Home
Sign Up
Allison McHenry from Daily Kos
Sep 14, 2026
Allison McHenry from Daily Kos
15
30
00:00
27:43
SUBSCRIBE FOR UPDATES
Subscribe
Listen on
Apple Podcasts
Apple Podcasts
Spotify
Spotify
YouTube
YouTube
Transcript
0:00
It would've been a two-year project, uh, in a, in a different world. From discovery to launch, uh, our migration was about nine months. We executed this with about three engineers on the XWP side- Yeah...
0:13
one engineer on the Daily Kos side. That would've been a team of 25 people- Yeah... uh, two or three years ago.
0:18
AI has kind of taken over the part of a software engineer's job that they really enjoy doing most, which is writing code. Identify with, yeah.
0:24
[laughs] And then given them back the part of the software development life cycle that they hate, which is reviewing code.
0:30
I mean, we built a feature to allow users to save articles on Daily Kos in about two and a half weeks. Yeah. That would've been a six-month project- Yeah... you know, before.
0:38
This is Not Just Faster Horses, conversations about media companies deploying AI inside product and engineering and people. My name's Ben May. I'm founder and MD at The Code Company.
0:49
Today I'm speaking with Alison McHenry, CTO at Daily Kos, a progressive media site with 20 years of content and community. Uh, uniquely, it's a purely, um, community-funded media organization as well.
1:05
Alison, to set the scene today, what does the inside of Daily Kos look like when it comes to product engineering and editorial, and how the sorta functions of the business operate today? Hi, Ben. Uh, nice to be here.
1:16
So currently Daily Kos is a small company. Um, we've unfortunately reduced in size. Uh, we were over 100 people when I started at Daily Kos in 2024. We're currently only 26.
1:28
So there's been, um, a big reorientation, uh, of product and engineering to the needs of the business in this contracting media market.
1:37
Um, so currently we have one in-house engineer, myself, um, I'm essentially functioning as the product lead.
1:45
Um, we also are supported by, uh, XWP, which is a agency that does WordPress, um, big bu- builds for different, different media companies, and we, uh, additionally contract, um, with some editorial freelance, uh, people, and then we have an in-house editorial staff of about 10.
2:03
Um, so we're looking to think about, um, as we're thinking about staffing and we're thinking about, you know, the needs of the business, we're trying to just get more done with less, which is where AI comes in. Yeah.
2:14
How have you... How do you think about on a day-to-day basis now that you're sort of in the new world, um, of ideation of products,
2:21
um, what gets, what gets pushed through to sort of the backlog to get worked on, uh, who says no, um, and how's AI sort of making that process more efficient as you sort of, yeah, think about how to be, like a lot of media companies, do more with less?
2:33
Yeah. Well, we have just a few high-level initiatives that we know we want to complete in 2026. Um, the WordPress migration, uh, as you alluded to, was the first major initiative.
2:43
Um, we completed that in April, and from April to, you know, the end of Q2, um, we've been basically reorienting to making sure the bugs are fixed, making sure that we have a good user experience on the site.
2:56
Um, driving to the end of the year, we want to build out a native app, and we'll be using Pug Pig as our partner for that, and we will be, um, working on a data initiative, um, to, uh, unify our data pipelines to make sure that we have a 360-degree view of our users when they come to the site.
3:14
Um,
3:15
and then we are also re- uh, rebuilding our homepage and streamlining it to make it more apparent to people who are new to Daily Kos what Daily Kos is all about, because we really are a unique property in the media world, uh, because we do have this community focus.
3:30
So those are the, the top three things.
3:33
Um, when we're building out products, we usually, you know, follow a pretty normal, uh, prioritization matrix of how difficult is this to do and what is the potential business revenue impact of the work.
3:46
So something that's a large initiative, um, but also has a large impact, like building out a native app, um, we view that as an investment.
3:53
So this is something that's we hope will pay dividends, uh, over time, versus something like the homepage redesign, which is, you know, less of an effort, um, but also, you know, a little narrower.
4:03
The executive team works as a unit to say no, so if we're looking at the cost of, you know, building something out, there just is a point where we have to draw a line and say, "You know what? This has to be next year."
4:13
Yeah. You know, we have to, you know, work out some revenue stuff first. Um...
4:16
How do you think about the split between sort of what, what you build internally and sort of keep for yourselves to almost hold any partner accountable for the work that they're doing? Where do you...
4:25
Yeah, when you, when you're building out these AI, AI initiatives internally. Yeah.
4:31
Well, that's always the tricky question, right, because, um, the pros of using an outsourced partner are, y- you know, usually lower cost, um, sometimes faster. You can scale up and down as needed.
4:41
Um, the cons are, you know, you do have this team that can roll off at any moment, and they take that institutional knowledge with them. Um, so with our single engineer and me, um, as the, uh, the tech team,
4:55
we have to make sure that, uh, you know, as the, um, agency partners are supporting us, that we understand what they're doing, why they're doing it, and we, you know, have transitioned all of our code to our own GitHub repos, that we own it.
5:06
Yeah. Um, and we, you know, obviously can use AI to tell us what it does if we don't already know. Um- Yeah... that's one of the, the, the great things about AI.
5:14
Um, as we're building out the initiatives, you know, thinking about documentation, thinking about knowledge transfer, you know, we've definitely had several meetings with the agency to just talk about, you know, how they think about building products to make sure that, that we are, um, doing engineering the same way that they're doing it.
5:30
So we're not trying to, uh, have our own code review process that's different from theirs or whatever, um, to make sure that we are actually working as a unified team. Being as efficient as possible. Exactly, yeah. Yeah.
5:41
Um, when you think about sort of the whole, uh, l- the life cycle of approving work, where are you in that sort of journey, uh- Yeah... to, to label the, the term, but yeah.
5:51
Well, AI is great at some things and less great at other things. So one thing that we really, uh, have enjoyed using it for is sharpening our product requirements process.
5:59
So at the beginning of a project, um, AI is really useful for a dialogue where, you know, what have we not thought of? What are edge cases that maybe we've missed? Um, how should we be thinking about this?
6:08
You know, is this going to actually move the metrics that we're trying to move? Um, in the build process, you know, AI is definitely supportive, um-
6:16
It's difficult because AI has kind of taken over the part of a software engineer's job that they really enjoy doing most, which is writing code. Identify with, yeah.
6:24
[laughs] And then giving them back the part of the software development life cycle that they hate, which is reviewing code. Um, so, you know, we were trying to make sure that the pull requests stay a manageable size.
6:33
Um, we are using AI for, uh, building tests. They do the first pass at PR review.
6:38
Um, we have a tool called CodeRabbit.ai, which does the, um, the initial, uh, code review pass, making sure that we are following our own linting rules and also looking for, you know, potential security holes or edge cases that maybe a, a human reviewer wouldn't im- immediately observe.
6:53
Um, and then of course everything that goes to production is approved by a person.
6:57
Um, so we make sure that every engineer understands that they have ultimate ownership over the code that goes to production, that they are responsible for.
7:05
Um, keeping pull requests to a manageable size is really important. Making sure that, you know, we understand what the AI is doing is really important.
7:12
[laughs] You know, waking up to 50 pull requests in the morning- Yeah, exactly. [laughs]... and figuring out how, how is this all gonna... 'Cause that's, we've seen that as well someplace- Yeah...
7:18
where all of a sudden the bottleneck's just moved to... And a bottle- Yeah. There- So many pull requests and then the human doing that is just sort of skimming, yeah, approved. Clicking approve. Yeah. Looks good to me.
7:27
Like, what's the value of that pull request if you're just sort of rubber stamping it at that point? Yeah. Yeah. No, so we have to, you know, making sure that, that everything is, is manageable and done by people.
7:35
But this is where the seniority of the engineers really is, is important as well because, um, we need to make sure that, you know, the engineer looking at it is thinking, you know, "Oh, does this make sense?
7:47
Does this, you know, is this how I would have approached the problem? You know, do I understand what the AI is doing?" All of those, you know, uh, bringing their experience to bear on the problem, um- Yeah...
7:55
is really important. How, uh, just digging in on the product requirement stuff, is AI allowing you to do that more thoroughly, or... And is that through like a, a skill or a little discipline or? Yeah.
8:07
No, I mean, the way that we write product briefs in a, using a template, um, so they're always the same and we always know, you know, what it is, um, that we're supposed to be doing.
8:15
And, you know, you typically, like, if I'm doing a product brief, I will write it myself. I'll give it to Claude and I'll say, you know, "What are the risks that I've missed?
8:25
You know, is this going to move the metrics that I think? You know, is there any side effects that I've, you know, uh, not observed or whatever?"
8:31
Um, it's really a partnership between the person and the AI, um, to try to make sure that like, you know, we're bringing our human intelligence and our understanding and empathy for our community. Yeah.
8:40
But also, you know, the comprehensive knowledge that AI has, um, to make sure that we are making... There's nothing that we're missing. Yeah. There's nothing that we've left out. Yeah.
8:50
When you were, uh, doing the migration, you sort of used AI as well to sort of analyze the old code base. What was unique about that process?
8:58
'Cause, you know, your, your experience, you've been in this sort of world a long time with media and engineering. Yeah.
9:02
Using AI in the engineering process for a migration like this, what were the sort of things that stood out to you, um, that were, yeah, transformative? Transformative.
9:10
I mean, it would've been a two-year project, uh, in a, in a different world. You know, it... To- from discovery to launch, uh, our migration was about nine months. Mm.
9:21
And, I, I mean, that just would not have been possible before AI. Um- So legacy code base. Yeah. Custom build. Everything. Custom data models. Yeah.
9:29
And some of what happened was, you know, the Perl code base in particular was not instrumented. There are no tests.
9:34
Um, there are no analytics, so we didn't even know how users were using the code, um, the, the pages that were live on the site. In some cases there were duplicated pages between Ruby on Rails and Perl.
9:44
So, you know, you might have a, a user profile, say, in Ruby on Rails that had certain, you know, data that was featured, and slightly different data featured in the Perl code base.
9:52
Um, the Perl search engine just ha- didn't work half the time. Like it would, just would throw out an error to the user, um, and just not even return results.
10:00
So w- we were in a position where there w- we really felt like this code base was so brittle that we needed AI to look at it and say, "This is what it's doing and this is how it's doing it." Mm.
10:10
Um, just to even understand what the requirements were for the migration and what we had to bring over and what we could just say, you know, "You know what?
10:16
That's not important," or, "We can build this later," or, you know, version three. Yeah. Um, and then in doing the, uh, the data an- analysis, you know, we needed to look at, um,
10:29
you know, groups and like which groups were used and not used, and which groups were written, uh, created by staff members. We had 67 user roles at the time of migration.
10:37
We needed to consolidate those into five, um, which we did, but like the AI helped with that.
10:43
Um, we had to do a deep dive into, you know, how are the different parts of the editorial interface being used and who's using them, and which permissions are actually, you know, a differentiator versus not.
10:54
Um, all of our images, we had, um, different crops saved as individual images in our- Set of files. Yeah... you know, S3 buckets.
11:02
Um, you know, could we unify these and then let WordPress do the, um, the different, you know, s- uh, size display to the user?
11:10
So just the complexity of the project, um, we executed this with about three engineers on the XWP side. Yeah. One engineer on the Daily Kos side. Um, that would've been a team of 25 people- Yeah...
11:23
uh, two or three years ago. In terms of WordPress itself, uh, and like with the editorial team and content team, are you using AI in that system yet?
11:29
Like obviously WordPress as of, you know, this year it's, it's quite AI native and you can connect in with MCP and all these other connectors and stuff.
11:37
WordPress does have some native AI features such as, um, summaries that, you know, just saves time for users, um, for our editorial team to n- not have to write those themselves.
11:46
But, um, that is something that I think we're gonna do more of a deep dive into now that we've launched. Yeah.
11:51
In terms of, um, the, the launch and, and going live, was there anything that sort of, uh, that, that AI kind of caught anyone off guard, that, uh, looked like was the right answer in the build or, um, you know, and was a, was a, a blind spot?
12:05
Well, one thing that we did with the migration was that we did a two-week beta for our users. Um, and that was a fully standalone...
12:12
We s- we set up a completely separate site with- 500 or so users who built, uh, unique user log-ins just for the beta, and they put content on the site just for the beta, and they wrote comments that were going to be torn down after the beta.
12:27
So in terms of user acceptance testing, um, that was a critical piece. And during that process, we definitely found, um, like we left, uh, our WP admin, unfortunately.
12:39
Uh, if you h- if you knew the URL, you could just get into it. Right. Um, you know, that was a, a blind spot that we missed. Um, there was some stuff with the commenting system where, um,
12:49
we thought things would work a certain way, and they didn't, and we had to, you know, kind of retrieve that and go back to the commenting system provider.
12:55
Some people had duplicate, uh, email addresses in the system, or they were pulled over twice. Um, we, we didn't really catch that before the beta.
13:03
Um, so there was a bunch of stuff that happened in the beta that was, that was really helpful in terms of catching that. I mean, that's a great way to test that sort of stuff- Yeah...
13:09
and limit the blast radius of a, of a launch, um- Exactly.
13:12
And then right around launch, you know, um, I was writing a blog post every day on the site, and the users were responding in the comments telling me about bugs or things that they saw that didn't work or things that, you know, I thought weren't a big deal, but they were a really big deal to the community.
13:26
One example of this was comment threading. Um- Mm...
13:29
we chose Viafore as our commenting system provider, but for most media sites, um, the comments, you know, it was fine to just have them be a single flow, a, a one column- Yep... um, commenting system, uh, output.
13:43
But for our users, they like to have conversations with one another, so that just wasn't gonna work. Like, you needed to have threaded comments where you could see who you were responding to individually.
13:51
Um, and that was something that I was like, "Oh, it's not that big of a deal," and then, like, you know- Yeah... it just turned into chaos, uh- Yeah...
13:56
on the site live, and I was just like, "Oh my gosh, we have to fix this." Yeah. So XWP actually put a gloss, uh, CSS gloss over the Viafore output to create threads for users.
14:05
Oh, [laughs] that's a good way to do it, yeah. Because we couldn't wait for Viafore to prioritize- To put threading in... uh, threading. Yeah. Yeah. Yeah.
14:12
So I wanna jump to, uh, and, and you were here in Phoenix this week talking about your re-platforming to WordPress. Yeah.
14:17
Uh, that you were on a 20-year-old, um, I think Pearl, um, uh, CMS that had been probably built over many years. Um,
14:26
what was the sort of decision-making to jump to both, I guess, like an open source platform like WordPress, um, again, when you have a lot of people now talking about that we could've vibe coded a new CMS, uh, in a week?
14:37
[laughs] Um, so yeah, j- I'm curious about that as a starting point. Um, and then you have... Yeah. Sure.
14:43
Well, the total cost of ownership question is something that, you know, we have to keep in the top of our minds always.
14:49
And so our former CMS was partly Ruby on Rails, partly Pearl, um, and the total cost of ownership, like, thinking was not only, um, there's this very elaborate system, you know, nobody has a mental model of how the whole thing works.
15:03
Uh, we are... It takes forever to make any product changes.
15:07
Um, we can't do, uh, keeping the lights on work such as upgrading our Node.js dependencies or our Ruby versions because the product work takes so long, and- Yeah...
15:16
you know, it's, it just is a drag on the whole organization, and it's expensive. You know, our AWS bill was the, the highest bill at the company at the time that we did the migration.
15:25
So overall, um, you know, everything about the system was expensive and slow, and we had gotten to the point in the beginning of 2025 where, um, you know, our run rate was just outpacing our ability to raise money.
15:39
Um, and it was very clear that something had to change, uh, unfortunately, so that made the decision for us really.
15:46
Um, the fact that we have, you know, this AI integration made it so that we could do more with a smaller team, which was really important and effective.
15:54
Um, but keeping it going on the old system just wasn't really gonna be possible. Yeah. So you've worked at a lot of different media companies like New York Times, Vox, like a lot of different ones over the years.
16:05
What do you think is unique, I guess, in media companies when it comes to, like, thinking about product, um, product and engineering cycles, how stuff gets out?
16:12
Um, I guess you're at a smaller publisher doing, you know, trying to do a lot more with a lot less than The New York Times with thousands of people in their- Yeah... you know, engineering team and things like that.
16:20
But, um, yeah, as you, like, look o- over different ones and I guess how that relates to media companies as they're sort of re- reorging, um, or rethinking about what their teams look like, uh, for what's sort of, you know, next to come, where one or two engineers can do what four or five could've done before AI tooling and stuff like that.
16:37
Well, I think every media company is aware of that, you know, and I think that if we're looking across the industry at the layoff patterns, um, you know, that efficiency gain is showing up in human, uh, loss, right?
16:51
Um, you just don't need as many people to do- Right... the same work as you needed, you know, a few years ago.
16:55
Um, one thing that does change between a big and a small media company is the build versus buy calculus, where if you... You know, it's, it's always nice to be able to build your own thing, right?
17:06
Because then you can build it exactly the way you want it for your users. You fully understand how it works. You can customize everything about it. So an example of this is The New York Times paywall. Mm-hmm.
17:16
Um, it, it is a fantastic piece of engineering. Yeah.
17:18
And they built it custom in-house, and they are able to tweak it, you know, based on their, you know, user population, and it's very, very fine-tuned, and they are constantly working on it to make it better.
17:29
If Daily Kos ever put a paywall on their site, we would be buying one, right? Right. You know, like, we are not building it- Yeah... from scratch.
17:35
And, um, commenting system was another s- example of that where, you know, previously we had our own commenting system in-house that we had b- written, and our users loved it, and it had lots of, you know, features.
17:45
There were some performance, uh, issues with it where it was fully built in React, and, um, for some users with older browsers, it was just too heavy. It was very slow. Um,
17:55
and we needed a team of engineers to keep it going, and when we, you know, were looking at the options to buy a commenting system, it... there just weren't a lot of choices. Yeah.
18:03
Um, and especially not choices that wanted to deal with the migration of 88 million comments.
18:08
Um, so our options were really OpenWeb, Viafore, and Disqus, and, um, you know, we had to, had to pick the commenting system that we thought was best for our users, but it wasn't a perfect fit.
18:19
And if we were building it in-house, you know, we would've been able to, to make it- More perfect, you know- Yeah... whatever that means. Yeah.
18:25
Um, so, you know, if you're at a bigger company, you're just able to, to deal with these, um, kind of edge case product requirements a little bit better.
18:33
When you're like rolling out these AI tooling and like any change management piece, I guess, how have you thought about like measuring, um, sort of any of the impact?
18:41
You know, any bean counter's gonna come and say, "Well, hang on, you're spending 50 grand on, on tokens this year. Um, what did we get for that?" Yeah.
18:48
Which is a very hard, if you're in product and engineering and tech, to justify that, 'cause no one's gonna go and turn it off- Right. Right... um, and go back to doing it by hand.
18:55
But yeah, how, how have you thought about, like how do you measure the, the, the value creation as well?
19:01
This is something that, you know, honestly, like the, the time is coming when, you know, Anthropic just like turns up the cost per token and it's just like, you know, now that you're dependent on us- Yeah, that's right...
19:12
good luck, right? Um, for sure that's coming. Um, in terms of measurement though, of product features, the hope is that every product feature that you deliver to your user drives incremental value.
19:26
So you could think about that as like a direct, you know, paywall or, you know, something, a, a donation ask, you know, something that's directly driving revenue, a subscription request, a sub- a newsletter signup box, you know, whatever it is that actually is directly, um, h- users giving money or attention from their pocket to you.
19:42
Or you could think about it as a feature that keeps the user on your site for longer, increases the amount of, you know, ads that they see or their, you know, the value, the relationship with your brand.
19:50
Um, so if you're deploying features to your site that aren't doing either one of those things, then why are you deploying that feature, you know, is the question. Yeah.
19:57
Um, and hopefully there's metrics that you've established, um, whether it's, you know, monthly active users or sessions or dollars, um, to measure the impact of all the features that you're deploying.
20:07
And if I'm de- you know, if it takes me six months to deploy a feature, my example that I gave in my talk was, um, it took us six months to put a Sign in with Google button on the old site, and it took us about a day and a half using a WordPress plugin.
20:19
Right. Um, in that six months, we're not getting the value of people being eve- able to easily create- Yeah, opportunity costs. Yeah... their account. Yeah. So that has a cost as well.
20:29
Um, so the idea is that we'll, we'll look, you know, holistically at, um, the value of, you know, the site as a whole, uh, at the end of the year versus pre-migration and, and- Yeah... see where we are.
20:42
So you obviously went through a journey of, you know, this, this sort of, uh, I think it was a 20-year-old CMS, like you'd been working on it for a long time, and obviously a lot of blood, sweat, and tears go into this, a lot of emotions.
20:52
If you were doing it again, if you got dropped into another media org that was running a custom.NET CMS that's running, you know, five different languages and AWS and all this sort of stuff, how would you sorta start thinking about it and sorta convince a, a leadership team?
21:06
Yeah, I mean, given this is my second, uh, convincing of a leadership team in, uh, as many jobs, I feel like, uh, I'm well positioned [laughs] to answer that question.
21:16
Um, you know, the, and this also kind of brings me into the, the SaaS fallacy where, you know, there's so many media companies that have been like, "We've spent all this time and money building this content management system.
21:25
We should just sell it to other media companies," and then they haven't like fully thought through what it takes to build a SaaS company- Mm...
21:32
which, you know, a multi-tenant databases and customer support, and how do you prioritize one customer's needs over your own internal team's needs, and like, like there's a, it's a whole thing.
21:41
Um, and I've seen this happen so many times, it just like makes me laugh.
21:45
Um, so in terms of like the build versus buy, like calculus, I mean the first question is to me always, um, is what you've built a business differentiator?
21:53
Is your CMS doing something unique or special that would be hard to replicate in a commercially available content management system? For Daily Kos, um, the answer was very clearly no. Um, it was text on a screen.
22:06
That's a solved problem. Uh, the internet knows how to do this. Yes. You know, this is not something that is very special to Daily Kos.
22:12
Um, second question is, you know, what is the total cost of ownership on an ongoing basis? So it's not just like how much does the CMS cost, um, commercially to buy in year one.
22:23
It's, you know, over time what is the cost, and then what is the cost to maintain it? Um, and with an in-house CMS, especially an older one,
22:34
there's a risk that is associated with the fact that, you know, as I was talking about my, my Perl CMS, um, if something's not tested, if there's no analytics, if there's no o- observability,
22:45
you've got a situation where it's just a matter of time before, you know, there's something that happens where, you know, it's just no longer compatible with life, and like that can happen very quickly. Mm.
22:55
And if that happens, you're on your back foot. Now it's an emergency and you have to do a migration, you know, under duress, which is never a position you wanna be in.
23:03
Um, so all of these different, you know, pieces, um, I feel like are, are elements.
23:10
Um, and then, you know, there, there's the New York Times example, which is like the New York Times has their own CMS and they have 500 engineers. Um, and they have very specific business rules.
23:18
They are never allowed to delete an article ever- Mm... because they are the paper of record. So, you know, that is, you know, a special- It's in their DNA. Yeah. Yeah.
23:27
A special, you know, case for theirs, um, where when I was there and we were looking at, you know, could we put cooking into the, the new CMS because recipes aren't such different data from a news text, whereas a game is very different data.
23:40
You know, a crossword puzzle, you know, has a across and down and, you know, clues and like, you know, it, it's a very different data shape than a- Yeah... a text article.
23:48
Um, but a recipe is not, and we looked into it and we're like, no, because there are so many unique business rules that the, the news team has to follow, um, that this just really isn't a good fit for something that is, um, you know, less specialized.
24:01
So I think, you know, your business is a unique use case, and you kind of know like- Yeah... you know, when you've gotten to the point where it's like- We can't keep, keep going this way. You know, like this [laughs]
24:13
the team can't support it, we can't support it, we can't pay for it. You know, it's, uh, it's time, you know.
24:17
Yeah, it's, and it's weird as, as, a, as a vendor that works with media companies as well, like to come in and say, "What you're actually doing is inherently not very complicated." Like, it's a solved problem.
24:25
Uh, g- like- Exactly. Yeah... again, uh, it would be nice to sell you a really complicated solution, but you're actually trying to do a very simple thing, uh, so you shouldn't be overcomplicating it.
24:32
Uh, and then also I guess the other thing is like opportunity cost, right? Like- Yeah... if you're running your own custom CMS, and then someone's like, "Well, how do we connect Claude to our CMS?" Like, well, you can't.
24:39
We have to, we have to now go figure out how to build an MCP adapter. Right. Right, right. Yeah.
24:42
How do we do that, and how does it not break everything, and, you know, all these sort of things that s- are solved problems in, in open source. Yeah.
24:49
And then you're just outsourcing, you know, like if you don't have an AI engineer on your team, you can either go hire one or you can, you know, use a system where these people have already employed these AI engineers, and they've already- Yeah...
24:59
built the MCP servers and, you know, you don't have to reinvent that particular wheel. Exactly. So, I mean, there is that like specialist piece of it as well. You know? Yeah. Yeah. Finally, um, where are you...
25:11
where's your head at in terms of excitement, um, the future of, of engineering? If, would, you know, would you be telling one of your kids to get a job as a software engineer- [laughs]... um, working in a media company?
25:21
Um, yeah, where, where... what are you, what are you getting out of bed for in the morning? What's exciting of like where this is going? Yeah. I mean- Or are you excited? Are y- is it all doom and gloom and, uh- Yeah.
25:30
[laughs]... people should just quit while they're ahead, uh? Right. [laughs] I mean, look, there's, there's some pros and cons to everything, right?
25:36
Um, and any time there's a new technology, especially one that's as disruptive to an industry as AI, you know, um, there are pros and cons. I think, um,
25:47
you know, I have children, and I am very concerned about them, like learning to think before they like let- outsource their thinking to AI. Yeah.
25:55
Um, and making sure that, you know, we haven't lost the ability to, you know, do the craft of software engineering, to know what good looks like. Yep.
26:02
Um, to understand what the AI is doing and why it's making the choices that it's making, and like are those the correct choices? You know, sometimes something's like way too verbose.
26:09
You know, we can really dry this up, you know? Yeah. Sometimes it's, um, oh, we missed a test. You know, w- what about this? You know, a, a person would never do this, you know?
26:19
And retaining that critical thinking piece, um, that's why I still write my product briefs myself as a first pass. You know, it's just like I wanna make sure that this is coming from a place of like, I understand this.
26:29
I know what I'm trying to achieve. Um, I know, like before I send it to AI, like this is, you know, the, the basic premise is coming from a person. Yeah.
26:38
Um, but excitingly, you know, we can do so much more so much faster. Yeah.
26:43
And there are initiatives that Daily Kos is gonna be undertaking in the next year where like it wouldn't have been possible for us to even- Even when you had the huge team... think about it.
26:52
Like you had a huge team- Yeah... and now you've got half the resources, but you could actually... And that's- Right...
26:55
I think that's one common link with so many media companies, is that they are, you know, it's a, it's a industry in, uh, existential threat f- since the beginning of the printing press, you know?
27:02
It's always, uh, a, a, business in change as well.
27:05
And the amount of times that you have to deprioritize features, where- wh- whichever side of the fence you sit on, uh, think, "Oh, we, we would love to do that feature for the commenting.
27:13
We'd love to do that," but there's only so many hours, dollars, whatever that is, that, yeah, I've, uh, that there's a whole bunch of opportunity that can be done now, that you can build an app and it's just like this, you know, instead of it being a half a million dollar project, um, to pull an app together.
27:28
Yeah. I mean, we built a feature to allow users to save articles on Daily Kos in about two and a half weeks. Yeah. That would've been a six-month project- Yeah...
27:35
you know, before with a, you know, an, an engineering team.
