Track 1: [0:21] Hi, and welcome to Teaching Python Podcast. This is Kelly, and I'm a teacher who codes. Track 3: [0:27] And I'm Julian. I'm a coach who codes. Track 1: [0:32] And we got it right this time. And this is episode 159, Big Lessons from Small Models. And today we're here with Gwyneth Peña-Sequenza. I met her at PyCon Ed Summit. And I was just telling her briefly on the show when she was talking, her voice like draws you in. And I kept smiling while she was doing this talk. And I said, I must have looked like a weirdo. But I was sitting in the front row and I was like, oh, my gosh, I got to get her on the show. And that was the first thing I said to, I texted Jillian. I was like, remind me, we got to get this girl on the show. And so I'm really excited to have you here. Thank you for coming. Track 2: [1:07] Well, thank you for having me. Track 1: [1:09] Anytime. So you're a cloud advocate at Microsoft. Yes. A good friend of Pamela Fox, right? You guys kind of work together. Track 2: [1:18] I'd like to consider myself a friend of Pamela Fox, yes. Track 1: [1:22] Yeah, and then in your spare time, you build open source courseware, I read. Track 2: [1:28] Yes, I do. I do. Track 1: [1:30] Excellent. So we're going to talk all about that. But before we get in, we're going to start with where we normally start the show. I'm going to start with the wins of the week. This is something that is something that happened in work, at home, something in your life that just kind of made you smile and yay, we won. And we always start with the guests. So we're gonna let you go first. Track 2: [1:50] Yeah, I'm going to have to say that my win of the year might actually be the New York Knicks winning the, NBA finals after 50 something plus years going to the parade tomorrow. It's going to be awesome. And yeah, that energy is keeping me going. Track 1: [2:11] That's great. I think my son, I think, was not really that happy, but I don't know why. I don't really follow. He's normally a soccer person, so him flipping back and forth to this basketball stuff, I don't really get. So I was a volleyball player, so that's my sport. What about you, Julian? Good news of the week? Track 3: [2:30] Yeah, a couple of things. But now that you mentioned sports, Gwyneth, the World Cup, Australia actually won a match on Sunday, which was awesome. I saw that. Or Monday, whatever day it was. We were very happy. We were watching it at my sister's place. So that was a fantastic win. On a personal note, I'm pretty excited. This is the first place I'm saying it. I'll probably do a LinkedIn post and stuff. I closed a contract. I'm a fractional CTO at a small financial firm starting in another month or so. So I'm very excited to be jumping in there and supporting them. So that's the first of a few, but I'm pumped. There's a lot of good stuff coming. Yeah. Track 2: [3:13] That's awesome. Track 3: [3:13] Thank you. Track 2: [3:14] Congrats. Track 3: [3:14] Thank you. Track 1: [3:15] Congrats. Now you have to go back to really working again. Track 3: [3:18] So that's okay. Track 1: [3:20] Oh, yeah. Track 3: [3:23] All right. What about you, Kel? Track 1: [3:24] Well, I was trying to think. I was going to say, you know... Track 1: [3:31] It's summer, and I was doing something. I was not working. I'll claim last week. I didn't work at all last week. Last week was the first time in about a year and a half where I put an out-of-office, email, and everyone was shocked because during the summertime, I do all the class link apps, and then I. Track 1: [3:53] Make sure all the books are integrated, and I just said, I'm just going to take off for the whole week. I ended up working about five hours, which was impressive. But my boss, she texted me and she's like, I know I'm not supposed to text. I said, you're the one who told me to put the not out of office. Why are you texting me? Well, it's just a little emergency. I'm like, oh, my gosh. But it was really nice. We went to North Carolina and I actually wrote a post. I was trying not to think anything about work. I didn't even open my computer. And I'm sitting here on this riverbank in North Carolina watching my kids throw this ball. And then all of a sudden I was like, oh, this is like AI. It's like systems, everything that's hidden underneath. And there are these currents. And I was just like, oh, my God, stop it, stop it, stop it. But it was cool just to kind of like shut down. And so I was really excited. And then last week we had that little hiccup where we couldn't meet with Quintinth. And I was going to tell you all about my thoughts on a river and how it equates to a small, you know, language model. But I'll let that slide. I was very nerding out. I'm sorry. I make these big connections that probably make no sense to anyone else, but... Track 3: [5:03] That's a good win. This is what happens when you switch off. You wouldn't have done that if you didn't have that out of office. Yeah. Track 1: [5:09] Yeah, probably not. Probably not. So anyways, I want to go real quick. I would just jump in because a lot of people didn't get to hear about the Ed Summit. We haven't released the videos yet. And I want you to do like a quick summary. And the reason why I was texting Julie and saying we're getting got to get you on the show. I want you to kind of summarize what you were talking about. I think it's really important for teachers to hear about the small language models and kind of self-contained issue of them or not issue, but benefit of them. Track 2: [5:41] Yeah. Issue and benefit in one. Yeah. So over the last year with Pamela Fox, my colleague on our Python developer advocacy team, we have been doing a lot of education around various generative AI topics. And we, of course, are very committed to making sure these resources are accessible to as many people as possible. And in the era of AI, that means access to the model that runs and that we'll interact with. Right. And it's it's tricky because especially now with price changes happening, you know, perhaps, you know, a hundred dollar subscription for someone is something simple. But for someone else, that is like not possible out of like you wouldn't even imagine to be able to to do something like that. So we. Track 2: [6:40] Did a couple of things with our resources so they would work as much as possible, with SLM's small language models. When I say small language models, I mean something that would run on like a, pretty standard laptop. I'm thinking like 16 gigs of RAM, you know, like 500 gigs of storage and nothing. Something you could buy off the shelf at like your Best Buy or something like that. And yeah, we did some things. So our models or the examples would work with models that you could run locally. So SLMs with models that are running in GitHub models, models that are running in OpenAI, on Azure AI and things like that. So basically, I just kind of did a quick summary on what we did, what the challenges were, where the lessons were. And overall, I think it was a pretty, I think it was well received. And I think a couple of people took some things into consideration for their teaching resources and things like that. Track 1: [7:38] Yeah, I, well, I'll be honest, I started on Pamela's October, you guys, the week thing that you were doing. And it took me about four days to get to through two of them. Don't stop, pause, rewind, write down. I like to put things up on my wall. But you know when you were speaking about it I was thinking about two of our teachers that are in school and they're starting an AI class for our upper school students, and they wanted to go big you know they I have this really powerful machine in my classroom and we have it set up with olama and all that good stuff with models and they were just kind of like, didn't know where to go. And I kept saying, just put it on a Raspberry Pi. Let's just put one of these small models on a Raspberry Pi, just to kind of start digging into the parameters and seeing what's wrong and seeing how you have to really think about the words that you're putting in. And as you were speaking, I was just like, yes, yes, I'm sending all this stuff. Because I think there's a lot to learn from a model that's not so powerful. We don't learn that much when when the model does everything for you. And that's kind of what I was taking away from. Track 2: [8:55] Yeah, the constraints are the lessons. From the prompt engineering, the context engineering that you get out of the habit of doing to an extent with the larger models, you really got to know your stuff when you're working with the smaller ones. Track 3: [9:09] Yeah, so I was just going to say, I've been playing with some agent stuff at home and I've been hitting hardware limits with the GPU that I have. And when you wrap the LLM with the agent, I run out of RAM very quickly and I hit that exact issue that you're talking about like three days ago where I went to a smaller model and suddenly the responses I got were very, very plain compared to what we're used to. So it's going to sound like an odd question, but do you feel like there's, we've almost given too much to the larger models like chat GPT 5.5 and everything like that, that we kind of have to bring ourselves back again to where it was a couple of years ago, where we really have to think about what we're asking. I didn't even catch your talk, by the way, because I wasn't at the conference and haven't seen the recording. So yeah, any insight into that? Track 2: [10:07] Yeah, I like to think of it in the way that I learned to drive. So I learned to drive in this 1970 something really old pickup truck. It was manual transmission. And it had literally it was it was like the car we had on the farm. And I had the worst teacher. I love my dad, but he's not the most patient teacher. So you know i am like learning to drive and in this really terrible car that would stall all the time and you have to do like this technique where you have your foot on the clutch and on the brake because he would make me stop, as we're going uphill so i could take off uphill so i'd have to do the little foot on the clutch and the brake and anything and then he taught me how to like kickstart the car in reverse. So anyway, I learned it was a bit of a traumatic experience, but I learned a lot. And when I think about, you know, learning with SLMs, it reminds me of very much learning how to drive. Track 2: [11:09] In this really bad car, very limited car. It was still a car and I was very happy that I had access to it and that I could use it. But there were so many challenges that like a perhaps more modern, like automatic car, like I would never have experienced. So now I know I live in New York City. I can pretty much drive anything anywhere I want. I've driven in various different countries. Like there's no, Like I could probably drive anything that I wanted to. Nothing really scares me when it comes to that type of stuff. And it's only because I learned through this same, this, you know, through this really bad car experience. And I think of that with like the experiences that we have now with like the Opus models, the GPT 5.5 models, like all these things like where you can say something very basic and it will make the assumptions and oftentimes correct assumptions of what you want. And then we were just getting into that habit of like, oh, the model is powerful enough. I can sort of abstract what I'm trying to say, what I'm trying to do, my skill set, and just kind of offload that to the model. Whereas like with an SLM, if you were to start learning with an SLM, you'll be building like some pretty good fundamentals from the start. So later on, you're going to get even more out of like the bigger, more powerful models. And I just thought it was just like the most hilarious analogy that my life has like turned into like, oh, it's the truck again. Like it won't leave me alone. Track 2: [12:35] Yeah, thanks to my dad. He knew, he knew in 15 years, you're going to need this analogy. Track 1: [12:43] Isn't that something? Some of our greatest teachers actually make us do the worst things to make us learn the biggest lessons, no? Track 2: [12:51] Yeah. Track 1: [12:52] I think that's something. I've been working on this talk. I'll just go ahead and say it because this is the only thing that's in my mind right now and I'm trying to clear it. I've been working on this talk that I have next week in D.C. I'm the last person of the conference right before they hand out lunch. Track 3: [13:09] Tough spot. Track 1: [13:10] Worst spot ever. On top of it, I'm the only teacher, With, you know, a majority of them being lawyers and business people and government officials. And I'm coming in to share what I have about AI. And all I can think of is y'all are screwed. I think these kids won't have that struggle, will not be driving the manual cars there. You know, we have four year olds who are don't really know that they're using AI, but they're already using AI. So I'm thinking, they're going to get up there in this work in five years, and they've never experienced maybe these small language models. And I just keep thinking back to, I don't know about you, Julian, I never asked you this, but I was reading about your past and how you were self-taught, Gwyneth. And like, I'm self-taught, and we went through this whole struggle of trying to learn and that that pain of processing and taking longer like what are your thoughts of how that's helped you, adjust in this ai era and i think that's like a big question for everyone you know. Track 2: [14:27] Yeah and i don't mean to get like this deeper reflective but when i think back at my. Track 1: [14:33] Get reflective i love that i get too meta and i think julian is like very very meta so bring it on yeah. Track 2: [14:41] It's it's almost that um everything that made my life very challenging is what is helping me accelerate in this new era so like I mentioned I so I'm born here in the U.S. In Connecticut and then when I was around nine I moved to Ecuador we lived on a farm there was there was electricity and a phone like landline but no and no internet no cell service, no TV. And this is 20... Like 2008 so you know these things were available it's just the area that i happened to live in and my parents wanted, my brother and i to have this life of like we want you to have these contrasts we want you to have lived the very comfortable life and we want you to go live i don't know why you know and also when you're like nine years old you're like why what the heck is happening right so anyway. Track 2: [15:30] So you know i go through you know middle school high school and, you know, when you're a teenager, you get very like, oh, what am I going to do with my life? So I had some interest in engineering from my dad, but I didn't really know. So I would just like download, like I would stay after school at the library and download like playlists of YouTube videos like Java programming. And when I would go home, that's all I could watch. Like I couldn't watch any other videos. So it was almost like I was putting like a screen time limit on myself, or like blocking you know sometimes now you have like browser plugins to like make youtube less distracting and stuff like that but i i had that like naturally like built into, into how i was living so i learned from that to really consume content like over and over and over and over again because that's all i could do so now, you know through my self-taught journey it's just the way I just learn like I, will limit myself to a book a course or like a task and just over and over again just because that was kind of like was built into me from a young age, and I think it's so helpful now because you know we have clawed code we have chat gpt we have so many ways to customize our learning journey but that also gives us the opportunity to switch all the time and. Track 2: [16:55] You know this as you know as teachers as educators like the most important thing is like showing up a little like consistency and not changing a lot and like following a curriculum following a plan but, it's thanks to like that discipline that was like forced into me that now and i can just stick through things and i can figure okay it's not working this problem's not working all right let's try this let's try that and yeah like having those limitations kind of built in has like helped me a lot, and especially now when we have infinite possibilities and choices. Track 1: [17:27] Yeah, that's a problem for me. Track 3: [17:29] Do you find it? Like, it's interesting, right? I love the amount of access we have to these technologies and this ability to learn from pretty much anywhere. It's awesome, right? You've got YouTube, you've got podcasts, you've got ChatGPT or LLMs, you've got everything. Do you find that in your experience and what you've seen, do you find that it affects attention and focus? Because to go from one YouTube video that you might watch, say, I don't know, while you're in the shower, you might be listening to a podcast or watching a YouTube video, and then you'll come out and you'll be like, oh, I'm going to context switch to learning about this topic through talking with an LLM. Or do you find that that complements each other having these different forms of learning? Track 2: [18:13] I think there's no one perfect solution for like a variety of people. But I also think depending on what you're trying to accomplish, there's no perfect solution either. And if you are honest with yourself and you kind of like you watch a video or whatever and you ask yourself, am I actively and intentionally consuming this? Then I don't really think there's any medium that's incorrect. But if you are just listening to something in the background and it's just kind of like distracting you, then, you know, you know, you're not doing yourself a service to do that. Right. And I think it's quite awesome now. Like I can I so two books that I did another talk at PyCon about, you know, kind of just like my learning journey in Python last year and two books that I love. And I have them like right here. So Fluent Python. Track 2: [19:07] Freaking love that book. Oh, my goodness. And then Effective Python. And the reason I love these two books, and I keep them very close to me, like if sometimes I travel, and they're massive books, but sometimes I travel to conferences and I take them with me because they are shaped in ways that work so well to learn with all the tools that we have now. So, for example, Effective Python is chapters, but also each chapter has various items. So, you know, an item will be two to three pages. And from that, you can spend hours on that. You are not you're in that item, then you're building it out. And then you, you know, perhaps you open up, I don't know, GitHub Copilot and you're like, OK, I'm building this out. Further explain this line, further explain this line on top of all the information I already got from the book. So I think now more than ever, what I'm looking for is, all right, how do I learn already? What am I using already and how am I sort of amplifying that? But sticking, trying to stick within the same thing, not like jumping here to the next one to the next one. And I think that's been working actually pretty well for me. Track 1: [20:09] I would love to just stay in one thing. I don't, I feel like, I feel like I'm constantly droning as an educator. I was like, what's the next new thing that's coming out that I have to train somebody on? And because I train teachers as well. It's just insane. And this, this question's been like itching. This is what I've been itching to ask you. And Sean would probably be so upset that he wasn't here to explain it all. So you do cloud stuff, right? Learning, like learn the cloud or you did or what? Tell me a little bit about Track 1: [20:43] that before I ask my question and sound stupid. Track 2: [20:45] Yeah, so my background is in cloud engineering operations, more of like the DevOps side of things when we think of like software and then, you know, ops, software and ops, right? But throughout my career at Microsoft, five years now, I've been moving across, Linux, cloud engineering, operations, .NET, and most recently Python. And, you know, who knows where I'll end up in. So I kind of like to think of myself as more of a technical storyteller, and I apply that to whatever topic my team happens to be focused on. Track 1: [21:14] Okay. So I'll go with this. So if, I was a teacher who was starting to teach like small language models a little bit and I wanted to bring in a little bit of cloud operations in the back, say, say we wanted to talk to the kids about deploying something or making something. At school that's contained, what would be a good place to start or teach us? Go ahead. Just tell me. Track 2: [21:42] Yeah, for sure. So there are ways that you can host SLMs on different clouds and like a variety of different options that you can use. So I think that would be an interesting experience because the most important part of cloud engineering is never necessarily what you are deploying, what you are hosting, but everything that goes around it. So are you selecting the correct type of architecture? You know, we have platform as a service, software as a service, we have infrastructure as a service. And, you know, you need to understand like budgets and alerts and perhaps some networking. So with any project that you have in mind to kind of deploy into any cloud, there's plenty of areas around it that you can, you know, really teach. And I think a lot of students and teachers should understand. Like, okay, I have this project. I have this code base. I have this thing. I want to be able to deploy it into a container. Awesome. This is how I could do it. I want to deploy it into a virtual machine. Awesome. This is how I can do it. I want to deploy it into a serverless function. You don't have to go crazy deep, but understanding the different types of compute and the different options that exist out there, Especially now where you can create, there's a lot of creating happening now. There's a lot of building. There's not a lot of deploying because the cloud and operations stuff. Track 2: [22:59] Is harder. And I wouldn't say it's harder than software development. It's just harder in terms of agents don't have as much context and have not been trained as much just because that information does not exist. Like there's so much text content on Python, on programming out there that agents have been trained on, that models have been trained on. And there's just less of that. Traditionally from like the operations side. So there's just so much more manual stuff that needs to be done. But yeah, taking your code base from, okay, it runs on my machine to actually, okay, how do I deploy it to the cloud? I think there are endless learning opportunities there. Track 1: [23:33] I think it's a good idea that that doesn't, hasn't been trained that much though on the cloud, right? Because that's one of the security, right? Because we were talking about this and I was trying to explain to this very eager kid that just graduated. I was like, yeah, we have... If we have a model, if we have something running on our computer, and it's only on our computer, and we're not on the internet, and we're containing it, you know, it's safe. It's private to an extent, right? But once we start connecting and deploying and serving it to a cloud, there becomes more risks. And there's a lot of things that we have to consider. So, because part of my job is to tell everybody no. It's like no that's not safe no that's not private no that model is not good um and so that was one of the things and i and i think i think it's a good thing that, ai doesn't have a lot of knowledge yet there's no question that sorry i'm processing no no it's really interesting because. Track 3: [24:38] I was building for that exact sake of how do i build a model and an agent and everything that stays fully self-contained. And as I have these conversations with clients and with people, the question always comes up about security and, well, what about prompt injection? What about it just going haywire and deleting your emails for that famous story, right? So where in the learning process would you suggest people start considering that? And I preface this by saying that I know some people who don't even want to get started with learning because they're intimidated by the security layer or at least that safety layer of going, oh, this is just going to ruin everything I have if I install this on my computer. So what do you think about that? Track 2: [25:28] Mm-hmm. Yeah, it's funny because when that whole, was it Claude? No, what is it? Claudebot? Track 1: [25:38] Yeah, Claudebot. Track 2: [25:40] Open Claw, yeah. Open Claw is the name now, right? And it was Claudebot, yeah. When it came out, I read about it and I was like, I don't think I will ever want to try this. I just think the first level of security and even for people learning is just a little bit of common sense, just a little bit of trust your gut. If you don't feel like you want to use this or should use it, you don't necessarily have to. And to this day, and maybe I'll get in trouble for saying this, I have not tried OpenClaw. You know, I've not, it's not one thing that has interest me. There are plenty of other areas that I can focus on. But yeah, just running an agent that has full access to absolutely everything you want doesn't sound like a good idea because it's not a good idea, right? So like, you know, common sense is always like the first level there. Track 2: [26:33] But this is the issue with security and why security will always be... Track 2: [26:39] A good area to be in is that it's always the afterthought it's never built in and to an extent sort of cloud engineering and DevOps best practices are always after that like software engineering has always had the prestige and always been like the software devs need to be able to do everything they want get out of their way, and that's why you know the ops has always been like MacGyvering and band-aiding around like you know we have this, I swear like when I was a sysadmin we had this virtual machine that was like windows 2006 r2, and it was just throw more ram at it throw more memory of it but please don't ever restart it and it had all sorts of errors that we needed to upgrade it and but no you can't because the the i think, it was like the data team needs access to it 24 7 and that machine is probably still running there's probably still throwing ram and things at it but like that's the reality of the ops and like the, non-software and side it's like everything else is an afterthought but i think now with ai where we're it's just accelerated towards the vulnerabilities towards the issues that we have so we have to push all of those things, more up front more immediate we need more people who are going into cloud engineering we need more people who are in security like and i think we're starting to see a little bit more of that and more like it's more widespread it's like oh we need better people educated in these areas as soon as possible. And as soon as you start something. Track 2: [28:02] You need to think about security. Like we have in our educational resources made with like Pamela, we're always trying to push for, you know, safe practices with tokens, safe practices for like, we try to use keyless auth when possible. And, you know, try to, even in documentation, if we're doing something that's insecure, we, you know, make note of it. Things like that can get you very far. But like the biggest thing is like, the minute you think of doing something with AI, think of how to do it securely as well. Yeah. Track 1: [28:31] Yeah, it was funny because I have been on a major learning journey about security. I didn't even know what OWASP was. I mean, I don't even know what it stands for. I just know it's risks. I'm sure you can tell me what OWASP is. So I learned about all these things and I know all the models and where the subprocessors are and what country they're in and where the pipelines are. And my favorite word that I hate saying is URL obfuscation. It doesn't roll off of my... I can never say it. I'm like, you know, that one where the URL comes in and it hacks us. Can you say the word? Do you know what word I'm talking about? No. Track 2: [29:12] Obfuscation? There's an F in there. Track 1: [29:14] Isn't it? Track 2: [29:14] Obfuscation? Track 1: [29:15] It's an O-B-F-U. Track 3: [29:17] Excuse. Excuse you. Track 1: [29:18] I know the word. Track 2: [29:19] Excuse me. Track 1: [29:21] Family-friendly podcast. Yeah, the word's horrible. But there are a lot of risks. And I think that's what makes... I think that's what makes everything about small language models more attractive to me. I think getting them into the schools is really cool. And I love the fact that I always say to the kids, oh, and you know what? The reason why I talk about AI so much is because it was made, guess what language it was made? It's like Python. Oh, see, that's how we learned it. But I love that you can, get everything running in just a short amount of lines. And then that's where, like, you get to learn everything about personas and temperatures and token and tokenization. And that's been a fun journey. That's what started me. And that's what I started with in August. And then I just went crazy. But anyways, I want to hear about your Python journey real quick. Track 1: [30:16] How's that going? You've been in it for a year now? Track 2: [30:19] Yeah, I joined this team January of last year. So, like, a year and a half. And our team is so stacked. Like we have some of like some pretty big people in the community, but also people who do a lot for the community and have been doing for years. And I'm never one to to doubt myself, but I'm always very like aware of, OK, I know where I stand. I know what I need to do. So I just like, you know, leveraged my my colleagues. Hey, what resources do you recommend books? And then I locked in on those two books that I mentioned. Track 1: [30:51] And you never did Pie by Goody Bites? Track 3: [30:54] Oh, come on. Track 2: [30:54] No. Track 3: [30:54] Come on. Track 1: [30:59] Julian tells really dumb dad jokes Oh yeah. Track 3: [31:01] You go through those exercises There's some really lame jokes and it's great. Track 2: [31:07] I'll have to add that to the list. I love a dad joke. Track 3: [31:09] We're going to get along. Track 2: [31:10] Sounds like my cup of tea there. Yeah. The biggest thing has been just working with a colleague that really forces you to get on. For me, that's Pamela. I was a great mentor in my career. But Pamela runs at like 100 miles an hour all the time. Like, let's do this. All right, I finished this. Now let's do that. So at the beginning of like last year, I was barely being able to like contribute. I'm like, okay, I could do this, this task. And then towards the end, it was so cool to like reflect on my own progress. And now, yeah, Pamela is fully like assigning tasks to me and like not only from like a documentation or consecration, but like from a technical perspective, like, okay, yeah, I can trust Gwen to do this. And yeah, it was really cool to kind of see my growth be in, you know, in the perspective of like her mentorship as well. So, yeah, it's been a good, very difficult, but good past year for me. Track 3: [32:04] Was there... Track 1: [32:05] I don't know if you've... Go ahead. Track 3: [32:07] I was just going to say, was there any... So, were you learning Python from scratch like you hadn't used it before? Track 2: [32:14] No. I used a lot of Python when I was a cloud engineer, but from, like, the classic scripting perspective, like, basically, when Bash couldn't get the job done, you know, Python. But it was never my primary language. which I wasn't doing any object-oriented stuff, quote, object-oriented Python. Track 3: [32:31] So was there any particular aspect of Python? Just out of curiosity, I love asking this question, is there anything in particular that you sat there and said, oh, hell no, I don't like this, or you struggled to learn it, or it just didn't make sense. It was the harder, more friction-filled piece of Python. Is there anything like that? Track 2: [32:50] Well, a lot of the like Pythonic stuff at the beginning, I was like, what the heck is Pythonic? And like, list comprehensions. And I'm like, why? You fit all of this in two lines. Why are you doing this? But then, you know, again, to credit, I think it was Fluent Python that really helped me with this comprehension. And I was like, OK, this makes sense. OK, yes, this is beautiful. But a lot of time when you look at it, like the first, like I had learned pretty much how to program back when I was in like a cloud engineering role. We had a bunch of like .NET super senior devs and I would just volunteer to help on projects. And I kind of gained enough rapport with them where they would be like, yeah, sure, you know, I'll teach you. So it was great. So I learned, but it was .NET, very different from Python. So I had the basics. I knew. So I pretty much could like read code and understand what was going on. But there was a lot of Python stuff for like all the sorts of comprehensions that exist. Oh, my goodness. That was that was probably the biggest thing for me. I was like, what is going on here? Track 1: [33:49] I love that. It's because every AI always brings out the list comprehension. So you can always trap a kid when they've used AI because it's it's like 4x in X. If if X is doing X and it will return X. And I'm just like, what is this saying? Read it back to me. I don't know. I was like, I don't even know what that means. Why don't you tell it to put it in simple for loops for you? But yeah, that never really came clear to me. I never taught them that because I feel just getting them to understand what's happening with the nested loop was... Was big enough. Just getting them to actually want to type out the nested loops now is like the biggest concern. All right. Anything else before I switch gears again? Track 3: [34:37] No, no, no. I'm good. I'm just, I love this. And I just wanted to say you were on the same, you and Anthony Shaw were on the same team, weren't you? Then. Okay. Track 2: [34:47] Yeah, we were. Anthony moved recently, I believe, but we were on the same team. Anthony did a great job at like tossing me tasks like, Gwen, do this. Gwen, do that. So he threw you under the bus, is what you're saying. Track 3: [35:00] Okay, that's good to know. Yeah, pretty much. I'm going to make him listen to this timestamp. He lives down the street from me. He's one of my best friends. Track 2: [35:10] Oh, cool, cool. Track 3: [35:11] I will. He texted me. We're interviewing you today. And he said, oh, say hi from me. So I'll fit it in. There you go. All right, Kel. Track 2: [35:19] All right. Well, we can list off a couple of negative traits of Anthony right now. Track 1: [35:24] Oh, please, because I think he's the smartest guy in the whole wide world. And every time I talk to him, I feel dumb, especially when he came and told me how he taught his daughter about binary code and everyone needs to learn about binary. And I'm like, why? Why? That's why I never got into computer science in the beginning, because people like you. He's like, no, but then you turn on the light and it pushes it. And I'm just like, why? So, yeah, go ahead. Tell me one bad thing about Anthony, because next time I see him, I didn't see him at PyCon this year. He didn't come, I don't think. Track 2: [35:55] Yeah. He has this way of talking to me where... I would say like, yeah, I'm working on this. And he's such, he thinks a lot before we're talking. And, you know, it could be like maybe 30 seconds, but he's, you know, thinking. And then in those 30 seconds, I'm like questioning my entire life. I'm like, oh my God, what did I say? Did I do this wrong? And then he'll just be like, yeah, it's good. Why don't you try? Like, you know, the most casual thing ever. So that's my only thing, my only thing of Anthony that drives me crazy. Track 3: [36:32] Yeah you hear that too thoughtful let's see. Track 1: [36:36] Maybe we should just have another episode of just like love love bashing anthony the roasting put on netflix. Track 3: [36:43] The roasting of. Track 1: [36:44] Anthony p shaw yeah anyway sorry okay i'm not gonna do that all right sorry we'll digress i mean lifeguard amazing coder come on, nice guy saves people's lives saves people's lives all right yeah, so tell us about Well, I want you guys, I want to give a little, I don't know if you're still working on that project, but you wrote, I read it somewhere, the Learn to Code platform was your kind of process for everybody of your self-taught, motion, and then you, open-sourced it for everyone. Track 1: [37:18] So give a little spiel on that for people who want to learn about it. Track 2: [37:22] When I joined Microsoft, it was a very big, big moment for me. I think, you know, with a non-traditional background and path into tech, you, anyone who goes through it, kind of doubts of like where you could go, what you could do, what your limits are. So when I joined five years ago, I knew that I wanted to be able to sort of open source all this knowledge that I could as much as possible and reach the masses. So I wrote a markdown file. It was just like a readmeum. Track 2: [37:53] And then I was writing, I realized, oh, I'm going to need a lot more than I could read me. So then I just kept writing and writing and building and building. And until this, I kind of update it all the time. But what it's turned into now is a, well, we fully moved to like a fully like Python platform in the beginning of this year. Track 2: [38:17] And we now have the platform itself, which has the curriculum. People will go, it's divided into phases. So we start with Linux, then networking, then programming, then cloud platform, DevOps security. So I think it's five phases total. And we have curriculum people go through and then we have a verification system. So we have hands-on labs. People need to work through them. And there's ways that they can get tokens that are issued to their GitHub username. They plug that into the platform and it verifies it. Or we also have some AI working on some of the steps as well that verifies code quality or did you follow the exact process that we told you? So, yeah, it's a way for me to practice the stuff that I'm learning with AI, but also kind of make this platform available for other people. And it's free. And I pay fully for all the hosting stuff and obviously all the time that I put into developing it. I do it alongside one of my closest friends. His name is Rishab Kumar. He works at Twilio at the moment and also has like a self-taught, non-traditional background into tech. And yeah, I get messages all the time like, thank you. I landed this job. Thank you. Like, I can't afford like these expensive boot camps or anything like that. But like this really helps me. And like every week now. And yeah, it's been the best thing I've ever done in my career. And when I retire eventually, it's most likely going to be the best thing that I have left as my goodbye to the field. Yeah you had to put like. Track 1: [39:45] A patreon on it you know buy me a coffee yeah. Track 2: [39:48] Okay yeah i should i sometimes i do feel like because last year we did like, i don't know 100 000 uh site views all organic like we don't pay for anything like if you if you search learn cloud computing i think we're one of the first things that pop up after the sponsored ones it's all organic literally but we don't we don't really do anything have. Track 1: [40:08] You checked your. Track 2: [40:08] Ai ratings on it no i should i I should, though. Track 1: [40:13] I bring that up. Michael Kennedy showed that to me at PyCon. I keep telling every time I go on the show, the AI product rankings. We'll see if you hit the top. I got to check that out. But I think that's great. I think it's kind of like my LinkedIn now. It's been a journey of where I started with Python as well because I'm self-taught. I'm a biology person. I hated computer science. And I typed the DOS things a long time ago, but I just copied from the book and I had no idea what I was doing. And every time we would talk about this with Sean, he's like a CMU guy, totally geeky. Him and Julian can nerd out on Star Wars or Star Trek or something. Track 3: [40:58] Listen, you've offended me in three different ways and you don't even know how. Track 1: [41:07] But I think there is like something about, you know, seeing where you were, you know, looking back and having that process and being able to help other people because there's a different way of teaching when you're self-taught. And I see that a lot between a computer science major person who teaches, and they go in and they're like, here's everything you need to learn about list, comprehension, and everything's in here. We're going to spend two months on list comprehension, and you might write a program that has six lines of code. Track 2: [41:41] Mm-hmm. Track 1: [41:43] And it's just like, no, that's not how people really learn. So I do have a small appreciation for self-taught people. Track 1: [41:51] Anything you want to ask? Track 3: [41:52] Oh, go ahead, Julian. But this also comes back to, Gwyneth, your experience. It wasn't just, I mean, you put in the hard yards and you built and you learned and you wrote about it and you were growing. But you also had this amazing network of people. You didn't learn in a silo by yourself. You opened yourself up to mentorship with Pamela. And I mean, you were stuck with Anthony, but you still made the most of the situation, right? And I think when you learn with other people, learning all of those things, like you were saying, Kel, it's not just, hey, here's everything you need to know about list comprehensions. If you're learning with people who have that experience in a community, you get that exposure to things like, hey, here's a list comprehension and when you would use it. And here is where I used it and it caused a problem in readability because the, people in the code base didn't understand it or something like that. So really adding to what you're saying, I think self-taught with people around you as well is super important too. Yeah, I had Sean. Track 1: [42:59] He left me. He's still around. Track 3: [43:01] Come on. Track 1: [43:02] I'm just kidding. Go ahead. It's like a Wednesday night. Sorry. Go ahead, Gwyneth. Track 2: [43:10] No I was I was just gonna say like you know reflecting on learning with AI it feels I always feel like I'm kind of doubting it but whenever I ask my colleague hey what about this and they reply I'm like yeah I'm so sure about that and it just, makes learning with people learning from people learning from you know people who spend their times to put together like again these like books that i cherish so much, more important than ever more like the community aspect is is you know stuff like pi con is like so so important and becoming more and more as we have even though we have more and more tools. Track 1: [43:49] Yeah yeah that's true yeah, sorry i got stuck in a thought hey what what do you guys know because you make me go down this thought, and then I'm going to try to formulate it into a question when it's not even formulated in my brain. What do you think is going to happen to kids who have this access? Answer my end of my talk for these business people that I have no clue what I'm going to say. What do you think is going to happen? If you have a four-year-old starting to use AI, every answer is there, no struggle, no self-taught. What do you wish or hope or predict is going to happen? Track 2: [44:26] Yeah, so I have a 13-year-old sister, and I am the person in charge of controlling all sorts of tech life that she has. I've been assigned that from my parents. And I think, you know, I don't want to tell anybody how to, you know, do what they want to do, but... I feel like I have to put in so much effort to really get her to learn things and sort of introduce struggle. Like we have, you know, stuff on screen times that are hard. And every day I get a Snapchat, hey, can I get more screen time? And then I ask, why do you need more screen time? And sometimes it's like, hey, I need to do this paper. Okay, cool. But sometimes it's just like, oh, I was snapping. I'm like, no, you're done. There's no. And it's hard. It's really hard. It is so hard to be the person who enforces that. But I think more than ever, we need to be as guardians, as parents, whatever, informed ourselves on what these things are doing. That way we can understand even a little bit what it could be doing to kids who are growing up natively. You're born and you have chat gpt is such an insane thought like we have to be as hands-on as possible i think another thing like my. Track 2: [45:51] I never did well in school which like i i tried doing college and i and i dropped out but i think my one of my biggest issues with schools is their lack of flexibility to reshaping a curriculum depending on you know what happens to be, different right there's there's still probably a lot of computer science curriculums out there that don't include anything with cloud engineering, even though cloud has been a thing for over a decade, right? Almost close to two. So we need to inform ourselves, educate ourselves, and, be the, I don't want to, the mean parents. I don't want to say mean as in like, you're actually mean, but you know, kids will see like, you're so mean. Like we have to like, we have to force that as, as teachers and as, as parents, like more than, more than ever, I think so. And be active in their, in their like part in their journey. Like my sister really loves playing violin. So then, you know, I'm researching violins with her. I'm like, okay, what if we try this? What if we do that? Like, and, and I have no interest in violin. Like, to be honest, most of the time when we're talking about violins or she's showing me the stuff she's doing. I'm sure there's plenty of other things that I could do, but I got to be there for her. I got to help her out with the stuff that she does have interest in. And then maybe there is some kind of AI aspect to it that would be interesting to involve. Is this an app or something that could help her? Like, yeah, let's take a look at that. But I got to be active in it. Track 1: [47:09] I have a really fun thing you guys could do. It's called EarSketch. You can have her play her violin and you can use code to remix it. Track 2: [47:17] Oh, that's cool. Track 1: [47:18] Yep. There you go. You're going to practice your Python. Track 2: [47:22] Talk to other people who are going through this. You know, we're never going through stuff alone. And then maybe you can get ideas like this. Track 1: [47:28] Anytime. Anytime. I got an idea for everyone. I just need to follow suit and be tough on my kid as my son's playing PlayStation while I record a podcast. Track 3: [47:36] You know what, though? I think you're bang on. And the whole attention thing, I think saying I didn't know is not an excuse anymore or I don't know this. One of the things I always hear at school and from a bunch of other parents, is just, oh, yeah, I'm not a computer person. I don't know these things. I'm not good with tech. And I'm like, you can't say that anymore. So I really like that recommendation. And if I was to suggest you, you gave me an idea. What if we wrapped screen time extension around problem solving? So you want that extension for YouTube? You're going to have to code me an app, You know, no vibe coding. Track 2: [48:15] You're going to have to. Track 1: [48:17] That's a definite way to get people loving how to code. Track 2: [48:21] Give them a list comprehension. You want 15 minutes? What is this list comprehension? Track 3: [48:26] Well, my son, my 11-year-old, it was his birthday the other day. He asked me for, I don't know what he asked me for last night. He wanted to play, oh, no, he wanted to watch some Netflix after dinner. I said, go and cure cancer, then come back to me. And he just walked out of the room, went and asked his mother. Track 1: [48:46] That's funny any questions that you want to ask us or anything. Track 2: [48:51] Yeah of course i mean i would love to know when it comes to ai what worries you the most from any aspect do. Track 3: [48:59] You want to go kill. Track 1: [49:02] Go you go first go ahead because i i don't want to sound like a doomsday i think it's very easy to. Track 3: [49:07] Sound like doomsday. Track 1: [49:08] I don't. Track 3: [49:09] Know i think look the easy one is just the erosion of skills, right? People relying too much on it. But for me, I actually think it's more of an, societal existential thing where people are becoming too reliant on LLMs for human connection. I think that's what worries me the most. I mean, we'll always have people who have the deep skills. We'll have people who have just a deep interest in topics that they'll just inherently want to know the nuts and bolts of how things work, right? Just like mechanics for a car, right? I don't know how to do everything in my car but i got a mechanic for that so i think that's okay, i don't know i just if there's a worry about ai for me it's just people becoming too reliant people choosing it over social connection because it's just easier, so that's not so much of a technical cody thing but more of a i don't know relationship, human humanity thing. Track 1: [50:09] Yeah, mine's a trifold. Dependency, capability, and judgment. And I kind of sum it up, and this is something that I'm talking about. It's like, you know, you have the one kid, everything's outsourced AI, whatever. That's the kid that was probably cheating or didn't want to learn, you know, didn't want to learn that thing. Not saying they didn't want to learn something, but didn't want to learn that thing. They outsource it. And then you have the kid that doesn't want to use it at all because they don't really need it. They're already a great writer. They love it. But then you have, and I'll use me as an example, I think it's me, where I'm doing things I never could have done before. I mean, I probably could have, but I couldn't do it like I'm doing it now. The fact learning so much about security and I'm watching videos and I'm, you know, using AI to get more websites and then double checking against like 12 different AIs, verifying these things. And I feel that, you remember that one day, November 25th, when it like cloud fair went down and I was like, I can't work. I mean, perplexity is gone. ChatGVT is gone. And I was just like, what the heck am I going to do? I have to do a presentation. I don't have a video and I need a picture and I need this. That's my biggest fear. Yeah. Track 1: [51:26] We have become so good doing things that we never could do before that I don't really want to go back. But at the same time, I don't know if that's a good thing. And so I'm like, I don't know. I don't know if that's a good thing. Like, what's going to make us different? I have one last question. Track 2: [51:48] And you can just say yes or no. If you had a switch, turn off all AI everywhere, would you yes or no? Track 3: [51:55] Yes. Track 1: [51:56] Yes. If I didn't have to work, if I didn't have to work. Track 2: [51:59] Yes. Track 1: [52:00] Yes. In fact, my house in North Carolina has horrible internet. So yes. Track 3: [52:05] That's it. You know, that was surprisingly a challenging question. But then my humanity came back in. Life was okay before it. We survived. Like the world's not going to end. So yeah. What about you, Gwyneth? What do you think? Track 2: [52:20] Yeah, I mean, honestly, I look at my career now, and it's so funny because I was just reflect, Rishabh, who created Learn a Cloud with me, we sync every week. I was just talking to him before, and I was telling him, you know, honestly, when I lived on the farm, I couldn't wait to escape that life. But now, I'm actually looking forward to escaping back to the farm. So, yeah, I would turn that switch. Track 1: [52:47] Yeah. Yeah. Remind me one time to tell you about my first move to Peru in 2010. And I was told, oh, we've got like, it was real quick. I came from London, came from London and we had like internet that was great. And everyone was iPad one to one. And then I went to Peru and the guy that was the head, he was like, I laid the pipes here at this school for the internet. We have, and I don't know what he said, like five gigs. And I was like, five gigs? I think I had that on my phone. What? I had it on my phone. And I'm just like, we can't do anything. And I remember just in that five years in Peru, how we expanded the Internet. And they, I don't know, because you only had the circle pipe around South America. And we had that one traffic going to the United States. And everything was so brutally slow and zoom. And but you know what? It wasn't that bad. Track 2: [53:43] I bet life is pretty good. Yeah. Track 1: [53:44] I love her. I wish I was there. It was a whole different life, but not because of the internet. But yeah, so that's good. Well, I could probably talk for another hour, but, you know, Julian knows I get up at four o'clock in the morning. Track 3: [53:58] It's like 8 p.m. Track 1: [53:59] It's like 8 p.m. Track 2: [54:00] Are you? Track 3: [54:01] Yeah. Track 1: [54:05] And I'll be texting you tonight after. Track 3: [54:07] How much later are you staying up before you go to bed? Just to compare it with Kelly. Track 2: [54:13] A couple hours at least. I try to hit, you know, 11-ish. Track 1: [54:17] I try to get seven hours. What time do you get up? Track 2: [54:20] Like seven. Track 1: [54:21] Oh, see? Track 3: [54:22] Well, AI told me that if I go to bed at 11.47. No, I'm joking. Track 1: [54:30] You know you can write a whole AI and put your health records in there and then get a whole medal. You don't have to have a doctor anymore. Track 3: [54:36] But who's going to put the glove on? Never mind. Never mind. Track 1: [54:42] All right. It's just 8 o'clock. Track 1: [54:47] If anybody wants to contact you, where's the best place to reach you? Track 2: [54:52] LinkedIn. Gwyneth. I'll share it somewhere. LinkedIn. Shoot me a message. Happy to chat about anything, especially if you want to use SLMs for teaching. We'd love to help you out there. Yeah. Track 1: [55:03] I love that you're on LinkedIn because I message you and you answer right away. And Julian, I message. Track 3: [55:08] You have my WhatsApp. You can text me. Call me. I'm waiting for your call. Yeah. Track 1: [55:16] All right. And Julian, anything? Any news? Anything that you want to share? Track 3: [55:20] Just, I'm just excited. This is great. I'm happy to be here and I'm building and it's good. Yeah. Track 1: [55:27] It's awesome. All right. Well, I have nothing else. So for teaching Python. Track 3: [55:31] And this is Julian. Track 1: [55:32] This is Kelly. Nailed it. Signing off. You were supposed to say signing off after. Crap, you say it. Track 3: [55:38] That's fine. Track 1: [55:40] No, we did it. That's it. We're killing it. It's going to go air off. Signing off. Track 3: [55:44] It's awesome. Track 1: [55:46] Thank you. Track 2: [55:48] Thank you. Pleasure to be here. Track 1: [55:50] Bye.