00:00:00 David Egts: So, Gunnar, in episode 273, Gunnar Hellekson: Mhm. David Egts: I think it all began there where you were talking about the the delight of abundance with AI where where AI could write all the code. Gunnar Hellekson: Yes. David Egts: You you were no longer constrained uh by anything anymore. you you are you've become infinite and and you know but in the meantime as far as like all the and and also we talked about the with open source as far as with you know that this is going to be great for the open source community where open source is going to have all kind of people doing poll requests and everything but you know a couple weeks later after we come out with that uh or months uh you know people are saying that uh what is it like the curl people or like we're not taking any AI pull requests it just we're over or the bug bounties, right? They they don't want any more of it and everything. Gunnar Hellekson: Yeah. David Egts: And uh you know, Gunnar Hellekson: Yeah. David Egts: we're seeing that uh there's too much AI going on, right? 00:01:03 David Egts: And and people are overwhelmed. Gunnar Hellekson: Mhm. David Egts: Uh other people are saying that open source is dead because we just got AI for everything. Uh and and one of the things that you mentioned uh we were talking afterwards is that we ought to get Mike McGrath on the line and and have him uh provide his opinions on it because he's he's been working in this space and uh and here we are. We got Mike. Gunnar Hellekson: Yeah. Hi, David Egts: So, Gunnar Hellekson: Mike. David Egts: welcome Mike. Mike McGrath: Well, hey guys. It's uh it's great to be Gunnar Hellekson: So, Mike, Mike McGrath: here. David Egts: Yeah. Gunnar Hellekson: do do you mind uh introducing yourself for folks who may not be familiar with you or your Mike McGrath: Yeah. Gunnar Hellekson: work? Mike McGrath: Uh so my name's Mike McGrath. I've been at Red Hat for 19 years just a couple weeks ago in various uh modes. I was a volunteer in the Fedora project before that and then I joined to work on Fedora full-time. I'm a founding member of Open Shift here and uh am I think one of the the few people that spend some cycles every once in a while really thinking about this intersection between open source and business. 00:02:05 Mike McGrath: It's an unsavory group uh trying to draw that that vin diagram but uh much to my si surprise I found myself in this area and today I lead uh rail and uh open stack and satellite and a bunch of those related uh technologies. The most recent of which is the invehicle operating system that we're working on which has been super Gunnar Hellekson: That's great. That's great. Mike McGrath: fun. Gunnar Hellekson: And I and I should say that Mike is one of my my favorite and only engineering leader uh associated associated with real. David Egts: It's in the top three. Mike McGrath: Thank David Egts: Yeah. Gunnar Hellekson: Yeah. Easily top David Egts: Yeah. Mike McGrath: you. Gunnar Hellekson: three. David Egts: Yeah. Mike McGrath: Gunnar, you're one of my favorites, Gunnar Hellekson: Oh, Mike McGrath: too. Gunnar Hellekson: thanks. I appreciate that. I appreciate that. Well, I So, Mike, um maybe spending a little bit of time just to table set on like what where your expertise comes from. Um you mentioned your Fedora background and you were managing the rail business, the engineering business long before um long before AI rolled around, but also you you spent the last year on uh on uh on our internal AI kind of assimilation efforts. 00:03:13 Gunnar Hellekson: You want to just give a quick brief on that because I think that's where a lot of your experiences come Mike McGrath: Yeah. Gunnar Hellekson: from. Mike McGrath: uh you know uh last year uh was I we can call it the year of AI you can call it whatever you want but for me at least last year was the year that I really spent some time adopting AI figuring what it can and can't do all the good and bad that comes with it. Uh and on our team we did the same. We pushed it in ways uh that we knew were going to fail so we could figure out where those failure points were. And in some cases it far exceeded our expectations. David Egts: Mhm. Mike McGrath: In some cases we were surprised that it was ever released for public and uh and everything you can imagine in between. Um just to give you a couple of examples uh of things that we were looking at. Um, one of them was, you know, just with general package automation and and we, as you can imagine, at Red Hat, we do quite a lot of packaging. 00:04:08 Mike McGrath: Um, that was probably my favorite of the projects that came in. We also looked at how AI could change and help managers manage things. And so, one of my favorite things from that was something for both managers and associates that we're still working on and releasing internally. But it's like, what's all the stuff that we did in the last quarter? because you know at every company they've got performance reviews and all the rest of it and we're all human and when the quarter ends you don't tend to get to them for another few weeks and by then it's hard to remember what somebody did three months ago and so uh having a scraper that can go through Jira and GitHub and GitLab and all these different places to summarize it so that the associate and manager can look at it and say actually yeah that's what I did I forgot I even did that or this thing's missing or whatever you can go through and do that that's the kind of thing that actually takes quite a bit of calories and is not the most fun thing to do at work uh manually. 00:05:03 Mike McGrath: Makes it very easy to do automatically with with AI. And then more just generally um we've you know I've been creating little bots and things. I've got an Emograph bot that has been very popularly adopted by a lot of my staff to ask questions to. It knows quite a bit of what I know and much to my surprise has gotten as good at this as I am at some things. And so now even I use my own bot for things uh when I'm I'm working on something and uh yeah I mean that that has just continued through to this year and uh the models have really gotten to a space where um even if you're not coding uh they are quite useful and I I don't know how many people understand that right now and it's it's really amazing. Gunnar Hellekson: Yeah, I David Egts: So te tell me about the your bot like what what what does your bot do Gunnar Hellekson: think David Egts: that? Yeah. Mike McGrath: Yeah. So, uh, my bot was originally designed to help answer some of the common questions I've got. 00:06:02 Mike McGrath: So, I'm a vice president at a company. I get a lot of random questions about like, I'd like to relocate this person or this person got a job offer. Should we counter it? like just a lot of you know general operational things that we have policies that are written for but often those policies either are open to interpretation or and sometimes contradict each other you know maybe you want to relocate somebody but they're an internal person versus an external hire and you know the policies aren't necessarily written that way and so I started just writing these things down and uh it got to a point where you know I literally kind of explained to I was using opus I kind of explained what I was trying to do I just like what questions do you think are going to come up because AI is excellent at prediction. So why not use it to come up with the questions that I should answer. And so I I put that together and I've got uh Opus inside of cursor is actually generating a Gemini gem that is sharable within my org. 00:07:00 Mike McGrath: It's kind of going back and forth a little a little craziness there. Um but I've probably got at this point we just did our CY26 planning. It's February 16th as of the writing of this and so I just last week finalized the 2026 updates for it. I probably got about 10 to 12 hours in it right now. Uh and it just has a ton of context and it's it's even things like hey what happened to this project and uh you know it's well it'll say well this got folded into this new project name. It's now being led by this person. If you have questions go ask them. and it knows about all of these links and it's got a lot of data about the org that I'm running and the org sizes and you know what we're trying to do in terms of hiring and all kinds of things that uh once it got going became extremely useful and uh I've found a lot of people using it they're integrating it into their own workflows now and I've started to integrate it into my status reports and all kinds of things that when you want a strategic look at uh what's going on it's a really great reference for that And it's uh probably my probably my favorite project that I've worked on so far just because it's been applicable in ways that I hadn't 00:08:12 David Egts: So, how do you make it? Mike McGrath: considered. David Egts: How do people discover it? Or or you it's a gem that's out there and and you somebody like asks you something, you just say, "Hey, that's a great question for my gym. Go ask Mike McGrath: Yeah. So, uh, in, uh, David Egts: it." Mike McGrath: sometime last year, kind of late last year, I think it was in the fall, uh, Google added the ability to share Gemini gems with a link just like you would a Google doc. David Egts: Yep. Mike McGrath: And so, David Egts: Yep. Mike McGrath: uh, I literally just share that around. It's part of some announcements that I've sent and they can go use that anytime they want. The uh, the cursor version that has some like really nice agentic workflows as part of it is just on an internal repo. I mean it does have some you know salary targets and you know stuff you generally wouldn't share with the public but things that are certainly like redhead internal where we are and it's you know labeled properly and they know not to share it but for the people internally that are using it um they've started to integrate it and they just know where to find it and occasionally I get an update but for the most part it seems like people are just letting me update it on its own and that's 00:09:14 Mike McGrath: been working out fine. Gunnar Hellekson: That's great. So Mike, tell me about the So I think a lot of people by this point I think are comfortable making gems for example and so when you say like I made a gem they kind of understand okay I wrote a prompt and then I attached some knowledge documents to it. Um and now I can answer questions about the knowledge documents and okay pretty straightforward. Um and I think you and I have both been on this kind of agentic journey over the last the recent few months. Do you want to talk a little bit about like what does it mean to have to make it agentic or to have agentic workflows associated with the associated with Mike McGrath: Yeah. Yeah. Gunnar Hellekson: it? Mike McGrath: That that agentic workflow uh once you start having these sort of AI systems talk to each other suddenly becomes very interesting. And so I'll give you a perfect example. Um, pretty much anytime I've got a uh a presentation coming up now, I've got Emrathbot that can go through and evaluate that presentation and just say whether or not the information in it is correct or sometimes I can say, hey, you know, slide seven, I need to add my attrition rate to it and I can just type it in and then get back to work and by the next time I go look at the slide, it is properly it's got my attrition rate, 00:10:24 Mike McGrath: whatever in there, whatever information that, you know, lots of data that I just forgotten and it's in there. Um, the other thing that's nice is, you know, I've got a status report that I do uh every uh every week and there are things that I know are important to my boss. So, uh you know, my boss has uh is the executive sponsor for some of our very some specific customers and I get I don't know how many pages of status reports come my way uh every week, but sometimes I may forget, oh, he's you know, this bank whatever and I needed to to update that. So, I've got a a fairly I think it's elaborate. AI doesn't think it's that elaborate, though. Scoring system for those things that come in. And I still read all of my status reports, but I'm no longer in a situation where I'm constantly context flipping between copying and pasting and trying to adjust it and this and that. Um, you know, I still read each one. I respond to it them all in one step. 00:11:17 Mike McGrath: And then I run my status report generator. And I'm not kidding, honestly, as of maybe mid January, it does a better job with my status reports than I ever did. Um, it is much better about filtering out fluff. Um, it is much better about targeting the customers that are on our scope right now and uh might be on my boss's scope because he may be in ready to head into a meeting with them. Um, and this is another one of those examples where our strategy adjusts over time and as long as I keep McGrathbot up to date, anything that might be based around or interested in strategy can get that update at the same time. And so my status report generator knows this. my slides generator and updator knows this. And and Gunnar, I know you and I have talked through different ways to kind of create fake personas for, you know, you're going to be talking to the CEO, you're going to be talking to your CTO. What are the things that they're likely to care about? 00:12:09 Mike McGrath: Uh you can start having your McGrathbot, have a conversation with them, and then you can go watch that and say, "Actually, this was a really good point. Um he's going to need some budget numbers on this that I hadn't considered or there's some startup that was just doing this. Why don't I pull in some of that information, you know, competitively so that he can stay up to date? Those are all things that basically just happen on their own once you get agentic workflows connected. And, you know, my another good example with the SAS support generator is it goes and it crafts the whole email for me. I'm not copying and pasting between emails. I'm not really even copying and pasting that much between slides anymore or slide notes. It just knows to go and insert things here and there and I just sit back and watch. And when it doesn't do the thing quite the way I want it to, I just tell it that and it goes and updates it. Yeah, the agentic workflows are really 00:12:58 Gunnar Hellekson: Yeah. And you're managing these. Mike McGrath: something. Gunnar Hellekson: You don't have like a fancy schmancy uh interface for creating agentic workflows where you're like got a mirror board where you're like wiring up geometric figures to each other or anything like that. Like you're literally just sitting at a cursor window typing the instructions in. Mike McGrath: Yes. Gunnar Hellekson: Right. Mike McGrath: Yes. And for those that haven't used cursor, Gunnar Hellekson: Yeah. Mike McGrath: it's an ID. It's VS Code kind of ID and much to my surprise uh because I'm a command line guy. I'm a home row person. Anytime my hands have to leave the home row, I hurt in my heart just a little bit. And therefore, I should love Claude. Claude is where I should be spending most of my time. But for whatever reason, uh I have just had a really good experience with cursor even for these non-coding tasks. Um, I should mention uh one of the key insights for me was just how good AI is with the markdown, just a standard markdown file format. 00:13:52 Mike McGrath: It's in text. It's it's formattable. It looks nice. But AI can also use some of those headers as visual cues as to what is and more and less important. Um, almost all of them a graphbot is written in that in that markdown format. And that was kind of a big key insight for me last year was like, oh, now we're talking in a language that both AI and I understand very well. And so all of my uh Google docs are written that way and go back and forth in between markdown and not all my emails are written in markdown before I convert them to a formatted thing. And that was I know just you know some of those little things you pick up as you get used to to using these tools. Gunnar Hellekson: Yeah. Yeah. So, so Micah David mentioned in the in the beginning I I I shot my mouth off about this this abundance idea. And so the the idea here was that um it is now much easier to do hard 00:14:33 Mike McGrath: Perfect. Gunnar Hellekson: things um with the help of these AI tools. And history kind of teaches us that the benefits of this mostly acrew to the bosses, right? And not to the people actually doing the actual work, right? Um, and and I'm wondering if you have because what you just described was in some cases like improvements on work that you were already doing, but also it sounded like you were actually able to do more or improve the quality of the work that you were doing as a result of using the tools. So, I'm wondering if you think if there's a how do you think about how do you think about it in terms of like um is it mostly a quality improvement or is it like a quantity improvement or some mix of both? Mike McGrath: Yeah, it is a mix of both. And I think this is where, as with anything, it's really important that when you go in, you have clarity of what you're trying to do. Um, you know, we've got lots of people inside Red Hat proposing lots of different things for our legal team to go review and just make sure we're not breaking any laws and, you know, all the normal stuff that you do at a company. 00:15:50 Mike McGrath: And one of the questions they ask as part of that process is like how much money is this going to save us? And that's a fine thing to ask. Absolutely. We should go look at that where we can. Um, and there's been a lot of times where making slides, doing updates, finding information has been orders of magnitude faster because of this process. And so when I when I'm looking at speed and accuracy and I don't have to go correct somebody or a question doesn't even come up to me to be asked because Emma Grathbots's already fixed it. That's all savings as far as I'm concerned. They got their information. They got to they got to trial a whole thing without scheduling a meeting with anybody and just see what Emrathbot said and Emma Grathbot does a pretty good job of saying, you know, this is you feel free to send this proposal to Mike, but he's likely to reject it because and if they feel really strongly about it, they still can. If not, they know, hey, here we are. 00:16:41 Mike McGrath: But there's a flip side of that, and I'll talk a little bit about my performance development uh discussion with my direct. So, most of my directs are senior directors um or highly skilled individual contributors. And this is one of those examples where um I used AI to help me craft some of their goals uh for next year. And in no world can I say that that was uh faster or cheaper than what I had done before. But I think every one of them would agree that my goals were better this year for them or this quarter for them than they were in the past. And my favorite example of this is a smart goal. It's an SMART goal. I forget what all the actionable measurable whatever. I forget exactly what it is. Gunnar Hellekson: Measurable extra. Yeah. Right. Yeah. Mike McGrath: Smart goal. Gunnar Hellekson: All that stuff time bound. Mike McGrath: Uh this is this is something that I I've always found st maybe some human beings are good at this but I've always found it stressful 00:17:30 Gunnar Hellekson: Yeah. Mike McGrath: to take like directionally I have very clear idea of what I want you to do and and for you to go execute it but it's hard to put that in a smart goal AI turns out to be excellent at this just excellent at it and uh I had a you know my team if they ever listen is gonna going to hate this comment but you know anytime you're typing stuff into your HR systems you know just a regular company we're doing all this stuff um they hate all that. And what I realized was in the it was the process of this that was so useful to me. What do I really want them working on this quarter? And I realized that for everybody that's complaining about uh the HR processes and copying and pasting things here and there, I realized that for me that was such a small part of the overall time that I spent really thinking about what I want my team to be doing and what I want them to be doing. and having this sort of AI coach kind of take a critical eye at what was being worked on and why and and ask me questions about what uh what was going on and whether or not these goals were actionable and clear enough. 00:18:39 Mike McGrath: Um I think landed all of us in a much better space and was a much better use of my time even though I think objectively it was more expensive from a dollars and cents point of view. the long term, you know, certainly I think it has payoffs, but in the short term, I could not honestly go and say, "Yeah, this saved us a lot of money." Instead, I think it it it uh helped me spend needed calories uh to to bring clarity to a process that uh I think sometimes goes uh underused. David Egts: Yeah. And I I I've seen it too like at Salesforce, we've talked internally about and and I've seen it in other places too. I've seen Microsoft site this and there are other studies are out there that it's like you could focus on using AI to reduce the bottom line. Um or you could use it to increase the top line. But if you decrease the bottom line, you'll do no better than zero, right? If if I cut and make everybody super efficient. 00:19:37 David Egts: But if I focus on the top line and growth, um, it's unlimited. Mike McGrath: Yeah. David Egts: See you later. Mike McGrath: Yeah. And I think the other thing we didn't mention or too is just once you it takes a while to set up an AI environment. If you're somebody that just installed it and tried to reinvent the Linux kernel and it didn't work and so now you're you're done with AI. Uh my my analogy has been setting up your AI environment should probably take about as long as when you get a new computer. like however long it's going to take you to sit down at the computer install and configure all your things and do all that like it's going to take you about that long and it's about that process to set up AI but once you do topline or bottom line you do see great quality of life improvements that that for me has also been you know as an assistant the quality of life improvements I've had have been just immense it's it's hard to imagine going back to not using AI at this 00:20:28 Gunnar Hellekson: And you know, Mike, one of the one of the things that you taught me was, you know, rewind a year ago and everybody was worried about AI replacing me, right? And I think you had the intuition and then I followed it and I agre and I had the same results where actually the very first thing I did was try to replace myself. Um, and that had that completely colored the rest of my experience in a positive way. Um, because it made me realize um, exactly how much benefit could be every individual person could get. Um, so I built Gunnarbot inspired by McGrathbot and then encouraged everybody who reports to me to go create bots of their own. Um, we even set it up so that all my directs shared their bots with each other. So now we can have all the you know somebody has a proposal they want to send to the staff meeting well have all the bots review the review the meeting beforehand right um and that is that has been a that has been a real benefit so I think the the irony here is like well you don't want to replace yourself with AI it's like no no no the very first thing you replace yourself with this the very first thing you do is replace yourself with AI um because the dividends of that being able to like you say once it's smart enough where you're 00:21:40 Mike McGrath: Yep. Gunnar Hellekson: actually asking your own bot questions. Um because you're right, like in a lot of cases, it's got a much better memory. It is much better at pattern matching. It's much better at kind of un intuiting what it is that you need. Um I've had a great result from that. Um Mike McGrath: Yes. And and it's very interesting too. Gunnar Hellekson: yeah. Mike McGrath: You can get these bots chatting with each other. You know, I think we've done a few experience experiments like this. And for me, we've had a couple of major decisions over the last uh couple of couple of months. And it's fascinating to watch your different bots get a consensus score in terms of what you should do. You've got a 90% consensus to move forward on this. you can see the concerns of each one, where they're going. And it's not perfect. It's not, you know, not human perfect, but it does give you a pretty good idea of where of where things are and and and also like what information might be needed. 00:22:35 Mike McGrath: If the bots were confused about something, why were they confused about it? Like, oh yeah, we actually do need to go get this information from somewhere because the real life people are going to need this, Gunnar Hellekson: Yeah. Mike McGrath: too. So, Gunnar Hellekson: Yeah. David Egts: So, how do you like how do you like with your your peers and all that? Is is everybody uh using AI to the same level or like I could imagine there there could be some that are just like, oh, I'm not going to bother with it and I'm just going to do things the way I've always done it and everything. Is do you think that's holding them going to hold them back? Mike McGrath: We I think we've got the full mix at Red Hat, both of my peers and down in the org. We've still got some people that are skeptical of AI and and concerned about its future and its impact on the economy and the environment. Uh we've got some that are fully into it like I am. Um I think my peers are a pretty good mix. 00:23:28 Mike McGrath: I think that there are some things that they've been very interested. I've had a lot of interest in Emma Graphbot because they think they like to chat with it. A lot of interest in the status report generator. I just it's like one of those weekly things that happens every week and they want to they want to know how I did it. Um, but I also, you know, I think you run into that same thing that we're in this large transition into AI. And so it's it's just a matter of I think a lot of people wish they had more time to do it. And, you know, I think for me, I've had a lot of uh a lot of time to play with it in my personal life as well. And so it just kind of helped me come up to speed. But also, you know, as as Gunners kind of mentioned, we had a major my team had a major AI project last year that allowed me to spend my time at Red Hat really, like I said, pushing things to their limit to see what would break and then and you know, laughing about it and enjoying it and and and learning from it. 00:24:19 Mike McGrath: And all the stuff that succeeded though is just now been added to my daily repertoire and we use it all the time. David Egts: So what about like so that's awesome for for you guys both to say that as people leaders and you know usually when you think about oh AI it's going to be used by the you know everybody all all of the people that work for you are going to be vibe coding and it's it's at the individual contributor layer of of everybody being more efficient at that layer. Um, but if we look at like swinging this over to open source a little bit and everything about, oh, this is this is going to kill open source and all that and you know, there was an article or a Medium post that I saw that uh that it was entitled the day open source realized code didn't matter anymore. And and the premise of the Medium post was that people would from an open source project standpoint, they would much rather not have poll requests and they would much rather have well-formed issues. 00:25:27 David Egts: And where do you fall on that, especially with your engineers working with the open source communities? Are they feeling overwhelmed by being flooded with random poll requests? It could be slop or how how do you see that happening from an open source perspective? Mike McGrath: Well, I'm happy to announce here on the Dave and Gunnar show that on January 16th, 2026, open source died. That it's done. We're done. I'm kidding. Of course, I'm kidding. Uh I actually am in the camp of open source is about to see a renaissance. And let let me explain why. Uh and we can dig into any of these you want to at any depth you want to. uh first of all is that any of the models just because of the way that they're trained today um and I believe in the future as well will always know more about an open source program than they will a proprietary program both in terms of the ability to change them but also the ability to understand what they can do not just what they were intended to do in the docs and other examples that's that's an example number one and I think this uh this article open source uh realize code doesn't doesn't matter anymore I I think that it outlines a significant change for us which is that now code is cheap. 00:26:42 Mike McGrath: Uh maybe maybe code is ubiquitous and code is everywhere but now more than ever context is important and curation is going to be important and certainly at Red Hat curation is a word that resonates with us. We know what that means. When I look at what is going on with open source contributions today, I actually don't see any different than that between that and the RPM Forge and Wild Wild West RPMs repos of the early 2000s. It's the same problem to me. All of this stuff just kind of exists. There's forks of it everywhere. Maybe I made a fork, maybe not, but it's just there and it's available and 90% of it uh is not of unknown origin. And that's kind of what we're seeing with some of the AI bits. And certainly I've heard people say, well, why don't we just fork every AI project and I'll just, you know, we'll have AI maintain it within my my own company. I think at least in my own experience, I think other people will run into this as well. 00:27:42 Mike McGrath: Uh, which is as with anything, AI doesn't change the fact that the last person that touched it owns it. And I don't think AI can own things. So, whoever is running that AI is going to own that thing. And that's just that's going to be a fact of life. And for anybody out there that truly thinks that uh we're going to be running our own forks, each company's going to have its own fork. I have to assume that you are someone that has heard of CI but has never run CI in a production environment because the fact is with AI or anything, everything's fine until it's not. And then it's expensive to go fix. And I don't think AI is going to be able to. It's going to be a helper, a great assistant in that scenario, but AI is going to be lacking in the context of understanding what it was trying to do, what its intent for that for the decisions it made were. The other thing I think is going to be very important here is that AI does its best work when it has firm boundaries and an understanding of what it is and is not allowed to do, which which is another way to say standards. 00:28:47 Mike McGrath: And I know NIST has recently come out with some AI coding standards. There's all kinds of this coming on. A uh open source is a very natural place to write those standards around. And I think it's really critical for all of us in open source to understand and embrace this because sure maybe not all of the code is going to be written by AI but our intentions and context is still needed. This is where you get into that, you know, AI as an assistant versus AI as an automation tool. And uh uh I think uh for the for the the communities that fully embrace AI, uh I think they're going to see uh benefits in ways that they had not considered. Uh because I think a lot of people just assume, well, AI is going to replace you in coding. like well you know I remember what it was like to learn the Fedora packaging guidelines and the tome of information that was and just how much faster that would have been for me to learn had I had AI available to help me with that because some of that stuff is seems pretty arcane if you don't understand why you're doing something and AI can help with that any other number of coding related tasks outside of coding including coding but like you know things that things that all of us developers have to do around the 00:30:04 Mike McGrath: coding So I where where would you guys like to go next? Because any one of these topics I think is very meaty. Gunnar Hellekson: Well, I think what um I'm particularly interested in, are you worried at all about set aside the coding for a second um the avalanche of PRs that people are getting, right? Uh the merge requests for example. I think I'm giving you a whole bunch of garbage MRS on some of your projects. Um which I'm a little bit self-conscious about. Um but but uh there is this worry that there's a there's an asymmetry now. Um, and I think there was a very good article, I think it was about GitHub, where they were saying like GitHub was built for a world where it was expensive to build code. And so a poll request was like a good gate or a good throttle on the amount of information coming into the project. Um, and it seemed to work for everybody, but if you in a world where you could get a 100 PRs in a single day, you could easily overwhelm a small development team. 00:31:02 Gunnar Hellekson: So I'm wondering if first of all, have you seen this in the wild? And um and also have you thought about any how do you compensate for that given the tools that we have available today? Mike McGrath: Yeah, I mean I think I saw was it Okamel had like a they were they they rejected a request because it was 13,000 lines and they just couldn't consume it. Now a 13,000 line PR is a bad idea for a human to send along too but definitely a bad idea for AI to send along. Um yeah, we we see this in a few in several different areas and I think different, you know, every community should come up with their own plan for this kind of thing. And certainly in a very large or high um high uh what do you call throughput requests that are coming out there just too many for the humans to to keep up with. They can try to throttle it. They can try to prioritize it, but this is one of those meet fire with fire moments. Um I think uh one of the things I wish that we had done sooner is you know this sort of context of the uh of the persona. 00:32:04 Mike McGrath: There's absolutely nothing stopping any of those uh uh communities from creating a persona that is to go and rate both the quality and content of these contributions, their strategic value or or just their general value to the the the org itself or to the to the the project itself. and you can rate them. And so this is an example where at no point in time in what I've described here has an AI merged the request or anything like that, but it can rate them. And if you you know this is one of those things where uh the Fedora community had a very big conversation about whether or not to and how to allow AI contributions. You should hold AI contributions to the same standards as you do anything else. Just because it's AI doesn't mean you have to let it through or or that you should let it through. Um, but you can meet those those sort of denial of service attack style contributions. You can meet that with other AI and uh depending on what your project needs, maybe it needs more quality, maybe it needs fewer uh merge requests, maybe you need somebody to go help and find the duplicates and AI is pretty good at that kind of thing. 00:33:12 Mike McGrath: and it can go and rate them and you can if you get it good enough that you trust it and it is you know if you've got a a success rate of 90% with your human contributors and you get AI to 95% well I mean that you can track that over time and trust it go ahead if you're not there then you can you know find the things that it's recommending go look at them and just take another eye to it you know it's very good at picking these things out uh and help out that way what I but I also but I I will also say this too. If you're somebody that is generating these contributions, I think anybody that's been involved in open source for any amount of time knows how painful and how unhelpful a driveby per merge request is. And if these are just drive by merge requests because you needed a thing and you really don't have any interest in that community, you're not going to be sticking around in the long term, like that does not actually help anybody. 00:34:04 Mike McGrath: I know it seems like it might and maybe you get to put it on your resume that you've contributed code, but in an AI world or before an AI world, uh that's I think that's bad form and should be avoided. And so if you want to get involved in a community, get involved with it. Um but you've get, you know, if you have a bunch of these thousand drivebys, um in terms of merge requests, it's not good. But I will say one other thing, this does kind of prove the point. I think of this increase in in contributions that we are greatly lowering the barrier to contribution for new contributors and that also to me is evidence of an open source renaissance coming our way. I I think it's I think it's going to happen. Gunnar Hellekson: You're reminding me of a I've got an old joke slashlaw which is Helix's law, right? This is back when my government days when we used to say um if you have a rule about the use of open source software, it can be improved by removing the words open source, right? 00:35:04 Gunnar Hellekson: In other words, like any rules that you put in place for open source are probably rules you want in place for your software in general. And in the same way, Mike McGrath: Yeah. Gunnar Hellekson: like any rules you have about AI contribution are probably just contribution rules, right? Um, Mike McGrath: Yes. Gunnar Hellekson: yeah. Yeah, I agree with you. That makes sense. Mike McGrath: I still I still remember and those same arguments that were made against open source in the early days are being it's exact same argument of concern being made about AI. very similar and it'll prove itself just like open source did. Gunnar Hellekson: Yeah. Mike McGrath: I David Egts: So, Mike McGrath: think David Egts: and and that reminds me too of like as how do you see it for like like early career people like I love how you said that that AI can help lower the barrier to get people to contribute to open source and and I think that for a lot of the computer science students that are it's they're having a hard time finding a job or differentiating themselves on their resumes. 00:36:02 David Egts: I I would tell the the students I mentor, it's like get involved in an open source project. You you know, you don't have to be employed to make a a visible impact and contribution. And uh but when h how do you both look at it when you're hiring people of the like the fighting fire with fire of AI bots flooding job wrecks and then all of a sudden the the HR bot filtering the flood and and finding like what do you look for in quality people? How do you get involved? And and and and what you know what advice would you give to that early career student, that computer science student that is like how do I break into Mike McGrath: Yeah, David Egts: this? Mike McGrath: I I think some of those fundamentals are just as true today as they have always been. You can gain experience and you can network with people and meet people in open source in ways that you can't uh while you're job hunting. Um, I still think probably one of the best things I ever did back before I worked at Red Hat was join the Fedora infrastructure team. 00:37:09 Mike McGrath: I'm an operations guy by trade and uh being able to have being able to talk with like-minded people at any point in time about what they're doing. um when I run into an issue at work is you get a different context than you would if you opened up a support ticket or if you Googled on you know you get this experience from people that you know and trust and open source communities remain I think the best place to go especially if you're early career to meet these people and come up with new opportunities and and just learn and stay sharp on your skills. I think having said that, when it comes to hiring somebody new at this point, my stance is if anybody tells you they know what the world is going to look like in three or five years, they're selling you something. I have no idea where this is all going to end up or how. But the key thing is that you uh show a proven ability to learn new things uh to be adaptable to whatever your company might need or wherever you're going to go work and then to deliver on that adaptability uh which can be very stressful at times when we're in a time of change like this. 00:38:15 Mike McGrath: Um and uh I forget the exact numbers. I I tried to inspire my team uh at the beginning of last year with some stories about uh the carrier pigeon and how long it took us to go from uh horsedrawn delivery to carrier pigeon delivery to when the telegraph came out. Every time one of these disruptive technologies came out, it was shorter and shorter and we were you for thousands of years we were using horses. Um, and then the the car came out and it was only it was like less than a generation of people that then just adopted the cars. You flash forward to the iPhone coming out. That total disruption from the time the iPhone came out to the time people basically stopped having landlines commonly. It was eight years something like that. AI is going to be, you know, on the on the AI timeline. We're already halfway done with an with a a an iPhone level disruption and this is probably even bigger than that. We're probably already almost done with it uh to the point where it will be the norm. 00:39:19 Mike McGrath: And uh I think we all need to prepare ourselves, but especially those early career people. Not trying to bash anybody's uh uh I'm not trying to bash anybody's uh credentials or or your degree, but whatever you you know, let's say you just graduated. Four years ago is when you started that process. The world looked very different four years ago than it does today. David Egts: four years ago. Yeah. Mike McGrath: And so the fact is some of those things you learned may or may not be relevant anymore, but certainly four years from now it's going to be even more different. And so just rely on don't rely on what you've learned. Rely on how you've taught yourself to learn. And then uh I think that's that's going to be your best bet going Gunnar Hellekson: Yeah, that Mike, I agree with you. Mike McGrath: forward. Gunnar Hellekson: I think the um everything you said is true about being able to be agile and kind of move like water um as we continue to get disrupted is more important now. 00:40:15 Gunnar Hellekson: It has always been important and it is becoming kind of the defining characteristic I think that I'm that I'm I know I'm looking for. Um I know you and I have talked about this um at work, but we might as well talk about here as well. Um there's a big conversation going on, especially in our industry, about whether AI is going to favor the uh uh the experts, the 30-year veterans, or whether it's going to favor the kids, uh the the the new talent coming up. Um I know, do you have a do you have an opinion either way, or do you have a more nuanced view? Mike McGrath: Well, it's a it's a good question. I've se first of all, I've firsthand seen both be very successful at it. You know, if you've got a very big change adopter, they're they're ready to go get it. Um I think as with any major disruption, you know, for me, you know, I've got more gray in my beard than I did four, five years ago. Just how it is. 00:41:16 Mike McGrath: Um at 27, I've just come to accept. I'm kidding. Um, but uh I I think I think the big thing is we tend to get set in our ways. I've got I've got a really good example of this. You know, for me, I said I'm a home run guy. I'm a VI guy. Never really got into Emacs. I like the tools that I like and I can go do them very quickly. And oftentimes I can go do that do what I find what I need more quickly than if I used AI to do it. Uh Perry Myers is a is a guy on my on my team that has been doing a lot of this AI adjustment. And one of the things that he uh he pointed out to me that I hadn't even realized was happening with me was that when you when you set AI to go do a task, uh it does tend to use some of the newer tools and things. And so it was using commands that I did not even know existed that were newer commands because I came up the early 2000s on this. 00:42:11 Mike McGrath: I was a Linux slack or Linux Bible guy and uh and so you know if if if I were to have fought that and said no I'm good with vi I'll pipe this to SSH and DD and you know whatever I need to do to to make this back up um you know I would have been worse off but because I watched the way AI was doing it to help me you know build a virtual machine or go get some logs or this or that um you know now this is this really dates me but like I only recently learned to use T-Max. I've been using screen mostly because of muscle memory. And so, you know, I think that there isn't there is an amount of for us long-timers, David Egts: Mhm. Mike McGrath: there's a mount there's an amount of effort we have to spend in unlearning that some of their newer kids don't have to spend. David Egts: Mhm. Mike McGrath: And that's kind of painful because I liked screen. Screen was just fine. T-max is great. But uh you know I think that you know when it comes to who it favors I think being able to come in fresh is probably going to be easier for you than the cows you have to learn kind of unlearning and becoming comfortable with the uncomfortable 00:43:18 David Egts: And one one thing though you mentioned about, you know, getting involved in the Fedora infrastructure community and all that. How did that happen? And was that like, hey, I'm I'm looking like why did it happen? How did it happen? Were you in college? Was it you were working and it was an itch you wanted to scratch? And you know, just to to help for the listeners at early career of like breaking into open source, what how did you do it? Mike McGrath: Yeah, I uh I was working at a company that was using uh Fedora for all kinds of stuff and uh I hadn't I wouldn't say I had participated in any uh uh distros or had any major I probably opened a few bugs but I used a lot of open source and at some point in time I remember you we had a Fedora release and it was there was a big you know back then we called it the slash dot effect when you denial of services system and you can't get your downloads and I think in just trying to look into that I was like oh I figured out where the the that community hangs out I found their mailing list and I found their Slack channel and it just so happened as so many things do I just you know found myself with kind of a bored couple of weeks at work you just done a major release and it went well and that was fine. 00:44:31 Mike McGrath: So I I kind of showed up. I don't know for for the the listeners out there, Seth Vidal, may he rest in peace, uh was the first person I ever talked to in the Fedora community and uh I had volunteered to set up Dagios to monitor their systems because they didn't have a monitoring system set up and that's what I used. And he very critically said, "Well, why do we need this?" and he demanded uh he demanded I think about why and give him an answer and I did and they invited me in to kind of set up this monitoring system for them and uh that you know that's any little thing that you get into that's where it starts and Fedora at the time was uh you know mostly red haters but there were a few volunteers and uh prior to me getting hired on that entire group became almost entirely volunteers all the redheaders had gone on to do other things or this or And uh you know that happened to be one of those situations where without the right people at the right time and not just me but there were others back then. 00:45:31 Mike McGrath: I think that community would have really floundered for a bit. Um and there's I've seen countless examples of that over time. But that that was how I got started was just offering the set of Nagios for for them to monitor because they just had a rough release uh right around Fedora Core 3. David Egts: Nice. Mike McGrath: So, David Egts: And then then that uh how did it go from participating in that community to full-time at Red Hat? Mike McGrath: well, this is this is actually one of my favorite stories, so thank you for asking. I'll definitely share. David Egts: Yeah. Mike McGrath: Um, I think it was around Fedora Core 4 or five. It was before we had, you know, split, you know, Fedor and Extras were still separate from each other. Um, I was a volunteer and we did a release and it happens that we did that release on the same day that Red Hat needed to do their quarterly finish financials and the slash effect ever present basically locked Red Hat out of their own systems. We did not properly segregate the community uh tools from the enterprise tools. 00:46:35 Mike McGrath: It was all on the same data center on the same networks and rightly so. the chief uh I was still volunteer but rightly so the the CFO at the time Charlie Peters said guys this can never happen again so he demanded a meeting get set up and the Fedora project lead at the time was Max Spivac you're going to want Mike McGrath at that meeting because I was the Fedora project lead at the time and uh Charlie's assistant while attempting to set this up could not find me in the system and you know said hey uh Max we can't find Mike in Oh yeah no he's a volunteer to which apparently the the executive assistant said, "You want us to invite a community volunteer to a meeting with Charlie Peters about how we close our company books at the quarter David Egts: Only at Red Hat. Mike McGrath: end?" And uh uh David Egts: Only at Red Hat will you experience this. Mike McGrath: I think I'm assuming Max said, "Oh, actually when you put it that way, no, David Egts: Yes. Mike McGrath: I'll attend." And out of that meeting, 00:47:33 Mike McGrath: they funded a full-time job working on Fedora because at the time there weren't I don't think I think Max might have been the only full-time employee working on Fedora and I would have been number two. And they really didn't know what to do with me. I reported to the marketing department. It was not part of engineering at that time for various reasons. And uh my favorite part of this was uh I won't say who but uh somebody in it had reached out to me uh congratulating me for starting work at Red Hat and to never mention the fact that I both had a budget and access to the data center uh because they had thought I was a Red Hat employee already because I always used my Fedor project email address and I guess they just didn't they did and a lot of Redhatters did at that time and they just did not realize that uh did not realize then. So, yeah, you know, it was a different time that it was 19 years ago, much smaller company. You could get away with that kind of stuff. 00:48:26 Mike McGrath: Um, but, uh, you know, we've grown up since then. But, yeah, it was a very, very funny hiring process. And that was a a nice little exclamation point on it because I guess as a community, I was not supposed to have a a Red Hat budget to spend, but I did. We had servers, we needed networking equipment, and so I, you know, I went and spent it. David Egts: That was awesome. Gunnar Hellekson: as a classic Red Hat story. It's great. Mike McGrath: Yes. Gunnar Hellekson: All right. David Egts: Yeah. So, well, we covered a lot. Uh, and and so, Mike, if if people want to uh, you know, get a copy of the show notes, Mike McGrath: you. David Egts: get back to re replay episode 273 where where Gunner was riffing on abundance and and to learn more about you, the Fedora Project, and everything else. Where should we send Mike McGrath: You go to djshow.org. David Egts: them? Mike McGrath: Couldn't be easier. It's all right David Egts: Dan Walsh's homepage. Mike McGrath: there. David Egts: All right, Mike. Well, this was great. I I really appreciate it. I love the the historical aspects about this and then also uh providing a lot of insight of of somebody that has uh you know, really is made a massive career about, you know, starting off as, you know, volunteering in open source and uh and growing to to where you are now. this is such a wonderful success story and I appreciate you sharing that and also how you're using Mike McGrath: Of David Egts: AI to make yourself a better people leader too. So that that was some really good lessons learned there. So thanks for being on the show. Mike McGrath: course. I love the topic and really enjoyed my time. Thank you for having me. Gunnar Hellekson: Okay. David Egts: Yeah. Gunnar Hellekson: All right. Well, thanks, Mike. Mike McGrath: Thank you. Gunnar Hellekson: Thanks, David Egts: Yep. Bye. Mike McGrath: All right. Do I hang out now? Is that it? We're This editable transcript was computer generated and might contain errors. People can also change the text after it was created.