สำเนาบทสนทนา 🔗
WEBVTT
00:00.000 --> 00:02.160
Oh, it's gonna be controversial.
00:03.101 --> 00:15.301
I think pair programming as we're used to one person is driving, commands, the other one's
looking over the shoulder, making sure it's correct and learning from it, that way kind of
00:15.301 --> 00:17.933
broke when you started using multi-agents, right?
00:17.933 --> 00:19.033
However,
00:19.440 --> 00:25.713
very important to have senior engineers sit with more junior engineers to show how you're
working with AI, right?
00:25.713 --> 00:28.224
You just assume like, everyone knows all these tricks.
00:28.224 --> 00:33.824
I got up my sleeves And also like writing the AI workflows together, I think is also very
important.
00:33.824 --> 00:35.165
Because they're so critical.
00:35.165 --> 00:46.314
But the traditional pair programming paradigm where you just sit and watch over each
other's shoulder, I think that to be replaced different flavor of it at least.
00:46.614 --> 00:50.096
I don't think that's sitting side by side for eight hours a day.
00:50.096 --> 00:51.236
it's not productive anymore
01:08.760 --> 01:15.466
Jonas Claesson is a principal engineer at Provision Analytics, which makes cloud software
to streamline food safety and quality assurance.
01:15.466 --> 01:18.428
He's also a Navalia alumni and a ThoughtWorks alumni.
01:18.428 --> 01:21.371
He and I worked together at Navalia, actually, not that long ago.
01:21.371 --> 01:30.938
And recently, he's been posting stuff on LinkedIn that I really love, the sort of concrete
stuff that just makes me smile because it isn't just sort of platitudes or predictions or
01:30.938 --> 01:31.469
forecasts.
01:31.469 --> 01:35.886
It's like, no, this is directly how AI is changing what I'm doing.
01:35.886 --> 01:40.846
talking about wasted providing SQL fixes and how work trees are better than branches.
01:41.046 --> 01:47.767
And probably my favorite was him talking about having his laptop next to him in the car
and running Claude Code as he was driving home.
01:47.767 --> 01:49.687
So that finish in big refactoring.
01:49.727 --> 01:51.347
I am just loving all of this stuff.
01:51.347 --> 01:53.721
And I've been working on getting him on the podcast.
01:53.721 --> 01:57.003
Today, Jonas is our guest on the Engineering with AI podcast.
01:57.003 --> 01:59.714
So, Jonas, thank you so much for being on the show today.
01:59.714 --> 02:00.983
We really appreciate it.
02:00.983 --> 02:02.795
Thank you so much, I'm very excited to be on.
02:02.795 --> 02:07.781
I've been listening to these episodes all along, so it's really great to be on here.
02:08.212 --> 02:09.173
Amazing amazing.
02:09.173 --> 02:10.926
No, I mean, it's funny.
02:10.926 --> 02:20.358
I'm used to talking to you in this exact situation I'm this is my home office and that's
your office and we're always talking to each other from the same spot ah
02:22.382 --> 02:33.982
Well, and now that this is my studio it's literally on there every time and
Load full transcript
it's too bad
about last night But I don't know like we'll see maybe maybe Hopefully be I will come back
02:33.982 --> 02:45.302
and it will be okay But anyhow, let's let's get past that and into the into the software
development Like I said in the bio and in the intro I've been loving reading what you've
02:45.302 --> 02:50.242
been talking about, you know, so so much really good stuff We're gonna link to some of it
in the show notes
02:50.396 --> 02:51.678
What's one of your favorites?
02:51.678 --> 02:53.332
What's the one that stands out for you?
02:55.110 --> 03:00.436
Of the articles, m I think they kind of follow a similar theme.
03:00.436 --> 03:05.483
It's like how AI, everyone talks about AI just makes code generation cheaper, right?
03:05.483 --> 03:08.547
But it also makes engineering rigor cheaper.
03:08.547 --> 03:13.913
There's so many things that you can now do to make sure the quality of your software is
better.
03:14.007 --> 03:14.978
Yeah, yeah.
03:14.978 --> 03:18.492
That you could still do before, but most of the time you didn't have time for it.
03:18.492 --> 03:19.623
It was just too much effort.
03:19.623 --> 03:23.928
But now you can have AI do all these extra checks and tests and validations and
everything, right?
03:23.928 --> 03:26.741
So I think that's one thing that a lot of people are not talking about.
03:26.741 --> 03:28.933
Because now...
03:29.179 --> 03:31.350
And this is a really good time for you too.
03:31.350 --> 03:36.464
mean, you've been at Provision now, it feels like just yesterday, but it's actually been a
while.
03:36.645 --> 03:38.296
Tell me about what's going on there.
03:38.296 --> 03:40.968
Like maybe start with a little bit about the business and your role.
03:40.968 --> 03:45.112
ah But then, like I'm guessing this has been timely for you guys.
03:45.112 --> 03:47.453
So let's start at the top with Provision.
03:47.627 --> 03:58.922
Yeah, so it's a big SaaS company that are just in the process of rewriting their software
from scratch, make it more scalable and easier to use and everything, right?
03:58.922 --> 04:06.636
uh So I came in a really good time when we were just starting this process and uh we were
starting it with Co-Pilot, right?
04:06.636 --> 04:09.427
About a year ago, that was kind of the only game in town.
04:09.427 --> 04:12.789
Claude was slowly coming along and...
04:13.181 --> 04:20.263
It's been just interesting to be able to experiment when you're a small company and you
have a CEO is really bought into AI and it's very supportive.
04:20.263 --> 04:22.763
We've been able to just go all in on this.
04:22.763 --> 04:28.405
And, um, I think I mentioned earlier to you, I think it was November, December last year.
04:28.405 --> 04:29.185
just made the decision.
04:29.185 --> 04:31.415
I'm not going to write a single line of code anymore.
04:31.415 --> 04:35.386
I'm just going to go all in on this and cloud's going to write all the code for me.
04:35.386 --> 04:38.787
And, uh,
then the whole team kind of adopted this.
04:38.787 --> 04:42.628
So now we have an engineering team of six developers that are.
04:42.846 --> 04:45.071
just using agents to write code.
04:45.143 --> 04:45.804
Wow, cool.
04:45.804 --> 04:52.938
Okay, so briefly, mean, not in exhaustive detail, but like, what's the stack, what
language and that kind of thing.
04:52.938 --> 04:58.793
Yeah, so it's frontend, react, tanstack query kind of what most people do.
04:58.793 --> 05:12.385
Backend is .NET with a Postgres database, a couple of other external services for sending
out emails and notifications, it's clerk for auth uh So pretty standard, simple, not
05:12.385 --> 05:14.387
simple, it's got a lot of complexity to it, but.
05:14.387 --> 05:15.693
uh
05:15.693 --> 05:18.598
also my first guest who's using Clawd with C Sharp.
05:18.639 --> 05:19.932
I suspected it would be good.
05:19.932 --> 05:25.092
I've had it look at like one C Sharp code base so far, but yeah, I'll be interested to get
your take on that.
05:25.092 --> 05:25.844
That's cool.
05:25.844 --> 05:30.848
Yeah, that's one thing I've noticed that the more established technologies to use, the
better the AI is.
05:30.848 --> 05:37.764
If you use some kind of off weird library, there isn't that much code out there for the AI
to kind of learn from.
05:37.764 --> 05:41.295
So if you use these established languages, this does a really good job.
05:41.697 --> 05:42.438
Yeah, for sure.
05:42.438 --> 05:43.839
Okay, cool.
05:43.840 --> 05:45.282
Okay, so great.
05:45.282 --> 05:47.324
We sort of already have to do it.
05:47.324 --> 05:48.105
Let's start with you.
05:48.105 --> 05:49.667
We'll get to the team and all that.
05:49.667 --> 05:51.289
But let's start with you.
05:51.510 --> 05:54.893
What, and I think you've already said it.
05:54.934 --> 05:57.097
You're saying you're having Claude write all of the code now.
05:57.097 --> 05:58.829
So I guess I know the answer.
05:58.829 --> 06:00.521
What's your favorite tool so far?
06:00.992 --> 06:05.044
I mean, I use Codex as well once in a while, but Claude really fits my workflow.
06:05.044 --> 06:13.910
especially I use the CLI and having multiple tabs and just quickly go between them and
check on your agents and all the shortcuts and everything.
06:13.910 --> 06:16.681
It just works really well for me, but I'm very tool agnostics.
06:16.681 --> 06:21.624
I constantly try out different tools and try not to be too married to one of them.
06:21.795 --> 06:24.116
Yeah, I wish I did.
06:24.397 --> 06:31.924
I have like a productivity uh quirk about me where I don't, you know that comic where the
guys are like, hey, look, it's a wheel.
06:31.924 --> 06:35.847
And the other guys have like a wagon that has square wheels.
06:35.847 --> 06:37.068
I'm the guy with the square wheels.
06:37.068 --> 06:43.073
But I make sure that I have smart friends around who will tell me on occasion, like there
really is a better wheel now, Kyle.
06:43.254 --> 06:46.016
But personally, I'm terrible at actually making myself do it.
06:46.016 --> 06:47.737
So yeah, cool.
06:47.978 --> 06:48.780
So I'm.
06:48.780 --> 06:52.776
I'm glad you've tried codecs and those kinds too with you.
06:52.990 --> 06:56.865
But like you say, like, Claude's probably one of the better ones right now.
06:56.865 --> 06:58.346
That's still holding true?
06:59.349 --> 06:59.773
Yeah.
06:59.773 --> 07:02.767
changing so fast, you gotta stay on top of this.
07:02.767 --> 07:04.578
Well, you know, that's a question I've got.
07:04.578 --> 07:07.529
Somebody once said like, but you're switching costs.
07:07.849 --> 07:08.650
I don't know.
07:08.650 --> 07:19.906
Like if I were a Fortune 500, right, uh legacy business, and I'd spend a lot of time with
the salesperson, you know, at Anthropic getting the master services agreement just right
07:19.906 --> 07:28.531
and the statement of work just right and making sure that like, you know, all the language
I wanted in there about my intellectual property, maybe I have high switching costs around
07:28.531 --> 07:29.722
the contract.
07:30.114 --> 07:31.516
Are there any other switching costs?
07:31.516 --> 07:35.612
Like I've had like, well actually it wasn't Codex, it was Gemini CLI.
07:35.612 --> 07:40.428
Check out a code base that I've been writing with Claude and it just read the Claude MD.
07:40.428 --> 07:42.211
Like these are all text files, right?
07:42.211 --> 07:45.606
So it feels like switching costs are gonna be really low in the end.
07:45.606 --> 07:47.808
How does that feel to you, is that right?
07:47.808 --> 07:48.418
Yeah, definitely.
07:48.418 --> 07:52.589
can't even, I mean, we got a lot of custom Claude workflows, but they could be easily
recorded.
07:52.589 --> 07:58.761
There's just some things how a Claude deals with work trees and how they enter different
work trees.
07:58.761 --> 07:59.982
you start them, sure.
07:59.982 --> 08:06.414
We've got to figure out how to do that in Codex, but in general, it would be fairly easy
to switch to anything else.
08:06.414 --> 08:08.061
And we're actually using Codex as well.
08:08.061 --> 08:14.036
And we'll come to that later when we talk about the PR review, because we use both Claude
and Codex and then we have them.
08:14.336 --> 08:15.557
battle each other.
08:15.799 --> 08:17.740
Nice, nice, nice, nice.
08:18.421 --> 08:31.749
Okay, in one of your other blog posts, you talked about using the Ralph Wigum And this
reminds me, I played around with Gastown, which is kind of similar, it's different, but
08:31.749 --> 08:36.092
it's kind of trying to get you to that like uh unattended operation.
08:36.092 --> 08:41.936
And I think that's what the Ralph Wigum was trying to be all about too, is give it some
exit criteria and let it just churn.
08:41.936 --> 08:44.417
So that's cool that you've been playing with that.
08:44.417 --> 08:45.642
A lot of people are, you know,
08:45.642 --> 08:47.948
reading about it but not
necessarily doing it.
08:47.948 --> 08:51.529
Yeah, I was just waiting for an example where I could apply it, to be honest.
08:51.529 --> 08:58.277
I had a perfect scenario for it where it would fit the sweet spot of when to use it.
08:58.277 --> 09:03.221
So we have all these integration tests for all our API endpoints, hundreds and hundreds of
them, right?
09:03.221 --> 09:12.048
And they all insert seed data, calling the API, inserting the database, and then calling
these endpoints to list and filter after they've inserted data.
09:12.048 --> 09:13.570
Obviously not very optimized, right?
09:13.570 --> 09:16.872
Because you have to do a lot of inserts and you...
09:17.640 --> 09:24.165
Not that great because you won't have that much data in the database either for all your
list operations and filtering and stuff like that.
09:24.165 --> 09:31.930
So the obvious solution to this is just, hey, let's just load a database with some
read-only seed data for one organization.
09:31.930 --> 09:39.735
And then anytime you need to test like a GET operation, you just work against the
organization, like the read-only data.
09:39.735 --> 09:41.316
um
09:41.482 --> 09:46.115
Before AI, that would have been sitting on the tech debt column a long time.
09:46.115 --> 09:47.270
No one wanted to touch that.
09:47.270 --> 09:55.752
That's like a week of misery of someone just figuring out every test Okay, I need to move
this into this seed data as read-only.
09:55.752 --> 09:58.595
So what the Ralph Wiggum technique does, it's...
09:59.060 --> 10:02.141
It basically takes one test at a time and it starts with a fresh context.
10:02.141 --> 10:05.145
So you're not going to start hallucinating because this was running for hours.
10:05.145 --> 10:09.080
So we just take one test and then keep a list of all the tests it had to do.
10:09.080 --> 10:17.616
And every time it runs in this loop, takes the next test that it hasn't processed and
figure out how we can uh add more seed data to satisfy this test.
10:17.616 --> 10:19.358
So it didn't have to do all these inserts.
10:19.358 --> 10:26.464
So it probably cut off like a 30, 40 seconds off every PR that had run all these
integration tests.
10:27.907 --> 10:31.041
Wait, wait, wait, wait, okay, so I followed a lot of that.
10:31.041 --> 10:38.953
the Ralph Wiggum loop had like a stack of tests that it had to go and figure out what seed
data to create how did that shave time?
10:38.964 --> 10:43.098
manual inserts as part of the test, it just had like in the database already.
10:43.098 --> 10:47.659
Like it was just a big insert in the beginning and then all the tests reused that data.
10:47.659 --> 10:51.784
so this was about accelerating your continuous delivery build times.
10:51.784 --> 10:56.617
Yeah, just refactoring all these tests, which is such an unpleasant thing to do manually.
10:56.617 --> 11:02.991
But with AI and Ralph Wiggum, that's another thing that's very important, this feedback
loop, by telling it like, when is the AI done?
11:02.991 --> 11:05.042
Just don't tell it like, hey, go do this.
11:05.042 --> 11:06.252
What's your definition of done?
11:06.252 --> 11:08.923
How do you verify yourself that you're done?
11:09.149 --> 11:11.346
So how much time did it shave off the build?
11:12.058 --> 11:15.417
of maybe 30-40 seconds off a 2 minute build.
11:15.417 --> 11:17.050
So it's pretty significant.
11:17.050 --> 11:20.495
cow, so that's a big percentage-wise decrease.
11:20.502 --> 11:21.360
And to your point.
11:21.360 --> 11:22.593
in like less than an hour.
11:22.593 --> 11:24.628
And then it just, that was a Friday evening.
11:24.628 --> 11:27.817
just, Ralph just kept going and going.
11:27.817 --> 11:35.159
I can't imagine facing down a product manager or engineering manager and try to say, hey,
we've got this thing we want to do.
11:35.159 --> 11:37.783
It's going to take a week, but it's going to save 30 seconds per build.
11:37.783 --> 11:40.235
It's never going to get greenlit.
11:40.235 --> 11:42.357
But now that it's down to an hour, you don't even ask.
11:42.357 --> 11:43.448
You just do it.
11:43.448 --> 11:44.669
So yeah.
11:44.669 --> 11:45.632
OK, cool.
11:45.632 --> 11:52.781
like our tech debt is so small because you don't have those things that you don't want to
touch like anything is pretty doable now like
11:52.781 --> 11:54.594
Right, right, right, right, right.
11:54.594 --> 12:03.639
Well, I think that's smart what you're inferring there is that you use the speed, the time
that you get back from the speed, you're not using it to just make more product faster.
12:03.639 --> 12:05.592
You're using it to pay down tech debt too, right?
12:05.592 --> 12:07.760
So I think that's an important part of what you're saying there.
12:07.760 --> 12:10.380
where people get it wrong and it's very tempting.
12:10.400 --> 12:11.840
What used to take four days, right?
12:11.840 --> 12:14.500
If it takes a day now, you've freed up three days.
12:14.500 --> 12:15.320
It's very tempting.
12:15.320 --> 12:18.957
just like, I'll just grab some more stores and just get more functionality out.
12:18.957 --> 12:19.937
That's what gets you in trouble.
12:19.937 --> 12:31.199
You have to take that time to invest in the product, pay down tech debt, write more tests,
figure out any way that you can just make things more stable and smooth in your...
12:31.199 --> 12:35.905
deployment pipelines wherever you need right just you have all this extra time like use it
wisely
12:36.035 --> 12:40.995
Yeah, okay, so one of the questions I ask everybody is like, what guardrails are you
putting in place?
12:40.995 --> 12:47.772
And you kind of hinted at something around Codex and PRs, maybe this makes sense to tell
this story now.
12:47.776 --> 12:54.465
Yeah, I hear a lot of people complain about agents doing PR reviews and to some extent
they're correct.
12:54.465 --> 13:01.012
Like if you use GitHub Copilot and it's reviewing any PR in Git for you.
13:01.012 --> 13:03.957
We always joke in a team, it's almost like GitHub Copilot is this guy.
13:03.957 --> 13:04.713
I found this issue.
13:04.713 --> 13:05.329
I found this issue here.
13:05.329 --> 13:06.690
It just tries to be so helpful.
13:06.690 --> 13:09.244
And like every little issue is like top priority, right?
13:09.244 --> 13:14.430
It doesn't have this broad context and really narrowed down on like these are the things
you really need to fix.
13:14.430 --> 13:15.712
These are a lot of nitpicky things.
13:15.712 --> 13:17.988
So we have this workflow now where.
13:17.988 --> 13:22.971
Every developer runs a custom Claude command called PR review, where we have the very
opinionated.
13:22.971 --> 13:23.831
I think this is the key.
13:23.831 --> 13:25.452
You need to write a prompt that's very opinionated.
13:25.452 --> 13:26.943
This is the things we care about.
13:26.943 --> 13:30.064
Don't just give it like, Claude, you figure out the review.
13:30.064 --> 13:31.206
This is what we care about.
13:31.206 --> 13:31.887
We care about
13:31.887 --> 13:37.089
Low complexity, don't duplicate code, don't try to introduce new frameworks or patterns.
13:37.089 --> 13:37.889
Keep it simple.
13:37.889 --> 13:40.541
Look at the rest of the code base and be consistent, right?
13:40.541 --> 13:45.264
So it's this prompt that is super opinionated on how you build software.
13:45.264 --> 13:46.425
So we run that first.
13:46.425 --> 13:49.625
it always categorizes like a must fix, should fix.
13:49.625 --> 13:51.336
or like nitpick stuff, right?
13:51.336 --> 14:02.074
So, developer runs that, the entire team runs it, and you go ahead and you fix those
issues, then you open a PR in GitHub, and then we have a GitHub action that calls Codex,
14:02.074 --> 14:04.376
and it kind of does the same thing, but a little bit different.
14:04.376 --> 14:10.540
So Codex runs the same thing then, And surprisingly enough, it finds stuff that Claude
misses.
14:10.540 --> 14:13.559
It's more focused bugs and stuff like that.
14:14.389 --> 14:18.350
And I can imagine that you also put this in your Claude.md to avoid these kinds of things.
14:18.350 --> 14:22.335
So it missed it the first time and it missed it the second time in the Pulled Request
review.
14:22.335 --> 14:22.595
Yeah.
14:22.595 --> 14:24.366
And then we have another command, right?
14:24.366 --> 14:29.370
That pulls the Claude comments and the Codex comments and it kind of gives you this
unified view.
14:29.370 --> 14:33.783
then also super opinionated, it tells it, okay, be honest, now Claude.
14:33.783 --> 14:37.707
Go through your comments, go through Codex comments, which one should be fixed?
14:37.707 --> 14:39.328
And it's surprisingly good.
14:39.328 --> 14:40.730
And I usually agree with it.
14:40.730 --> 14:48.127
Wow, so it's reading its own work, it's reading Codex's work, and then it's coming to you
with a final opinion, which you then validate and say, like, yes.
14:48.127 --> 14:49.107
OK.
14:49.128 --> 14:49.646
Huh.
14:49.646 --> 14:50.886
an interaction with it.
14:50.886 --> 14:51.966
Okay, tell me more about this one.
14:51.966 --> 14:52.546
What do you think?
14:52.546 --> 14:55.046
Like, and then I tell you, okay, go and fix this.
14:55.046 --> 14:58.546
So then we have two rounds of code reviews by AI.
14:58.546 --> 14:59.687
Then we assign a human.
14:59.687 --> 15:01.687
So, cause that's always, that's the bottleneck right now.
15:01.687 --> 15:02.747
It's the human review.
15:02.747 --> 15:05.667
So we're trying to just make that as smooth as possible.
15:05.667 --> 15:08.147
So it's more of a quick sanity check, right?
15:08.147 --> 15:09.328
If I look at a PR.
15:09.328 --> 15:12.853
I might look at which data migrations, which new endpoints, like stuff that matters.
15:12.853 --> 15:17.571
I'm not going into the nitty-gritty things and like, you used a for loop there instead of
a for each.
15:17.571 --> 15:19.564
Claude already took care of that, right?
15:19.564 --> 15:22.117
So I'm more laser focused on certain things.
15:22.117 --> 15:30.961
So it sounds like a fair amount of the guardrails here are getting rid of false positives,
like trying to prevent it from telling you stuff that doesn't matter.
15:31.752 --> 15:34.811
good is it at finding stuff that does matter?
15:35.114 --> 15:37.075
I've been pretty impressed by it so far.
15:37.156 --> 15:45.412
way more than like we used Copilot originally, like we still had that Claude view and then
it went to Copilot and Copilot we just find all these like little things that didn't
15:45.412 --> 15:45.903
really matter.
15:45.903 --> 15:47.774
It was just so much noise, very little signal.
15:47.774 --> 15:48.864
But yeah,
15:48.864 --> 15:51.547
It's always about tuning these context files as well, right?
15:51.547 --> 16:00.989
Like this, prompt if there's things that's missing or if you think it's focusing too much
on one thing, we'll go and tweak it a little bit to make sure it aligns with what we value
16:00.989 --> 16:02.710
as good quality code.
16:02.717 --> 16:05.040
Yeah, okay, but it's still finding stuff, right?
16:05.040 --> 16:08.945
If it weren't finding anything, you would have like turned it off by now, I imagine.
16:08.945 --> 16:09.691
Yeah.
16:09.691 --> 16:14.041
ship a PR without Claude finding stuff and Codex usually finds one or two items as well.
16:14.041 --> 16:15.038
cool, cool.
16:15.038 --> 16:15.759
a huge difference.
16:15.759 --> 16:21.985
Like everyone on the team, when I look at their PR, even if they're more junior engineer,
like before it would be like a learning experience, right?
16:21.985 --> 16:24.687
You find things and try to teach them as a whole back and forth.
16:24.687 --> 16:27.530
Now I look at a more junior PR and it's solid.
16:27.530 --> 16:34.958
It looks like a senior engineer wrote it because they've been guided by all these context
files and, AI automation workflow that they're using, right?
16:34.958 --> 16:39.238
To make sure that their code is like aligns with exactly how we want it written.
16:39.238 --> 16:42.158
Cool, cool, okay, you're talking me into it.
16:42.158 --> 16:45.338
I'm actually at the point where I'm like, I just don't do them at all.
16:45.338 --> 16:46.618
I'm not doing PR reviews.
16:46.618 --> 16:52.758
I mean, for me, it's a one-person show and it's a pre-revenue startup, so that's an easy
decision to make.
16:53.018 --> 16:55.598
This is a side gig, it's not even my day job.
16:55.658 --> 16:57.838
We're obviously not doing that in enterprise.
16:57.838 --> 16:59.758
In enterprise, there's still PRs going, right?
16:59.758 --> 17:03.159
So I'm thinking about teams that I know at work, right?
17:03.159 --> 17:04.580
That could use this,
17:04.580 --> 17:07.201
Actually, you know the team I'm thinking of too.
17:07.201 --> 17:14.408
So one of the developers on this team, developed a similar kind of bot that is all aimed
at Snyk vulnerabilities.
17:14.408 --> 17:19.311
So it goes through whatever Snyk is reporting, and it finds any vulnerability and fixes
it.
17:19.311 --> 17:26.372
And one of the things that they sort of tacked on is they made sure to run their
end-to-end test, which you know the E-to-E test I'm talking about.
17:26.372 --> 17:28.452
It was kind of broken all of the time.
17:28.452 --> 17:30.892
But now that we have Claude, it isn't.
17:30.892 --> 17:32.713
And so it's actually kind of useful.
17:32.713 --> 17:35.094
And so it's been a nice tie-in to have it.
17:35.094 --> 17:39.134
reduces the effort with keeping us up to date and its new vulnerabilities, which is kind
of cool.
17:39.134 --> 17:41.134
But we don't have your PR wizardry yet.
17:41.134 --> 17:43.274
So I'm going to tell them about that.
17:43.274 --> 17:44.031
Awesome.
17:44.031 --> 17:45.111
I play with it.
17:45.111 --> 17:52.150
I've been surprisingly impressed with it because you think like you have Claude write all
this code and then at the end of it, you tell him, okay, now go review it.
17:52.150 --> 17:57.451
Like it already told you, like, he's not going to find anything, but it does because it
kind of takes a step back, right?
17:57.451 --> 17:58.831
It gets a new prompt.
17:58.831 --> 18:02.291
It looks at things from a different perspective and it finds stuff all the time.
18:02.291 --> 18:05.951
And a lot of times it will find bugs that I would have never found, right?
18:05.951 --> 18:11.289
Like little things that are very easy to miss when you manually look at them.
18:11.712 --> 18:18.485
I think superpowers, which I'm using right now, one of the Claude plugins is it is finding
a few things.
18:18.485 --> 18:23.499
I think it does a little bit of what you're doing, but it's not nearly as elaborate as
what you're describing.
18:23.499 --> 18:23.819
Okay.
18:23.819 --> 18:24.539
Fine.
18:24.539 --> 18:27.062
I'm convinced I'm going to try it and I'm going to tell the teams to try it too.
18:27.062 --> 18:27.422
Okay.
18:27.422 --> 18:29.244
So let's get back to you guys.
18:29.244 --> 18:33.527
so we talked a little bit about you and we've talked about some of the experiences you
guys have had.
18:33.527 --> 18:35.579
Let's get into the teams that you're serving, right.
18:35.579 --> 18:37.720
you've probably been there for the arc here.
18:37.720 --> 18:40.842
How has product management and story breakdown changed?
18:40.842 --> 18:49.367
When you think about how provision works, do you have product managers that write the PRDs
and then with Claude and then Claude takes the PRDs and turns them into stories?
18:49.367 --> 18:49.918
Like what?
18:49.918 --> 18:50.908
Probably not, right?
18:50.908 --> 18:55.720
Like so, but what has changed with both product management and story breakdown and things
like that?
18:55.905 --> 18:59.474
The whole company uses Claude across the board, even product, customer service,
everything.
18:59.474 --> 19:01.384
So it's very ingrained in the company.
19:01.384 --> 19:09.155
One thing I've been thinking about related to story refinement, which we don't do right
now, the first thing we do when we start a story is we have like a custom Claude command,
19:09.155 --> 19:17.058
start PR, you give it the ticket number, it goes to linear, it grabs the whole
description, any figma designs, any attachment, anything it needs, right?
19:17.058 --> 19:19.219
And then it starts querying you about gaps.
19:19.219 --> 19:26.102
Like if this thing's not in the story, because it had access to the story and the code
base, so it knows like, well, you didn't tell me about these three items.
19:26.102 --> 19:28.612
And that's the developer doing that right now.
19:28.612 --> 19:35.188
I will tell it the gaps in the technology, but if there's like business requirements that
are missing, you will kind of tell me about that.
19:35.188 --> 19:36.952
And I have to go back to the product owner.
19:36.952 --> 19:40.263
It was usually super fast and will tell me right away when I, when I slack him.
19:40.263 --> 19:43.548
But it'd be cool to move that up before the developer touched it.
19:43.548 --> 19:51.236
So the product owner could run some Claude command that access to code base and help him
refine, like tell them, well, developer can't start working on this because you haven't
19:51.236 --> 19:53.128
defined what happens in these scenarios, right?
19:53.128 --> 19:58.033
if you archive this employee here in the system and then you go here and let's see, should
it show up in the dropdown or not?
19:58.033 --> 20:02.047
Like this, all these things that could be refined before it even hits the developer.
20:02.047 --> 20:03.333
Right now we get it on the developer.
20:03.333 --> 20:07.694
It's not a huge thing, but it would be nice if it was just like completely refined.
20:07.694 --> 20:12.878
We pick it up and Claude has already filled in the gaps before the developer even touches
it.
20:13.075 --> 20:13.986
That's fantastic.
20:13.986 --> 20:22.075
Yeah, honestly, that's the kind of thing that I've like waiting for and it you're one of
the first people that's actually called something like that, you know, to have it as glue
20:22.075 --> 20:23.082
between us.
20:23.082 --> 20:23.533
What about?
20:23.533 --> 20:25.554
to go work on another half of this to be honest.
20:25.554 --> 20:33.460
Some like a refined ticket, just give it a story and it will basically go and update your
linear based on your answers when it's kind of telling you what's missing.
20:33.540 --> 20:34.362
Nice.
20:34.362 --> 20:36.769
What about architecture and planning?
20:36.769 --> 20:45.209
To be honest, in a startup, I mean, it helps a lot with just prototyping different
architectures and validating them because it's so quick to just spin up different
20:45.209 --> 20:46.441
solutions in architecture.
20:46.441 --> 20:50.906
Working for smaller company, it's not as big of challenge as working for big enterprise
where you have...
20:50.906 --> 20:53.906
a zillion microservices all over the place and different things, right?
20:53.906 --> 21:00.787
Like we're a pretty lean team but we recently had to figure out how to do analytics on our
product.
21:00.787 --> 21:05.808
And it was very good to just kind of like, hey, show me all the ways in which you can do
this.
21:05.961 --> 21:14.621
We actually have a very complicated analytics problem because all our data for all our
customers, all our customers, they can create their own forms and they can create whatever
21:14.621 --> 21:17.601
fields and grids and tables they want, right?
21:17.647 --> 21:22.112
So nothing is predefined and you can have multiple versions of that and every customer is
different.
21:22.112 --> 21:27.279
And then we need to get that into a warehouse and expose that as like a flattened out
structure.
21:27.279 --> 21:29.371
we actually met with our Microsoft representative.
21:29.371 --> 21:32.844
He was going to demo fabric and he was like, look at how good this handles JSON.
21:32.844 --> 21:37.749
And we told him the challenges we were facing and he like, almost felt like he was
starting to cry.
21:37.749 --> 21:40.090
Like, have no idea how this all this.
21:40.131 --> 21:42.402
But with Claude you can tell it is problem.
21:42.402 --> 21:46.895
like, it frankly told it, there's no product out there that can really solve this for you.
21:46.895 --> 21:49.457
But here's a couple of ways you can do it.
21:49.457 --> 21:59.263
And we actually came up with a solution fairly quickly that was able to like extract all
these completely unstructured JSON that are different for every client into like a unified
21:59.263 --> 22:05.797
data warehouse that is being updated and depending on the structure and the JSON and the
schema.
22:05.983 --> 22:07.307
Wow, cool.
22:07.307 --> 22:11.148
Did you end up using AI agents as part of that pipeline?
22:11.477 --> 22:18.599
More just to come up with the solution and to build like the custom code to
make this
happen, which would have been a lot of effort.
22:18.900 --> 22:22.431
it was also really good to have AI there because it's very easy to test with AI.
22:22.431 --> 22:27.693
can just tell that, okay, you built all these new views in the database that are like the
flattened out structure.
22:27.693 --> 22:32.524
You just go and compare every single view with the original JSON document.
22:33.064 --> 22:37.027
know it's going to, like, that would be horrible to do manually, but.
22:37.027 --> 22:38.108
yeah, yeah.
22:38.215 --> 22:46.176
this is the kind of thing that software developers would have thought about for, well not
thought, but debated about for months endlessly about like, wow, I can't do it that way,
22:46.176 --> 22:55.305
it'll never work because I read in a book once and somebody else will have, oh, I saw
something in a magazine about such and such and well, it be a magazine.
22:55.305 --> 22:56.529
But you get the idea.
22:56.529 --> 23:02.060
It's the kind of thing that people would have debated endlessly and never started because
we couldn't agree on how.
23:02.060 --> 23:05.201
But to your point, you can just sort of like wade in and say, like, what do you think?
23:05.201 --> 23:05.754
Just go.
23:05.754 --> 23:06.939
it's not a lot of cost.
23:06.939 --> 23:08.630
Yeah, just the cost of prototyping.
23:08.630 --> 23:09.461
Just do it this way.
23:09.461 --> 23:10.002
Let's try it.
23:10.002 --> 23:10.882
Let's see how that works.
23:10.882 --> 23:12.551
Oh, let's try it this way.
23:12.551 --> 23:13.905
You can do multiple prototypes.
23:13.905 --> 23:17.557
You just didn't have that luxury before AI, to be honest.
23:19.219 --> 23:21.290
And just all the the QA can do for you.
23:21.290 --> 23:28.676
I just like, just spend the week and just go and like do extensive testing on all these
things I built, like, chug through the data.
23:29.728 --> 23:32.131
What other changes are you seeing in software development?
23:32.131 --> 23:36.207
mean, this is kind of like all about software development, I know, but like at the actual
like dev time.
23:36.207 --> 23:37.787
mean, it's really changed how you work, right?
23:37.787 --> 23:40.627
Like it's, you're an AI coordinator now.
23:40.627 --> 23:41.867
You're not a software engineer anymore.
23:41.867 --> 23:51.289
You're more guiding your agents and having them do the work, reviewing their work,
changing their context file to make them work exactly what you want.
23:51.289 --> 23:54.769
Yeah, so that's essentially what we've become now, you know?
23:54.963 --> 23:56.904
Right, it's changed the role, to your point.
23:56.904 --> 23:58.244
Yeah, absolutely.
23:58.244 --> 24:04.787
Yeah, I think the rigor and the engineering now isn't about the for loop or the for each
loop or the while loop.
24:04.787 --> 24:05.547
It's about this.
24:05.547 --> 24:08.468
It's about how do you make sure that it doesn't make that mistake again.
24:08.468 --> 24:11.519
Okay, what have you seen in testing?
24:11.519 --> 24:13.125
How does it change how testing works?
24:13.125 --> 24:14.257
that's a big one as well.
24:14.257 --> 24:16.901
let's start from just locally when you're testing, right?
24:16.901 --> 24:20.584
It's so important to tell AI what is considered done, like I mentioned before.
24:20.584 --> 24:24.640
If I write a new endpoint in API, I always tell, okay, don't just write the code.
24:24.640 --> 24:28.684
This is essentially the integration test I want you to write.
24:28.754 --> 24:35.970
and it's just going to send this data, it's going to do these assertions when it comes
back and just keep working until this passes, right?
24:35.970 --> 24:38.442
And like, don't just tell it to go right and then I'll test it.
24:38.442 --> 24:40.494
Like you have to tell it when to stop.
24:40.494 --> 24:42.144
And the same thing on the UI side.
24:42.144 --> 24:43.837
I do this all the time.
24:43.837 --> 24:48.221
It's like, I could get AI to write some code and then open the browser and click around,
make sure it works.
24:48.221 --> 24:52.065
But I always tell it like, hey, launch the Playwright MCP.
24:52.065 --> 24:53.276
You already know the story.
24:53.276 --> 24:54.236
You have all the context.
24:54.236 --> 24:55.718
You know what you have to test.
24:55.718 --> 24:57.159
just go and do it.
24:57.159 --> 24:58.919
And people think, oh, well, it's really slow, right?
24:58.919 --> 25:02.958
When you run the player at MC through Claude, it's kind of slow because it needs to
navigate.
25:02.958 --> 25:04.420
needs to figure out which buttons to click.
25:04.420 --> 25:07.820
But it saves a lot of time because you don't have to constantly tell, oh, it's not
working.
25:07.820 --> 25:09.720
I'm not seeing the button, right?
25:09.720 --> 25:14.580
When Claude is navigating MCP and it's going through the DOM, it knows it has all that
context.
25:14.580 --> 25:20.220
So you can quickly fix it and then go and click through and figure out the UI again.
25:20.408 --> 25:23.962
So one question about this, and this is like a literally naive question.
25:23.962 --> 25:26.533
This is not like me trying to be a smart interviewer here.
25:26.533 --> 25:34.558
When you say, like, come up with the specifications ahead of time, like, this is what I
want, you know, that test, that API to return.
25:34.558 --> 25:43.073
Are you sitting down, like in Visual Studio, and, you know, and writing out your JSON
response that you want to see, or are you working with it to develop those specs in the
25:43.073 --> 25:44.113
first place?
25:44.144 --> 25:44.995
Yeah, I'm working with it.
25:44.995 --> 25:47.473
I'm just telling like, hey, I want to send these parameters.
25:47.473 --> 25:49.707
It's all in the prompt usually.
25:50.330 --> 25:51.632
Same thing with the UI.
25:51.826 --> 25:57.589
I think when I hear people say, like, you just need to give a good exit criteria, I'm
like, yeah, but I don't know what that is.
25:57.589 -->
26:00.451
So I just keep going with my square wheels.
26:00.451 --> 26:06.195
But to your point, you're sitting down with Claude to come up with what the exit criteria
would be for it.
26:06.195 --> 26:09.656
Yeah, so that kind of covers the local testing that you do as a developer.
26:09.656 --> 26:16.016
I mean, it's also said that it was so much cheaper to have Claude just go and do extensive
testing because you don't have to constantly babysit it.
26:16.016 --> 26:20.795
You can just go off and try every single scenario, right?
26:20.795 --> 26:24.276
Like manual testing, and you're going to give access to your local Postgres database.
26:24.276 --> 26:26.997
They can go and check records there, you know?
26:26.997 --> 26:29.297
So that helps a lot.
26:29.490 --> 26:30.190
Yeah, 100%.
26:30.190 --> 26:33.804
The second piece is just writing all our E2E tests and integration tests.
26:33.804 --> 26:35.985
It's just become so much easier now.
26:37.127 --> 26:38.203
I think integration...
26:38.203 --> 26:41.329
things stable and trying to keep them not flaky, 100%.
26:41.329 --> 26:42.804
It's still a problem, to be honest.
26:42.804 --> 26:44.116
Still a problem.
26:46.424 --> 26:56.417
Even back in the Selenium days, Cypress, we're using Playwright right now, but it's still
a challenge that sometimes when it runs on GitHub Action, it's just running on a slower
26:56.417 --> 26:59.504
machine and there's a timing issue and you have a complex UX.
26:59.504 --> 27:09.286
So that's still an issue, but we need those things that run on every PR to make sure our
system works, but we can't obviously have E2E tests for Playwright for every single thing
27:09.286 --> 27:10.430
in the app.
27:10.540 --> 27:15.420
So then we have a nightly job that runs and covers the other things so it doesn't hold up
every PR.
27:15.763 --> 27:19.209
But now we're also experimenting with having these like...
27:19.322 --> 27:26.054
QA bots that just run on laptop like all day long and they're not using like written
playwrights.
27:26.054 --> 27:26.875
So they're slower.
27:26.875 --> 27:30.036
So they're more kind of discovering things in real time.
27:30.036 --> 27:34.767
So they're real slow, but they're very reliable because they can like kind of self heal.
27:34.767 --> 27:38.686
So if you move a button, it can figure out, the create button is not there anymore.
27:38.686 --> 27:41.279
Like whatever you did, a test ID gave it, but.
27:41.279 --> 27:44.731
it can kind of figure out in real time, oh, it's not great, it's new now.
27:44.731 --> 27:46.553
You just changed the name of it or something, right?
27:46.553 --> 27:53.708
So it can, it runs really slow, but it's a lot smarter and more robust because you can
kind of self heal.
27:53.708 --> 27:55.869
A lot of these QA AI tools are out there.
27:55.869 --> 27:58.200
They all claim, it can self heal.
27:58.201 --> 27:59.952
Yeah, they can, but they're all there.
27:59.952 --> 28:04.986
It's more marketing than actual reality, to be honest.
28:04.986 --> 28:06.887
They're still very cumbersome to use.
28:06.887 --> 28:08.499
The ones I've looked at.
28:08.499 --> 28:11.530
Okay, so I get like MCP in a terminal, 100%.
28:11.530 --> 28:12.974
I think we've all done that one.
28:12.974 --> 28:22.085
I get MCP with Playwright running in the daily to double check everything and running in
the GitHub action for every build.
28:22.085 --> 28:23.176
I totally get all of that.
28:23.176 --> 28:29.643
This last bit, like when you say like running on a laptop, do you a developer's machine or
some other way?
28:29.643 --> 28:31.301
We're setting up like a separate laptop.
28:31.301 --> 28:32.759
We're just experimenting with this.
28:32.759 --> 28:38.873
So it's just you're checking out the latest build and spinning it up and just error
guessing, sort of exploratory.
28:38.873 --> 28:39.263
testing.
28:39.263 --> 28:47.989
It's like a compliment, right, to our E2E where this bot can just go through our app and
kind of look at the story and kind of like figure out if there's anything missing.
28:47.989 --> 28:50.003
So it's not like, it's more.
28:50.003 --> 28:51.084
by story.
28:51.322 --> 28:55.402
Yeah, but it's more like we don't tell it exactly what to look for, right?
28:55.402 --> 28:59.062
It just looks at the story and it tries to figure out what to test based on that.
28:59.062 --> 29:02.742
Super slow, we're kind of just looking into it now and see if we can make that work.
29:02.742 --> 29:05.902
Because this is really nice compliment to what we have already.
29:07.662 --> 29:11.822
So I don't know if it's going to work, but we've done some experiments.
29:11.822 --> 29:15.182
seems to, it's just the speed that's involved.
29:16.122 --> 29:16.829
Yeah.
29:16.829 --> 29:18.137
What about debugging?
29:18.137 --> 29:20.596
Oh, that was the last time I opened that debugger.
29:20.940 --> 29:21.803
I can't even remember.
29:21.803 --> 29:23.007
uh
29:23.765 --> 29:25.180
That's what I mean.
29:25.180 --> 29:26.352
That's really good at it.
29:26.352 --> 29:31.755
Yeah, it's incredible, especially things that were really hard to troubleshoot back in the
day.
29:31.755 --> 29:38.059
You have some race condition, you You could spend hours just pulling your beard thinking,
what's going on here, right?
29:38.059 --> 29:40.450
You just give Claude as much information as can.
29:40.450 --> 29:44.824
Any logs, any information that happened at this time with these two users were used in the
system.
29:44.824 --> 29:46.436
This was the data that was in the database.
29:46.436 --> 29:49.140
Just throw as much information at it as you can.
29:49.140 --> 29:54.077
And most static code analysis, it can figure out, oh, I found this race condition right
here.
29:54.077 --> 29:55.339
And bugs, same thing.
29:55.339 --> 29:58.564
It's incredible if you just give it all the logs, all the data.
29:58.564 --> 30:01.167
It can usually just figure it out for you.
30:01.534 -->
30:03.396
I feel like it's faster at reading than I am.
30:03.396 --> 30:10.302
so that's, usually if I've got like a production issue, I'll queue up Claude and say, go
start to look at this, this is what I'm seeing.
30:10.302 --> 30:13.507
And then I'll try to like race it and I never win.
30:13.507 --> 30:19.173
But sometimes I will catch the occasional detail that I'm like, oh, I should tell it this,
it might not have seen it yet.
30:19.173 --> 30:20.534
But it will still win.
30:20.534 --> 30:24.017
But I still work with it, you know, but it's like I'm its little sidekick.
30:24.095 --> 30:26.029
Yeah, it was most of the time.
30:26.029 --> 30:30.601
I was so happy the other day, or was a while ago, when I did actually beat Claude to the
punch.
30:30.601 --> 30:32.902
I was like, I found a bug while he was looking at it.
30:32.902 --> 30:33.923
Missed it.
30:34.080 --> 30:35.240
Nicely done.
30:35.418 --> 30:35.998
OK.
30:35.998 --> 30:37.299
What about integration?
30:37.299 --> 30:47.184
And here, I'm thinking about not like your CI, CD pipeline, you know, and that kind of
integration, but more like, OK, there's a new partner API or a new product that you're
30:47.184 --> 30:52.020
getting access to an API or a department you don't normally talk to, like that kind of
integration.
30:52.020 --> 30:53.487
What are you seeing the changes there?
30:53.678 --> 30:57.489
I mean, since we're all one startup, don't have too much experience with that.
30:57.489 --> 30:58.402
related to that.
30:58.402 --> 31:06.268
Like when we deploy a new version of the API and then the new version of the UI, we always
have to make sure that the UI works with the old version and the new version, right?
31:06.268 --> 31:09.343
Cause people are not gonna instantly refresh their browser to get the new version.
31:09.343 --> 31:11.486
So that would be kind of similar where you need to make sure.
31:11.486 --> 31:13.929
what you would use contract testing for and things of that in the past.
31:13.929 --> 31:20.416
There's a lot of things you can do to just tell the AI just like, okay, this is the
version running in prod, this is version running in staging.
31:20.416 --> 31:27.902
Just basically go look through the commits and just figure out, hey, is this new version
gonna break the UI that's currently running?
31:28.061 --> 31:29.653
It's found it incredibly helpful.
31:29.653 --> 31:33.745
It's caught so many issues where it did break it.
31:34.947 --> 31:36.179
Yeah, like, just go through,
31:36.179 --> 31:38.491
Like all our services has like a slash build endpoint.
31:38.491 --> 31:41.263
And it'll tell you this is the Git commit that's running currently.
31:41.263 --> 31:48.270
So AI will go and fetch that and then it can just go into your Git repo and figure out,
these are the commits that are new in the new build, right?
31:48.270 --> 31:53.135
Versus what's currently running, what new endpoints have been added, are any of those
going to break, right?
31:53.135
--> 31:57.298
Which is tedious process to do because you need a lot of tests around that.
31:57.298 --> 31:59.601
And if you just have like one UI, you're not going to have.
31:59.601 --> 32:04.913
that extensive UI test that you test every single functionality for every version and
making sure nothing breaks.
32:04.913 --> 32:11.695
So it's caught so many issues where we were like, okay, no, someone just renamed something
here or had to do some change here and it's gonna break the UI.
32:11.695 --> 32:13.185
So, which is fine.
32:13.185 --> 32:14.986
Maybe that was what we wanted.
32:14.986 --> 32:20.027
We don't want to, we wanted to kind of clean up, but at least we have to do the deploy off
hours.
32:20.447 --> 32:23.989
Because it might be split second where things are not working properly.
32:24.258 --> 32:24.973
Awesome.
32:24.973 --> 32:26.074
guys are going really deep.
32:26.074 --> 32:27.726
I'm loving your answers.
32:27.787 --> 32:28.267
Okay.
32:28.267 --> 32:30.050
um What about refactoring?
32:30.050 --> 32:30.682
What do you see in there?
32:30.682 --> 32:34.763
I mean, a lot of people saying refactoring is dead, can just regenerate.
32:34.763 --> 32:37.343
I would say refactoring is incredibly cheap now.
32:37.343 --> 32:39.697
I think my only opinion on that.
32:39.697 --> 32:43.776
I'm not in a camp that you should just regenerate from spec all the time.
32:44.660 --> 32:46.791
Maybe we'll get there.
32:46.791 --> 32:48.394
a maybe we get there kind of thing.
32:48.394 --> 32:50.018
That seems very aspirational to me.
32:50.018 --> 32:52.994
how good a spec would you have to be writing to get that to be, yeah.
32:52.994 --> 33:00.578
Like every time, even though you have a really good spec, you're still making decisions,
telling it what, cause the spec is usually very business focused.
33:00.578 --> 33:06.691
And then you have, you make design decision and coding decisions with Claude back and
forth telling it, okay, this is how I want to structure it.
33:06.691 --> 33:08.202
I want to use this kind of pattern.
33:08.202 --> 33:08.932
Right.
33:08.932 --> 33:11.554
That usually doesn't go back into the spec.
33:11.554 --> 33:15.005
So when you regenerate the next time, you're going to answer those same questions, right?
33:15.005 --> 33:16.206
Maybe one day.
33:16.916 --> 33:17.956
yeah, yeah, yeah, yeah, yeah.
33:17.956 --> 33:20.536
But to your point, you find it cheap now doing refactoring.
33:20.536 --> 33:22.376
So you're doing a lot of it, I guess.
33:22.924 --> 33:23.586
Yeah, I mean...
33:23.586 --> 33:26.069
Especially, yeah, I mentioned tech that before, right?
33:26.069 --> 33:28.807
When there's things we missed or things over time.
33:28.807 --> 33:31.171
not sort of academic refactoring that I'm hearing.
33:31.171 --> 33:34.394
I'm not hearing like, I think we can improve this kind of stuff.
33:34.394 --> 33:37.118
I'm hearing more like, yeah, there's a tech debt over there that slowed us down.
33:37.118 --> 33:37.609
Kill it.
33:37.609 --> 33:39.020
Make it never slow us down again.
33:39.020 --> 33:42.174
Very sort of pragmatic tech debt approach is what I'm hearing.
33:42.174 --> 33:43.225
Yeah.
33:43.225 --> 33:44.145
Yeah.
33:44.217 --> 33:44.578
find.
33:44.578 --> 33:52.972
Like, you start one way and it kind of gets bigger and bigger and then, wow, we gotta redo
how we did this, just to make it perform.
33:52.972 --> 34:00.264
And it's just very easy to do that and then still have it like, okay, does it produce the
identical result to what you had before, right?
34:00.264 --> 34:07.006
It's so easy to have it do all those tests afterwards to verify that, your refactor is
good.
34:07.110 --> 34:15.463
Yeah, and like with debugging, I find that it's good at finding what the performance
problem is really quickly and quickly putting together quick shim tests to see if it
34:15.463 --> 34:16.175
actually helped.
34:16.175 --> 34:20.021
yes, we could have done all of this before, but like it just gets faster.
34:20.021 --> 34:20.881
Yeah.
34:21.163 --> 34:21.465
Yeah.
34:21.465 --> 34:25.529
I was mentioning before with the whole SQL performance that I got into.
34:25.529 --> 34:27.494
I that's always been a tool available.
34:27.494 --> 34:33.178
I used to always read like explain plans and stuff like that for really critical things,
right?
34:33.178 --> 34:34.319
Cause it takes a long time.
34:34.319 --> 34:39.213
Like you need to go get these SQLs and get it run for the query plan and then figure out
what all these things mean.
34:39.213 --> 34:40.544
now you can do it so easily.
34:40.544 --> 34:43.837
And also to explain the query plan, right?
34:43.837 --> 34:45.758
That takes a bit of talent too.
34:45.884 --> 34:46.746
Yeah.
34:46.746 --> 34:47.597
Let's go into that one.
34:47.597 --> 34:48.910
That was one of the articles you wrote.
34:48.910 --> 34:54.836
And if I remember right, you were talking about having built like a skill or an MCP,
something that sort of.
34:54.836 --> 34:55.768
custom command.
34:55.768 --> 35:02.752
You basically give it an endpoint name and it knows how to test that because every
endpoint has an integration test.
35:04.187 --> 35:07.651
and it's not too so much about like, okay, run this with millions and millions of records
in the database.
35:07.651 --> 35:10.273
It's more just about like, do static code analysis.
35:10.273 --> 35:16.135
So if you have your entity framework code, you can very easily go and say, okay, you're
joining all these tables.
35:16.135 --> 35:19.887
and then you can even kind of give it, okay, this is how many rows are in these tables
usually.
35:19.887 --> 35:31.162
so what the command does, it runs this endpoint in your branch and it records uh the
static analysis on how you interact with the database and it checks the query plan and
35:31.162 --> 35:35.939
then it checks out your main branch and then it kind of does the same thing and then it
kind of compares the two and it says,
35:35.939 --> 35:38.744
You are overfetching by this many thousands of records.
35:38.744 --> 35:40.928
You're joining all these tables that are not necessary.
35:40.928 --> 35:48.623
ah And it just gives you a really good side by side comparison in like a minute, which
would have been so tedious before, right?
35:48.623 --> 35:52.862
You basically would have waited for performance to get bad, to run like one of those
tools.
35:52.862 --> 35:53.930
you, we all.
35:53.930 --> 35:57.374
the database with tons of records and you would just test it like, it feels quicker.
35:57.374 --> 35:57.926
Done.
35:57.926 --> 36:01.672
Yep, And that's another one that feel like I really ought to try out.
36:01.672 --> 36:04.963
uh God, we're going through the list really fast.
36:04.963 --> 36:09.057
How do you feel like it's changed pair programming, mob coding, these kinds of things?
36:09.562 --> 36:11.722
Oh, it's gonna be controversial.
36:12.663 --> 36:23.826
I think pair programming as we're used to it, one person is driving, commands, the other
one's looking over the shoulder, making sure it's correct and things like that and
36:23.826 --> 36:29.510
learning from it, that kind of way kind of broke when you started using multi-agents,
right?
36:29.510 --> 36:30.610
However,
36:31.017 --> 36:37.290
very important to have senior engineers sit with more junior engineers to show how you're
working with AI, right?
36:37.290 --> 36:39.801
You just assume like, everyone knows all these tricks.
36:39.801 --> 36:42.842
I got up my sleeves and work for some of what not.
36:42.922 --> 36:44.203
I think that's important.
36:44.203 --> 36:48.882
And also like writing the AI workflows together, I think is also very important.
36:48.882 --> 36:50.422
Because they're so critical.
36:50.422 --> 36:57.122
Like everything you do now is around a couple of different commands that we always run as
part of our process.
36:57.122 --> 36:58.782
So I think that would be very important.
36:58.782 --> 37:09.931
But the traditional pair programming paradigm where you just sit and watch over each
other's shoulder, I think that to be replaced different flavor of it at least.
37:10.231 --> 37:13.713
I don't think that's sitting side by side for eight hours a day.
37:13.713 --> 37:17.049
it's not productive anymore with with Gentoo AI.
37:17.264 --> 37:17.944
Right.
37:18.304 --> 37:22.944
So listeners may have heard me talk about how the ThoughtWorks approach to pair
programming.
37:22.944 --> 37:23.884
It was literally that.
37:23.884 --> 37:25.104
It was like 80%, 90%.
37:25.104 --> 37:28.484
You sit down for three hours in the morning and pair program together.
37:28.484 --> 37:29.484
And you go have lunch.
37:29.484 --> 37:32.704
And then you come back and pair program for another three or four hours together.
37:32.704 --> 37:34.603
So you're, yeah.
37:34.603 --> 37:37.383
to be honest, I thought it was exhausting.
37:37.543 --> 37:45.403
As an introvert, I like putting on my headphones and like get into flow and code away and
to sit and talk to someone for eight hours a day.
37:45.403 --> 37:46.863
I was always exhausted.
37:46.883 --> 37:47.843
I have all the value though.
37:47.843 --> 37:52.343
Like it really brings a lot of value sharing the context, right?
37:52.343 --> 37:55.343
Amongst two people and constantly have a sounding board.
37:55.623 --> 37:58.063
But that's the thing, now AI is your sounding board, right?
37:58.292 --> 37:58.919
Right.
37:58.919 --> 38:03.801
I asked uh so many questions every day just to learn and challenge me.
38:03.801 --> 38:10.589
Like, this is what I think, like come up with a uh counter argument to how I see the
world, right?
38:10.589 --> 38:11.711
Yeah, yeah.
38:11.711 --> 38:12.225
I love that.
38:12.225 --> 38:13.507
I love that.
38:13.550 --> 38:14.101
Wait a minute.
38:14.101 --> 38:16.879
What are you asking it for counter arguments about?
38:17.631 --> 38:22.751
I mean, anytime where I'm just like, hey, I built this feature a gazillion times in the
past.
38:22.751 --> 38:24.831
This is how I would do it.
38:24.831 --> 38:25.411
Right.
38:25.411 --> 38:26.491
Challenge me on this.
38:26.491 --> 38:27.831
Is there a better way now?
38:27.831 --> 38:28.171
Right.
38:28.171 --> 38:30.931
Cause I, cause I've been doing it the same way for the last decade.
38:30.931 --> 38:32.651
Doesn't mean that you're it.
38:32.651 --> 38:32.831
Right.
38:32.831 --> 38:33.972
Cause sometimes you have that, right.
38:33.972 --> 38:36.652
was working on a impersonation feature the other day.
38:36.652 --> 38:36.852
Right.
38:36.852 --> 38:37.692
I built that one.
38:37.692 --> 38:40.012
That's the third time I built that one in a SaaS app.
38:40.072 --> 38:41.912
Like I just like, I'm almost on autopilot.
38:41.912 --> 38:44.932
I know exactly what to do and what challenges to face.
38:44.972 --> 38:45.645
So.
38:45.645 --> 38:48.556
I always want to make sure that Claude is keeping me honest there.
38:48.556 --> 38:52.197
This is how I've always done it, but it's a better way now.
38:52.468 --> 38:53.089
OK.
38:53.089 --> 39:01.171
So impersonation here, I think you mean like software as a service where it's like, OK,
I'm customer service and I want to be able to log in as Kyle.
39:01.391 --> 39:02.271
Yeah.
39:03.131 --> 39:03.611
Yeah.
39:03.611 --> 39:03.902
Yeah.
39:03.902 --> 39:08.653
And that way I can service your account or maybe I'm trying to figure out a bug or
whatever.
39:08.653 --> 39:09.053
OK.
39:09.053 --> 39:09.383
Right.
39:09.383 --> 39:15.165
So that's the kind of thing where it's not like, oh, you know, it's such and such method
in the base class library that does that.
39:15.165 --> 39:17.135
You're looking for an entire approach.
39:17.135 --> 39:18.095
Yeah.
39:19.144 --> 39:21.489
So you guys have been doing this for a while, you said.
39:21.489 --> 39:24.194
Like what size of team are we talking about?
39:25.338 --> 39:26.159
Yeah.
39:26.251 --> 39:32.975
started first kind of like evaluate this, how feasible is this to just like go all in on
this, right?
39:33.195 --> 39:37.478
And I quickly realized that this is really working.
39:37.478 --> 39:44.403
Like, especially if you use that time you gain to reinvest and put these guardrails in and
make sure the quality is there.
39:44.624 --> 39:45.424
And...
39:45.570 --> 39:47.881
The rest of the team has been really great.
39:48.201 --> 39:53.096
know the whole team has just picked it up and everyone is just using Claude and multiple
agents.
39:53.096 --> 39:59.088
And we have all these shared AI workflows and context files and everything.
39:59.088 --> 40:03.820
it's, I think it made it easier that the rest of the team was very, was sold pretty
quickly.
40:03.820 --> 40:05.271
didn't have much research.
40:05.271 --> 40:06.211
Yeah.
40:06.643 --> 40:07.795
yep, got it.
40:07.795 --> 40:11.577
it sounds like you guys are working on developing these things together, right?
40:11.577 --> 40:12.798
Which makes a lot of sense.
40:12.798 --> 40:17.581
I mentioned superpowers earlier, but when I added it to my Claude setup, it started doing
weird stuff.
40:17.581 --> 40:19.161
It started suggesting pull requests.
40:19.161 --> 40:22.443
And I'm like, there's just one of me, I've never done a pull request, we're not doing
that.
40:22.443 --> 40:23.884
So where did that even come from?
40:23.884 --> 40:27.849
So I can imagine that a team working together, you'd really wanna do a
40:27.849 --> 40:31.849
ways of working session almost around like what do we want it to do?
40:31.849 --> 40:33.429
how does that work for you guys?
40:33.429 --> 40:38.209
Is that like a weekly tech huddle where you're kind of going through Claude MD together?
40:38.212 --> 40:39.528
how does it work?
40:39.855 --> 40:42.295
We have a weekly tech meeting.
40:42.455 --> 40:43.995
We could probably be better.
40:43.995 --> 40:45.295
A lot of this is coming from me.
40:45.295 --> 40:50.995
I'm pushing a lot of these things, but other developers are adopting it and using it.
40:51.321 --> 40:54.855
But I'm trying to get other team members too to contribute more to these files.
40:54.855 --> 40:56.986
Especially people are really good at certain things.
40:56.986 --> 40:59.624
we got a UX or UI engineer on our team.
40:59.624 --> 41:01.280
They're really good React developer.
41:01.280 --> 41:05.584
I really like tell them, please go and add things in the PR review command.
41:05.584 --> 41:07.025
Go and add things in the Claude MD.
41:07.025 --> 41:12.469
Like we need your expertise in here because you're really strong on this.
41:13.222 --> 41:14.921
So you're sort of the tool smith on some level.
41:14.921 --> 41:17.422
You're the one, the agent herder, in a way.
41:17.422 --> 41:27.086
Yeah, I'm trying to, but that's, where I think, yeah, like I'm a weekly mob session or
something where we just like review all these tools on a continuous basis.
41:27.086 --> 41:29.747
That's probably a good idea.
41:30.167 --> 41:31.708
We just been running so fast here.
41:31.708 --> 41:36.505
Like we sometimes we just got to take a step back and like, uh yeah.
41:36.661 --> 41:37.022
okay.
41:37.022 --> 41:39.262
So it definitely sounds like you're bought in.
41:39.262 --> 41:43.003
There's no hesitation in the way you're talking about it.
41:43.003 --> 41:47.823
What kind of improvement do think you're getting out of all of this?
41:48.294 --> 41:49.075
It's a lot.
41:49.075 --> 41:55.860
I mean, we started this project the first four or five months, we were just co-pilot a lot
of manual coding, right?
41:55.860 --> 42:00.826
um So we have the velocity back then compared to now.
42:00.826 --> 42:09.462
I don't have the numbers in front of me, but I'm saying it's like probably three to four
times the number of points iteration easily.
42:09.462 --> 42:10.242
Wow, wow, wow.
42:10.242 --> 42:14.262
OK, and using a similar points currency that you do today.
42:14.262 --> 42:14.661
Yeah.
42:14.661 --> 42:14.941
it.
42:14.941 --> 42:20.201
We switched from target process to linear and linear doesn't have 0.5 points on stories.
42:20.201 --> 42:22.981
We had to like switch everything up a little bit.
42:22.981 --> 42:28.121
So that affected our scoring, but still it's a huge improvement.
42:28.128 --> 42:31.872
Yeah, And that lines up with what I'm hearing from other people as well.
42:31.872 --> 42:34.874
And it definitely feels like what I'm seeing personally too.
42:35.034 --> 42:36.476
What do you see this do to ratios?
42:36.476 --> 42:42.020
Like six developers, like what's the complement of testers, product people, business
analysts?
42:42.020 --> 42:45.814
Like do you have much of that, UI designers, or is it all devs?
42:45.814 --> 42:48.035
Yeah, we have a UX designer, which is really helpful.
42:48.035 --> 42:52.596
She does all the Figma designs, which then Claude can pull in when it builds a story.
42:52.596 --> 42:53.378
Super helpful.
42:53.378 --> 42:58.741
We have a product owner, no QA, so we have to make sure we write a lot of tests ourselves.
42:58.741 --> 43:06.005
I think it would be helpful to have like a QA that's outside of the team and can use AI,
whatever tools they want to just...
43:06.005 --> 43:10.582
separately from the team, right, all these tasks and figure out whatever ways AI can be
used.
43:10.582 --> 43:17.636
But it's been working pretty well so far, even with just the tremendous amount of features
we're pumping out every week.
43:17.636 --> 43:23.243
Okay, so that ratio then, six devs, one product owner, is working out.
43:23.243 --> 43:26.608
Do you feel like you could add two more devs without adding another product owner?
43:26.608 --> 43:30.533
Or you could add a product owner first?
43:30.533 --> 43:33.444
How do you feel about how that part works out?
43:33.836 --> 43:34.528
That might be hard.
43:34.528 --> 43:39.689
think we feel like we're a very big team at six developers with everyone
using agentic AI.
43:39.898 --> 43:40.679
Yep.
43:41.007 --> 43:41.438
Got it.
43:41.438 --> 43:44.332
So to your point, you wouldn't necessarily want to get anything bigger.
43:45.115 --> 43:45.751
Yep.
43:45.751 --> 43:45.882
so.
43:45.882 --> 43:48.753
It's a lot of value of having a small team, to be honest.
43:48.753 --> 43:52.905
just keeps the cohesion and adding more developers is not going to make us go faster.
43:52.905 --> 44:00.199
You can actually make us go slower because then there's just so much, like keeping
everyone on the same page gets harder the bigger the team you get.
44:00.199 --> 44:03.720
If you got a small, tight-knit team, it's very powerful.
44:03.954 --> 44:04.626
Yeah, 100%.
44:04.626 --> 44:06.369
And it kind of allows you to stay small.
44:06.369 --> 44:10.798
Like it takes away any pressure to get bigger because you can go so much faster.
44:10.840 --> 44:11.384
Yep.
44:11.384 --> 44:11.745
a startup.
44:11.745 --> 44:16.425
We're replacing the original product and then we're also building all these new features.
44:16.425 --> 44:21.185
we're in this phase now where we just like we're building so many new features.
44:21.185 --> 44:24.325
We're going to get to a phase where it's going to be more maintenance, right?
44:24.325 --> 44:30.685
And we're six developers and we're reaching all our goals in terms of like deadlines and
timelines easily.
44:32.529 --> 44:34.134
So yeah, six developers.
44:34.134 --> 44:35.917
a good team size, I think.
44:36.103 --> 44:37.768
Yeah, yeah, yeah, I'm with you.
44:37.768 --> 44:39.052
Okay, cool.
44:39.052 --> 44:42.822
Has this changed how you hire or staff, how you train or mentor?
44:42.822 --> 44:46.771
we haven't really hired anyone since we kind of went all in with AI.
44:46.771 --> 44:48.484
The hires happened before that.
44:48.484 --> 44:57.617
like how we mentor people is interesting because it's so easy for anyone now to learn any
part of the system just by having a conversation with AI, right?
44:57.617 --> 45:03.507
And we try to keep a lot of these context files with very domain specific things in our
code base.
45:03.507 --> 45:10.567
not just like you ask about the code, we can also ask about how the product worked and all
these more areas of the application.
45:10.567 --> 45:19.196
More if you can just sit there and like have a conversation about it's kind of reduced the
need for constantly be asking other developers how the system works.
45:19.089 --> 45:21.010
So another question I ask
45:21.010 --> 45:22.110
How do you enable teams?
45:22.110 --> 45:28.190
mean, that sounds like you're part of your role, at least, is to sort of figure out how to
use these tools.
45:28.191 --> 45:30.891
do you just go to somebody's desk and say, like, hey, this is how we're going to do it
now?
45:30.891 --> 45:34.551
I mean, looks like you guys are remote because you're in the same remote office you were
always in.
45:34.551 --> 45:39.971
So how does it actually work when it
comes to it's time to bring a team to speed with a
new approach?
45:39.971 --> 45:45.011
We were actually in an office, but people are only in Mondays and Wednesdays.
45:45.011 --> 45:47.632
And it's a long commute for me, so I'm here the other days.
45:47.632 --> 45:56.472
a lot of it is coming up with these shared AI workflows that we can all agree upon, and
then just getting the team members to use it.
45:56.792 --> 45:58.412
And that's been the biggest part.
45:58.412 --> 46:01.632
hasn't been that big of a challenge, be honest.
46:01.632 --> 46:06.932
It's just a couple of Claude commands that we're just using religiously.
46:07.907 --> 46:12.054
Yeah, just having all these like, color details.
46:12.054 --> 46:18.400
of like push the command into the code base and then like what show up on Slack and be
like, hey y'all, there's a new command.
46:18.597 --> 46:19.817
Pretty much.
46:20.138 --> 46:20.759
Yeah.
46:20.759 --> 46:21.579
And not forcing it.
46:21.579 --> 46:23.221
There's a couple of things like the PR views.
46:23.221 --> 46:25.002
That's like everyone has to run that.
46:25.002 --> 46:26.353
But there's tons of other commands too.
46:26.353 --> 46:27.614
Like you can use it if you want.
46:27.614 --> 46:29.226
If you want to come up with your own commands.
46:29.226 --> 46:29.946
Just go ahead.
46:29.946 --> 46:32.318
Like I don't want to force it on anyone either.
46:32.318 --> 46:38.933
Like if you don't want to run multiple agents at the same time in different work trees,
I'm not going to force you to do that.
46:38.933 --> 46:42.797
But if you want to, there's a start PR command and it will do all these things for you.
46:43.058 --> 46:43.829
Set it up for you.
46:43.829 --> 46:45.199
Copy all the files.
46:46.132 --> 46:49.132
Right, but yeah, it's optional.
46:49.402 --> 46:57.191
it came time to choose, mean, you've already talked about how Claude is really working for
you guys.
46:57.191 --> 46:58.792
Was that a conversation?
46:59.092 --> 47:04.916
did everybody else want to go try out Codex or try out Gemini CLI or WindSurf for
something?
47:04.916 --> 47:07.117
Or how did that conversation go?
47:07.636 --> 47:08.917
there wasn't a lot of pushback.
47:08.917 --> 47:12.821
mean, especially at the time, Claude was ahead of Codex, I think.
47:12.821 --> 47:13.883
Now they're more equal.
47:13.883 --> 47:17.556
And people just saw me working with Claude and what I was able to accomplish.
47:17.556 --> 47:18.787
It was a pretty easy sell.
47:18.787 --> 47:23.071
So the entire team now has a Claude Max subscription, which I think is incredibly cheap.
47:23.071 --> 47:25.814
I don't know how they can sell it for like 140 bucks a month.
47:25.814 --> 47:30.098
I run multi-agents all the time and I barely never hit any limits.
47:30.098 --> 47:38.038
it's probably just a matter of time before increase the fees for these tools, because
we're gonna get so addicted to them and then they're gonna charge $1,000 a month and we're
47:38.038 --> 47:41.598
like, okay, it's still worth it, just take my money.
47:42.020 --> 47:51.024
My hope is that Genton can make the GPUs cheaper and so the cloud fees can, yeah, I mean
if we can reduce their cost structure then their profit margins are already gonna be huge.
47:51.024 --> 47:54.666
It's just a matter of making sure that the costs aren't too bad.
47:54.666 --> 48:02.889
so we've sort of covered off all the questions that I ask people around like how Gen.AI is
changing their work and we're kind of getting into the more
48:02.889 --> 48:04.371
questions that could be about anything.
48:04.371 --> 48:08.046
Could be about work, could be about personal life, could be about just anything you want
to cover.
48:08.046 --> 48:10.210
What is something you've had to change your mind on?
48:10.210 --> 48:11.932
Something that wasn't intuitive at first.
48:11.932 --> 48:15.117
You thought it was going to be this way and it turned out, nope, was something else.
48:15.117 --> 48:21.432
I think, I mean, obviously just having AI generate the code, thought that would never
like, no, it can't generate all it.
48:21.432 --> 48:24.994
Especially using like cursor early on, I was really turned off by that experience.
48:24.994 --> 48:26.736
It would just go and generate all these files.
48:26.736 --> 48:29.237
It poor quality and I didn't like it at all.
48:29.237 --> 48:30.688
I like I had no control.
48:30.688 --> 48:34.291
So for AI to just write all my code, I thought that's never going to happen.
48:34.291 --> 48:35.323
I'm never going to allow it.
48:35.323 --> 48:41.840
And it's hard for anyone, spent all your life figuring out to be a good engineer and a
coder.
48:41.840 --> 48:46.100
And then you have this tool that just does everything you spent decades and decades
learning.
48:46.100 --> 48:48.820
It's, I'm not going to lie, right?
48:48.820 --> 48:54.155
It's a very hard thing to accept, it doesn't help being denial about it.
48:54.155 --> 48:56.675
And this is the way things are now.
48:57.132 --> 48:59.273
What's something that you're most proud of?
48:59.960 --> 49:03.280
I usually recall this story because it's kind of interesting.
49:04.080 --> 49:11.340
How I got my first job was when I was in university and it was right during the big IT
bubble burst.
49:11.340 --> 49:17.436
It was in 2001 I believe, it was my second year I somehow managed to land a job,
49:17.436 --> 49:21.688
when there was like people with 10 plus years of experience looking for a job.
49:21.688 --> 49:24.790
And I'm still puzzled how I managed to do that.
49:24.790 --> 49:34.061
And that then turned into like a 20 year engagement with this company where I essentially
took over entire software stack and I was the only developer and I ran this SaaS business
49:34.061 --> 49:36.271
as a side project for 20 plus years.
49:36.271 --> 49:39.713
I thought that was a pretty interesting story because I still puzzled.
49:39.713 --> 49:43.847
I think I had very high
confidence back then.
49:43.847 --> 49:50.963
Because I remember the hiring manager was just, I have guys here with 10 plus years
experience, why would I hire some university student, second year of university.
49:50.963 --> 49:55.427
I said something to him, just give me your hardest problem, I'll just go and solve it for
you.
49:55.427 --> 49:58.270
And I ended up doing that and I showed him something.
49:58.270 --> 50:01.092
They were so impressed, they yeah, we'll hire you.
50:01.526 --> 50:03.143
that engagement for 20 years.
50:03.143 --> 50:04.711
was pretty interesting.
50:04.711 --> 50:05.366
wow.
50:05.366 --> 50:07.020
Yeah, I just gave it up a couple of years ago.
50:07.020 --> 50:10.177
was a fairly large SaaS app.
50:10.177 --> 50:19.269
It worked out really well because later on, this was back in Sweden where I went to
university and I then moved to Canada and I couldn't work for the longest time because it
50:19.269 --> 50:20.903
took so long to get my work permit.
50:20.903 --> 50:28.488
It took like seven months where I was just sitting in limbo, running out of money and they
approached me and asked me like, hey, could you like rewrite our whole app it was desktop
50:28.488 --> 50:28.852
back then.
50:28.852 --> 50:31.633
So they wanted like a SaaS app and like, sure, I'll do it.
50:31.633 --> 50:33.023
And it just worked out so perfectly.
50:33.023 --> 50:35.942
I could work for them while I was waiting for my work permit.
50:35.942 --> 50:37.563
I would have run out of money.
50:37.956 --> 50:49.365
And then I maintained this app for 20 years, which was a very interesting experience
because very few users and still work on other full-time projects during the day.
50:49.365 --> 50:52.337
Just to see a codebase evolve over such a long time.
50:52.337 --> 50:56.269
And I had to do a complete rewrite of the UI, I think, two times.
50:56.269 --> 50:59.092
Just because like every seven years, think you got to do a UI rewrite.
50:59.092 --> 51:00.544
That's always something I keep in my head.
51:00.544 --> 51:02.182
But the backend was kind of the same.
51:02.182 --> 51:03.788
It was a Java backend.
51:03.788 --> 51:09.047
It just kind of got tweaked over time, but it was essentially the same code base as 20
years ago.
51:09.047 --> 51:11.966
Okay, hopefully you moved off of struts onto something else at some point.
51:11.966 --> 51:13.021
Just just guessing.
51:13.021 --> 51:18.385
Yeah, so it's small updates on the backend, but like big rewrites on the UI, because UI
changes so quickly.
51:18.385 --> 51:22.955
I think you have to do like almost a quick, complete rewrite every seven years.
51:22.955 --> 51:23.977
That's what I have in mind.
51:23.977 --> 51:24.257
that.
51:24.257 --> 51:26.077
So every seven years you just gotta rewrite the UI.
51:26.077 --> 51:27.397
That's good advice.
51:27.777 --> 51:29.417
Yeah, I'm sticking with that.
51:29.557 --> 51:32.833
Okay, so what's something that's giving you joy right now?
51:32.862 --> 51:33.703
Outside of work or?
51:33.703 --> 51:38.120
oh
51:38.120 --> 51:38.841
workaholics.
51:38.841 --> 51:40.670
So, you know, it is what it is.
51:40.670 --> 51:43.370
Yeah, I've kind of fit in that camp a little bit too.
51:43.370 --> 51:45.890
I need to work a little bit less.
51:46.686 --> 51:50.386
One thing that's really bringing me joy, I know you've seen these pictures behind me.
51:50.386 --> 51:55.166
I put these up a long time ago because I wanted to go up one of the mountains here called
Mount Temple.
51:55.166 --> 51:57.166
It's one of the tallest mountains in Canada.
51:57.166 --> 52:00.226
So I've been training for that and I finally did it this summer.
52:00.226 --> 52:01.466
It took me two attempts.
52:01.466 --> 52:05.926
I failed the first one because there was too much snow, even though it was middle of
summer.
52:05.926 --> 52:11.366
We just got hammered with snow and it got way too dangerous once we were like 70 % up the
mountain.
52:11.566 --> 52:16.046
But then I came back a of years, sorry, a couple of weeks later and it was a beautiful
day.
52:16.046 --> 52:16.640
It was no snow.
52:16.640 --> 52:28.820
just a little bit on the top and yeah it was a 10-hour hike to get up and down but it was
was incredible that brings a lot of joy and I just come back from Utah where I was last
52:28.820 --> 52:36.820
week hiking in Zion National Park which was an incredible experience so yeah the outdoors
bring me incredible joy
52:36.958 --> 52:41.457
oh I have not been to Zion yet, but Monument Valley in southern Utah.
52:41.457 --> 52:43.230
It's just beautiful down there.
52:43.230 --> 52:44.351
Holy cow.
52:45.414 --> 52:46.255
Yep.
52:46.356 --> 52:47.278
Awesome.
52:47.278 --> 52:48.801
Jonas, thank you so much.
52:48.801 --> 52:49.892
This has been great.
52:49.982 --> 52:52.227
Thank you for having me on, it's been really fun.
53:05.867 --> 53:07.930
I'm going to pause for a second.
53:07.930 --> 53:10.392
Can you hear my dog barking in the background?
53:10.415 --> 53:14.803
I used to record in a studio So now I'm doing this out of my home studio.
53:17.500 --> 53:18.360
am I gonna do?
53:18.360 --> 53:20.020
How am I gonna my dog to stop barking?
53:21.920 --> 53:22.880
Okay.
53:24.640 --> 53:26.680
They're both rescues.
53:26.820 --> 53:28.960
But it's always the one that's barking.
53:28.960 --> 53:32.140
She's kind of got a Jack Russell terrier kind of thing to her.
53:33.300 --> 53:34.260
No.
53:34.780 --> 53:35.673
She stopped.
53:36.316 --> 53:38.276
So I'm going.
53:41.040 --> 53:42.540
think my wife just got home.
53:42.540 --> 53:45.260
I think that's part of why this is going so.
53:45.260 --> 53:47.431
But I'm gonna try to just keep going.