TRANSKRİPT 🔗
WEBVTT
00:00:00.000 --> 00:00:03.839
Most engineering leaders spent years drifting
00:00:03.839 --> 00:00:07.960
further away from the code. My guest today is
00:00:07.960 --> 00:00:11.640
a CTO who says he's written more code this year
00:00:11.640 --> 00:00:15.640
than in the rest of his career combined. Not
00:00:15.640 --> 00:00:18.960
production code for his team, he's built his
00:00:18.960 --> 00:00:23.160
own software factory. A customer files a bug,
00:00:23.379 --> 00:00:27.300
an agent opens the pull request, tests it, scores
00:00:27.300 --> 00:00:30.929
it, and ships it if it passes. At his company,
00:00:31.309 --> 00:00:35.670
pull requests are up almost 80% this year, and
00:00:35.670 --> 00:00:39.789
quality went up too. But he also says he doesn't
00:00:39.789 --> 00:00:43.909
need junior engineers to fix bugs anymore. So
00:00:43.909 --> 00:00:46.950
who trains them? I'm Brian Teller from Teller's
00:00:46.950 --> 00:00:49.392
Tech, and this is Ship It Weekly.
00:01:06.287 --> 00:01:10.230
Welcome back to Ship It Weekly, where I filter the noise and
00:01:10.230 --> 00:01:13.049
focus on what matters when you are the one running
00:01:13.049 --> 00:01:16.730
infrastructure and owning reliability. Most weeks,
00:01:16.849 --> 00:01:21.469
it's a quick DevOps, SRE, platform, cloud, and
00:01:21.469 --> 00:01:25.269
security news recap. In between those, I do conversation
00:01:25.269 --> 00:01:28.670
episodes with people building and operating these
00:01:28.670 --> 00:01:32.170
systems. Ship It Weekly turns one this month.
00:01:32.409 --> 00:01:35.430
To celebrate, through mid-December, you are
00:01:35.430 --> 00:01:38.569
getting two conversations a week, Mondays and
00:01:38.569 --> 00:01:42.290
Wednesdays, with the news every Friday. Then,
00:01:42.329 --> 00:01:46.640
we're back to our usual rhythm. Today, I'm joined by
00:01:46.640 --> 00:01:51.359
Ryan Packer, CTO at BambooHR and a former VP
00:01:51.359 --> 00:01:55.140
of Engineering at Adobe. Ryan has watched engineering
00:01:55.140 --> 00:01:59.599
organizations change from Omniture to Adobe to
00:01:59.599 --> 00:02:03.840
BambooHR. And right now, he's writing more code
00:02:03.840 --> 00:02:08.000
than he has in years. We talk about why a CTO
00:02:08.000 --> 00:02:12.520
building in his own sandbox helps his team and
00:02:12.520 --> 00:02:16.979
where it doesn't. What happens to CI/CD when every
00:02:16.979 --> 00:02:20.620
engineer is running a fleet of agents? How BambooHR
00:02:20.620 --> 00:02:25.770
handles HR and payroll data with AI? What
00:02:25.770 --> 00:02:28.650
changes for mid-career and junior engineers?
00:02:28.969 --> 00:02:32.569
And why he thinks every engineer will end up
00:02:32.569 --> 00:02:36.267
with a team of their own. All right, let's jump in.
00:02:40.817 --> 00:02:42.949
Ryan, thank you for joining me. Thank you
00:02:42.949 --> 00:02:45.150
so much for having me. Excited to be here. You
00:02:45.150 --> 00:02:47.050
spent a lot
Load full transcript
of your career moving from engineer
00:02:47.050 --> 00:02:50.129
into larger and larger leadership roles. Are
00:02:50.129 --> 00:02:52.229
you actually writing more code today than you
00:02:52.229 --> 00:02:55.300
were like five or 10 years ago? Yeah, absolutely.
00:02:55.759 --> 00:02:58.479
I think that's one of the most important changes
00:02:58.479 --> 00:03:01.680
for senior engineering leadership. If you think
00:03:01.680 --> 00:03:05.439
about it, most of us, whether it's a CTO position
00:03:05.439 --> 00:03:08.960
or a VP of engineering position, most of us grew
00:03:08.960 --> 00:03:13.000
up writing software. We had done the job of the
00:03:13.000 --> 00:03:16.000
people that we're managing. And that's not really
00:03:16.000 --> 00:03:18.580
the case anymore. The tools that we're using,
00:03:18.800 --> 00:03:21.139
the way that we're building software is totally
00:03:21.139 --> 00:03:24.039
different. So I think it's really important for
00:03:24.039 --> 00:03:27.060
software leaders to, even though I'm never going
00:03:27.060 --> 00:03:30.360
to go back into writing full production code
00:03:30.360 --> 00:03:33.960
for my team, it's really important to me that
00:03:33.960 --> 00:03:36.719
I'm using the same tools and learning the same
00:03:36.719 --> 00:03:38.620
new techniques that they're trying to learn.
00:03:38.740 --> 00:03:41.860
So yeah, I've spent more time coding probably
00:03:41.860 --> 00:03:45.240
in 2026 than I have in the rest of my career
00:03:45.240 --> 00:03:47.460
combined. Awesome. What kind of things are you
00:03:47.460 --> 00:03:50.300
working on or coding? A little bit of everything.
00:03:50.830 --> 00:03:53.090
I have several different hobby projects that
00:03:53.090 --> 00:03:55.550
I will build. Some of the more interesting stuff,
00:03:55.590 --> 00:03:58.449
though, is I actually have kind of my own version
00:03:58.449 --> 00:04:03.050
of BambooHR, an HRIS system. And so when there
00:04:03.050 --> 00:04:05.770
are features and things that I'm interested in,
00:04:05.889 --> 00:04:08.430
I'm not necessarily trying to ship them to production,
00:04:08.550 --> 00:04:11.689
but I will go in and actually build out features
00:04:11.689 --> 00:04:16.230
and functionality in a way that is really incredible
00:04:16.230 --> 00:04:19.959
to be able to then share the full vision, rather
00:04:19.959 --> 00:04:23.100
than writing a PRD or just sending a Slack message
00:04:23.100 --> 00:04:24.639
to somebody like, hey, what about this feature
00:04:24.639 --> 00:04:26.860
that I've been thinking about? I can actually
00:04:26.860 --> 00:04:29.699
go in and build out the functionality and say,
00:04:29.879 --> 00:04:32.560
hey, take a look at this. I know this isn't exactly
00:04:32.560 --> 00:04:35.740
how we would do it for Bamboo, but I can convey
00:04:35.740 --> 00:04:39.740
fully the ideas through code. And that's incredible.
00:04:40.360 --> 00:04:43.300
Where do you think a CTO writing code
actually
00:04:43.300 --> 00:04:46.500
helps an organization versus where do you feel
00:04:46.500 --> 00:04:47.839
like, okay, I need to get out of the way of the
00:04:47.839 --> 00:04:49.639
engineers and let them work on this specific
00:04:49.639 --> 00:04:53.079
project? Yeah, I think one of the things that
00:04:53.079 --> 00:04:57.060
you have to be really careful of as a CTO who's
00:04:57.060 --> 00:05:01.300
now jumping in and coding is thinking that you're
00:05:01.300 --> 00:05:04.100
doing the same thing that they're doing. Even
00:05:04.100 --> 00:05:07.660
though I just said you want to be trying to do
00:05:07.660 --> 00:05:09.699
similar things, you want to be using the same
00:05:09.699 --> 00:05:12.480
tools, learning the technologies. But you've
00:05:12.480 --> 00:05:14.740
got to be honest about the fact that building
00:05:14.740 --> 00:05:17.279
the example that I just gave, where I'm building
00:05:17.279 --> 00:05:20.699
a feature in my own little sandbox, I don't have
00:05:20.699 --> 00:05:23.720
to think about the same level of security, the
00:05:23.720 --> 00:05:26.800
same level of scalability. I don't have to think
00:05:26.800 --> 00:05:29.939
about a CI/CD pipeline in the same way. I'm trying
00:05:29.939 --> 00:05:32.139
to incorporate all of those best practices into
00:05:32.139 --> 00:05:35.699
everything that I do. But the reality is the
00:05:35.699 --> 00:05:37.980
software that I'm building most of the time is
00:05:37.980 --> 00:05:41.550
going to be used by... 5, 10, 20, maybe 100 people
00:05:41.550 --> 00:05:43.769
if I'm building something that's kind of interesting
00:05:43.769 --> 00:05:47.389
across some subsection of our organization. Whereas
00:05:47.389 --> 00:05:50.709
the software that my team is building is going
00:05:50.709 --> 00:05:53.310
to be used literally by millions of people. And
00:05:53.310 --> 00:05:56.889
the scale is completely different. And so I think
00:05:56.889 --> 00:05:59.269
we just have to be careful that we don't come
00:05:59.269 --> 00:06:02.889
across, even though we can now build and show
00:06:02.889 --> 00:06:05.129
things and turn things around in a really fast
00:06:05.129 --> 00:06:07.810
and exciting way, we've got to be really careful
00:06:07.810 --> 00:06:10.810
that we don't come across to our teams as thinking
00:06:10.810 --> 00:06:12.790
that we're doing the same thing that they're
00:06:12.790 --> 00:06:16.339
doing. Okay, you talked a little bit about this,
00:06:16.339 --> 00:06:17.620
but I'm curious, are you building things that
00:06:17.620 --> 00:06:19.660
are customer focused? Are you building things
00:06:19.660 --> 00:06:22.560
that are like internal engineering focused? Is
00:06:22.560 --> 00:06:24.680
that a little bit of both? Yeah, a little bit
00:06:24.680 --> 00:06:27.199
of everything. I mean, I think at this point,
00:06:27.220 --> 00:06:30.579
who hasn't written a recipe application for their
00:06:30.579 --> 00:06:33.129
family?
So, like, I've done that, too. Like,
00:06:33.129 --> 00:06:35.310
I've built the thing that lets my family vote
00:06:35.310 --> 00:06:38.990
on what they like to eat and orders the groceries
00:06:38.990 --> 00:06:42.649
from Walmart or Costco. So, you know, I've done
00:06:42.649 --> 00:06:45.050
stuff like that where it's kind of family-facing.
00:06:45.810 --> 00:06:48.810
And then internally, I've built things like,
00:06:48.810 --> 00:06:53.269
you know, organization planning in a way that
00:06:53.269 --> 00:06:55.769
we would never put into the product. It's very
00:06:55.769 --> 00:07:01.430
specific and tailored to my organization, but tools that would allow leaders
00:07:01.430 --> 00:07:04.889
to go in and create a hypothesis on what a new
00:07:04.889 --> 00:07:07.730
org structure might look like, drag and drop people
00:07:07.730 --> 00:07:10.550
around and see what the pros and cons and the
00:07:10.550 --> 00:07:14.370
like, how the workload would be distributed, how
00:07:14.370 --> 00:07:17.769
it connects to the kind of key areas of of our
00:07:17.769 --> 00:07:20.709
source code repository. So things, everything from
00:07:20.709 --> 00:07:25.009
kind of internal tooling to small kind of family-facing
00:07:25.009 --> 00:07:29.610
apps to prototyping features that would
00:07:29.610 --> 00:07:33.370
go out in the product, but I'm not really doing
00:07:33.370 --> 00:07:36.430
it for the purpose of prototyping for customers.
00:07:36.670 --> 00:07:39.050
It's for the purpose of learning in a common
00:07:39.050 --> 00:07:41.889
language and a common experience so that we can
00:07:41.889 --> 00:07:45.290
have more rich conversations about how we're
00:07:45.290 --> 00:07:48.579
building. And how has that helped you? Have you
00:07:48.579 --> 00:07:50.540
noticed a change, like talking with engineering
00:07:50.540 --> 00:07:54.060
leaders about like your experience with AI and
00:07:54.060 --> 00:07:55.939
how does that translate down to like the individual
00:07:55.939 --> 00:08:01.240
ICs? Yeah, I think it has been, for me personally,
00:08:01.540 --> 00:08:06.360
it has been game changing because again, the
00:08:06.360 --> 00:08:08.980
way that we build software is so different today.
00:08:09.100 --> 00:08:11.879
A big portion of our team is really thinking
00:08:11.879 --> 00:08:13.959
about, you know, what we would think of as the
00:08:13.959 --> 00:08:16.800
software factory, like we're building the factory.
00:08:17.550 --> 00:08:20.610
That's as important, if not more important, than
00:08:20.610 --> 00:08:22.990
the software that we're producing. The cars,
00:08:23.089 --> 00:08:26.009
I use an auto manufacturing metaphor a lot with
00:08:26.009 --> 00:08:28.970
my team. The vehicles that are coming off the
00:08:28.970 --> 00:08:32.269
assembly line may change and will change quickly.
00:08:32.850 --> 00:08:36.450
What's most important to me is getting better
00:08:36.450 --> 00:08:40.370
at getting better around how we build. And so
00:08:40.370 --> 00:08:43.830
for me to be able to speak credibly to that.
00:08:44.409 --> 00:08:47.629
I've built my own kind of software factory. And
00:08:47.629 --> 00:08:50.929
again, I know that it's not as big and robust,
00:08:51.029 --> 00:08:53.690
scalable, as complex as what we have to build
00:08:53.690 --> 00:08:57.809
at BambooHR to scale our overall R&D efforts.
00:08:58.149 --> 00:09:02.950
But I know what it means to build software where
00:09:02.950 --> 00:09:07.809
a customer can submit a bug in the product that
00:09:07.809 --> 00:09:09.889
captures all of the information in the product
00:09:09.889 --> 00:09:13.110
that then is automatically picked up by an agent
00:09:13.870 --> 00:09:17.850
who creates the PR, who then tests the PR with
00:09:17.850 --> 00:09:21.070
various security, performance, all of these things,
00:09:21.169 --> 00:09:25.289
scores the PR, and releases it back to production
00:09:25.289 --> 00:09:28.070
if it meets certain criteria. So having seen
00:09:28.070 --> 00:09:31.269
that process end-to-end, while acknowledging
00:09:31.269 --> 00:09:34.789
that it's not the same, it does give you, I think,
00:09:34.809 --> 00:09:39.549
a much more credible view on what's possible.
00:09:39.950 --> 00:09:43.620
And I hope that's appreciated by my team. You
00:09:43.620 --> 00:09:45.559
know, you kind of alluded to this before. There
00:09:45.559 --> 00:09:48.570
is a little bit of a double-edged sword. You
00:09:48.570 --> 00:09:50.250
don't want to get in the way. You don't want
00:09:50.250 --> 00:09:53.409
to come across as thinking that your sandbox
00:09:53.409 --> 00:09:58.370
experiments equate to a full blown production
00:09:58.370 --> 00:10:01.809
environment. But I think, you know, I hope it
00:10:01.809 --> 00:10:06.570
helps me to stay current and to be able to just
00:10:06.570 --> 00:10:10.649
understand that I'm struggling. And I guess I'll
00:10:10.649 --> 00:10:14.470
say it this way. I hope that my team can see
00:10:14.470 --> 00:10:17.169
that I'm struggling to keep up with it all just
00:10:17.169 --> 00:10:19.269
as much as they are. Because that's one of the
00:10:19.269 --> 00:10:22.750
hardest things right now. The pace of change
00:10:22.750 --> 00:10:26.710
is relentless. And there are plenty of days that
00:10:26.710 --> 00:10:30.049
I feel completely overwhelmed. If I'm just sitting
00:10:30.049 --> 00:10:31.970
back and saying, like, my team's going to figure
00:10:31.970 --> 00:10:34.730
that out, I don't feel like I'm in it with them.
00:10:34.990 --> 00:10:39.830
And so if nothing else, by being in the trenches
00:10:39.830 --> 00:10:42.789
together, learning these new technologies. I
00:10:42.789 --> 00:10:46.470
spent the weekend playing with Jev. And, you
00:10:46.470 --> 00:10:48.190
know, a week ago, nobody knew what that was.
00:10:48.850 --> 00:10:52.429
Every week there's something new, right? And
00:10:52.429 --> 00:10:56.669
so I hope
it gives me more credibility with my
00:10:56.669 --> 00:10:58.590
team. What have you even seen in your organization
00:10:58.590 --> 00:11:02.350
as far as how has it changed development or software
00:11:02.350 --> 00:11:04.750
engineering in general over the last, like, let's
00:11:04.750 --> 00:11:08.370
say three years? Yeah, I mean, three years is
00:11:08.370 --> 00:11:11.789
like an eternity these days. I'm measuring time
00:11:11.789 --> 00:11:15.370
in like days and weeks. But even since the beginning
00:11:15.370 --> 00:11:18.950
of this year, I'll give an example. The number
00:11:18.950 --> 00:11:21.429
of pull requests that a team produces is certainly
00:11:21.429 --> 00:11:23.850
not the most important metric in the world. And
00:11:23.850 --> 00:11:27.169
there are lots of reasons why it's flawed. But
00:11:27.169 --> 00:11:31.690
it is a measure of productivity. And as we've
00:11:31.690 --> 00:11:34.840
moved to more and more agentic engineering, the
00:11:34.840 --> 00:11:37.460
number of PRs has steadily increased throughout
00:11:37.460 --> 00:11:41.179
the year. So we're up almost 80% from where
00:11:41.179 --> 00:11:44.120
we were at the beginning of the year. So there's
00:11:44.120 --> 00:11:47.679
definitely an increase in velocity. Now, the
00:11:47.679 --> 00:11:50.159
question you have to ask is, okay, what is that
00:11:50.159 --> 00:11:53.539
doing to quality? And so we've invested a lot
00:11:53.539 --> 00:11:57.159
in automating as much of the quality process
00:11:57.159 --> 00:12:00.419
as well so that we can make sure that we're not
00:12:00.419 --> 00:12:03.929
just shipping, you know, AI slop. And, you know,
00:12:03.929 --> 00:12:07.850
I'm super excited because the quality by all
00:12:07.850 --> 00:12:10.769
of the metrics that we use has gradually improved
00:12:10.769 --> 00:12:13.789
throughout the year as well. And so to see both
00:12:13.789 --> 00:12:17.629
the increase in output along with, you know,
00:12:17.649 --> 00:12:21.370
slight improvements in quality is really exciting.
00:12:21.549 --> 00:12:23.870
Now, there are a couple of ways that you can
00:12:23.870 --> 00:12:27.750
look at that. One is just you can produce more
00:12:27.750 --> 00:12:30.509
features. And, you know, I think there's something
00:12:30.509 --> 00:12:33.629
there. There's like every software product in
00:12:33.629 --> 00:12:35.389
the world. There's lots more that we want to
00:12:35.389 --> 00:12:38.090
do. There's lots that we want to add. I think
00:12:38.090 --> 00:12:41.210
the part that is more exciting, though, is we
00:12:41.210 --> 00:12:43.490
oftentimes will refer to it as getting more at-bats
00:12:43.490 --> 00:12:47.429
with our customers. Being able to go from
00:12:47.429 --> 00:12:50.909
an idea to a working customer-facing prototype
00:12:50.909 --> 00:12:55.250
in, you know, a day. And then being able to get
00:12:55.250 --> 00:12:57.769
the feedback means that the thing that we ultimately
00:12:57.769 --> 00:13:02.909
release is better. So even if the number of features
00:13:02.909 --> 00:13:06.970
that we released stayed flat, the number of PRs
00:13:06.970 --> 00:13:09.090
that actually make it into the production version
00:13:09.090 --> 00:13:13.450
of our software, if that stayed flat, I still
00:13:13.450 --> 00:13:17.289
am very confident that the quality has gone up.
00:13:17.330 --> 00:13:20.779
Not just the, you know... The quality in terms
00:13:20.779 --> 00:13:22.379
of number of bugs and defects and those kinds
00:13:22.379 --> 00:13:24.399
of things, but the quality of the experience.
00:13:24.799 --> 00:13:28.159
How well are we hitting the mark in terms of
00:13:28.159 --> 00:13:30.919
getting the right feature for the right customer?
00:13:31.080 --> 00:13:33.320
That, to me, is actually the more exciting part.
00:13:33.679 --> 00:13:36.039
Yeah, it's not just more features, trying to
00:13:36.039 --> 00:13:38.580
actually make sure it's the right features. Internally,
00:13:38.580 --> 00:13:42.659
are you noticing more robust CI/CD because platform
00:13:42.659 --> 00:13:45.419
teams have this tool now? Are we able to put
00:13:45.419 --> 00:13:48.600
more checks in place? How is it helping internally,
00:13:48.990 --> 00:13:51.509
on the actual engineering side? Yeah, if I'm
00:13:51.509 --> 00:13:53.789
honest on that one, I would say that that has
00:13:53.789 --> 00:13:56.769
been more of a challenge than a help. One of
00:13:56.769 --> 00:13:59.250
the things, so I've been at Bamboo for about
00:13:59.250 --> 00:14:01.710
three and a half years. And one of the things
00:14:01.710 --> 00:14:05.210
that surprised me the most when I joined, people
00:14:05.210 --> 00:14:08.690
have been talking CI/CD for what, 10, 15, 20 years.
00:14:09.129 --> 00:14:12.149
The reality in a lot of the organizations that
00:14:12.149 --> 00:14:14.690
I've been a part of, that I've observed, there
00:14:14.690 --> 00:14:17.490
are varying degrees of what people mean when
00:14:17.490 --> 00:14:21.429
they say, we've got a robust CI/CD pipeline. One
00:14:21.429 --> 00:14:23.850
of the things that was really surprising to me
00:14:23.850 --> 00:14:27.950
is we were releasing hundreds of times a day
00:14:27.950 --> 00:14:32.429
before I ever got here. The number of releases
00:14:32.429 --> 00:14:35.690
and the degree to which we had truly a really
00:14:35.690 --> 00:14:40.690
robust CI/CD pipeline was, I think, ahead of where
00:14:40.690 --> 00:14:43.429
most companies were. And like I said, this was
00:14:43.429 --> 00:14:45.250
all before I got here. I can't take any credit
00:14:45.250 --> 00:14:49.090
for this. What I've found is that as we have
00:14:49.090 --> 00:14:52.450
become more and more agentic, you're starting
00:14:52.450 --> 00:14:56.269
to see strain on the CI/CD pipeline and particularly
00:14:56.269 --> 00:15:00.029
on test environments, the ability to spin up
00:15:00.029 --> 00:15:03.490
enough resources.
I think we all know at this
00:15:03.490 --> 00:15:07.710
point, one of the most important factors in agents
00:15:07.710 --> 00:15:11.210
doing a good job building software is their ability
00:15:11.210 --> 00:15:14.779
to go in and test and run that software and
00:15:14.779 --> 00:15:17.759
see for themselves, validate themselves. Well, that
00:15:17.759 --> 00:15:21.159
requires us to create multiple environments. In
00:15:21.159 --> 00:15:25.379
the early days of agentic coding, usually that
00:15:25.379 --> 00:15:27.779
looked like every engineer having one or two
00:15:27.779 --> 00:15:31.019
environments on their machine. As we scale to
00:15:31.019 --> 00:15:33.360
doing more and more in parallel, and every engineer
00:15:33.360 --> 00:15:36.980
maybe has, you know, 10, 15, 20, we get to 100, 100
00:15:36.980 --> 00:15:40.000
agents working in parallel, having all of those
00:15:40.000 --> 00:15:43.750
environments that are able to be spun up really,
00:15:43.809 --> 00:15:46.750
really quickly and spun down equally as quickly,
00:15:46.929 --> 00:15:49.769
backed with real data, particularly when we're
00:15:49.769 --> 00:15:53.149
talking about use cases that really involve large
00:15:53.149 --> 00:15:56.990
amounts of customer data, that's not trivial.
00:15:57.269 --> 00:16:00.460
And so I would say, you know, I don't want to
00:16:00.460 --> 00:16:02.720
sound like I think our CI/CD process has gotten
00:16:02.720 --> 00:16:06.740
worse, but I would say AI is definitely challenging
00:16:06.740 --> 00:16:10.649
the volume that we can do. And so there's been
00:16:10.649 --> 00:16:14.450
a lot of effort to improve the CI/CD pipeline
00:16:14.450 --> 00:16:18.629
so that we can scale to the volume of full agentic
00:16:18.629 --> 00:16:21.370
engineering. Are there data privacy concerns?
00:16:21.529 --> 00:16:23.809
I'm assuming like Bamboo, you guys are dealing
00:16:23.809 --> 00:16:28.669
with PII. Is that a concern when using AI? Yeah,
00:16:28.690 --> 00:16:30.710
no, it's a huge concern. It's something that
00:16:30.710 --> 00:16:33.809
we take really, really seriously. You know, HR
00:16:33.809 --> 00:16:36.370
data is some of the most sensitive data out there.
00:16:36.990 --> 00:16:41.269
And we also have a payroll product. So thousands
00:16:41.269 --> 00:16:43.850
and thousands of people get their paychecks through
00:16:43.850 --> 00:16:47.269
us. There's not a lot that's more sensitive than
00:16:47.269 --> 00:16:50.929
your direct deposit information for your bank
00:16:50.929 --> 00:16:55.129
account. And so having that customer trust, having
00:16:55.129 --> 00:16:58.789
a really, really high... premium on security
00:16:58.789 --> 00:17:01.409
and data privacy has always been really foundational
00:17:01.409 --> 00:17:04.670
to us. We've been really fortunate to build a
00:17:04.670 --> 00:17:06.849
brand that people trust over many, many, many
00:17:06.849 --> 00:17:10.809
years, but it doesn't take a lot to, you could
00:17:10.809 --> 00:17:14.190
lose that trust quickly, I guess. So we, there
00:17:14.190 --> 00:17:15.630
are lots of different things that we do, but
00:17:15.630 --> 00:17:18.430
first of all, the way that we contract and engage
00:17:18.430 --> 00:17:22.410
with each of the frontier labs, we do a lot to
00:17:22.410 --> 00:17:25.309
protect the data in that regard. So we're very
00:17:25.309 --> 00:17:28.009
careful about what we send to them, and what
00:17:28.009 --> 00:17:31.210
terms we agree to with any particular provider
00:17:31.210 --> 00:17:34.910
in order to make sure that the obvious things,
00:17:34.990 --> 00:17:38.009
of course, that they can't train on our proprietary
00:17:38.009 --> 00:17:41.029
data and things like that, but also data retention
00:17:41.029 --> 00:17:43.890
policies and those types of things are really
00:17:43.890 --> 00:17:48.470
critical. And so we have our AI council that
00:17:48.470 --> 00:17:51.970
meets on a regular basis to think through all
00:17:51.970 --> 00:17:57.130
of these types of issues that, you know, maybe some of the things that are a challenge
00:17:57.130 --> 00:18:00.910
as you adopt AI, but it's a great question because
00:18:00.910 --> 00:18:04.730
we need to make sure that that trust stays front
00:18:04.730 --> 00:18:08.269
and center for us. Okay. So now I'm curious for
00:18:08.269 --> 00:18:11.109
like the senior engineer, if producing code itself
00:18:11.109 --> 00:18:16.160
becomes dramatically easier, what actually becomes
00:18:16.160 --> 00:18:20.380
more valuable in a senior engineer? Yeah. You
00:18:20.380 --> 00:18:24.099
know, first of all, I don't think any of us know
00:18:24.099 --> 00:18:26.140
exactly what the world is going to look like
00:18:26.640 --> 00:18:29.000
in a few months, let alone a few years, but I'll
00:18:29.000 --> 00:18:33.660
pontificate. Um, I have always said the actual
00:18:33.660 --> 00:18:37.099
writing of the code has never been the most important
00:18:37.099 --> 00:18:40.400
part of an engineer's job. I think the decisions
00:18:40.400 --> 00:18:43.079
that we make about what we're going to produce
00:18:43.079 --> 00:18:46.200
and how we're going to produce it are absolutely
00:18:46.200 --> 00:18:50.619
critical. And so, you know, will AI continue to
00:18:50.619 --> 00:18:54.200
to kind of do some of those things? Yeah, I I think
00:18:54.200 --> 00:19:01.180
it will. But our role has to be around what is
00:19:01.180 --> 00:19:03.460
it that we want to put in front of our customers.
00:19:03.900 --> 00:19:08.160
People talk a lot about taste, and I think there's
00:19:08.160 --> 00:19:10.799
a lot of validity there. It's kind of become
00:19:10.799 --> 00:19:13.819
an overused term, but I do think that there's
00:19:13.819 --> 00:19:17.480
still a huge role for people to play in deciding
00:19:17.480 --> 00:19:21.029
what to build. And so, okay, there's a lot of
00:19:21.029 --> 00:19:23.630
anxiety around like mid-career
engineers right
00:19:23.630 --> 00:19:25.750
now. Like someone who spent like, let's say like
00:19:25.750 --> 00:19:28.349
10 or 20 years getting really good at something
00:19:28.349 --> 00:19:31.170
and then AI can like suddenly do it in seconds.
00:19:31.390 --> 00:19:33.670
Do you think that that anxiety is justified?
00:19:34.390 --> 00:19:39.609
I think it's a really fair question. I definitely
00:19:39.609 --> 00:19:44.849
would say that mid-career, I think mid-career
00:19:44.849 --> 00:19:48.119
engineers are certainly in facing the hardest
00:19:48.119 --> 00:19:51.400
change. I think early career folks, they're at
00:19:51.400 --> 00:19:54.799
the beginning. They aren't set in their ways.
00:19:54.819 --> 00:19:57.339
They haven't invested a bunch of time and energy
00:19:57.339 --> 00:19:59.660
and oftentimes money into learning particular
00:19:59.660 --> 00:20:02.700
things. Late career, like, okay, they're late
00:20:02.700 --> 00:20:06.250
career. So I understand and I'm really sensitive
00:20:06.250 --> 00:20:09.309
and empathetic for folks that are in that middle
00:20:09.309 --> 00:20:11.450
portion of their career. What I would say is
00:20:11.450 --> 00:20:16.250
this, people who are resilient, people who love
00:20:16.250 --> 00:20:20.829
to learn, people who want to do things differently
00:20:20.829 --> 00:20:23.869
and are willing to embrace change, I don't think
00:20:23.869 --> 00:20:26.529
they have anything to worry about. But that's
00:20:26.529 --> 00:20:29.250
hard, and that's not everybody. I think we all
00:20:29.250 --> 00:20:33.349
want to say that we embrace change, but like
00:20:33.349 --> 00:20:36.829
human nature, we're creatures of comfort, and
00:20:36.829 --> 00:20:40.329
change is hard for a lot of us. And so I think
00:20:40.329 --> 00:20:44.410
those middle career folks... who are maybe reluctant
00:20:44.410 --> 00:20:48.650
to try things differently and are letting a very
00:20:48.650 --> 00:20:51.549
realistic fear like, okay, am I going to have
00:20:51.549 --> 00:20:55.130
a job? The thing that I'm paid to do can now
00:20:55.130 --> 00:21:00.230
be done by AI faster and if not already in the
00:21:00.230 --> 00:21:05.809
very near future can be done better by an LLM.
00:21:05.990 --> 00:21:11.799
Like, yeah, that's scary. Now, an engineer in
00:21:11.799 --> 00:21:14.279
the middle of their career who takes all of the
00:21:14.279 --> 00:21:18.119
systems thinking, the design, problem-solving
00:21:18.119 --> 00:21:21.980
thinking, takes all of that and learns how to
00:21:21.980 --> 00:21:26.339
add LLM and other AI capabilities to it, do I
00:21:26.339 --> 00:21:30.000
think they have anything to worry about? I don't.
00:21:30.000 --> 00:21:33.980
I'm in the camp of we will continue for the foreseeable
00:21:33.980 --> 00:21:37.400
future to need more software, not less software.
00:21:37.680 --> 00:21:41.259
And even though AI will dramatically magnify
00:21:41.259 --> 00:21:44.779
the amount and the pace of software, there's
00:21:44.779 --> 00:21:49.039
still tremendous need for great software engineers.
00:21:49.339 --> 00:21:52.720
A good point. Okay, so let's look at the other
00:21:52.720 --> 00:21:56.299
side of that. Junior engineers. One thing I keep
00:21:56.299 --> 00:21:58.559
coming back to is that a lot of the boring work
00:21:58.559 --> 00:22:02.420
AI is eliminating, but it's also the kind of
00:22:02.420 --> 00:22:04.640
the work that has taught us how systems behave.
00:22:05.279 --> 00:22:08.019
So what happens to junior engineers? Who trains
00:22:08.019 --> 00:22:12.980
them? Yeah, I this is this is I'll start by saying
00:22:12.980 --> 00:22:15.099
I don't have a great answer. This is something
00:22:15.099 --> 00:22:16.980
that I have been thinking about for a couple
00:22:16.980 --> 00:22:19.599
of years now. And I remember like even at a really
00:22:19.599 --> 00:22:23.369
tactical level, maybe the first thing that a
00:22:23.369 --> 00:22:25.970
junior engineer oftentimes does when they join
00:22:25.970 --> 00:22:29.630
a new company is you might hand them some bugs
00:22:29.630 --> 00:22:33.670
to go investigate. I don't need junior engineers
00:22:33.670 --> 00:22:37.509
to go fix bugs anymore. I can, by and large,
00:22:37.609 --> 00:22:40.150
have AI go take a look at those things and do
00:22:40.150 --> 00:22:43.109
as good a better job. And so I think your point
00:22:43.109 --> 00:22:46.390
is a really good one. How do we develop those
00:22:46.390 --> 00:22:50.329
engineering skills, if not through the practice
00:22:50.329 --> 00:22:54.190
of doing what the AIs have done? And again, I
00:22:54.190 --> 00:22:56.769
don't think I have a perfect answer here. Things
00:22:56.769 --> 00:22:59.289
that we talk about internally and things that
00:22:59.289 --> 00:23:02.109
we're thinking about, there's still plenty of
00:23:02.109 --> 00:23:06.519
room for the systems thinking, the designing,
00:23:06.660 --> 00:23:10.220
the architecting, and the interaction with the
00:23:10.220 --> 00:23:13.619
LLMs can still be really, really instructive.
00:23:13.660 --> 00:23:15.920
And so one of the things that I like to talk
00:23:15.920 --> 00:23:18.880
about with people that are relatively new to
00:23:18.880 --> 00:23:22.359
AI and relatively new to leveraging AI in a professional
00:23:22.359 --> 00:23:26.460
setting, make sure you understand what... maybe
00:23:26.460 --> 00:23:29.960
not 100% of the time, but make sure that you
00:23:29.960 --> 00:23:32.759
take time to understand what the LLM is doing
00:23:32.759 --> 00:23:35.900
and why it's doing it. If you completely hand
00:23:35.900 --> 00:23:39.279
it over and just say, you know, go do it, you
00:23:39.279 --> 00:23:41.940
are limiting yourself in the number of opportunities
00:23:41.940 --> 00:23:43.880
that you're going to have to learn. But that's
00:23:43.880 --> 00:23:45.799
something that I have loved as I've been building
00:23:45.799
--> 00:23:48.579
out software. Are there places where I completely
00:23:48.579 --> 00:23:51.319
hand it over to the LLM and just let it go wild
00:23:51.319 --> 00:23:53.720
and I'm shipping production code without looking
00:23:53.720 --> 00:23:56.950
at it? Yeah, there are. But there are other times
00:23:56.950 --> 00:24:00.750
when I'm very intentionally saying, okay, I want
00:24:00.750 --> 00:24:04.470
to go learn about this new technology, or I've
00:24:04.470 --> 00:24:08.329
seen the LLM do a certain thing 15 different
00:24:08.329 --> 00:24:11.369
times. Let's do it again, but this time take
00:24:11.369 --> 00:24:14.329
me through it. Tell me why you made this decision.
00:24:14.549 --> 00:24:17.970
Tell me why you picked this technology. And I
00:24:17.970 --> 00:24:21.910
have found, again, this isn't a novel idea, but
00:24:21.910 --> 00:24:26.339
the idea of asking the LLM to confront itself
00:24:26.339 --> 00:24:30.460
or creating an adversarial agent to say, come
00:24:30.460 --> 00:24:33.420
up with the five things that were done incorrectly
00:24:33.420 --> 00:24:36.259
here that you would do differently. Having the
00:24:36.259 --> 00:24:39.200
adversarial conversations amongst the agents,
00:24:39.299 --> 00:24:42.349
I think is extremely enlightening. Will that
00:24:42.349 --> 00:24:47.049
fully close the gap? I don't know. My hope is
00:24:47.049 --> 00:24:51.519
those skills, a lot of them will move further
00:24:51.519 --> 00:24:53.759
down the stack and we'll be operating at a higher
00:24:53.759 --> 00:24:56.059
level and the skills that we need to be successful
00:24:56.059 --> 00:24:59.859
will still be there. You'll have engineers that
00:24:59.859 --> 00:25:03.240
still want to write compilers and they'll work
00:25:03.240 --> 00:25:06.059
with LLMs to write the compilers, but that's
00:25:06.059 --> 00:25:08.519
going to be a subset and what they need to learn
00:25:08.519 --> 00:25:10.700
is going to be different than the engineer who
00:25:10.700 --> 00:25:14.339
is comfortable living at what we now have as
00:25:14.339 --> 00:25:17.839
this really, really high-level abstraction layer
00:25:17.839 --> 00:25:21.960
of conversational coding. Yeah. Is there something
00:25:21.960 --> 00:25:25.000
that you wouldn't hand AI now that maybe you
00:25:25.000 --> 00:25:27.759
would in three months? Yeah, that's a great question.
00:25:27.940 --> 00:25:32.339
I think the way I think about it is the biggest
00:25:32.339 --> 00:25:34.559
changes that I've seen over the last three months
00:25:34.559 --> 00:25:36.880
and that I anticipate over the next three months
00:25:36.880 --> 00:25:40.940
is kind of the size of work that we can confidently
00:25:40.940 --> 00:25:45.400
give to an LLM. So if I go back a full year,
00:25:47.210 --> 00:25:50.329
the level that I would specify my intentions,
00:25:50.609 --> 00:25:54.470
whether that be in a PRD or a spec, and spec-driven
00:25:54.470 --> 00:25:57.369
development was big
for a minute, it
00:25:57.369 --> 00:26:00.869
was quite granular. I was handing off, okay,
00:26:00.970 --> 00:26:07.490
Claude, go work on item 3.4.5. Now I'm saying,
00:26:07.569 --> 00:26:11.970
go tackle item 3, that top-level thing, like
00:26:11.970 --> 00:26:15.569
the size of the chunk of work that can be done
00:26:16.140 --> 00:26:19.380
is just so much larger. I'm sure there are other
00:26:19.380 --> 00:26:23.240
ways and types of things that LLMs will continue
00:26:23.240 --> 00:26:26.279
to get better at. I am really interested to see
00:26:26.279 --> 00:26:32.309
how Jev changes kind of the system one thinking.
00:26:32.750 --> 00:26:35.589
Most of the LLMs to date have been really system
00:26:35.589 --> 00:26:40.390
two focused. And so to see... how that changes
00:26:40.390 --> 00:26:43.569
things. I love the way that that team is talking
00:26:43.569 --> 00:26:47.529
about those really fast classification and kind
00:26:47.529 --> 00:26:51.829
of data analysis, MapReduce-like functionality.
00:26:52.529 --> 00:26:55.710
LLMs have been pretty bad at that, pretty slow,
00:26:55.789 --> 00:26:58.329
it's pretty expensive, and you see teams throwing
00:26:58.329 --> 00:27:02.289
problems at LLMs because they can, but they're
00:27:02.289 --> 00:27:04.329
not particularly good or efficient at that type
00:27:04.329 --> 00:27:07.009
of work. You know, it's obviously been out for
00:27:07.009 --> 00:27:09.609
less than a week, so it's pretty early to say
00:27:09.609 --> 00:27:12.670
that that's going to be the new thing that's
00:27:12.670 --> 00:27:15.369
going to change our world. But I'm excited to
00:27:15.369 --> 00:27:20.130
see how models like Jev will transform the type
00:27:20.130 --> 00:27:23.190
of work that you can hand off to an agent. Yeah,
00:27:23.230 --> 00:27:25.829
absolutely. Okay, so extrapolating that out,
00:27:25.869 --> 00:27:27.970
I guess, a little bit more. If these tools keep
00:27:27.970 --> 00:27:30.250
improving at anything close to the pace that
00:27:30.250 --> 00:27:32.250
we're currently at, what does the engineering
00:27:32.250 --> 00:27:34.470
organization look like in three or five years?
00:27:34.569 --> 00:27:38.950
How does it change? Yeah. I think the main thing
00:27:38.950 --> 00:27:40.789
that we see and that we're already starting to
00:27:40.789 --> 00:27:45.869
move toward is smaller teams that are then managing
00:27:45.869 --> 00:27:50.490
larger groups of agents. So it's a little bit
00:27:50.490 --> 00:27:52.990
of an unfair way to think about it, but if we
00:27:52.990 --> 00:27:57.369
started out in a world where every engineer had
00:27:57.369 --> 00:28:00.019
a tool that helped them a little bit, And then
00:28:00.019 --> 00:28:02.640
we kind of move on this journey to every engineer
00:28:02.640 --> 00:28:05.740
has a pair programmer that they're working with.
00:28:05.900 --> 00:28:07.839
And I think that's where a lot of people actually
00:28:07.839 --> 00:28:11.559
live
today. I think the future, and we've got
00:28:11.559 --> 00:28:13.180
some people that are there and some people that
00:28:13.180 --> 00:28:15.559
are still on the journey, is where every engineer
00:28:15.559 --> 00:28:19.299
has a team that works for them. And so the role
00:28:19.299 --> 00:28:22.259
of a software engineer, kind of across all of
00:28:22.259 --> 00:28:24.079
the different disciplines of software engineering,
00:28:24.339 --> 00:28:27.700
starts to look a little bit more like the role
00:28:27.700 --> 00:28:31.230
of a software engineering manager. They're getting
00:28:31.230 --> 00:28:36.730
better at delegating and consuming reports, whatever
00:28:36.730 --> 00:28:39.269
that means, from their agentic team so that they
00:28:39.269 --> 00:28:42.829
understand how the work is going. They're communicating
00:28:42.829 --> 00:28:46.349
more and more with customers. They're communicating
00:28:46.349 --> 00:28:49.329
more with their product peers. But we're definitely
00:28:49.329 --> 00:28:52.170
bought into the idea of a full-stack developer,
00:28:52.529 --> 00:28:55.829
somebody who can think across all of the different
00:28:55.829 --> 00:28:59.950
functions. But we're not there today. We have
00:28:59.950 --> 00:29:03.650
product designers who are able to pump out incredible
00:29:03.650 --> 00:29:08.269
prototypes. And if they ask the LLM to tell them
00:29:08.269 --> 00:29:11.509
how to make their prototype more secure, it can
00:29:11.509 --> 00:29:14.289
do it. But there's so many questions that we
00:29:14.289 --> 00:29:17.490
as engineers have just grown up understanding
00:29:17.490 --> 00:29:21.130
that they don't even know to ask. And so I think
00:29:21.130 --> 00:29:24.630
we still have a pretty big gap between what the
00:29:24.630 --> 00:29:29.740
LLM can do and what it will naturally do. That's
00:29:29.740 --> 00:29:33.200
where these kind of full-stack developers will
00:29:33.200 --> 00:29:36.000
continue to emerge. We think it starts out more
00:29:36.000 --> 00:29:40.059
as a pod. You've got a small team that maybe
00:29:40.059 --> 00:29:43.019
has a product manager and a designer and an engineer.
00:29:44.009 --> 00:29:47.369
And they're still working together, but they're
00:29:47.369 --> 00:29:51.430
able to do what previously a team of maybe 10
00:29:51.430 --> 00:29:56.359
to 15 R&D employees would be able to do. So
00:29:56.359 --> 00:29:58.599
that's the first thing. I think we see those
00:29:58.599 --> 00:30:01.700
teams getting smaller. One of the biggest questions
00:30:01.700 --> 00:30:05.619
that we're grappling with is how many people
00:30:05.619 --> 00:30:09.740
and agents can you have working in one code base
00:30:09.740 --> 00:30:12.480
at a time? And I think that's going to be one
00:30:12.480 --> 00:30:17.269
of the scaling limits. So I think, you know,
00:30:17.309 --> 00:30:20.130
we've already known that for humans. If you put
00:30:20.130 --> 00:30:23.250
300 software engineers on one small area of the
00:30:23.250 --> 00:30:26.410
code, it's just not going to be productive. And
00:30:26.410 --> 00:30:28.349
we can talk about the mythical man-month and
00:30:28.349 --> 00:30:31.509
all those things. But the same thing is true
00:30:31.509 --> 00:30:33.930
for agents. The ceiling might be a little bit
00:30:33.930 --> 00:30:36.549
higher, but we still have those same challenges.
00:30:36.789 --> 00:30:39.130
And so I think a lot of organizations have to
00:30:39.130 --> 00:30:42.690
think, okay, do we have opportunities to broaden
00:30:42.690 --> 00:30:46.250
the surface area such that we can bring the full
00:30:46.250 --> 00:30:50.910
force of agentic engineering to bear? I think
00:30:50.910 --> 00:30:53.210
that's where a lot of people get really worried
00:30:53.210 --> 00:30:54.910
because they say like, oh, well, I can do the
00:30:54.910 --> 00:30:57.910
same amount with less people. Aren't we going
00:30:57.910 --> 00:31:01.589
to have to lay off a bunch of people? My supposition
00:31:01.589 --> 00:31:04.769
is we're in this really fortunate position where
00:31:05.230 --> 00:31:08.029
there is nearly unlimited work to do. There's
00:31:08.029 --> 00:31:11.430
so many adjacent areas that we can go into. And
00:31:11.430 --> 00:31:14.549
if we're building a culture of incredible engineering,
00:31:14.970 --> 00:31:17.890
whether that's using LLMs or anything else, whatever
00:31:17.890 --> 00:31:21.490
the technology is to make us efficient, we can
00:31:21.490 --> 00:31:25.289
still have a really high ROI on the investment
00:31:25.289 --> 00:31:27.509
that we're making in our R&D organization. And
00:31:27.509 --> 00:31:32.789
so I actually don't see us shrinking the number
00:31:32.789 --> 00:31:36.210
of engineers across BambooHR. If we were to sit
00:31:36.210 --> 00:31:38.009
down a year from now, I expect that we would
00:31:38.009 --> 00:31:41.369
have the same or more engineers as we do today
00:31:41.369 --> 00:31:44.309
because there's just so much opportunity. But
00:31:44.309 --> 00:31:46.670
there is this interesting constraint that we're
00:31:46.670 --> 00:31:49.509
going through right now where you can't just
00:31:49.509 --> 00:31:52.890
throw an unlimited number of agents at the existing
00:31:52.890 --> 00:31:56.829
code base and expect to get the full leverage.
00:31:57.190 --> 00:32:00.670
That's a good point. Okay, so wrapping up, what's
00:32:00.670 --> 00:32:03.609
something about AI or software engineering that
00:32:03.609 --> 00:32:05.650
you believe today that you would think a lot
00:32:05.650 --> 00:32:10.079
of engineering leaders still get wrong? Oh, gosh.
00:32:11.400 --> 00:32:15.519
That's a tough one. Because I tend to think that
00:32:15.519 --> 00:32:17.819
most engineering leaders are probably smarter
00:32:17.819 --> 00:32:21.059
than I am and are doing it pretty well already.
00:32:22.019 --> 00:32:27.400
I do think that there is a reluctance to push
00:32:28.109 --> 00:32:32.109
the boundaries of what's possible. I think a lot of us
00:32:32.109 --> 00:32:35.589
have grown up in a culture of accountability
00:32:35.589 --> 00:32:41.190
and that's great. I think a lot of us are struggling
00:32:41.190 --> 00:32:45.009
to fully adopt kind of a non-deterministic workflow.
00:32:45.640 --> 00:32:50.940
And so for those of us that are maybe the stereotypical
00:32:50.940 --> 00:32:55.099
engineer, the overachieving type A, I want to
00:32:55.099 --> 00:32:57.740
build stuff and I want predictable outcomes.
00:32:58.910 --> 00:33:03.420
I think it can be difficult to maybe loosen the
00:33:03.420 --> 00:33:06.200
reins a little bit. And I'm not saying that people
00:33:06.200 --> 00:33:10.019
should go out and immediately turn over their
00:33:10.019 --> 00:33:15.220
entire organization to an LLM. But are we creating
00:33:15.220 --> 00:33:19.880
enough space within our organizations to have
00:33:19.880 --> 00:33:23.140
experimentation that is pushing the boundaries
00:33:23.140 --> 00:33:26.519
of what an LLM can do? A lot of times I'll talk
00:33:26.519 --> 00:33:30.819
about if every project that we're doing has greater
00:33:30.819 --> 00:33:34.880
than 50% chance of success, we're not being
00:33:34.880 --> 00:33:37.420
ambitious enough. There's got to be something
00:33:37.420 --> 00:33:41.039
in our portfolio that we think only has a 10%
00:33:41.039 --> 00:33:44.559
chance of success. And that's where we see
00:33:44.559 --> 00:33:46.799
the real breakthroughs and the real changes in
00:33:46.799 --> 00:33:51.700
innovation. And so are we giving the LLMs big
00:33:51.700 --> 00:33:55.210
enough tasks? Are we turning enough over? Again,
00:33:55.329 --> 00:33:57.809
not across the board, not recklessly, not all
00:33:57.809 --> 00:34:02.069
at once, but are we creating those areas of real
00:34:02.069 --> 00:34:05.750
frontier experimentation to push the limits and
00:34:05.750 --> 00:34:08.510
see what's possible? Very well said. Where can
00:34:08.510 --> 00:34:12.059
people find out more about BambooHR? Yeah, definitely
00:34:12.059 --> 00:34:15.460
check us out on social media. We're on all of
00:34:15.460 --> 00:34:17.800
the different social media platforms, but bamboohr.com.
00:34:17.800 --> 00:34:21.150
Also, check out the careers page. We're
00:34:21.150 --> 00:34:23.610
continuing to hire, looking for great talent,
00:34:23.710 --> 00:34:26.190
people that are excited about where this space
00:34:26.190 --> 00:34:29.449
is headed. We've got a lot of great opportunities.
00:34:30.210 --> 00:34:31.909
Awesome. Ryan, thank you so much for joining
00:34:31.909 --> 00:34:33.869
me. Really, really appreciate it. Yeah, thank
00:34:33.869 --> 00:34:36.590
you for having me. It was fun. That was my conversation
00:34:36.590 --> 00:34:41.110
with Ryan Packer from BambooHR. The biggest
00:34:41.110 --> 00:34:44.690
takeaway for me is why a CTO should be coding
00:34:44.690 --> 00:34:48.750
again. It's not to ship features. It's to use
00:34:48.750 --> 00:34:53.349
the same tools his team is learning. And to show
00:34:53.349 --> 00:34:58.489
an idea in working code instead of a PRD. I liked
00:34:58.489 --> 00:35:01.989
his software factory framing. The cars coming
00:35:01.989 --> 00:35:05.909
off the line will change. What matters is getting
00:35:05.909 --> 00:35:08.949
better at getting better. Their pull requests
00:35:08.949 --> 00:35:13.369
are up almost 80% this year with quality improving.
00:35:13.809 --> 00:35:17.820
But the part he's most excited about is more
00:35:17.820 --> 00:35:21.340
at-bats with customers. For the platform folks,
00:35:21.679 --> 00:35:26.820
AI is putting real strain on CI/CD. So when every
00:35:26.820 --> 00:35:31.860
engineer has 10, 20, or 100 agents going, each
00:35:31.860 --> 00:35:34.800
one needs an environment you can spin up fast
00:35:34.800 --> 00:35:39.380
with real data and tear down just as fast. On
00:35:39.380 --> 00:35:42.820
junior engineers, Ryan doesn't claim a perfect
00:35:42.820 --> 00:35:46.989
answer. His advice? Make the LLM explain its
00:35:46.989 --> 00:35:50.590
choices instead of handing everything over. And
00:35:50.590 --> 00:35:54.769
try an adversarial agent that lists five things
00:35:54.769 --> 00:35:58.409
it would have done differently. Last one. If
00:35:58.409 --> 00:36:01.809
every project in your portfolio has better than
00:36:01.809 --> 00:36:06.389
a 50% chance of success, you are not being ambitious
00:36:06.389 --> 00:36:10.769
enough. Put something in there with a 10% chance.
00:36:11.469 --> 00:36:16.409
You can find BambooHR at bamboohr.com, including
00:36:16.409 --> 00:36:20.550
their careers page, plus Ryan on LinkedIn through
00:36:20.550 --> 00:36:23.769
the links in the show notes. Follow or subscribe
00:36:23.769 --> 00:36:27.170
to Ship It Weekly wherever you listen and find
00:36:27.170 --> 00:36:30.809
previous episodes at shipitweekly.fm. Thanks
00:36:30.809 --> 00:36:33.269
for listening, and I'll see you later this week.