:12
very active demo
1:22:13
SPEAKER_00: scene,
1:22:14
SPEAKER_00: and you couldn't really have a successful demo.
1:22:17
SPEAKER_00: And I think even nowadays, this is where the size was limited.
1:22:20
SPEAKER_00: Back then,
1:22:21
SPEAKER_00: Poor physical limitations, today it's artificial.
1:22:23
SPEAKER_00: You're limiting it because you want to create a challenge,
1:22:26
but you could only win those demo challenges if you knew how to
1:22:29
do Assembler,
1:22:29
SPEAKER_00: and you couldn't win that in BASIC.
1:22:32
SPEAKER_00: So agents will work as one of the tools that developers have available.
1:22:38
SPEAKER_00: Just like we have compilers, linkers, linters, and whatnot.
1:22:41
SPEAKER_00: They will pick up tasks that I don't want to do, like writing test cases.
1:22:45
SPEAKER_00: I don't know many developers that want to write test cases,
1:22:47
SPEAKER_00: and certainly not more than they've already written.
1:22:51
SPEAKER_00: Unit testing is always a, how many do I write until...
1:22:55
SPEAKER_00: I can't get away with it when I go into code review.
1:22:57
SPEAKER_00: Because there isn't like a right answer.
1:22:59
SPEAKER_00: It's like not a dozen and you're good.
1:23:00
SPEAKER_00: It's like how many you need is actually a very subjective thing.
1:23:05
SPEAKER_00: And those have tried to holistically test all the edge cases.
1:23:09
SPEAKER_00: They're never shipping actual valuable software.
1:23:12
SPEAKER_00: Writing documentation.
1:23:14
SPEAKER_00: fixing security vulnerabilities,
1:23:15
finding these security vulnerabilities when I didn't see them
1:23:18
because I had a long
1:23:18
day and I have obviously always my own bias that the code that
1:23:24
I'm writing does
1:23:25
exactly what I want it to be until you have even something
1:23:27
simple like a pull
1:23:28
SPEAKER_00: request description,
1:23:29
SPEAKER_00: right?
1:23:29
SPEAKER_00: Like you have the
1:23:31
SPEAKER_00: AI describe the work that you have done,
1:23:33
SPEAKER_00: and all of a sudden it shows,
1:23:34
SPEAKER_00: well,
1:23:34
SPEAKER_00: Thomas broke the behavior.
1:23:35
SPEAKER_00: It won't say it like that, but you get the idea.
1:23:39
SPEAKER_00: Thomas changed the behavior of the checkout flow where I'm like,
1:23:42
SPEAKER_00: wait a second,
1:23:42
SPEAKER_00: I didn't touch the checkout flow.
1:23:44
SPEAKER_00: Why is this now changed?
1:23:46
SPEAKER_00: Because when it describes the code changes,
1:23:48
it doesn't have the confirmation bias that you self had when
1:23:51
you would write your
1:23:52
SPEAKER_00: own description.
1:23:52
SPEAKER_00: So I think we're heading into that world,
1:23:54
SPEAKER_00: and now you have these,
1:23:55
SPEAKER_00: let's say,
1:23:56
SPEAKER_00: I start four agent flows.
1:23:57
SPEAKER_00: I think that's what...
1:23:59
SPEAKER_00: Jessica Dean did in the day two keynote.
1:24:02
SPEAKER_00: But now we have all this code generated by all these agents.
1:24:05
I'm the one now having to validate that what the agent did is
1:24:08
actually the thing
1:24:09
SPEAKER_00: that I'm willing to merge into my code base.
1:24:11
We all know that this gets harder the more open pull requests
1:24:14
you have in different
1:24:15
branches that might have merge conflicts by the time you're
1:24:18
merging those things.
1:24:19
SPEAKER_00: So you've got to understand all that code.
1:24:21
SPEAKER_00: You've got to understand how the system is designed.
1:24:23
You need to know what the agent did to have the level of
1:24:27
confidence,
1:24:27
trust.
1:24:28
SPEAKER_00: We talked about security, availability, all these things earlier.
1:24:31
SPEAKER_00: Scale.
1:24:32
SPEAKER_00: At the end of the day,
1:24:33
SPEAKER_00: most companies are for-profit organizations,
1:24:35
SPEAKER_00: and even those that are non-profit operating under budget.
1:24:38
And so you need the agent to write code that solves the
1:24:40
business problem and still
1:24:41
SPEAKER_00: makes you a healthy margin.
1:24:43
Sounds like keep an open mind and experiment and see how they
1:24:47
can help you,
1:24:48
right?
1:24:48
SPEAKER_00: The other thing that is,
1:24:50
SPEAKER_00: I think,
1:24:50
super important to this is that we're already seeing this
1:24:53
paradox because now we
1:24:54
have all the AI requirements that come on top of what a
1:24:58
full-stack engineer does,
1:24:59
SPEAKER_00: right?
1:24:59
Like no longer this backend and database and infrastructure and
1:25:02
content and how the
1:25:03
SPEAKER_00: content communicates with the backend.
1:25:05
SPEAKER_00: And, you know, at GitHub to give you one last data point is...
1:25:09
SPEAKER_00: we are seeing 120,000 API requests per second on an average day.
1:25:15
SPEAKER_02: Oh, wow.
1:25:15
SPEAKER_00: That's seven per minute.
1:25:17
SPEAKER_00: That's 430 million per hour or 10 billion per day.
1:25:22
SPEAKER_00: That's API requests.
1:25:24
And now that doesn't say,
1:25:25
is it a cheap API,
1:25:26
like listing the first 10 repos that you have?
1:25:29
Or is it a GraphQL API that can take lots of complex queries to
1:25:33
actually get another result?
1:25:35
SPEAKER_00: If you just take that example, you've got to have engineering skills.
1:25:39
SPEAKER_00: You've got to have developed craft.
1:25:41
SPEAKER_00: You need senior people that know how to build large-scale systems.
1:25:44
You need people that take large complex problems,
1:25:47
break them down into smaller problems.
1:25:48
SPEAKER_00: Maybe they use AI along the way and say, what's a good new SQL database?
1:25:53
SPEAKER_00: What's a good infrastructure?
1:25:55
SPEAKER_00: How do I do data residency?
1:25:56
SPEAKER_00: How do I do GDPR?
1:25:57
SPEAKER_00: That job is going to be break it down to the layer,
1:26:01
but then now I can sign it to the agent that will provide high
1:26:04
quality code without
1:26:06
SPEAKER_00: me having to reprompt it 15 times.
1:26:08
SPEAKER_00: And then I can merge that into my code base and move to the next task.
1:26:11
SPEAKER_00: That's what engineering is all about.
1:26:13
SPEAKER_00: The coding skill will be part of that engineering skill set,
1:26:16
but ultimately engineering means I can build a really large
1:26:19
complex system and then
1:26:20
SPEAKER_00: evolve that into even larger system next week in today's world.
1:26:26
SPEAKER_01: Love this takeaway.
1:26:27
SPEAKER_01: Abstractions and being able to take on more complexity.
1:26:30
SPEAKER_01: This was a really good conversation.
1:26:31
SPEAKER_00: Yeah, awesome.
1:26:32
SPEAKER_00: Really great to meet you.
1:26:33
SPEAKER_00: Great person and looking forward to stay in touch.
1:26:35
SPEAKER_01: Thanks very much to Thomas for this conversation.
1:26:38
The thing that struck me most is how despite GitHub building AI
1:26:41
tools like Copilot,
1:26:43
Thomas very strongly believes that there's a lot of value in
1:26:46
hiring junior
1:26:47
SPEAKER_01: engineers.
1:26:48
And he thinks that the future of software engineering will not
1:26:50
be about autonomous
1:26:51
SPEAKER_01: AI agents,
1:26:52
but software engineers directing AI agents on what kind of work
1:26:56
to do and staying
1:26:57
SPEAKER_01: in charge of this work.
1:26:58
SPEAKER_01: For more practical reading and listening to on AI engineering,
1:27:01
check out Deep Dice from the Pragmatic Engineer linked in the
1:27:04
show notes below.
1:27:05
SPEAKER_01: If you enjoyed the show,
1:27:06
it would mean a lot if you left a rating on the show on your
1:27:08
podcast player.
1:27:09
SPEAKER_01: This helps more people discover the show.
1:27:11
SPEAKER_01: Thanks and see you in the next one.