WEBVTT
00:00:06.240 --> 00:00:08.320
Welcome to the Dashboard Effect Podcast.
00:00:08.400 --> 00:00:09.279
I'm Brick Thompson.
00:00:09.519 --> 00:00:10.480
And I'm Landon Oaks.
00:00:10.880 --> 00:00:18.000
Landon, last time we met, we talked about using AI tools in just in business but in programming.
00:00:18.160 --> 00:00:28.399
And one of the things that we touched on was that there can be some reticence for software engineers to want to use these tools for a number of reasons.
00:00:28.559 --> 00:00:31.920
One, maybe they tried them earlier and they didn't work very well.
00:00:32.159 --> 00:00:37.920
Or I don't know, they they they had an experience where it seemed like it took more time than it actually saved.
00:00:38.159 --> 00:00:44.240
But one of the things I've also hear is that there's some people that just love the art of programming.
00:00:44.640 --> 00:00:51.759
I think of myself as a programmer, I've always thought of it or have a relationship to it that's more like it's a tool.
00:00:51.920 --> 00:01:05.680
Like the gratification I get is out of building something that works and solves a problem, not necessarily figuring out the beauty of the code, but there are other people that really enjoy that figuring out the beauty of the code.
00:01:05.920 --> 00:01:11.280
And so I'm sure uh even on your team, and maybe you've had that experience yourself, those types of feelings.
00:01:11.359 --> 00:01:11.680
I don't know.
00:01:11.760 --> 00:01:14.000
I I'd just like to kind of talk through that a little bit.
00:01:14.079 --> 00:01:23.359
And how do you how do you support someone who's got that, but also you want them to take advantage of some of the productivity gains that you can get through some of these tools?
00:01:23.680 --> 00:01:23.920
Yeah.
00:01:24.159 --> 00:01:25.519
Yeah, it's a great question.
00:01:25.680 --> 00:01:30.239
I um I'm actually I lean a little bit more towards your camp personally, right?
00:01:30.640 --> 00:01:37.840
Um, where I I the fun for me is what happens at the end, not so much building it in the first place.
00:01:38.079 --> 00:01:38.480
Yeah.
00:01:38.719 --> 00:01:47.599
Um but I do have several, several guys on my team that are more in that camp of the art of the code, right?
00:01:48.079 --> 00:01:57.120
Um and so I think what you hit on early was was the first roadblock that I saw was okay, they tried it back in August or October, right?
00:01:57.200 --> 00:01:58.560
Um, of 2025.
00:01:58.719 --> 00:01:58.879
Yeah.
00:01:59.120 --> 00:02:00.959
It uh wasn't very good.
00:02:01.200 --> 00:02:04.719
Uh and so a lot of them just kind of stopped and never returned to it.
00:02:05.200 --> 00:02:09.439
I mean, they'll use it for like questions and help to do some research, et cetera.
00:02:09.599 --> 00:02:11.759
But they're just like, oh no, it's it's not good at code.
00:02:12.159 --> 00:02:15.280
Well, and they probably tried it prior to that, too, several times.
00:02:15.520 --> 00:02:17.520
So we sort of had people testing it.
00:02:18.240 --> 00:02:20.719
But what so what haven't after October then?
00:02:21.120 --> 00:02:22.719
Yeah, obviously it got way better.
00:02:23.280 --> 00:02:26.800
Um so now it now it's really impressive what it can do.
00:02:26.960 --> 00:02:27.360
Yeah.
00:02:27.680 --> 00:02:41.039
Um and so I've been, you know, we're trying just kind of trying to get those guys to get back in there and start using it so at least see that, hey, this is not the same model, you know, this is it works much better now.
00:02:41.280 --> 00:02:57.280
Um, which I, you know, I've I've shown them a couple of things that I've done with it, which have been pretty cool, taking like, you know, a couple hours when it would have taken a full day of work um to kind of help them get back into that mindset of, okay, this is this is a different game now.
00:02:57.759 --> 00:03:07.919
It is different because certainly with Opus 4.7, a GBT 5.5, Codex and Claude Code, you can write code much faster.
00:03:08.639 --> 00:03:10.960
There's still issues, though, that crop up.
00:03:11.120 --> 00:03:14.400
Like it it can write code that's not as efficient as it should be.
00:03:14.560 --> 00:03:27.919
You know, it's it's repeating code that um doesn't need to, or I guess it could you know introduce bugs and so on that you don't know about, although frankly you can use those tools for really thorough testing as well.
00:03:28.159 --> 00:03:32.960
Um so there's sort of a different mindset that has to come into play, I think.
00:03:33.199 --> 00:03:33.599
Yeah.
00:03:33.840 --> 00:03:40.879
You know, you're using it to to crank out sort of the rote code that you would spend a few hours writing it can just do quickly.
00:03:41.039 --> 00:03:50.719
But then you have to be really smart about how you prompt it, what your design is, um, maybe you know how you give the requirements to the tool.
00:03:50.800 --> 00:03:53.759
And then how are you verifying that it's really good?
00:03:54.560 --> 00:03:59.199
And I think maybe the art comes into comes back into it in terms of taste.
00:03:59.360 --> 00:04:04.240
Like there's a certain there's a certain thing about taste in software development.
00:04:04.639 --> 00:04:16.000
You can see code that works, but it's not nice to use, not nice to look at, um, just feels clunky, and then code that's very similar but is really smooth and easy to use, something like that.
00:04:16.160 --> 00:04:18.560
So maybe the art comes back in with that.
00:04:18.720 --> 00:04:19.279
I don't know.
00:04:19.600 --> 00:04:21.360
Yeah, yeah, no, I completely agree.
00:04:21.439 --> 00:04:26.240
That's that's where the background of coding is so huge with AI.
00:04:26.560 --> 00:04:34.160
Um, you can actually go look at it now and then tell it to make these changes, do something different, etc.
00:04:34.879 --> 00:04:44.480
Um, which that's actually how one of our one of like my better engineers on my team, well probably the best, um, he uh he kind of got over his.
00:04:44.560 --> 00:04:46.480
He was much more much more in the art camp, right?
00:04:46.720 --> 00:04:54.160
You know, he likes to do the code, he likes to make sure it's reusable, it's modular, it's clean, big proponent of that, and he's really good at it.
00:04:54.480 --> 00:05:00.079
Um how he kind of got over that hump is he uses it in Visual Studio Code now with the extensions.
00:05:00.480 --> 00:05:00.639
Okay.
00:05:00.959 --> 00:05:08.720
And he has it like basically when you when you use AI to do something, you can do yes, always do this, or I'm gonna improve every single thing you do.
00:05:08.879 --> 00:05:09.120
Yes.
00:05:09.439 --> 00:05:15.199
He is in the I'm gonna improve every single thing you do, because he actually reads it, looks at it.
00:05:15.439 --> 00:05:19.040
And if he doesn't like it, he'll tell it, no, you need to change it to do this.
00:05:19.199 --> 00:05:22.879
Or you know, this looks bad, like make this cleaner.
00:05:23.120 --> 00:05:30.079
Um and he still gets huge gains out of it, but he he still feels like he's kind of in control, you know, and really the orchestrator.
00:05:30.480 --> 00:05:32.319
Do you think he enjoys that mode of working?
00:05:32.639 --> 00:05:33.040
He does.
00:05:33.120 --> 00:05:33.279
Yeah.
00:05:33.680 --> 00:05:34.240
It's interesting.
00:05:34.399 --> 00:05:36.240
I I'm more in the YOLO camp.
00:05:36.399 --> 00:05:36.639
Yeah.
00:05:36.800 --> 00:05:39.040
Let it roll and then let's uh look at it afterwards.
00:05:39.360 --> 00:05:39.839
Yeah, yeah.
00:05:40.160 --> 00:05:41.519
I think for me, it it depends.
00:05:41.680 --> 00:05:49.360
If it's something that I know really well, I I like the way he does it, because I, you know, I can really make sure that it's gonna be really useful.
00:05:49.680 --> 00:05:52.160
If it's something that I don't work in, I'm like, just go for it.
00:05:52.240 --> 00:05:54.399
I'm not I'm not gonna give you any good feedback anyways.
00:05:54.560 --> 00:05:56.079
Like, it's interesting too, though.
00:05:56.160 --> 00:06:03.759
If I'm working on production code that, you know, is is critical to our business or our client's business, then I'm not YOLO at all.
00:06:03.920 --> 00:06:11.759
But if I'm building if I'm building some kind of an application, maybe it's to help me with some kind of business utility, or maybe it's a personal thing, yeah.
00:06:12.000 --> 00:06:15.680
I'm more apt to let it go and see how it does and then work on it from there.
00:06:15.839 --> 00:06:23.360
Um and and of course, if you're doing something in production, you've got to review all the code and and make sure it's actually safe and not doing something weird.
00:06:23.759 --> 00:06:24.000
For sure.
00:06:24.399 --> 00:06:29.600
So what do you think is the best way to help someone get over that hump?
00:06:30.160 --> 00:06:40.079
Not that not that they have to, but I do worry that people that don't start utilizing these tools that it could it could hurt them just in their professional career.
00:06:40.319 --> 00:06:40.720
Yeah.
00:06:40.959 --> 00:06:44.079
Um how do you how do you think about that?
00:06:44.160 --> 00:06:46.720
Aaron Powell Yeah, it's a it's a really good question.
00:06:46.879 --> 00:06:53.680
Um and we're we're still kind of trying to figure it out for so for instance on my team, I you know, we we use AI a lot.
00:06:53.759 --> 00:06:56.240
I think we could use it more and better.
00:06:56.720 --> 00:06:59.279
And we're still trying to foster that kind of culture.
00:06:59.439 --> 00:07:09.279
Um one of the things we're working on now is for lack of a better name, we're calling it AI enablement, um, where we have certain people in the company who have already fully adopted AI.
00:07:09.360 --> 00:07:13.920
They're they're using it in awesome ways, they're getting incredible gains out of it.
00:07:14.240 --> 00:07:25.279
Um so we're gonna have those people kind of lead a little bit of a group to uh show what they're doing, show others what they're doing, encourage other people who aren't in that group to also show what they're doing.
00:07:25.519 --> 00:07:28.959
So the idea is like, let's just show them the art of the possible.
00:07:29.120 --> 00:07:29.279
Yeah.
00:07:29.519 --> 00:07:36.879
Hopefully get the the, you know, get the machine turning upstairs to where they're like, oh, you know, I could use this on that problem I had yesterday.
00:07:37.040 --> 00:07:37.120
Yeah.
00:07:37.360 --> 00:07:38.399
I hated that problem.
00:07:38.560 --> 00:07:40.079
It'd be awesome.
00:07:40.639 --> 00:07:41.199
You know.
00:07:41.759 --> 00:07:42.160
Yeah.
00:07:42.319 --> 00:07:43.600
Yeah, it's interesting.
00:07:43.839 --> 00:07:47.279
I think uh I think it some of it is that.
00:07:47.360 --> 00:07:50.560
Like, let me just show you how fun this can be to be this productive.
00:07:51.360 --> 00:07:54.160
Some of it I think is uh mindset too.
00:07:54.399 --> 00:08:02.399
Um, you know, I I definitely talked to people that have, and I've had this, still have it from time to time, sort of the existential worry about AI.
00:08:02.560 --> 00:08:04.399
What does it mean in the longer term?
00:08:04.639 --> 00:08:11.120
Um, I think the last few months, people have been very focused on uh the possibility that AI is going to wipe out a lot of jobs.
00:08:11.759 --> 00:08:19.680
I think more recently, everything that I'm reading and talking to people, it looks like it's actually going to net create a lot of jobs.
00:08:19.920 --> 00:08:33.279
And um there will definitely be disruption, which is not great for the people that get disrupted, uh, but uh for society as a whole, and even for those people in the long run, there's more to do, I think, not less.
00:08:33.440 --> 00:08:45.200
And I think in software engineering, we sort of thought, well, AI is really good at writing code, so you don't need as many software engineers, but you know, the hiring of software engineering, software engineers continues to go up.
00:08:45.440 --> 00:08:53.440
And then I think there's sort of not really false flags, but there's there's indicators in the press, like when someone does a layoff and they blame it on AI.
00:08:53.679 --> 00:08:55.679
I'm not sure it's actually always AI.
00:08:56.080 --> 00:09:00.799
Um-COVID, maybe the business isn't doing that well.
00:09:01.039 --> 00:09:08.240
Um, but going to, you know, if you're a public company and go to the street and say, hey, we're doing this layoff to increase our efficiency thanks to AI.
00:09:08.399 --> 00:09:15.360
So we're an even better company than you thought, sort of masking maybe we made mistakes in the past that we're correcting now.
00:09:15.679 --> 00:09:29.440
So I think as people start to see, I hope, especially in the software engineering world, start to see that no, this isn't an Armageddon for software engineers, that people will be, will have less of that existential dread.
00:09:29.519 --> 00:09:37.279
Um, they may still have the sort of the feeling of, I I don't want to do something that's gonna take away this art that I love doing.
00:09:37.440 --> 00:09:50.639
If you're a craftsman and you know, you build uh let's say you build beautiful cabinetry by hand, um, it might be hard to switch to working in more of a an automated CNC type environment, um, which I totally get.
00:09:50.799 --> 00:10:01.600
So I'm not quite sure how you do that, but maybe there's a new art around prompting and writing requirements and how you interact and so on that that could be just as much fun.
00:10:01.759 --> 00:10:03.440
But you have to sort of get over that hump.
00:10:03.840 --> 00:10:04.080
Yeah.
00:10:04.240 --> 00:10:05.519
I mean, yeah, absolutely.
00:10:05.600 --> 00:10:06.639
I I agree.
00:10:06.799 --> 00:10:13.120
I think um, you know, it's you're like at least where we're at now, you're always still gonna need that human in the loop.
00:10:13.360 --> 00:10:23.440
And it becomes critical for that human to know kind of what the AI is doing to where we talked about, make sure that like the taste is there, the aspect the AI is not able to do yet.
00:10:23.679 --> 00:10:23.840
Yeah.
00:10:24.320 --> 00:10:27.039
So you're just gonna you know, people are just gonna be way more productive.
00:10:27.200 --> 00:10:31.840
We're gonna be able to do a lot more things, but yeah, you're gonna always need that person.
00:10:32.879 --> 00:10:38.720
I mean, if you if you literally turn everything over to the AIs, it's it doesn't go that well in in terms of coding.
00:10:39.039 --> 00:10:52.799
I mean you can have them, you know, commit and push to to uh to GitHub and and approve its own PRs and all that, but you're just gonna end up with a mess and probably more work or or business problems that you didn't expect.
00:10:53.039 --> 00:10:57.360
So there really is a still a tight integration with the person to get it right.
00:10:57.600 --> 00:11:09.679
And and maybe the maybe the hump is to see that, okay, that's actually really fun getting to orchestrate that, and you're still exercising your code mind as you're looking at code and approving PRs and so on.
00:11:09.919 --> 00:11:12.480
Um but it takes some work to get to that.
00:11:12.960 --> 00:11:19.120
And we're just literally probably the last five months at the point where the tools are actually really good.
00:11:19.360 --> 00:11:22.320
Up until then, they were helpful, but had a lot of problems.
00:11:22.480 --> 00:11:30.720
But around Christmas time, 2025, I think what was it, Opus 4.5, and then we got 4.6, I think, in February, 4.7 now.
00:11:31.039 --> 00:11:32.000
Or did that happen in February?
00:11:32.240 --> 00:11:34.799
Anyway, I've lost track of all happening so quickly.
00:11:35.039 --> 00:11:40.960
And the uh GPT 5.5 and then Codex, their harness for coding, is actually quite good.
00:11:41.039 --> 00:11:42.559
Um that just changed the game.
00:11:42.720 --> 00:11:44.960
I mean, it actually is a really good tool now.
00:11:45.360 --> 00:11:45.679
Mm-hmm.
00:11:45.919 --> 00:11:46.159
Yeah.
00:11:46.480 --> 00:11:46.720
Yeah.
00:11:46.879 --> 00:11:48.960
They're they're wonderful now at this point.
00:11:49.279 --> 00:11:52.159
I um I was a bit of a skeptic at the beginning as well.
00:11:52.399 --> 00:11:53.679
Now I'm all in.
00:11:54.080 --> 00:11:54.480
You're not.
00:11:54.559 --> 00:11:54.879
Yeah.
00:11:55.200 --> 00:11:55.360
Yeah.
00:11:55.440 --> 00:12:04.320
I know you're working on things like setting up frameworks for building certain things so that the AIs go down the right path to start with and sort of don't get off track.
00:12:04.559 --> 00:12:09.519
Probably makes it a lot easier then to look at the code and know whether it's making sense to Yeah.
00:12:09.679 --> 00:12:09.919
Yeah.
00:12:10.320 --> 00:12:16.960
And it also kind of gives it like it doesn't have to worry so much about the entire picture, which is part of can be a drawback, right?
00:12:17.039 --> 00:12:23.039
When it's looking at a single spot and it doesn't know the whole thing's happening over here, right?
00:12:23.200 --> 00:12:24.559
That can get you into issues.
00:12:25.039 --> 00:12:42.000
We're um we're playing around with the balance there to because the the other thing is if it has the whole context, you know, unless you're not actually reviewing what it's doing, there's nothing stopping it from changing something, you know, way upstream when you're wanting it to fix something in this one spot.
00:12:42.480 --> 00:12:42.960
That's the worst.
00:12:43.120 --> 00:12:47.840
When you have it mostly working, you're fixing the one bug and it breaks other than that.
00:12:48.080 --> 00:12:49.360
But they're getting better at not doing that.
00:12:49.440 --> 00:12:49.519
Yeah.
00:12:50.240 --> 00:12:50.399
Yeah.
00:12:50.639 --> 00:12:53.360
They are every day, it feels like at this point, anyways.
00:12:53.759 --> 00:12:53.840
Yeah.
00:12:54.000 --> 00:12:55.360
Well, it's an interesting discussion.
00:12:55.519 --> 00:13:00.399
Uh I want to check back in say a month or two and see see how things have progressed.
00:13:00.720 --> 00:13:01.360
There's something else.
00:13:01.440 --> 00:13:03.759
Uh you set up an open claw last week.
00:13:04.000 --> 00:13:04.159
Yeah.
00:13:04.480 --> 00:13:07.279
Uh yeah, uh got your Mac mini and uh set that up.
00:13:07.360 --> 00:13:08.879
I've had one for a couple months.
00:13:09.039 --> 00:13:15.200
Uh we're not using them in business just because there's still a lot of concerns around uh security and and so on there.
00:13:15.440 --> 00:13:23.360
But that's gonna be a whole nother level of when you're working with uh an even more agentic harness, really, I think.
00:13:23.600 --> 00:13:29.039
Um I think we're a ways from that, although certainly people just turn stuff over to their open cloud to build things.
00:13:29.200 --> 00:13:29.360
Yeah.
00:13:29.519 --> 00:13:30.879
Um I've tried tried that.
00:13:31.039 --> 00:13:35.200
It it it can work, but you lose a little bit of the connection, at least for me.
00:13:35.360 --> 00:13:38.879
Anyway, that's something we should probably talk about in the future is you get some more experience with it.
00:13:39.120 --> 00:13:39.679
Yeah, yeah.
00:13:39.759 --> 00:13:41.759
I'm I've just barely scratched the surface.
00:13:41.840 --> 00:13:44.480
It's crazy how deep that that thing is.
00:13:44.720 --> 00:13:47.200
There's so many options, so many things you can do with it.
00:13:47.360 --> 00:13:47.600
Yeah.
00:13:47.759 --> 00:13:50.480
So it's um, you know, I need to figure out where to start at this point.
00:13:50.960 --> 00:13:54.320
All right, we'll come back to that in a future episode, then when you've had a little time.
00:13:54.559 --> 00:13:55.200
Yeah, perfect.
00:13:55.360 --> 00:13:56.240
All right, good to chat.
00:13:56.399 --> 00:13:56.960
Talk to you soon.
00:13:57.120 --> 00:13:57.759
Yeah, thank you.
00:13:57.919 --> 00:13:58.080
Okay.
00:00:06.240 --> 00:00:08.320
Welcome to the Dashboard Effect Podcast.
00:00:08.400 --> 00:00:09.279
I'm Brick Thompson.
00:00:09.519 --> 00:00:10.480
And I'm Landon Oaks.
00:00:10.880 --> 00:00:18.000
Landon, last time we met, we talked about using AI tools in just in business but in programming.
00:00:18.160 --> 00:00:28.399
And one of the things that we touched on was that there can be some reticence for software engineers to want to use these tools for a number of reasons.
00:00:28.559 --> 00:00:31.920
One, maybe they tried them earlier and they didn't work very well.
00:00:32.159 --> 00:00:37.920
Or I don't know, they they they had an experience where it seemed like it took more time than it actually saved.
00:00:38.159 --> 00:00:44.240
But one of the things I've also hear is that there's some people that just love the art of programming.
00:00:44.640 --> 00:00:51.759
I think of myself as a programmer, I've always thought of it or have a relationship to it that's more like it's a tool.
00:00:51.920 --> 00:01:05.680
Like the gratification I get is out of building something that works and solves a problem, not necessarily figuring out the beauty of the code, but there are other people that really enjoy that figuring out the beauty of the code.
00:01:05.920 --> 00:01:11.280
And so I'm sure uh even on your team, and maybe you've had that experience yourself, those types of feelings.
00:01:11.359 --> 00:01:11.680
I don't know.
00:01:11.760 --> 00:01:14.000
I I'd just like to kind of talk through that a little bit.
00:01:14.079 --> 00:01:23.359
And how do you how do you support someone who's got that, but also you want them to take advantage of some of the productivity gains that you can get through some of these tools?
00:01:23.680 --> 00:01:23.920
Yeah.
00:01:24.159 --> 00:01:25.519
Yeah, it's a great question.
00:01:25.680 --> 00:01:30.239
I um I'm actually I lean a little bit more towards your camp personally, right?
00:01:30.640 --> 00:01:37.840
Um, where I I the fun for me is what happens at the end, not so much building it in the first place.
00:01:38.079 --> 00:01:38.480
Yeah.
00:01:38.719 --> 00:01:47.599
Um but I do have several, several guys on my team that are more in that camp of the art of the code, right?
00:01:48.079 --> 00:01:57.120
Um and so I think what you hit on early was was the first roadblock that I saw was okay, they tried it back in August or October, right?
00:01:57.200 --> 00:01:58.560
Um, of 2025.
00:01:58.719 --> 00:01:58.879
Yeah.
00:01:59.120 --> 00:02:00.959
It uh wasn't very good.
00:02:01.200 --> 00:02:04.719
Uh and so a lot of them just kind of stopped and never returned to it.
00:02:05.200 --> 00:02:09.439
I mean, they'll use it for like questions and help to do some research, et cetera.
00:02:09.599 --> 00:02:11.759
But they're just like, oh no, it's it's not good at code.
00:02:12.159 --> 00:02:15.280
Well, and they probably tried it prior to that, too, several times.
00:02:15.520 --> 00:02:17.520
So we sort of had people testing it.
00:02:18.240 --> 00:02:20.719
But what so what haven't after October then?
00:02:21.120 --> 00:02:22.719
Yeah, obviously it got way better.
00:02:23.280 --> 00:02:26.800
Um so now it now it's really impressive what it can do.
00:02:26.960 --> 00:02:27.360
Yeah.
00:02:27.680 --> 00:02:41.039
Um and so I've been, you know, we're trying just kind of trying to get those guys to get back in there and start using it so at least see that, hey, this is not the same model, you know, this is it works much better now.
00:02:41.280 --> 00:02:57.280
Um, which I, you know, I've I've shown them a couple of things that I've done with it, which have been pretty cool, taking like, you know, a couple hours when it would have taken a full day of work um to kind of help them get back into that mindset of, okay, this is this is a different game now.
00:02:57.759 --> 00:03:07.919
It is different because certainly with Opus 4.7, a GBT 5.5, Codex and Claude Code, you can write code much faster.
00:03:08.639 --> 00:03:10.960
There's still issues, though, that crop up.
00:03:11.120 --> 00:03:14.400
Like it it can write code that's not as efficient as it should be.
00:03:14.560 --> 00:03:27.919
You know, it's it's repeating code that um doesn't need to, or I guess it could you know introduce bugs and so on that you don't know about, although frankly you can use those tools for really thorough testing as well.
00:03:28.159 --> 00:03:32.960
Um so there's sort of a different mindset that has to come into play, I think.
00:03:33.199 --> 00:03:33.599
Yeah.
00:03:33.840 --> 00:03:40.879
You know, you're using it to to crank out sort of the rote code that you would spend a few hours writing it can just do quickly.
00:03:41.039 --> 00:03:50.719
But then you have to be really smart about how you prompt it, what your design is, um, maybe you know how you give the requirements to the tool.
00:03:50.800 --> 00:03:53.759
And then how are you verifying that it's really good?
00:03:54.560 --> 00:03:59.199
And I think maybe the art comes into comes back into it in terms of taste.
00:03:59.360 --> 00:04:04.240
Like there's a certain there's a certain thing about taste in software development.
00:04:04.639 --> 00:04:16.000
You can see code that works, but it's not nice to use, not nice to look at, um, just feels clunky, and then code that's very similar but is really smooth and easy to use, something like that.
00:04:16.160 --> 00:04:18.560
So maybe the art comes back in with that.
00:04:18.720 --> 00:04:19.279
I don't know.
00:04:19.600 --> 00:04:21.360
Yeah, yeah, no, I completely agree.
00:04:21.439 --> 00:04:26.240
That's that's where the background of coding is so huge with AI.
00:04:26.560 --> 00:04:34.160
Um, you can actually go look at it now and then tell it to make these changes, do something different, etc.
00:04:34.879 --> 00:04:44.480
Um, which that's actually how one of our one of like my better engineers on my team, well probably the best, um, he uh he kind of got over his.
00:04:44.560 --> 00:04:46.480
He was much more much more in the art camp, right?
00:04:46.720 --> 00:04:54.160
You know, he likes to do the code, he likes to make sure it's reusable, it's modular, it's clean, big proponent of that, and he's really good at it.
00:04:54.480 --> 00:05:00.079
Um how he kind of got over that hump is he uses it in Visual Studio Code now with the extensions.
00:05:00.480 --> 00:05:00.639
Okay.
00:05:00.959 --> 00:05:08.720
And he has it like basically when you when you use AI to do something, you can do yes, always do this, or I'm gonna improve every single thing you do.
00:05:08.879 --> 00:05:09.120
Yes.
00:05:09.439 --> 00:05:15.199
He is in the I'm gonna improve every single thing you do, because he actually reads it, looks at it.
00:05:15.439 --> 00:05:19.040
And if he doesn't like it, he'll tell it, no, you need to change it to do this.
00:05:19.199 --> 00:05:22.879
Or you know, this looks bad, like make this cleaner.
00:05:23.120 --> 00:05:30.079
Um and he still gets huge gains out of it, but he he still feels like he's kind of in control, you know, and really the orchestrator.
00:05:30.480 --> 00:05:32.319
Do you think he enjoys that mode of working?
00:05:32.639 --> 00:05:33.040
He does.
00:05:33.120 --> 00:05:33.279
Yeah.
00:05:33.680 --> 00:05:34.240
It's interesting.
00:05:34.399 --> 00:05:36.240
I I'm more in the YOLO camp.
00:05:36.399 --> 00:05:36.639
Yeah.
00:05:36.800 --> 00:05:39.040
Let it roll and then let's uh look at it afterwards.
00:05:39.360 --> 00:05:39.839
Yeah, yeah.
00:05:40.160 --> 00:05:41.519
I think for me, it it depends.
00:05:41.680 --> 00:05:49.360
If it's something that I know really well, I I like the way he does it, because I, you know, I can really make sure that it's gonna be really useful.
00:05:49.680 --> 00:05:52.160
If it's something that I don't work in, I'm like, just go for it.
00:05:52.240 --> 00:05:54.399
I'm not I'm not gonna give you any good feedback anyways.
00:05:54.560 --> 00:05:56.079
Like, it's interesting too, though.
00:05:56.160 --> 00:06:03.759
If I'm working on production code that, you know, is is critical to our business or our client's business, then I'm not YOLO at all.
00:06:03.920 --> 00:06:11.759
But if I'm building if I'm building some kind of an application, maybe it's to help me with some kind of business utility, or maybe it's a personal thing, yeah.
00:06:12.000 --> 00:06:15.680
I'm more apt to let it go and see how it does and then work on it from there.
00:06:15.839 --> 00:06:23.360
Um and and of course, if you're doing something in production, you've got to review all the code and and make sure it's actually safe and not doing something weird.
00:06:23.759 --> 00:06:24.000
For sure.
00:06:24.399 --> 00:06:29.600
So what do you think is the best way to help someone get over that hump?
00:06:30.160 --> 00:06:40.079
Not that not that they have to, but I do worry that people that don't start utilizing these tools that it could it could hurt them just in their professional career.
00:06:40.319 --> 00:06:40.720
Yeah.
00:06:40.959 --> 00:06:44.079
Um how do you how do you think about that?
00:06:44.160 --> 00:06:46.720
Aaron Powell Yeah, it's a it's a really good question.
00:06:46.879 --> 00:06:53.680
Um and we're we're still kind of trying to figure it out for so for instance on my team, I you know, we we use AI a lot.
00:06:53.759 --> 00:06:56.240
I think we could use it more and better.
00:06:56.720 --> 00:06:59.279
And we're still trying to foster that kind of culture.
00:06:59.439 --> 00:07:09.279
Um one of the things we're working on now is for lack of a better name, we're calling it AI enablement, um, where we have certain people in the company who have already fully adopted AI.
00:07:09.360 --> 00:07:13.920
They're they're using it in awesome ways, they're getting incredible gains out of it.
00:07:14.240 --> 00:07:25.279
Um so we're gonna have those people kind of lead a little bit of a group to uh show what they're doing, show others what they're doing, encourage other people who aren't in that group to also show what they're doing.
00:07:25.519 --> 00:07:28.959
So the idea is like, let's just show them the art of the possible.
00:07:29.120 --> 00:07:29.279
Yeah.
00:07:29.519 --> 00:07:36.879
Hopefully get the the, you know, get the machine turning upstairs to where they're like, oh, you know, I could use this on that problem I had yesterday.
00:07:37.040 --> 00:07:37.120
Yeah.
00:07:37.360 --> 00:07:38.399
I hated that problem.
00:07:38.560 --> 00:07:40.079
It'd be awesome.
00:07:40.639 --> 00:07:41.199
You know.
00:07:41.759 --> 00:07:42.160
Yeah.
00:07:42.319 --> 00:07:43.600
Yeah, it's interesting.
00:07:43.839 --> 00:07:47.279
I think uh I think it some of it is that.
00:07:47.360 --> 00:07:50.560
Like, let me just show you how fun this can be to be this productive.
00:07:51.360 --> 00:07:54.160
Some of it I think is uh mindset too.
00:07:54.399 --> 00:08:02.399
Um, you know, I I definitely talked to people that have, and I've had this, still have it from time to time, sort of the existential worry about AI.
00:08:02.560 --> 00:08:04.399
What does it mean in the longer term?
00:08:04.639 --> 00:08:11.120
Um, I think the last few months, people have been very focused on uh the possibility that AI is going to wipe out a lot of jobs.
00:08:11.759 --> 00:08:19.680
I think more recently, everything that I'm reading and talking to people, it looks like it's actually going to net create a lot of jobs.
00:08:19.920 --> 00:08:33.279
And um there will definitely be disruption, which is not great for the people that get disrupted, uh, but uh for society as a whole, and even for those people in the long run, there's more to do, I think, not less.
00:08:33.440 --> 00:08:45.200
And I think in software engineering, we sort of thought, well, AI is really good at writing code, so you don't need as many software engineers, but you know, the hiring of software engineering, software engineers continues to go up.
00:08:45.440 --> 00:08:53.440
And then I think there's sort of not really false flags, but there's there's indicators in the press, like when someone does a layoff and they blame it on AI.
00:08:53.679 --> 00:08:55.679
I'm not sure it's actually always AI.
00:08:56.080 --> 00:09:00.799
Um-COVID, maybe the business isn't doing that well.
00:09:01.039 --> 00:09:08.240
Um, but going to, you know, if you're a public company and go to the street and say, hey, we're doing this layoff to increase our efficiency thanks to AI.
00:09:08.399 --> 00:09:15.360
So we're an even better company than you thought, sort of masking maybe we made mistakes in the past that we're correcting now.
00:09:15.679 --> 00:09:29.440
So I think as people start to see, I hope, especially in the software engineering world, start to see that no, this isn't an Armageddon for software engineers, that people will be, will have less of that existential dread.
00:09:29.519 --> 00:09:37.279
Um, they may still have the sort of the feeling of, I I don't want to do something that's gonna take away this art that I love doing.
00:09:37.440 --> 00:09:50.639
If you're a craftsman and you know, you build uh let's say you build beautiful cabinetry by hand, um, it might be hard to switch to working in more of a an automated CNC type environment, um, which I totally get.
00:09:50.799 --> 00:10:01.600
So I'm not quite sure how you do that, but maybe there's a new art around prompting and writing requirements and how you interact and so on that that could be just as much fun.
00:10:01.759 --> 00:10:03.440
But you have to sort of get over that hump.
00:10:03.840 --> 00:10:04.080
Yeah.
00:10:04.240 --> 00:10:05.519
I mean, yeah, absolutely.
00:10:05.600 --> 00:10:06.639
I I agree.
00:10:06.799 --> 00:10:13.120
I think um, you know, it's you're like at least where we're at now, you're always still gonna need that human in the loop.
00:10:13.360 --> 00:10:23.440
And it becomes critical for that human to know kind of what the AI is doing to where we talked about, make sure that like the taste is there, the aspect the AI is not able to do yet.
00:10:23.679 --> 00:10:23.840
Yeah.
00:10:24.320 --> 00:10:27.039
So you're just gonna you know, people are just gonna be way more productive.
00:10:27.200 --> 00:10:31.840
We're gonna be able to do a lot more things, but yeah, you're gonna always need that person.
00:10:32.879 --> 00:10:38.720
I mean, if you if you literally turn everything over to the AIs, it's it doesn't go that well in in terms of coding.
00:10:39.039 --> 00:10:52.799
I mean you can have them, you know, commit and push to to uh to GitHub and and approve its own PRs and all that, but you're just gonna end up with a mess and probably more work or or business problems that you didn't expect.
00:10:53.039 --> 00:10:57.360
So there really is a still a tight integration with the person to get it right.
00:10:57.600 --> 00:11:09.679
And and maybe the maybe the hump is to see that, okay, that's actually really fun getting to orchestrate that, and you're still exercising your code mind as you're looking at code and approving PRs and so on.
00:11:09.919 --> 00:11:12.480
Um but it takes some work to get to that.
00:11:12.960 --> 00:11:19.120
And we're just literally probably the last five months at the point where the tools are actually really good.
00:11:19.360 --> 00:11:22.320
Up until then, they were helpful, but had a lot of problems.
00:11:22.480 --> 00:11:30.720
But around Christmas time, 2025, I think what was it, Opus 4.5, and then we got 4.6, I think, in February, 4.7 now.
00:11:31.039 --> 00:11:32.000
Or did that happen in February?
00:11:32.240 --> 00:11:34.799
Anyway, I've lost track of all happening so quickly.
00:11:35.039 --> 00:11:40.960
And the uh GPT 5.5 and then Codex, their harness for coding, is actually quite good.
00:11:41.039 --> 00:11:42.559
Um that just changed the game.
00:11:42.720 --> 00:11:44.960
I mean, it actually is a really good tool now.
00:11:45.360 --> 00:11:45.679
Mm-hmm.
00:11:45.919 --> 00:11:46.159
Yeah.
00:11:46.480 --> 00:11:46.720
Yeah.
00:11:46.879 --> 00:11:48.960
They're they're wonderful now at this point.
00:11:49.279 --> 00:11:52.159
I um I was a bit of a skeptic at the beginning as well.
00:11:52.399 --> 00:11:53.679
Now I'm all in.
00:11:54.080 --> 00:11:54.480
You're not.
00:11:54.559 --> 00:11:54.879
Yeah.
00:11:55.200 --> 00:11:55.360
Yeah.
00:11:55.440 --> 00:12:04.320
I know you're working on things like setting up frameworks for building certain things so that the AIs go down the right path to start with and sort of don't get off track.
00:12:04.559 --> 00:12:09.519
Probably makes it a lot easier then to look at the code and know whether it's making sense to Yeah.
00:12:09.679 --> 00:12:09.919
Yeah.
00:12:10.320 --> 00:12:16.960
And it also kind of gives it like it doesn't have to worry so much about the entire picture, which is part of can be a drawback, right?
00:12:17.039 --> 00:12:23.039
When it's looking at a single spot and it doesn't know the whole thing's happening over here, right?
00:12:23.200 --> 00:12:24.559
That can get you into issues.
00:12:25.039 --> 00:12:42.000
We're um we're playing around with the balance there to because the the other thing is if it has the whole context, you know, unless you're not actually reviewing what it's doing, there's nothing stopping it from changing something, you know, way upstream when you're wanting it to fix something in this one spot.
00:12:42.480 --> 00:12:42.960
That's the worst.
00:12:43.120 --> 00:12:47.840
When you have it mostly working, you're fixing the one bug and it breaks other than that.
00:12:48.080 --> 00:12:49.360
But they're getting better at not doing that.
00:12:49.440 --> 00:12:49.519
Yeah.
00:12:50.240 --> 00:12:50.399
Yeah.
00:12:50.639 --> 00:12:53.360
They are every day, it feels like at this point, anyways.
00:12:53.759 --> 00:12:53.840
Yeah.
00:12:54.000 --> 00:12:55.360
Well, it's an interesting discussion.
00:12:55.519 --> 00:13:00.399
Uh I want to check back in say a month or two and see see how things have progressed.
00:13:00.720 --> 00:13:01.360
There's something else.
00:13:01.440 --> 00:13:03.759
Uh you set up an open claw last week.
00:13:04.000 --> 00:13:04.159
Yeah.
00:13:04.480 --> 00:13:07.279
Uh yeah, uh got your Mac mini and uh set that up.
00:13:07.360 --> 00:13:08.879
I've had one for a couple months.
00:13:09.039 --> 00:13:15.200
Uh we're not using them in business just because there's still a lot of concerns around uh security and and so on there.
00:13:15.440 --> 00:13:23.360
But that's gonna be a whole nother level of when you're working with uh an even more agentic harness, really, I think.
00:13:23.600 --> 00:13:29.039
Um I think we're a ways from that, although certainly people just turn stuff over to their open cloud to build things.
00:13:29.200 --> 00:13:29.360
Yeah.
00:13:29.519 --> 00:13:30.879
Um I've tried tried that.
00:13:31.039 --> 00:13:35.200
It it it can work, but you lose a little bit of the connection, at least for me.
00:13:35.360 --> 00:13:38.879
Anyway, that's something we should probably talk about in the future is you get some more experience with it.
00:13:39.120 --> 00:13:39.679
Yeah, yeah.
00:13:39.759 --> 00:13:41.759
I'm I've just barely scratched the surface.
00:13:41.840 --> 00:13:44.480
It's crazy how deep that that thing is.
00:13:44.720 --> 00:13:47.200
There's so many options, so many things you can do with it.
00:13:47.360 --> 00:13:47.600
Yeah.
00:13:47.759 --> 00:13:50.480
So it's um, you know, I need to figure out where to start at this point.
00:13:50.960 --> 00:13:54.320
All right, we'll come back to that in a future episode, then when you've had a little time.
00:13:54.559 --> 00:13:55.200
Yeah, perfect.
00:13:55.360 --> 00:13:56.240
All right, good to chat.
00:13:56.399 --> 00:13:56.960
Talk to you soon.
00:13:57.120 --> 00:13:57.759
Yeah, thank you.
00:13:57.919 --> 00:13:58.080
Okay.
