WEBVTT 00:00:00.020 --> 00:00:03.480 Which of your Python tools would you still pick if you started fresh today? 00:00:04.020 --> 00:00:05.580 For a long time, the answer was easy. 00:00:05.730 --> 00:00:07.160 The stack felt settled. 00:00:07.860 --> 00:00:10.520 AI has reopened almost every one of those choices. 00:00:11.160 --> 00:00:12.100 Some tools got faster, 00:00:12.340 --> 00:00:14.100 others got replaced outright. 00:00:14.370 --> 00:00:14.900 On this episode, 00:00:15.120 --> 00:00:19.480 a panel takes stock of the state of the art in Python for 2026. 00:00:20.360 --> 00:00:23.760 Calvin Hendrix Parker is CTO and co-founder of Six Feet Up. 00:00:24.260 --> 00:00:26.480 Pat Decker works in DevOps at Hybu. 00:00:27.000 --> 00:00:30.060 Reuven Lerner has trained Python developers for 15 years, 00:00:30.240 --> 00:00:31.520 and Sidney Runkle does product 00:00:31.840 --> 00:00:34.180 management for open source at LangChain. 00:00:34.830 --> 00:00:36.600 One of them barely even opens an editor anymore. 00:00:37.260 --> 00:00:40.380 Listen and hear the surprising replacement they have. 00:00:40.720 --> 00:00:43.260 This is Talk Python To Me, episode 566, 00:00:43.640 --> 00:00:45.140 recorded October 1st, 00:00:45.440 --> 00:00:46.360 2026. 00:01:03.460 --> 00:01:07.800 Welcome to Talk Python To Me, the number one Python podcast for developers and data scientists. 00:01:08.320 --> 00:01:09.060 This is your host, 00:01:09.280 --> 00:01:09.700 Michael Kennedy. 00:01:09.900 --> 00:01:13.640 I'm a PSF fellow who's been coding for over 25 years. 00:01:14.220 --> 00:01:15.380 Let's connect on social media. 00:01:15.800 --> 00:01:17.960 You'll find me and Talk Python on Mastodon, Bluesky, 00:01:18.340 --> 00:01:18.860 and X. 00:01:19.050 --> 00:01:20.980 The social links are all in your show notes. 00:01:21.740 --> 00:01:25.260 You can find over 10 years of past episodes at talkpython.fm. 00:01:25.440 --> 00:01:28.680 And if you want to be part of the show, you can join our recording live streams. 00:01:29.000 --> 00:01:29.460 That's right. 00:01:29.670 --> 00:01:32.920 We live stream the raw uncut version of each episode on YouTube. 00:01:33.500 --> 00:01:37.840 Just visit talkpython.fm/youtube to see the schedule of upcoming events. 00:01:38.180 --> 00:01:41.800 Be sure to subscribe there and press the bell so you'll get notified anytime we're recording. 00:01:42.680 --> 00:01:43.000 Hey everyone, 00:01:43.220 --> 00:01:44.360 I have a really big announcement. 00:01:44.660 --> 00:01:48.340 For several years now, I've wanted to offer something premium to all of you out there listening. 00:01:48.680 --> 00:01:51.740 The obvious option is some kind of feed with no ads. 00:01:52.320 --> 00:01:53.560 And that just didn't sit right with me. 00:01:54.020 --> 00:01:55.560 No ads is not that much of a bonus. 00:01:56.040 --> 00:02:02.720 But over the last couple of months, I've been working really hard to make the Talk Python experience go beyond just the podcast and the courses. 00:02:03.480 --> 00:02:07.340 After a ton of work, I am thrilled to announce Talk Python membership. 00:02:08.060 --> 00:02:09.080 So what does this get you? 00:02:09.320 --> 00:02:09.700 In short, 00:02:09.979 --> 00:02:10.340 community, 00:02:10.740 --> 00:02:11.300 tech events, 00:02:11.680 --> 00:02:12.800 and episode previews. 00:02:13.600 --> 00:02:13.960 Community. 00:02:14.460 --> 00:02:16.780 Talk Python membership gets you a real community. 00:02:17.100 --> 00:02:22.660 Starts with a private Discord server where you can hang out with fellow members and well over 100 podcast guests. 00:02:23.160 --> 00:02:26.980 Many of the most renowned folks who helped build Python itself are hanging out there. 00:02:27.380 --> 00:02:33.160 We have rooms to discuss fun tools you found, places to post job openings, or even say you're looking for work, 00:02:33.640 --> 00:02:35.440 and other areas to just hang out. 00:02:35.880 --> 00:02:36.320 Tech events. 00:02:36.760 --> 00:02:41.060 Many of you tell me how Talk Python brings the Python community to your location and your world. 00:02:41.520 --> 00:02:44.940 Maybe you work at a company with just a few fellow developers or data scientists. 00:02:45.520 --> 00:02:49.940 Maybe you live in a small town where large meetups and conferences just don't exist. 00:02:50.500 --> 00:02:51.760 At Talk Python To Membership, 00:02:51.920 --> 00:02:55.840 we're running monthly tech meetups that you don't just attend but participate in. 00:02:56.240 --> 00:02:58.260 These are often run by well-known Python people. 00:02:58.820 --> 00:03:00.600 Talks run from 30 to 45 minutes, 00:03:00.920 --> 00:03:04.300 and some are followed by a hangout and mingle in a voice and video channel. 00:03:04.760 --> 00:03:06.780 I already have the first couple of tech events on the schedule. 00:03:07.440 --> 00:03:11.340 October 28th, I'll be kicking everything off with Talk Python in production, 00:03:11.770 --> 00:03:15.140 and I have some wild updates since the last time you've heard me discuss this. 00:03:15.920 --> 00:03:18.080 Coming up after that is Sebastian Ramirez. 00:03:18.560 --> 00:03:21.200 He'll take us through deploying to FastAPI Cloud, 00:03:22.099 --> 00:03:26.200 and sometime after that, I'll be presenting Succeeding with Agentic Coding, 00:03:26.800 --> 00:03:30.380 backed with a significant amount of real-world examples and data. 00:03:30.600 --> 00:03:35.340 You'll find times and details for these and new meetups at talkpython.fm/meetups. 00:03:35.960 --> 00:03:37.260 And it's not just live events. 00:03:37.900 --> 00:03:41.640 All of these events are recorded and available to watch and replay anytime. 00:03:42.620 --> 00:03:43.500 Episode previews. 00:03:43.600 --> 00:03:43.800 Now, 00:03:43.980 --> 00:03:48.320 I'm sure most of you listen to the entire episode and never miss a single one. 00:03:48.720 --> 00:03:49.500 At the same time, 00:03:49.680 --> 00:03:52.040 there may be some that you are unsure about the topic, 00:03:52.270 --> 00:03:52.640 whether it's a 00:03:52.650 --> 00:03:55.800 good fit and whether or not you want to take the time to listen to it. 00:03:56.360 --> 00:04:00.140 That's why I'm creating a members-only sidecar podcast feed. 00:04:00.640 --> 00:04:07.680 Each full episode comes with a short summary episode that covers what's discussed and interesting angles and surprises along the way. 00:04:08.200 --> 00:04:13.760 That way you can give the summary a quick listen and decide which Talk Python episodes you're going to take the time to dive into. 00:04:14.780 --> 00:04:15.720 An ad-free feed. 00:04:15.940 --> 00:04:19.340 Oh yeah, I guess members can also have an ad-free feed. 00:04:19.480 --> 00:04:21.560 I think people will enjoy having no ads, 00:04:22.079 --> 00:04:24.360 but obviously there's a lot more to membership than that. 00:04:25.340 --> 00:04:26.040 So there it is. 00:04:26.280 --> 00:04:27.740 Announcing Talk Python membership. 00:04:28.300 --> 00:04:28.620 community, 00:04:29.120 --> 00:04:29.600 tech meetups, 00:04:30.200 --> 00:04:31.140 mingle with episode guests, 00:04:31.680 --> 00:04:32.400 add free feeds. 00:04:33.240 --> 00:04:33.960 Please give it a look. 00:04:34.140 --> 00:04:36.840 Just visit Talk Python and click membership in the nav bar. 00:04:37.560 --> 00:04:39.860 I really hope we can build an awesome space together. 00:04:40.600 --> 00:04:41.740 I've got a great blueprint here, 00:04:41.860 --> 00:04:44.400 but it depends on everyone out there listening as well. 00:04:44.620 --> 00:04:46.760 So let's get together and build a community. 00:04:47.800 --> 00:04:49.580 Regardless of whether you can become a member or not, 00:04:49.620 --> 00:04:51.980 I really appreciate you listening to and supporting the show. 00:04:52.960 --> 00:04:53.220 Hey team, 00:04:53.740 --> 00:04:53.960 Calvin, 00:04:54.300 --> 00:04:54.980 Pat, Sidney, 00:04:55.140 --> 00:04:55.600 and Reuven. 00:04:55.880 --> 00:04:56.800 Awesome to have you all here. 00:04:57.020 --> 00:04:57.720 Thanks for having us, Michael. 00:04:57.900 --> 00:04:58.540 Excellent to be here. 00:04:58.760 --> 00:04:59.020 Thanks. 00:04:59.230 --> 00:05:01.180 Good to have three of you back. 00:05:01.570 --> 00:05:02.700 Always enjoy spending time with you. 00:05:03.100 --> 00:05:04.780 And Pat, welcome to Talk Python. 00:05:05.160 --> 00:05:07.380 You've been on Python Bytes before, but not on Talk Python, yes? 00:05:07.540 --> 00:05:07.860 Right. 00:05:07.890 --> 00:05:08.840 It was a few years ago. 00:05:08.910 --> 00:05:10.360 I'm not even sure how long ago that was. 00:05:10.500 --> 00:05:12.260 Yeah, back when we had guests on Python Bytes. 00:05:12.560 --> 00:05:15.160 Well, the reason we stopped having guests on Python Bytes, just sidebars, 00:05:15.320 --> 00:05:15.980 people would, 00:05:16.270 --> 00:05:19.840 when we have guests, we get into these conversations and we go down rat holes and we talk for a 00:05:19.840 --> 00:05:20.000 while. 00:05:20.120 --> 00:05:21.560 They ended up being hour long episodes. 00:05:21.680 --> 00:05:22.160 People are like, 00:05:22.480 --> 00:05:24.640 the bytes part is supposed to be a short little catch up. 00:05:25.000 --> 00:05:26.200 How is it over an hour now? 00:05:26.860 --> 00:05:27.440 Like, ah, yeah, 00:05:27.700 --> 00:05:28.120 that's true. 00:05:28.540 --> 00:05:30.120 Maybe we just need to do less. 00:05:30.540 --> 00:05:31.320 Keep it true to its nature. 00:05:31.520 --> 00:05:34.620 But Talk Python can be long and rambly, so happy to have Pat. 00:05:34.620 --> 00:05:35.980 Just talking is in the name. 00:05:36.140 --> 00:05:37.320 It literally is in the name. 00:05:37.440 --> 00:05:39.220 It tells you what you're in for. 00:05:39.620 --> 00:05:39.820 So, 00:05:40.040 --> 00:05:40.500 Pat, welcome. 00:05:41.200 --> 00:05:42.740 The rest of you all, great to have you back. 00:05:43.219 --> 00:05:47.080 We're going to be talking about an idea inspired by what tools should we be using? 00:05:47.500 --> 00:05:50.620 Things are changing fast, I think, at least as fast as they ever have, 00:05:50.850 --> 00:05:51.560 if not faster. 00:05:51.810 --> 00:05:54.480 It feels like the world is constantly accelerating. 00:05:55.080 --> 00:05:56.760 And that's true for DevTools as well. 00:05:56.800 --> 00:05:57.880 It's always been true for DevTools, 00:05:58.040 --> 00:05:58.460 but I think, 00:05:58.780 --> 00:05:59.720 especially in the AI angle, 00:06:00.120 --> 00:06:00.960 it's a little bit more true. 00:06:01.020 --> 00:06:04.420 And it's good that we have an AI specialist here, 00:06:04.620 --> 00:06:04.880 Sydney, 00:06:05.380 --> 00:06:08.080 for an inside look from that kind of, I know. 00:06:08.220 --> 00:06:09.140 We'd be lost without. 00:06:09.440 --> 00:06:12.680 Let's dive into basically a survey of the different areas 00:06:13.000 --> 00:06:13.840 of programming, 00:06:14.180 --> 00:06:14.380 Python, 00:06:14.960 --> 00:06:15.080 DevOps, 00:06:15.500 --> 00:06:16.260 all that kind of stuff, 00:06:16.620 --> 00:06:18.900 and what tools we're using, where we see trends going, 00:06:19.360 --> 00:06:22.880 and start that by doing a quick intro for everyone. 00:06:23.220 --> 00:06:24.940 Calvin, we'll just go around the squares here. 00:06:25.000 --> 00:06:27.000 Calvin, people know you, but tell me about yourself anyway. 00:06:27.220 --> 00:06:27.420 Sure. 00:06:28.240 --> 00:06:29.040 Calvin Hendryx-Parker. 00:06:29.140 --> 00:06:31.100 I'm CTO and co-founder of Six Feet Up. 00:06:31.260 --> 00:06:35.140 We are a Python consulting agency that loves solving hard problems. 00:06:35.980 --> 00:06:40.420 We've been in business now 27 years and doing Python for 26 of it. 00:06:40.640 --> 00:06:45.180 So I started using Python back when it was in version 1.5.2. 00:06:45.680 --> 00:06:46.680 That was Zope timeline. 00:06:47.160 --> 00:06:49.620 So I've been a historic Zope user. 00:06:49.960 --> 00:06:52.500 I know a lot of the listeners probably do not know what Zope is, 00:06:52.540 --> 00:06:56.780 but that was kind of the granddaddy of all web application servers out there. 00:06:56.780 --> 00:06:58.520 There was kind of a Java one, there was a Perl one, 00:06:58.600 --> 00:07:01.100 and then there was this quirky little Python one called Zope 00:07:01.220 --> 00:07:02.980 that uses an object database on the backside. 00:07:03.300 --> 00:07:04.880 So throughout the years, we've changed. 00:07:04.900 --> 00:07:06.060 We're now doing a lot more Django, 00:07:06.340 --> 00:07:07.960 doing a lot more FastAPI, 00:07:08.120 --> 00:07:09.240 doing all kinds of other technologies, 00:07:09.540 --> 00:07:13.000 but we've pretty stayed true to Python at our core 00:07:13.580 --> 00:07:16.260 and use other technologies around the periphery as needed. 00:07:16.620 --> 00:07:17.940 And that's changing fast too. 00:07:18.340 --> 00:07:20.280 So it'll be interesting to hear what everyone else has to say. 00:07:20.400 --> 00:07:20.740 Yeah, awesome. 00:07:20.900 --> 00:07:23.840 And you'll have a lot of different perspectives from all your client conversations. 00:07:24.200 --> 00:07:24.320 True. 00:07:24.540 --> 00:07:25.340 Pat, hello. 00:07:25.840 --> 00:07:26.380 Tell us about yourself. 00:07:26.580 --> 00:07:27.200 Out of college, 00:07:27.380 --> 00:07:33.500 wrote COBOL for about 15 years and then went into infrastructure and land 00:07:33.900 --> 00:07:34.080 management, 00:07:34.560 --> 00:07:36.200 worked there for another nine years. 00:07:36.460 --> 00:07:41.060 And now I'm at a company called Hibu for the last 18 years. 00:07:41.680 --> 00:07:43.740 They're kind of a Java shop for the most part, 00:07:43.940 --> 00:07:44.640 at least historically. 00:07:45.160 --> 00:07:48.240 But we're gradually adding more and more Python for, 00:07:48.620 --> 00:07:48.800 say, 00:07:48.980 --> 00:07:49.660 data analytics, 00:07:50.450 --> 00:07:51.720 a little bit of web development. 00:07:52.400 --> 00:07:55.280 And in the DevOps space where I live, for the most part, 00:07:55.440 --> 00:07:56.680 we use it a heck of a lot. 00:07:57.520 --> 00:07:58.340 And we're on AWS. 00:07:59.460 --> 00:08:02.180 So Boto3 is a daily thing for me. 00:08:02.300 --> 00:08:03.500 The DevOps side is really interesting. 00:08:03.800 --> 00:08:06.560 It's got a little bit of a data science story in that like, oh, 00:08:06.740 --> 00:08:09.460 we're not really a Python shop, but this Python stuff's working over here. 00:08:09.470 --> 00:08:12.300 And then it kind of can spread through the company. 00:08:12.700 --> 00:08:12.840 Also, 00:08:13.390 --> 00:08:15.540 Team AWS for the first two, right? 00:08:15.940 --> 00:08:17.080 Calvin, you're also AWS. 00:08:17.600 --> 00:08:18.420 I am an AWS hero, 00:08:19.060 --> 00:08:20.980 but we participate in all of them. 00:08:21.120 --> 00:08:24.940 We're doing GSpeak work, we're doing Azure work, we're doing stuff on Alibaba Cloud. 00:08:25.380 --> 00:08:26.700 So I am an AWS hero. 00:08:26.920 --> 00:08:27.660 Sydney, welcome back. 00:08:27.990 --> 00:08:28.660 Great to have you as always. 00:08:28.960 --> 00:08:30.120 Yeah, great to be back. 00:08:30.420 --> 00:08:33.620 I feel like I've taken each podcast from a different location. 00:08:33.919 --> 00:08:34.400 And it continues, 00:08:34.659 --> 00:08:34.740 yeah. 00:08:35.340 --> 00:08:35.460 Yeah, 00:08:35.719 --> 00:08:36.520 I'm now in San Francisco. 00:08:37.380 --> 00:08:39.479 I am a big open source guru. 00:08:39.719 --> 00:08:42.479 So kind of at the end of college and then after college, 00:08:43.169 --> 00:08:46.960 I was working at Pydantic leading maintenance of the Pydantic core package. 00:08:47.670 --> 00:08:48.880 But Pydantic, the company, 00:08:49.220 --> 00:08:51.520 is working on kind of AI observability and tooling. 00:08:52.010 --> 00:08:54.240 And then transitioned over to LangChain, 00:08:54.450 --> 00:08:58.400 where I also helped lead maintenance of kind of the open source suite there. 00:08:58.990 --> 00:08:59.740 So LangChain, 00:09:00.320 --> 00:09:00.680 LangGraph, 00:09:00.840 --> 00:09:01.820 and now DeepAgents. 00:09:02.280 --> 00:09:06.460 And then I'm now doing kind of product management for the open source team at LinkedIn, 00:09:06.760 --> 00:09:10.120 but still very excited to be in the like open source and AI agent space. 00:09:10.460 --> 00:09:11.460 What a cool early career, 00:09:11.700 --> 00:09:11.760 huh? 00:09:12.040 --> 00:09:14.860 We had you on when you were still in college talking about, 00:09:15.280 --> 00:09:19.680 you know, tips for younger people in college to jumpstart their career. 00:09:19.880 --> 00:09:20.920 And it looks like you're, 00:09:21.220 --> 00:09:21.840 you know, in practice, 00:09:22.040 --> 00:09:22.760 it's working really well. 00:09:22.900 --> 00:09:23.320 So congratulations. 00:09:23.760 --> 00:09:25.160 Yeah, that was a great first podcast. 00:09:25.280 --> 00:09:28.420 And then I think we talked both while I was at Pydantic and LinkedIn. 00:09:28.860 --> 00:09:29.900 So excited to be back on. 00:09:30.040 --> 00:09:30.400 Great to have you. 00:09:30.780 --> 00:09:30.960 Reuven, 00:09:31.360 --> 00:09:31.660 hello, hello. 00:09:31.860 --> 00:09:32.440 Hello there. 00:09:32.760 --> 00:09:33.620 Delighted to be back here. 00:09:33.930 --> 00:09:34.260 So yeah, 00:09:34.440 --> 00:09:36.860 so I've also been in business for quite some time. 00:09:37.110 --> 00:09:39.540 I just celebrated about 30 years in December, 00:09:39.950 --> 00:09:42.600 and I started off doing projects of various sorts. 00:09:42.890 --> 00:09:45.740 And pretty soon people started asking me not to do the project, 00:09:46.090 --> 00:09:47.280 but rather to teach their people 00:09:47.380 --> 00:09:47.860 to do the project. 00:09:48.320 --> 00:09:51.400 And so slowly but surely, I stopped doing projects. I started doing training. 00:09:51.790 --> 00:09:54.860 And so I've been doing only training for I think it is now 15 years, 00:09:55.280 --> 00:09:55.480 Python, 00:09:55.880 --> 00:09:56.360 Pandas, Git. 00:10:01.280 --> 00:10:01.720 And 00:10:01.760 --> 00:10:02.660 else is doing AI. 00:10:03.060 --> 00:10:05.140 So someone had to be the guinea pig. 00:10:05.390 --> 00:10:07.040 And so I'm having tons and tons of fun 00:10:07.340 --> 00:10:09.000 developing software now via, 00:10:09.620 --> 00:10:12.100 especially a lot of Claude Code and then helping other people do that 00:10:12.200 --> 00:10:12.420 as well. 00:10:12.840 --> 00:10:13.280 Yeah. And I just, 00:10:13.500 --> 00:10:14.640 I actually, I'll also say, 00:10:14.860 --> 00:10:17.060 and my newest book just came out in the last 00:10:17.070 --> 00:10:17.380 few days, 00:10:17.840 --> 00:10:20.400 which is AI assisted Python for non-programmers. 00:10:20.700 --> 00:10:22.020 So it's my non-programmers course 00:10:22.340 --> 00:10:27.700 in book form with lots of AI sort of companion exercises to help you learn better. 00:10:28.100 --> 00:10:32.660 the book idea. I think it's going to be big. But yeah, what an early bet you made on AI these days 00:10:32.810 --> 00:10:36.680 to be working with it. I mean, I know most of us are probably not doing that. So that'll be 00:10:36.800 --> 00:10:38.000 interesting to hear. Edgy for sure. 00:10:38.340 --> 00:10:39.040 Very edgy. 00:10:39.240 --> 00:10:39.720 All right. 00:10:40.080 --> 00:10:42.440 Speaking of edgy, let's get right into it. 00:10:42.660 --> 00:10:48.100 So I put together kind of a spectrum of different topics and some of them might be, 00:10:48.260 --> 00:10:48.440 oh, 00:10:48.600 --> 00:10:48.880 of course, 00:10:48.930 --> 00:10:50.260 we're all doing this. 00:10:50.500 --> 00:10:51.160 But at the same time, 00:10:51.380 --> 00:10:51.540 others, 00:10:52.000 --> 00:10:52.900 maybe not so much. 00:10:53.300 --> 00:10:53.680 Maybe it's new. 00:10:57.700 --> 00:10:58.060 So 00:10:58.080 --> 00:10:59.040 editors first. 00:10:59.540 --> 00:11:02.720 And I just kind of a bit of a historical thing. 00:11:03.280 --> 00:11:04.700 I used to ask when I started 00:11:04.700 --> 00:11:06.440 the podcast, I used to ask at the end of the show, 00:11:06.620 --> 00:11:08.080 a couple of things like, you know, recommend a 00:11:08.220 --> 00:11:12.420 package. And then what editor are you using? And this always gave me random answers from all these 00:11:12.540 --> 00:11:15.120 little weird things I've never even heard of, you know, 00:11:15.280 --> 00:11:17.480 over the last six or seven years, it was 00:11:17.580 --> 00:11:18.660 just like VS Code, 00:11:18.980 --> 00:11:21.360 ByCharm, VS Code, VS Code, ByCharm, ByCharm. 00:11:21.560 --> 00:11:22.500 I'm like, okay, this, I'm not 00:11:22.680 --> 00:11:23.460 asking this question anymore, 00:11:23.720 --> 00:11:27.500 but I'm not sure that that's actually still true because of the H1, 00:11:27.740 --> 00:11:30.960 you see on the VS Code, the open source AI code editor. 00:11:30.960 --> 00:11:32.920 And I think AI has scrambled this a bit. 00:11:33.320 --> 00:11:34.320 So I don't know. 00:11:34.960 --> 00:11:35.640 Let's start with you, 00:11:35.800 --> 00:11:35.880 Sydney, 00:11:36.360 --> 00:11:36.840 first here. 00:11:37.000 --> 00:11:37.080 Like, 00:11:38.060 --> 00:11:38.600 what are you doing? 00:11:38.840 --> 00:11:40.380 What are you using these days to write code 00:11:40.620 --> 00:11:43.900 to the extent that code is written and read code when it's not? 00:11:44.040 --> 00:11:48.560 I'll go back and say I was first a PyTerm user and then a VS Code user. 00:11:49.120 --> 00:11:52.780 And then I transitioned to Cursor when I wanted, 00:11:53.200 --> 00:11:53.320 like, 00:11:53.480 --> 00:11:53.660 kind of, 00:11:53.680 --> 00:11:56.779 I think they had, like, agents rolled into the IDE earlier 00:11:56.800 --> 00:11:58.480 than some of the other IDEs. 00:11:58.820 --> 00:12:00.000 But my slightly, 00:12:00.560 --> 00:12:01.800 this would have sounded very odd 00:12:02.000 --> 00:12:02.760 like two years ago, 00:12:03.080 --> 00:12:06.140 but honestly where I do most of my coding now is in Slack. 00:12:06.700 --> 00:12:10.060 So we have coding agents available via like Slack tag. 00:12:10.680 --> 00:12:12.220 So I can say at OpenSuite, 00:12:12.700 --> 00:12:12.920 you know, 00:12:13.320 --> 00:12:14.340 help me with this task. 00:12:14.640 --> 00:12:16.760 Slack is where we come up with a lot of ideas 00:12:17.020 --> 00:12:18.220 for what we want to change 00:12:18.540 --> 00:12:19.580 or like problems that we're having. 00:12:20.020 --> 00:12:21.260 So I certainly still like, 00:12:21.540 --> 00:12:22.600 we'll start something from Slack 00:12:23.120 --> 00:12:24.660 and then go look at it in GitHub 00:12:25.200 --> 00:12:26.760 and then maybe pull it down locally 00:12:26.780 --> 00:12:28.960 I like really need to get into the weeds with changes. 00:12:29.260 --> 00:12:31.600 But honestly, I would say like 90% of 00:12:31.600 --> 00:12:33.000 my work is now in Slack. 00:12:33.560 --> 00:12:35.540 Or if I like start from Slack, 00:12:36.030 --> 00:12:38.680 and then add comments in GitHub, 00:12:39.230 --> 00:12:41.540 my like Slack agent will pick those up and address. 00:12:42.050 --> 00:12:43.660 So yeah, that's my like very AI native 00:12:44.250 --> 00:12:44.820 answer these days. 00:12:45.050 --> 00:12:46.740 Did I say there would be surprises? 00:12:46.980 --> 00:12:48.560 There are already incredible. 00:12:48.840 --> 00:12:48.940 Like, 00:12:49.280 --> 00:12:51.120 that is so amazing and futuristic. 00:12:51.880 --> 00:12:53.040 And yet at the same time, 00:12:53.370 --> 00:12:54.040 I can see it. 00:12:54.400 --> 00:12:55.560 I've done some really 00:12:55.580 --> 00:13:01.220 interesting stuff with just talking to my phone that then translates over to Hermes. 00:13:01.740 --> 00:13:01.920 It's like, 00:13:01.920 --> 00:13:05.480 I need you to program this. And I came back later and through a chat, just like, 00:13:05.860 --> 00:13:06.200 how are we doing? 00:13:06.480 --> 00:13:11.680 And that was all, I never touched an editor or nothing. I just either voice typed or actual 00:13:11.860 --> 00:13:13.300 typed and it's wow. 00:13:13.540 --> 00:13:15.080 But Slack as a group, 00:13:15.240 --> 00:13:16.500 I guess it makes a lot of sense, 00:13:16.660 --> 00:13:16.940 but yeah, 00:13:17.140 --> 00:13:21.020 it's really interesting. What do you all think? I'm reeling from this because I thought, 00:13:21.420 --> 00:13:21.600 wow, 00:13:21.860 --> 00:13:25.600 Like I've gone through this like evolution of there's an Emacs person. 00:13:25.920 --> 00:13:28.040 And then like I did a little bit of PyCharm, 00:13:28.580 --> 00:13:30.080 but I realized that I did a lot of stuff in 00:13:30.220 --> 00:13:30.340 Jupyter. 00:13:30.920 --> 00:13:31.720 But lately in the last year, 00:13:31.880 --> 00:13:33.080 most of it's been in clawed code. 00:13:33.200 --> 00:13:33.880 And I was like, okay, 00:13:34.290 --> 00:13:37.020 so that's like clearly as modern and cool as it gets. 00:13:37.160 --> 00:13:37.220 Right. 00:13:37.420 --> 00:13:39.740 I tell my children all the time that I'm as cool as it gets. 00:13:39.880 --> 00:13:41.080 And one day they will believe me. 00:13:41.420 --> 00:13:44.120 And yet Sydney like massively upstages me here. 00:13:44.120 --> 00:13:44.360 I'm like, 00:13:44.560 --> 00:13:44.960 I'm like, 00:13:45.300 --> 00:13:48.360 my brain is melting thinking of Slack as your main coding editor. 00:13:48.520 --> 00:13:49.640 That is super wild. 00:13:50.040 --> 00:13:50.600 Or not even editor, 00:13:50.760 --> 00:13:52.360 like dispatching agents based on it. 00:13:52.640 --> 00:13:53.080 That's so cool. 00:13:53.160 --> 00:13:55.280 Well, I thought back in the day when I was doing Zope 00:13:55.280 --> 00:13:58.940 and we were coding into a text area on a web page, 00:13:59.280 --> 00:13:59.860 that was bad. 00:14:00.260 --> 00:14:02.280 And that drove me a little like kind of batty. 00:14:02.660 --> 00:14:05.780 But to do it into a Slack agent also kind of blows my mind. 00:14:06.000 --> 00:14:08.900 I will say I do not even have VS Code installed 00:14:09.340 --> 00:14:11.520 or any VS Code variants installed on my computer. 00:14:11.840 --> 00:14:14.200 I'm pretty well in Pi and Claude Code 00:14:14.560 --> 00:14:16.560 for all of my coding usages. 00:14:17.120 --> 00:14:18.240 And then when I need to view something, 00:14:18.580 --> 00:14:20.760 I'm using my favorite alternative cat, 00:14:20.930 --> 00:14:21.520 which is bat, 00:14:21.800 --> 00:14:22.200 B-A-T. 00:14:22.720 --> 00:14:23.480 And then I'll use, 00:14:23.650 --> 00:14:25.460 if I need a rich, really richer experience, 00:14:25.780 --> 00:14:27.020 I'll jump back into PyCharm. 00:14:27.300 --> 00:14:30.220 It's probably my like fallback, fallback, fallback thing 00:14:30.330 --> 00:14:31.620 when I need to really dive into something. 00:14:32.020 --> 00:14:33.460 But most of them I'm viewing Markdown, 00:14:33.810 --> 00:14:36.300 which I found on OS X there is, 00:14:36.520 --> 00:14:38.160 I'm dating myself by saying like that. 00:14:38.400 --> 00:14:39.080 On macOS, 00:14:39.270 --> 00:14:41.840 there is a viewer called Telescopo 00:14:42.180 --> 00:14:43.880 that is an excellent Markdown viewer, 00:14:44.380 --> 00:14:45.800 like very rich experience. 00:14:46.220 --> 00:14:48.860 So I can view like the superpower specs and plans. 00:14:49.620 --> 00:14:54.980 And that's where I spend a lot of my time actually is probably with the markdown pieces before I say go implement some stuff. 00:14:55.260 --> 00:14:55.460 Yeah, 00:14:55.640 --> 00:14:55.900 that's wild. 00:14:56.040 --> 00:15:01.100 You're very, I know you're very terminal heavy like Claude Code style or more of the open, 00:15:01.300 --> 00:15:02.600 open multi-model ones. 00:15:02.760 --> 00:15:04.460 VI user for life, 00:15:04.900 --> 00:15:05.120 Vim, 00:15:05.280 --> 00:15:05.840 NeoVim, 00:15:06.060 --> 00:15:07.640 you know, kind of follow that progression. 00:15:08.220 --> 00:15:12.520 So I guess I'm the guy on the beach kicking sand in the feeble Emacs user's face. 00:15:12.680 --> 00:15:13.000 I'm sorry. 00:15:14.520 --> 00:15:15.360 I'll take that stand. 00:15:15.760 --> 00:15:20.340 And when there's no VI key bindings available and I open up some kind of a window that doesn't have them, 00:15:20.600 --> 00:15:21.200 oh, am I lost. 00:15:21.380 --> 00:15:31.040 I got on stage once at PyCon to do a lightning talk and somehow VI mode got disabled on my PyCharm and I was absolutely petrified. 00:15:31.280 --> 00:15:31.520 Pat, 00:15:31.720 --> 00:15:32.460 how about you? 00:15:32.700 --> 00:15:33.800 Where's the editor story? 00:15:34.000 --> 00:15:35.240 It has evolved a lot. 00:15:35.460 --> 00:15:36.600 Way back at Iowa State, 00:15:36.920 --> 00:15:39.460 I took Pascal on punched cards. 00:15:39.740 --> 00:15:40.760 So that was a little different. 00:15:41.040 --> 00:15:44.540 Then ended up getting into mainframe programming. 00:15:45.040 --> 00:15:46.040 In the IBM world, 00:15:46.340 --> 00:15:47.720 they had their editor. 00:15:48.860 --> 00:15:50.600 And when we transitioned then to PCs, 00:15:51.000 --> 00:15:51.640 there was a company, 00:15:52.100 --> 00:15:53.520 Mansfield Software, that created 00:15:53.820 --> 00:15:56.660 K-Edit to mimic X-Edit from the mainframe. 00:15:56.700 --> 00:15:58.240 And they did a great job. 00:15:58.520 --> 00:16:00.980 And that was running under DOS pre-Windows. 00:16:01.380 --> 00:16:03.700 And then they did come out with a Windows version, 00:16:03.920 --> 00:16:05.200 which I still use today. 00:16:05.900 --> 00:16:06.140 Now, 00:16:06.660 --> 00:16:06.800 most, 00:16:07.000 --> 00:16:07.720 there it is right there. 00:16:07.800 --> 00:16:07.980 Yes. 00:16:08.240 --> 00:16:09.240 Most of the time, 00:16:09.570 --> 00:16:15.000 I attempt to use VS Code because you can remotely connect to all these AWS instances, 00:16:15.270 --> 00:16:17.960 or maybe they're even in-house running on VMware. 00:16:18.470 --> 00:16:21.400 But when you want to use some AI assistance, 00:16:21.800 --> 00:16:22.820 I'm in Devon, 00:16:23.040 --> 00:16:24.460 I'm in Kiro, 00:16:24.850 --> 00:16:25.920 which is from AWS. 00:16:26.860 --> 00:16:32.020 And then for the final edit, I may tweak something at runtime with Vim. 00:16:32.360 --> 00:16:41.500 That's usually my choice if I have once in a while Nano because the defaults for Vim sometimes on a particular instance prevent pasting or something. 00:16:41.860 --> 00:16:43.220 So I'll have to fall to Nano. 00:16:43.520 --> 00:16:51.280 Yeah, I do fall back to Nano every now and then on some kind of server or I just need to edit this thing and we're just going to use Nano is going to be here. 00:16:51.380 --> 00:16:51.920 It'll be fine. 00:16:52.380 --> 00:16:54.660 And I did used to do Emacs way back in the day. 00:16:55.160 --> 00:16:56.120 So sorry, sorry, Calvin. 00:16:56.640 --> 00:16:58.620 But so it feels a little bit like that. 00:16:58.820 --> 00:16:59.980 I've come to terms with it. 00:17:00.160 --> 00:17:01.700 I'm OK with people who use Emacs. 00:17:02.260 --> 00:17:07.520 just just for the record like when i was in school like we heard rumors that there were people who 00:17:07.680 --> 00:17:13.339 use this editor called vi and like i was like fine but we all sort of agreed that whoever used such a 00:17:13.500 --> 00:17:18.880 thing like clearly wasn't a serious programmer and it literally took me like a decade out of school 00:17:18.920 --> 00:17:24.660 to discover there are smart people and good programmers who use this vi thing and we should 00:17:24.760 --> 00:17:29.619 perhaps not look down at them so much so it's okay yeah i think it's a school by school thing 00:17:29.640 --> 00:17:34.660 I think it depends on what terminal they set you down to and what tutorial they told you to do. 00:17:34.810 --> 00:17:35.960 And that's what you got. 00:17:36.040 --> 00:17:37.200 100% brainwashing. 00:17:37.420 --> 00:17:37.560 Yes. 00:17:37.740 --> 00:17:45.540 Well, I think that brings us to a really nice next up section that I think has certainly changed a lot in the last few years. 00:17:45.570 --> 00:17:46.540 And that's terminals. 00:17:47.160 --> 00:17:50.840 So pretty much you all just open up Terminal on the Mac, right? 00:17:52.960 --> 00:17:54.680 Or CMD on Windows? 00:17:55.020 --> 00:17:55.740 I do CMD. 00:17:56.090 --> 00:17:56.800 I do WSL, 00:17:57.140 --> 00:17:57.200 though, 00:17:57.380 --> 00:17:57.540 too, 00:17:57.970 --> 00:17:59.020 the subsystem for Linux. 00:17:59.540 --> 00:18:01.540 But by and large, I'm just in Bash. 00:18:01.620 --> 00:18:05.420 So I don't do phish or any of these other command interpreters. 00:18:05.720 --> 00:18:08.540 The one thing I do like is something called Royal TS. 00:18:09.340 --> 00:18:17.260 And that allows me to have a whole list and subgroups of servers where I can remotely connect as whatever user I want. 00:18:17.320 --> 00:18:18.580 It saves credentials. 00:18:19.200 --> 00:18:20.420 You can make it a little bit colorful. 00:18:20.900 --> 00:18:23.060 It works for SSH, for RDP, 00:18:24.060 --> 00:18:24.380 HTTP, 00:18:25.360 --> 00:18:25.920 SFTP, 00:18:26.720 --> 00:18:27.340 VNC. 00:18:28.000 --> 00:18:28.960 It's pretty great. 00:18:29.340 --> 00:18:30.100 I'm a fan. 00:18:30.300 --> 00:18:31.160 Is that Windows only, 00:18:31.520 --> 00:18:31.980 the Royal TS? 00:18:32.300 --> 00:18:32.860 It is not. 00:18:33.040 --> 00:18:34.200 It's Mac and Windows both, 00:18:34.440 --> 00:18:34.540 yeah. 00:18:34.740 --> 00:18:35.360 That's very interesting. 00:18:35.800 --> 00:18:38.980 Yeah, because you do so much DevOps of managing remote machines, 00:18:39.100 --> 00:18:40.080 and especially if they're Windows, 00:18:40.620 --> 00:18:40.860 you know, 00:18:41.120 --> 00:18:42.340 you often need a GUI. 00:18:42.400 --> 00:18:43.660 Connect remotely and away you go, 00:18:44.020 --> 00:18:44.120 yeah. 00:18:44.540 --> 00:18:47.600 But I do use TMUX because you have some long-running tests 00:18:47.840 --> 00:18:49.740 and you never know when your credentials expire 00:18:50.140 --> 00:18:52.220 and you get disconnected or this or that. 00:18:52.620 --> 00:18:54.380 So your process continues to run. 00:18:54.660 --> 00:18:54.900 Calvin, 00:18:55.100 --> 00:18:56.140 you don't use Terminal? 00:18:56.380 --> 00:18:58.120 I am a, no, I do not use Terminal. 00:18:58.360 --> 00:19:02.320 big fan of a cross-platform terminal called Kitty. 00:19:02.780 --> 00:19:06.140 It's claimed to fame is a 3D accelerated 00:19:06.860 --> 00:19:07.100 terminal. 00:19:07.440 --> 00:19:08.480 So it uses your GPU. 00:19:08.760 --> 00:19:11.960 So it's a GPU accelerated terminal. So it'll scroll text faster 00:19:12.160 --> 00:19:13.320 than any terminal out there. 00:19:13.780 --> 00:19:15.680 And I like the fact that it's on Linux. 00:19:16.240 --> 00:19:16.720 Because for a while, 00:19:16.740 --> 00:19:18.540 I was running full-time Linux desktop. 00:19:18.760 --> 00:19:19.620 Like for three years, 00:19:20.040 --> 00:19:21.780 I ran only Linux because Apple 00:19:22.320 --> 00:19:24.960 insisted on this touch bar thing on their keyboard. 00:19:25.420 --> 00:19:27.140 And so I jettisoned my Mac, 00:19:27.500 --> 00:19:28.540 went to straight Linux. 00:19:28.780 --> 00:19:31.140 And then when they got back to a regular keyboard, I came back to the Mac, 00:19:31.280 --> 00:19:32.680 but I kept Kitty the whole time. 00:19:32.880 --> 00:19:33.860 And anytime I get on Windows, 00:19:34.020 --> 00:19:34.940 I can always get Kitty 00:19:35.420 --> 00:19:36.040 there as well. 00:19:36.420 --> 00:19:39.220 So I like that. I also have been a Tmux person historically, 00:19:39.780 --> 00:19:43.840 although I have switched over to herdr for window management, remote session management, 00:19:44.440 --> 00:19:45.560 AI agent management. 00:19:45.960 --> 00:19:50.340 I really love the fact that it is a kind of AI native multiplexer. 00:19:50.760 --> 00:19:54.539 Kitty also has a built in multiplexer, which I kind of stopped using Tmux when I adopted Kitty 00:19:54.560 --> 00:19:57.100 because of Kitty's native window management 00:19:57.320 --> 00:19:58.740 and Tmux-like features. 00:19:59.220 --> 00:20:01.840 But now I really like herdr for doing that kind of thing. 00:20:02.160 --> 00:20:03.420 Actually, I haven't had a chance to play with it. 00:20:03.540 --> 00:20:05.980 herdr now has features for remote agent sessions. 00:20:06.600 --> 00:20:08.940 And so I use Mosh all the time with a VPN 00:20:09.340 --> 00:20:10.220 from my various devices 00:20:10.520 --> 00:20:12.460 so that I can constantly be in connection 00:20:12.700 --> 00:20:15.880 with those agents that are running on any server, any place. 00:20:16.200 --> 00:20:18.500 Something I've seen a lot of traction on is herdr. 00:20:18.800 --> 00:20:21.140 Although I still like editors 00:20:21.900 --> 00:20:23.560 when I'm working with my AI stuff. 00:20:23.880 --> 00:20:28.060 And so it's not that I'm necessarily writing a ton of the code or always babysitting it, 00:20:28.140 --> 00:20:31.460 but I'd like to kind of interact with it from an editor side while it's running. 00:20:31.560 --> 00:20:33.720 So I haven't gone fully down this path, 00:20:33.900 --> 00:20:35.360 but yeah, it's super cool. 00:20:35.580 --> 00:20:37.080 If I was going to use a more rich client, 00:20:37.320 --> 00:20:41.260 I think the JetBrains have their Air AI agentic client, 00:20:41.580 --> 00:20:44.040 and then Codex is pretty reasonable. 00:20:44.160 --> 00:20:46.660 And so the same thing with the Cloud desktop, 00:20:46.880 --> 00:20:49.600 which has now Codework and Code in it. 00:20:49.760 --> 00:20:50.020 Sydney, 00:20:50.200 --> 00:20:52.400 what kind of terminal do you use to connect to Slack? 00:20:52.660 --> 00:20:53.000 Just kidding. 00:20:53.740 --> 00:20:57.460 I feel like I'm not as like tinkery here. 00:20:58.100 --> 00:21:02.600 I mostly just use like the embedded terminal in whatever IDE I'm using. 00:21:03.340 --> 00:21:06.540 I've heard a lot of good things about Ohmypy recently. 00:21:06.730 --> 00:21:10.580 And I know like OhMyZSH was popular as well. 00:21:10.970 --> 00:21:11.300 But yeah, 00:21:11.680 --> 00:21:11.780 honestly, 00:21:12.140 --> 00:21:13.400 my answer is boring here. 00:21:13.400 --> 00:21:16.740 Like just because I'm doing so much not in an IDE, 00:21:16.830 --> 00:21:20.140 I feel like I'm a little bit less tied to a specific terminal. 00:21:20.320 --> 00:21:20.800 Yeah, very cool. 00:21:21.180 --> 00:21:21.280 Reuben, 00:21:21.800 --> 00:21:22.700 Jupyter has a terminal, right? 00:21:22.860 --> 00:21:23.840 More than many other people, 00:21:24.000 --> 00:21:24.440 that's for sure. 00:21:24.690 --> 00:21:26.080 Like I've been using a lot of Marimo lately. 00:21:26.580 --> 00:21:31.240 And one of the things I totally miss is the like exclamation point allowing me to execute 00:21:31.440 --> 00:21:32.340 commands at the backend. 00:21:32.690 --> 00:21:34.540 Because I just got so used to doing that from Jupyter. 00:21:35.020 --> 00:21:35.660 For terminal stuff, 00:21:35.880 --> 00:21:38.560 I was using iTerm2 for many, 00:21:38.720 --> 00:21:39.060 many years. 00:21:39.600 --> 00:21:43.220 And at some point I realized I'm not using like 90% of its functionality, 00:21:43.600 --> 00:21:43.840 including all 00:21:43.900 --> 00:21:44.640 the TMUX stuff, 00:21:44.940 --> 00:21:45.340 which I just, 00:21:45.610 --> 00:21:46.560 I never figured it out. 00:21:46.860 --> 00:21:50.620 I never figured out how to get it to work or how to work with it or why I would want it. 00:21:50.740 --> 00:21:52.100 I'm sure someone here can explain to me, 00:21:52.430 --> 00:21:54.660 I would just open up new tabs and I was okay with that. 00:21:54.950 --> 00:21:56.760 And then I found out about ghost TTY, 00:21:57.180 --> 00:21:59.160 which I think I've just pronounced an extra T in there. 00:21:59.540 --> 00:22:00.800 And I switched out a few months ago. 00:22:01.100 --> 00:22:01.560 I've been super, 00:22:01.720 --> 00:22:03.760 super happy with it just because I don't 00:22:04.020 --> 00:22:05.800 need all the extra iterm2 stuff. 00:22:06.500 --> 00:22:09.280 And I had some problems with the Claude Code actually. It was like 00:22:09.400 --> 00:22:12.140 giving me some weird scroll back and flickering and so forth. 00:22:12.490 --> 00:22:14.820 So I put Claude Code into TUI full 00:22:14.960 --> 00:22:15.440 screen mode. 00:22:15.730 --> 00:22:20.660 And not only now when I load it up, does it give me full scrollable and searchable back 00:22:20.680 --> 00:22:21.420 my entire history, 00:22:21.740 --> 00:22:24.700 but it puts this diff window up in the right of my terminal. 00:22:25.060 --> 00:22:25.640 And I've been just 00:22:25.800 --> 00:22:28.100 like super, super happy as a result. That's super cool. 00:22:28.540 --> 00:22:31.600 I was for a long time working on warp, 00:22:32.040 --> 00:22:33.560 which I really like warp as a terminal, 00:22:33.780 --> 00:22:37.260 but they're kind of trying to be a pie or some, 00:22:37.460 --> 00:22:37.660 you know, 00:22:37.960 --> 00:22:38.520 Claude Code, 00:22:38.880 --> 00:22:39.780 but through there, 00:22:40.140 --> 00:22:42.400 which I don't know, kind of turned me off a little bit. 00:22:42.500 --> 00:22:46.120 And my friend Mark said, you need to be checking out CMUX. 00:22:46.640 --> 00:22:47.680 Are you familiar with CMUX? 00:22:47.840 --> 00:22:48.560 I'm familiar 00:22:48.580 --> 00:22:53.520 with it but not tried it yeah it's pretty nice it'll do a little bit of a you know multi-window 00:22:53.860 --> 00:22:58.840 split different ways type of things and what's cool is you have workspaces that are connected to 00:22:59.120 --> 00:23:04.640 maybe it remembers the layout and it'll restore ssh connections if you shut it down and open it back 00:23:04.740 --> 00:23:10.280 up across all the different things and so it's pretty neat it has a an iphone version as well so 00:23:10.580 --> 00:23:15.780 cmux and mark and i were at a coffee shop sitting there looking at something and someone across the 00:23:15.800 --> 00:23:19.020 way he goes excuse me i'm sorry i just i have to come over i just saw that you were talking about 00:23:19.120 --> 00:23:25.000 cmux and here's my cmux okay so apparently uh at least three of us use it now and it's i believe 00:23:25.100 --> 00:23:30.460 it's based on ghost tty as well rumin it's open source it's pretty cool can i just say how wild 00:23:30.640 --> 00:23:38.260 it is that in 2026 people are still developing new terminals and like i would have expected this was 00:23:38.370 --> 00:23:44.640 a long ago solve problem like feature complete and yet like i guess it's all the ai stuff that's 00:23:44.660 --> 00:23:49.920 like driving all this new rush of creativity and products and you have to rewrite it in rust too 00:23:50.120 --> 00:23:55.060 reuben 100 i mean it's just completely crazy like i'm so excited to set up my terminal 00:23:55.580 --> 00:24:03.120 to uh be able to ssh into or talk to my you know studio m5 max so i can get it to run a terminal 00:24:03.190 --> 00:24:07.660 like oh my gosh you know i really take an advantage of it right so it can shell out to the cloud for 00:24:07.840 --> 00:24:12.620 ai things or whatever it's doing or server work um incredible now this next one this next topic 00:24:12.640 --> 00:24:13.300 I think. 00:24:13.740 --> 00:24:14.660 Go ahead, Calvin. You got something to say? 00:24:14.660 --> 00:24:23.700 I was going to say that the terminal UI and the CLIs that have now followed are giving a really rich interface for all the LLMs and agentic harness pieces. 00:24:24.020 --> 00:24:27.420 So instead of relying on, say, like MCPs and things like that, 00:24:27.620 --> 00:24:29.660 you're getting actually really great CLIs. 00:24:29.660 --> 00:24:34.000 Like my email client now ships with a CLI that I can go use myself, 00:24:34.200 --> 00:24:34.340 too. 00:24:34.720 --> 00:24:40.100 I can tell my agent to use it, but I can also use it, which I found quite handy because I'm living on the terminal. 00:24:40.220 --> 00:24:41.260 so I can stay on the terminal, 00:24:41.860 --> 00:24:42.500 check in on email, 00:24:42.670 --> 00:24:43.020 for example. 00:24:43.090 --> 00:24:45.520 So I use Spark Desktop as my email client, 00:24:45.680 --> 00:24:46.700 and it has a Spark CLI. 00:24:46.860 --> 00:24:48.420 Awesome. Yeah, I use Spark as well. 00:24:48.490 --> 00:24:52.020 I don't use our CLI, but I do use Spark as well, which I like it a lot. 00:24:52.100 --> 00:24:55.620 I like it better than the whole MCP option for local stuff like that. 00:24:55.700 --> 00:25:00.040 I went so far as there's even a CLI for Talk Python these days. 00:25:00.540 --> 00:25:03.700 I just want to point this out as a segue to the next one here. 00:25:04.220 --> 00:25:05.120 UV tool install, 00:25:05.620 --> 00:25:05.740 okay? 00:25:06.260 --> 00:25:08.240 That's how you get the Talk Python CLI. 00:25:08.560 --> 00:25:09.520 You don't have to authenticate anything. 00:25:09.680 --> 00:25:11.000 It just lets you ask questions like, 00:25:11.280 --> 00:25:14.020 you know, what did Sydney say on her first episode? 00:25:14.300 --> 00:25:14.380 Like, 00:25:14.790 --> 00:25:15.760 give me the highlights or whatever. 00:25:15.760 --> 00:25:16.460 And it would literally, 00:25:16.740 --> 00:25:16.920 you know, 00:25:17.220 --> 00:25:18.920 do that kind of stuff if you plug it to your AI. 00:25:19.140 --> 00:25:23.780 But the only thing better is just typing uvx in the name of the package you want to run and not even thinking about it. 00:25:23.880 --> 00:25:25.660 I know there's a lot of tools and, 00:25:26.080 --> 00:25:26.200 you know, 00:25:26.480 --> 00:25:28.600 they've really done incredible stuff. 00:25:29.020 --> 00:25:29.220 You know, 00:25:29.560 --> 00:25:30.000 pipX, 00:25:30.600 --> 00:25:31.380 there's pip, 00:25:31.660 --> 00:25:31.820 obviously. 00:25:32.260 --> 00:25:33.780 There was a time when there was no pip. 00:25:34.270 --> 00:25:36.460 Python existed for a long time without a package manager. 00:25:36.860 --> 00:25:39.080 Partly why the standard library has so much stuff in it. 00:25:39.240 --> 00:25:39.880 Those are dark days, 00:25:40.080 --> 00:25:40.180 Michael. 00:25:40.630 --> 00:25:41.020 Oh my God. 00:25:41.240 --> 00:25:43.680 Like the amount of time I spent in my intro classes, 00:25:44.460 --> 00:25:48.200 helping them install pip so that they could install things. 00:25:48.360 --> 00:25:49.060 Like, oh my God, 00:25:49.380 --> 00:25:49.820 what madness. 00:25:50.120 --> 00:25:50.420 Exactly. 00:25:50.600 --> 00:25:52.040 I think uv saved a lot of that, 00:25:52.240 --> 00:25:52.860 especially with the Python, 00:25:53.220 --> 00:25:53.520 you know, uv, 00:25:53.770 --> 00:25:53.860 Python, 00:25:54.540 --> 00:25:54.600 three, 00:25:54.990 --> 00:25:55.080 whatever. 00:25:55.620 --> 00:25:57.160 I guess today we could say 3.15, 00:25:57.330 --> 00:25:57.460 right? 00:25:57.620 --> 00:25:58.800 Like it's October 1st. 00:25:59.140 --> 00:26:00.260 It's supposed to be out sometime. 00:26:00.310 --> 00:26:00.900 Did it launch? 00:26:01.280 --> 00:26:01.660 I hope so. 00:26:02.120 --> 00:26:02.960 That's an exciting release. 00:26:03.220 --> 00:26:03.780 I haven't checked. 00:26:03.830 --> 00:26:05.380 I just woke up and went to a podcast. 00:26:05.700 --> 00:26:08.220 So I'm disconnected from the world. 00:26:08.440 --> 00:26:09.300 So that's very interesting. 00:26:09.580 --> 00:26:12.140 Download is still 3.14.8 as of now. 00:26:12.320 --> 00:26:13.520 I mean, we had Poetry. 00:26:14.180 --> 00:26:15.160 We had a bunch of others. 00:26:15.700 --> 00:26:15.960 Flit. 00:26:16.460 --> 00:26:18.560 But uv definitely took the world by storm. 00:26:18.800 --> 00:26:19.200 I agree. 00:26:19.510 --> 00:26:20.640 And we went from Pip. 00:26:21.240 --> 00:26:23.060 And at some point in there, we tried PipX. 00:26:23.280 --> 00:26:24.980 But when uv came on the scene, 00:26:25.460 --> 00:26:27.860 and the speed was the true feature of that, 00:26:28.100 --> 00:26:30.920 and the consistency of how you ran the various tools, 00:26:31.200 --> 00:26:32.700 I don't like to disparage other tools. 00:26:32.860 --> 00:26:34.940 But I will say that when you type the word poetry, 00:26:35.160 --> 00:26:37.020 if you use the word poetry in our Slack, 00:26:37.660 --> 00:26:40.900 there was a Slack bot response with a haiku that was not favorable to poetry. 00:26:41.140 --> 00:26:43.800 Would it rewrite poetry and Russ using an AI if you said that? 00:26:43.870 --> 00:26:44.820 Or was it just a poem? 00:26:45.000 --> 00:26:48.940 I'll just say that poetry is a really good tool if you are all in on poetry. 00:26:49.420 --> 00:26:51.260 And I wasn't ready to be all in on poetry. 00:26:51.370 --> 00:26:54.780 There was a whole bunch of extra overhead of pieces it would do 00:26:54.940 --> 00:26:56.300 that I just didn't either didn't need, 00:26:56.480 --> 00:26:56.860 didn't want, 00:26:57.020 --> 00:26:58.200 or I didn't care to want to go learn. 00:26:58.660 --> 00:27:02.360 And so I feel like uv encapsulated a lot of my pre-existing opinions 00:27:02.760 --> 00:27:05.780 because I was using PyPI for managing Python versions. 00:27:06.440 --> 00:27:13.060 I was using pip with lock files and like some extra flags to make it very, very reliable for 00:27:13.740 --> 00:27:16.120 getting exactly what I wanted when and where for penning. 00:27:16.640 --> 00:27:17.840 What was the other opinion I had on that? 00:27:17.920 --> 00:27:18.020 Oh, 00:27:18.320 --> 00:27:21.460 using pipx for like the tools install. 00:27:21.880 --> 00:27:24.220 So I kind of had that opinion around how we would 00:27:24.420 --> 00:27:27.200 install like Python tooling into the machine. 00:27:27.760 --> 00:27:30.320 So uv just took all like the things I already love. 00:27:30.320 --> 00:27:33.540 So I guess I was biased. I was already biased on it because those are the ways I liked operating. 00:27:33.940 --> 00:27:35.960 and uv took my same opinions and tax. 00:27:36.080 --> 00:27:36.580 And so I was like, all right, 00:27:36.640 --> 00:27:37.840 I'm all about uv at this point. 00:27:37.980 --> 00:27:38.220 Let's go. 00:27:38.400 --> 00:27:38.480 Nice. 00:27:38.950 --> 00:27:39.060 Sydney, 00:27:39.580 --> 00:27:40.180 what do you all use? 00:27:40.280 --> 00:27:40.780 All uv, 00:27:41.180 --> 00:27:41.360 yeah. 00:27:41.640 --> 00:27:45.460 I feel like uv was like the speed up 00:27:45.640 --> 00:27:47.460 for package management 00:27:48.340 --> 00:27:51.740 that like rivaled what AI agents did for coding, 00:27:52.070 --> 00:27:52.180 right? 00:27:52.360 --> 00:27:53.440 Like the chart we're seeing here, 00:27:53.490 --> 00:27:57.060 just like the delta and like pain that it took away 00:27:57.100 --> 00:27:58.220 was like pretty radical, 00:27:58.420 --> 00:27:58.980 which is exciting. 00:27:59.200 --> 00:27:59.460 So yeah, 00:27:59.680 --> 00:27:59.980 all uv, 00:28:00.240 --> 00:28:01.300 and I'm sure we'll get to this later, 00:28:01.460 --> 00:28:04.440 but I'm a really big fan of all of the developer tooling 00:28:05.000 --> 00:28:06.680 that Astral has developed. 00:28:06.760 --> 00:28:08.680 I think they're rolled into OpenAI now, 00:28:08.900 --> 00:28:10.420 but yeah, just really great stuff 00:28:10.560 --> 00:28:11.800 in terms of like developer experience 00:28:12.060 --> 00:28:14.900 and like attention to speed and productivity, 00:28:15.500 --> 00:28:16.080 which is super great. 00:28:16.360 --> 00:28:17.840 And I think it's really interesting to see now 00:28:18.280 --> 00:28:21.800 like how easy it is for agents to use those tools. 00:28:21.960 --> 00:28:23.660 And I think they've really like geared towards 00:28:24.280 --> 00:28:27.040 making the tools agent native for coding as well. 00:28:27.140 --> 00:28:29.100 It's got an unexpected AI angle as well. 00:28:29.100 --> 00:28:30.340 A lot of the AI tools really like it. 00:28:30.600 --> 00:28:33.540 I'm going to come back to some of the astral tools in this section, 00:28:33.760 --> 00:28:34.740 but in another section, 00:28:34.850 --> 00:28:34.960 rather. 00:28:35.380 --> 00:28:35.520 Pat, 00:28:35.930 --> 00:28:36.980 I know uv works on Windows, 00:28:37.240 --> 00:28:39.620 but in the DevOps world, you got other things or is it 00:28:39.820 --> 00:28:40.440 all uv? 00:28:40.720 --> 00:28:41.600 Funny you say that. 00:28:41.760 --> 00:28:42.660 I am all uv, 00:28:42.940 --> 00:28:47.040 but in the corporate environment, they have policies applied to the machines 00:28:47.460 --> 00:28:49.440 and just straight installing Python. 00:28:49.920 --> 00:28:50.420 That's a no-go. 00:28:51.200 --> 00:28:53.120 And uv kind of ends up in the same spot. 00:28:53.420 --> 00:28:54.600 But under WSL, 00:28:55.300 --> 00:28:57.260 then we're more free to do what we want. 00:28:57.800 --> 00:28:59.320 And I'm more of a heavy, 00:28:59.540 --> 00:29:01.760 like a PEP 723 user, 00:29:02.080 --> 00:29:04.520 the section at the top to declare your dependencies, 00:29:05.220 --> 00:29:08.760 because almost 90 some percent of what I write, 00:29:09.020 --> 00:29:10.560 it's a single file script. 00:29:11.240 --> 00:29:14.060 And then for side projects, I do virtual environments, 00:29:14.240 --> 00:29:16.300 but I use uv now rather than them. 00:29:16.640 --> 00:29:18.840 Yeah, that's really cool with the stuff at the top. 00:29:18.920 --> 00:29:20.700 You can almost treat it as a bash script. 00:29:20.780 --> 00:29:22.940 Like here's a file, you just run it, it's going to work, right? 00:29:23.060 --> 00:29:27.640 I saw someone put an interesting shebang on the top of one with that same inline includes. 00:29:27.800 --> 00:29:29.240 where it's just like UBRun. 00:29:29.560 --> 00:29:30.420 I was like, oh, this is genius. 00:29:30.640 --> 00:29:32.140 So it's change mod 755, 00:29:32.300 --> 00:29:32.980 this Python file. 00:29:33.500 --> 00:29:34.700 It's now fully executable. 00:29:35.299 --> 00:29:36.160 Nothing else needed. 00:29:36.340 --> 00:29:37.260 The magic just happens. 00:29:37.540 --> 00:29:40.620 I keep having to prompt my DevOps agents, 00:29:41.280 --> 00:29:44.680 the various agents to use the 723 section up there. 00:29:45.120 --> 00:29:47.420 And then one of them just threw that in, the shebang. 00:29:47.940 --> 00:29:48.440 And I'm like, oh, 00:29:48.620 --> 00:29:49.080 that's genius. 00:29:49.360 --> 00:29:49.940 Let's do that. 00:29:50.100 --> 00:29:50.720 Very, very cool. 00:29:51.020 --> 00:29:52.760 Sydney said something about Astral and OpenAI. 00:29:52.880 --> 00:29:55.700 And I think it's good that I had kind of forgotten 00:29:55.840 --> 00:29:56.479 that even happened 00:29:56.500 --> 00:29:59.660 because things just seem to have carried on as they were, 00:29:59.820 --> 00:29:59.940 right? 00:30:00.040 --> 00:30:03.920 Yeah, I had Charlie Marsh on a couple of months ago to talk about that and the changes 00:30:04.480 --> 00:30:07.760 and sort of look after he'd been inside OpenAI for a couple of months. 00:30:08.160 --> 00:30:09.520 And yeah, that's a good point. 00:30:09.740 --> 00:30:11.760 There hasn't been an, oh gosh, what are we going to do now? 00:30:12.040 --> 00:30:16.080 That was so 2025 or until it was gone and it's still going strong. 00:30:16.520 --> 00:30:16.600 Reuben. 00:30:16.780 --> 00:30:18.360 I'm also all in on uv. 00:30:18.820 --> 00:30:22.280 It took me a little bit of time to realize it was not just a super, 00:30:22.460 --> 00:30:23.420 super fast pip, 00:30:23.620 --> 00:30:26.000 but that was like the gateway drug to start using it. 00:30:26.340 --> 00:30:26.780 But like, 00:30:27.220 --> 00:30:27.640 think of it, 00:30:27.860 --> 00:30:32.700 how many times over the years do we go to Python conferences and there were two, 00:30:32.960 --> 00:30:33.060 three, 00:30:33.260 --> 00:30:35.860 four talks trying to explain packaging, 00:30:36.300 --> 00:30:37.200 all the different tools, 00:30:37.480 --> 00:30:39.520 all the different ways to do it, different versions of Python, on and on. 00:30:39.920 --> 00:30:42.640 And the talk would inevitably have in it somewhere, 00:30:43.220 --> 00:30:46.200 oh, if only we could solve this mess to make it easier for people. 00:30:46.640 --> 00:30:57.940 And the core developers were like, we are not entertaining any suggestions unless you come and actually solve these problems because it's so complex and you can suggest things all day long and we don't have time to even entertain these ideas. 00:30:58.350 --> 00:31:02.560 And along comes uv and it solves all the problems for, 00:31:02.720 --> 00:31:03.200 you know, 00:31:03.460 --> 00:31:04.900 for, you know, 00:31:05.120 --> 00:31:05.580 versioning, 00:31:05.780 --> 00:31:06.580 for packaging, 00:31:07.000 --> 00:31:08.140 for uploading to PyPI, 00:31:08.480 --> 00:31:09.520 for Python versions. 00:31:10.140 --> 00:31:11.320 And it does so super fast. 00:31:11.660 --> 00:31:14.340 Like, it's not a surprise that everyone's on board with it. 00:31:14.920 --> 00:31:16.100 And when, you know, 00:31:16.300 --> 00:31:17.880 Astral was bought by OpenAI, 00:31:18.050 --> 00:31:19.600 I know there were some people worried about it, 00:31:19.790 --> 00:31:24.340 but there's a lot of corporate sponsorship of open source and we've lived with it for years. 00:31:24.590 --> 00:31:30.160 And I think Astral has proven themselves so far to be very reliable stewards of this super amazing tool. 00:31:30.290 --> 00:31:30.440 Yeah. 00:31:30.620 --> 00:31:34.480 And it's not hindered by the release cycle of Python, the language itself either, 00:31:34.900 --> 00:31:36.140 which is another big benefit here. 00:31:36.300 --> 00:31:37.040 Yeah, it's super interesting. 00:31:37.460 --> 00:31:38.360 And how long will it be? 00:31:38.730 --> 00:31:41.940 This, I think, is maybe one of the most important things that it's done. 00:31:42.040 --> 00:31:47.540 How long will it be until I could just uv Python install 3.15 a day? 00:31:47.920 --> 00:31:50.460 Will that work today or tomorrow with Python Builds Channel? 00:31:50.600 --> 00:31:55.460 It worked this week when I did the Python Bytes demo of the dungeon crawler for the What's New. 00:31:55.600 --> 00:31:55.980 That's right. 00:31:56.100 --> 00:31:57.040 Yeah, that used the RC, 00:31:57.800 --> 00:31:58.820 whatever the last RC was, right? 00:31:59.040 --> 00:31:59.940 I did this months ago. 00:32:00.000 --> 00:32:02.280 I had a playlist that I did on my YouTube channel of like, 00:32:02.440 --> 00:32:03.800 get ready for Python 3.15. 00:32:04.020 --> 00:32:06.140 And how did I tell people to install it with uv? 00:32:06.580 --> 00:32:07.480 And it just worked. 00:32:07.520 --> 00:32:08.220 It was extraordinary. 00:32:08.600 --> 00:32:10.940 So for all the benefits that you all extolled about it, 00:32:11.190 --> 00:32:14.860 I think one of its killer features is it lives outside Python. 00:32:15.400 --> 00:32:17.640 And it can be the way into Python. 00:32:17.980 --> 00:32:18.560 So for example, 00:32:19.080 --> 00:32:21.240 with Pat's way of I want to run this thing, 00:32:21.740 --> 00:32:23.560 this script that I built, and here's its dependencies, 00:32:23.730 --> 00:32:25.120 and it needs a VNV and all that sort of stuff. 00:32:25.400 --> 00:32:28.160 You don't even have to have Python on the machine at all. 00:32:28.490 --> 00:32:29.380 And that still works, 00:32:29.750 --> 00:32:29.860 right? 00:32:29.960 --> 00:32:31.760 And there was none of the other packaging tools, 00:32:31.880 --> 00:32:34.840 as much as I liked all of them, none of them could do that. 00:32:35.010 --> 00:32:36.300 It was always you set up Python, 00:32:36.510 --> 00:32:37.720 then you install this into it, 00:32:37.780 --> 00:32:40.120 and then you can start having tooling conversations, 00:32:40.500 --> 00:32:40.560 right? 00:32:40.700 --> 00:32:40.820 Yeah, 00:32:41.000 --> 00:32:41.820 just a few years back, 00:32:41.840 --> 00:32:45.200 I was really considering a container-only approach to this 00:32:45.400 --> 00:32:47.680 because installing Python and installing dependencies 00:32:47.760 --> 00:32:49.920 and installing packages has just gotten so difficult. 00:32:50.340 --> 00:32:51.860 But the containerization piece worked. 00:32:52.280 --> 00:32:53.320 But uv changed all that for me. 00:32:53.420 --> 00:32:55.760 Like it just now made it that same convenient 00:32:56.240 --> 00:32:57.960 kind of one-liner to get what I needed. 00:32:58.180 --> 00:33:00.000 I'll add that virtual environments, 00:33:00.580 --> 00:33:02.820 virtual environments have like been the solution 00:33:03.140 --> 00:33:04.900 to the conflicts between different projects. 00:33:05.400 --> 00:33:05.860 And everyone, 00:33:06.080 --> 00:33:07.740 I mean, I certainly hated them 00:33:07.760 --> 00:33:10.220 like dealing with them and activate and deactivate and where am I? 00:33:10.500 --> 00:33:11.360 And uv is like, 00:33:11.600 --> 00:33:14.980 oh, we'll just hide that from you. You don't like, we're going to take care of it, 00:33:15.280 --> 00:33:16.720 but you don't have to actually do anything. 00:33:17.100 --> 00:33:17.560 And for me, 00:33:17.740 --> 00:33:19.100 that was like magical. 00:33:19.280 --> 00:33:21.980 I mean, it's so bad that XKCD made a comic about the, 00:33:22.280 --> 00:33:24.340 my laptop is now a super fun site. 00:33:24.820 --> 00:33:25.900 For those of you guys out of the US, 00:33:26.120 --> 00:33:27.820 a super fun site is something so polluted, 00:33:27.940 --> 00:33:32.100 it requires extraordinary measures to like protect the public from it. 00:33:32.520 --> 00:33:37.720 Let's give Sydney the last word here. I think that the fact that uv can actually bring 00:33:37.720 --> 00:33:42.300 Python to the machine or to the environment is why AI is really liked it, 00:33:42.580 --> 00:33:42.660 right? 00:33:42.660 --> 00:33:45.720 And really make use of it because it just removes all those details, 00:33:46.460 --> 00:33:46.540 right? 00:33:46.580 --> 00:33:47.120 What do you think, Sydney? 00:33:47.260 --> 00:33:52.940 I think this like generally connects to this idea with agents and models that like they're 00:33:53.360 --> 00:33:54.900 really good at writing code. 00:33:55.320 --> 00:33:57.400 And the reason that is, is kind of twofold. 00:33:57.500 --> 00:34:01.840 One is like they're trained on a bunch of code and a lot of code is like verifiable, 00:34:02.060 --> 00:34:02.120 right? 00:34:02.240 --> 00:34:05.940 Like you can confirm that like something runs or it doesn't, a test passes or it doesn't. 00:34:06.180 --> 00:34:09.840 And then they're also like post-trained on code syntax. 00:34:10.470 --> 00:34:15.600 So these tools now are like very familiar at the like weights level for models, 00:34:15.810 --> 00:34:15.960 right? 00:34:16.520 --> 00:34:18.300 And I think uv, 00:34:18.940 --> 00:34:19.040 obviously, 00:34:19.649 --> 00:34:26.260 even before AI emerged or before like prolific AI emerged as like the leading package manager. 00:34:26.330 --> 00:34:27.940 And so I think AI tools, 00:34:28.659 --> 00:34:34.020 AI models are able to like use it effectively to control an environment where code is run. 00:34:34.419 --> 00:34:34.820 And ultimately, 00:34:35.040 --> 00:34:38.879 like models don't only need to be good at writing code for like the coding 00:34:39.020 --> 00:34:45.220 agent case, but like general agents also use code as like a means to achieve an end. 00:34:45.240 --> 00:34:45.659 And so, 00:34:45.899 --> 00:34:48.679 yeah, I think it's pretty exciting that like this performant tool came out and 00:34:48.679 --> 00:34:52.260 is now like a super fundamental part of how those models operate. 00:34:52.460 --> 00:34:54.620 There's a reason that OpenAI acquired Astro. 00:34:54.820 --> 00:34:54.940 Right, 00:34:55.100 --> 00:34:55.240 right. 00:34:55.500 --> 00:34:55.620 Exactly. 00:34:55.899 --> 00:34:57.300 It's not just that we need things, 00:34:57.600 --> 00:34:57.880 I'm sure. 00:34:57.960 --> 00:34:58.020 Yeah. 00:34:58.260 --> 00:34:58.440 Somehow. 00:34:58.660 --> 00:34:59.440 You might also note, 00:34:59.600 --> 00:35:02.579 like if you're using kind of the latest and greatest models like Astra 00:35:02.600 --> 00:35:04.420 or Opus 5.5, 00:35:04.600 --> 00:35:07.760 like they are really heavy on the code writing, 00:35:08.040 --> 00:35:10.040 like in your Claude Code session. 00:35:10.190 --> 00:35:12.060 If you're looking at like the tools that it's calling, 00:35:12.260 --> 00:35:14.340 it's no longer like read file, write file. 00:35:14.520 --> 00:35:15.300 It's like bash, 00:35:15.480 --> 00:35:15.540 bash, 00:35:15.780 --> 00:35:16.180 bash with, 00:35:16.460 --> 00:35:18.260 you know, all these like complex scripts and things. 00:35:18.820 --> 00:35:20.220 So definitely plays into that as well. 00:35:20.300 --> 00:35:23.000 Yeah, cool. Or you'll see a Python dash C and some giant string. 00:35:23.320 --> 00:35:25.740 Which I don't like because then it makes me kind of approve it. 00:35:25.900 --> 00:35:29.480 I think the harnesses need to grow some more Python native tools. 00:35:29.930 --> 00:35:31.260 So they just call those naturally. 00:35:31.620 --> 00:35:33.120 Yeah, then I can just call eval. It'll be all fine. 00:35:33.160 --> 00:35:34.680 I mean, look at what we're doing already. 00:35:36.440 --> 00:35:40.760 My one interesting kind of observability note here is like, 00:35:41.060 --> 00:35:46.180 or I guess I'm curious why you would prefer for them to call more like Python native tools. 00:35:46.580 --> 00:35:49.080 I think mostly for the reporting aspect of it. 00:35:49.120 --> 00:35:52.360 I've been using agents view as a tool to report on my sessions 00:35:52.500 --> 00:35:54.980 to kind of evaluate how well I'm performing. 00:35:55.120 --> 00:35:58.420 I'm performing with the agent to get better output. 00:35:58.700 --> 00:36:08.260 So I watch and when I see 85% of the session was just bash tool calls, I'm like, I wish it kind of knew more granularly what those things were at a glance. 00:36:08.860 --> 00:36:09.800 So I like this. 00:36:10.370 --> 00:36:12.000 Michael just brought that up on the agents feed.io. 00:36:12.680 --> 00:36:15.580 It is really interesting to kind of analyze how you're doing. 00:36:16.050 --> 00:36:18.740 You now are feeding back into the human part of this, 00:36:19.180 --> 00:36:22.600 how we can interoperate more natively with these tools. 00:36:22.840 --> 00:36:25.660 This is like definitely connected to what I think about a lot. 00:36:26.000 --> 00:36:26.240 I think, 00:36:26.530 --> 00:36:30.600 I don't know if it needs to be like specific tools that are attached to the agent 00:36:30.790 --> 00:36:33.540 that are like more discreet, like Python operations. 00:36:33.810 --> 00:36:36.100 But I do think having like observability into like, 00:36:36.300 --> 00:36:37.780 is this thing operating efficiently? 00:36:38.100 --> 00:36:41.740 Is that like a result of its structure or of my like, 00:36:41.920 --> 00:36:43.380 am I doing a bad job prompting it 00:36:43.440 --> 00:36:43.940 for this task? 00:36:44.010 --> 00:36:46.440 Or is it like not well equipped to handle this task? 00:36:46.600 --> 00:36:49.460 Like this one will give you a grade on your context window usage. 00:36:50.020 --> 00:36:53.940 And it'll score you on like bad behaviors and point them out. 00:36:54.040 --> 00:36:54.620 AgentP is great. 00:36:54.660 --> 00:36:55.880 This is totally new to me. 00:36:55.960 --> 00:36:57.300 and this is amazing looking. 00:36:57.560 --> 00:36:59.440 We're in serious danger of going down the hole 00:36:59.710 --> 00:37:01.300 and having a too long of a show. 00:37:01.979 --> 00:37:03.140 But Agent's View is amazing. 00:37:03.620 --> 00:37:04.860 Calvin and I kind of discovered it 00:37:04.860 --> 00:37:06.120 at the same time on Python Bytes. 00:37:06.560 --> 00:37:08.700 And I think you got it set this way as well, 00:37:08.820 --> 00:37:10.280 but I've got it set up where there's a 00:37:10.940 --> 00:37:12.620 daemon server sort of version 00:37:12.690 --> 00:37:13.920 so I can view it from the web. 00:37:14.260 --> 00:37:15.000 But more importantly, 00:37:15.560 --> 00:37:15.780 normally, 00:37:16.180 --> 00:37:17.960 if you install it, it just works on your one machine. 00:37:18.400 --> 00:37:19.240 But I've got my Mac Mini, 00:37:19.420 --> 00:37:20.160 I've got my Mac Studio, 00:37:20.320 --> 00:37:20.920 I've got my MacBook. 00:37:21.300 --> 00:37:24.100 I want a holistic view of all the AI things that I've done. 00:37:24.480 --> 00:37:28.660 And so you can get them all to push to some central little Postgres Docker container. 00:37:29.100 --> 00:37:31.780 And then you get this view over of everything. 00:37:32.120 --> 00:37:35.620 I have that set up so all my machines feed into that one database. 00:37:35.960 --> 00:37:38.240 And so I get a holistic view of all my AI usage. 00:37:38.480 --> 00:37:43.740 And Reuben, the most scary part is it shows you the cost of tokens if you did it all on true cost, 00:37:43.980 --> 00:37:44.620 which is... 00:37:44.660 --> 00:37:47.000 I fear we do need another podcast for this. 00:37:47.140 --> 00:37:48.280 This is super interesting. 00:37:48.560 --> 00:37:49.240 Y'all do some research. 00:37:49.340 --> 00:37:51.160 We might just circle back in a few weeks. 00:37:52.160 --> 00:37:52.480 Okay, 00:37:52.760 --> 00:37:55.140 so I thought we had nothing to say about uv other than just, 00:37:55.300 --> 00:37:56.180 yeah, you're right, Michael. 00:37:56.340 --> 00:37:56.880 We all agree. 00:37:57.340 --> 00:37:57.680 Web frameworks. 00:37:58.340 --> 00:38:00.680 Sebastian was supposed to be here, but he couldn't make it. 00:38:00.900 --> 00:38:06.000 So maybe we can start with FastAPI to sort of shout out his little project. 00:38:06.380 --> 00:38:07.120 Have you all heard of this? 00:38:07.280 --> 00:38:08.160 I've heard of FastAPI. 00:38:08.280 --> 00:38:10.400 I actually just gave a talk about it at DjangoCon. 00:38:10.680 --> 00:38:19.380 So Frank Wiles and I got up on stage and compared what using Django with, say, Django Ninja would be like compared to FastAPI and maybe when one's better for another. 00:38:20.020 --> 00:38:21.780 FastAPI is definitely still faster. 00:38:21.880 --> 00:38:24.660 but not by like a huge, huge amount. 00:38:25.400 --> 00:38:26.400 There's definitely instances 00:38:26.660 --> 00:38:28.700 where this is the better framework to reach for 00:38:29.240 --> 00:38:30.440 if you've got specific use cases 00:38:30.700 --> 00:38:31.920 because you don't have to bring along 00:38:32.040 --> 00:38:34.000 a whole big surface area 00:38:34.140 --> 00:38:36.180 of all the awesomeness that Django gives you. 00:38:36.520 --> 00:38:39.060 So we use both and we pick and choose 00:38:39.260 --> 00:38:41.640 which one fits the task best. 00:38:41.820 --> 00:38:43.080 If you have something that may grow 00:38:43.180 --> 00:38:43.700 into something bigger, 00:38:44.180 --> 00:38:45.680 Django Pi may be a better place to start. 00:38:46.300 --> 00:38:47.780 If you've got some quick and dirty, 00:38:47.880 --> 00:38:50.960 you need to get good quality APIs up 00:38:50.980 --> 00:38:52.780 for some specific use case. 00:38:53.300 --> 00:38:55.260 I use FastAPI for some personal projects like that 00:38:55.390 --> 00:38:57.360 because it's just, I wanted a quick way, 00:38:57.780 --> 00:39:02.060 the tooling around FastAPI and the skills that he provides. 00:39:02.200 --> 00:39:05.740 So Sebastian's also very big into library skills 00:39:05.850 --> 00:39:06.900 as his other repository, 00:39:07.680 --> 00:39:10.260 where he's built a package to produce skills 00:39:10.580 --> 00:39:11.900 for other open source packages, 00:39:12.460 --> 00:39:14.520 which means that this has a really rich set of skills 00:39:14.740 --> 00:39:16.380 that your agent can use when coding 00:39:16.810 --> 00:39:18.940 to be very effective on the best practices 00:39:19.200 --> 00:39:20.240 for how to use FastAPI. 00:39:20.400 --> 00:39:23.920 So he's approaching it from an AI-native coding perspective 00:39:24.060 --> 00:39:26.360 that this should be used with those harnesses 00:39:26.780 --> 00:39:29.800 and you'll get hopefully best practice quality code out of it. 00:39:29.860 --> 00:39:31.020 We've got to make sure that we keep going 00:39:31.300 --> 00:39:33.300 because at the end of our list, 00:39:33.340 --> 00:39:36.140 I have a documentation system that is AI-native 00:39:36.140 --> 00:39:38.920 and I absolutely think is underappreciated. 00:39:39.320 --> 00:39:39.640 But to me, 00:39:39.800 --> 00:39:42.020 FastAPI is not about fast performance. 00:39:42.580 --> 00:39:45.080 It's like fast development of doing things right 00:39:45.180 --> 00:39:48.940 and comprehensively with Pydantic and OpenAPI 00:39:48.960 --> 00:39:52.440 on all those different things that it brings and validation. 00:39:52.760 --> 00:39:54.160 I mean, if that's your true concern, 00:39:54.530 --> 00:39:56.600 I think you'd be really pleased with Django Ninja. 00:39:57.000 --> 00:39:58.580 You can basically change one import 00:39:58.920 --> 00:40:00.560 and the code looks almost identical 00:40:00.780 --> 00:40:02.760 from a FastAPI project to a Django Ninja project. 00:40:03.600 --> 00:40:07.100 So I'll still put in a big plug for Django in this case. 00:40:07.320 --> 00:40:09.720 100% Django is on this list, I think. 00:40:09.820 --> 00:40:11.020 I want to hand this off to Sydney 00:40:11.170 --> 00:40:13.300 because I think those two things that you brought up, Calvin, 00:40:13.680 --> 00:40:16.740 both have a central pillar and that's Pydantic, 00:40:17.220 --> 00:40:18.800 which I know you've done a lot with. 00:40:18.860 --> 00:40:25.240 I actually was first using Django and trying to get Pydantic to work well with it. 00:40:25.390 --> 00:40:28.060 When I first interacted with Pydantic open source, 00:40:28.270 --> 00:40:29.800 I had used Pydantic previously, 00:40:30.460 --> 00:40:36.300 but that's really how I got my open source start and then started using FastAPI very soon after that. 00:40:36.370 --> 00:40:37.540 This was years and years ago, 00:40:37.700 --> 00:40:45.440 but I think Pydantic has filled a gap that Python had in terms of it's not a strongly typed language. 00:40:45.760 --> 00:40:48.240 And there's a lot of challenges that come along with that. 00:40:48.620 --> 00:40:53.740 And so it's kind of this bridge into like, if you would like to opt into more strongly typed things, 00:40:54.260 --> 00:40:56.960 then you can use Pydantic for data modeling and validation. 00:40:57.640 --> 00:41:00.540 And it makes sense that it's kind of the backbone of FastAPI, 00:41:00.720 --> 00:41:04.240 just in that you need kind of structured data for requests and responses, 00:41:04.420 --> 00:41:05.760 which are really like the bread and butter there. 00:41:06.180 --> 00:41:13.340 And it's really exciting to see how Pydantic was first, I think really first popularized probably through FastAPI. 00:41:13.790 --> 00:41:23.600 And then has really been further popularized with the emergence of LLMs because they also kind of need a bridge into like more structured data, right? 00:41:23.740 --> 00:41:27.380 Like LLMs produce text and Pydantic is kind of the bridge back into like, 00:41:27.720 --> 00:41:32.940 let's validate that data and turn it into a predictable model that we can then use in code driven applications. 00:41:33.920 --> 00:41:35.940 Here I am again on my like AI rant, 00:41:36.180 --> 00:41:36.820 but yeah. 00:41:36.960 --> 00:41:38.080 There's some other angles as well. 00:41:38.110 --> 00:41:39.960 I think I want to save them for later. 00:41:40.030 --> 00:41:40.720 But yeah, I think this, 00:41:40.940 --> 00:41:43.720 I think Pidantic fits very well with AI things. 00:41:44.060 --> 00:41:44.620 So interesting. 00:41:45.160 --> 00:41:45.380 Pat, 00:41:45.700 --> 00:41:47.520 what's the web story from your world? 00:41:47.530 --> 00:41:53.100 I typically end up using Flask for anything at work just because there's so much source 00:41:53.210 --> 00:41:59.240 out there that the LLMs have ingested and can easily spit something out that works. 00:41:59.700 --> 00:42:01.620 And for me, it's more quick and dirty. 00:42:02.100 --> 00:42:05.900 an end user application that faces customers or anything like that. 00:42:06.130 --> 00:42:07.060 For personal projects, 00:42:07.170 --> 00:42:10.400 I use Robyn and I know that you kind of evaluated Robyn. 00:42:11.120 --> 00:42:12.500 What, what interested me is it's very 00:42:12.740 --> 00:42:18.200 flask like and written in Rust and supposed to be multi-threaded. 00:42:18.560 --> 00:42:20.960 And I found it to be very usable 00:42:21.120 --> 00:42:21.340 anyway. 00:42:21.670 --> 00:42:23.240 A couple of things bit me where the, 00:42:23.670 --> 00:42:25.640 there's not near as much code out there for the 00:42:25.660 --> 00:42:27.320 the language models to look at. 00:42:27.580 --> 00:42:27.700 So, 00:42:27.880 --> 00:42:27.980 uh, 00:42:28.360 --> 00:42:28.480 um, 00:42:28.800 --> 00:42:31.940 somehow you have to herd your agent to, 00:42:32.180 --> 00:42:35.540 to write the correct code. I have a whole AI section, but I don't know if we're even going 00:42:35.540 --> 00:42:38.080 to be able to get to it because we're just going down this little edge on each one, 00:42:38.200 --> 00:42:39.640 which I think is very interesting. 00:42:40.100 --> 00:42:41.520 But one thing I've been doing for projects, 00:42:41.860 --> 00:42:47.220 especially like Robyn and others that are not as well-known is I will literally download the 00:42:47.400 --> 00:42:48.000 entire repo, 00:42:48.260 --> 00:42:49.500 which often has the documentation. 00:42:50.200 --> 00:42:52.420 And then I'll open up a Claude Code session. 00:42:55.220 --> 00:42:55.620 I'll 00:42:55.640 --> 00:42:57.320 documentation and the quick starts, 00:42:57.960 --> 00:43:02.500 ground it in the code and generate me a skill or some kind of 00:43:02.780 --> 00:43:07.420 thing that I can then give to my Claude Code or whatever on the project I'm actually trying to 00:43:07.510 --> 00:43:12.940 build with this and have it kind of create a self-teaching thing and say, you always use the 00:43:13.000 --> 00:43:16.440 Robyn agent for building this because you don't know it well enough or something along those lines, 00:43:16.540 --> 00:43:18.760 which sounds a little bit like what Sebastian was doing. 00:43:19.200 --> 00:43:20.800 And that's been very successful for me. 00:43:21.140 --> 00:43:23.480 Reuben. So I don't do that much web stuff. 00:43:23.850 --> 00:43:26.460 And for a long time when I did, it was just like, well, 00:43:26.510 --> 00:43:29.200 throw something together with Flask just because small, 00:43:29.640 --> 00:43:29.820 simple, 00:43:30.090 --> 00:43:31.180 I didn't need all the overhead of 00:43:31.300 --> 00:43:31.480 Django. 00:43:32.180 --> 00:43:32.360 Lately, 00:43:32.920 --> 00:43:34.420 when I've been doing web stuff, 00:43:34.720 --> 00:43:36.480 like the back end of the practice system I have 00:43:36.560 --> 00:43:38.040 on my site, 00:43:38.260 --> 00:43:41.600 I ended up using FastAPI and that's fine too. 00:43:41.880 --> 00:43:43.500 But I tend to be sort of just neutral 00:43:43.720 --> 00:43:48.900 on this in part because it plays an important but small role in the applications that I'm running. 00:43:49.200 --> 00:43:50.440 FastAPI works fantastic. 00:43:51.010 --> 00:43:53.260 And the fact that there's a ton, a ton of, you were saying, 00:43:53.720 --> 00:43:54.820 like Argya says, Pat was saying, 00:43:55.130 --> 00:43:59.980 the fact there's a ton of existing code out there in both Flask and in FastAPI, 00:44:00.310 --> 00:44:01.040 and for that matter in Django, 00:44:01.240 --> 00:44:03.860 means that the LLMs are very good at generating and working with it. 00:44:03.920 --> 00:44:05.260 I'm on Team Flask as well. 00:44:05.820 --> 00:44:09.120 Most of my web apps are Flask, but I also have some FastAPI things. 00:44:09.460 --> 00:44:10.760 All right, we got to keep, go ahead. 00:44:10.920 --> 00:44:16.360 One sad moment there is I wish Pyramid would have made a bigger go and become more prominent. 00:44:16.400 --> 00:44:22.340 I really enjoyed the Pyramid framework, but sadly, it's not an option today, really. 00:44:22.480 --> 00:44:22.840 I agree. 00:44:22.980 --> 00:44:25.540 I was a believer and longtime user of Pyramid. 00:44:25.920 --> 00:44:29.020 All right, let's move on here just for the sake of time. 00:44:29.480 --> 00:44:31.880 I'm going to skip a couple and let's talk type hints. 00:44:32.400 --> 00:44:32.800 Type hints, 00:44:33.100 --> 00:44:34.940 to type hint or not to type hint. 00:44:35.100 --> 00:44:36.520 That is the type question. 00:44:36.760 --> 00:44:38.300 Well, I think Sydney needs to kick off this one. 00:44:38.500 --> 00:44:39.000 Yes, exactly. 00:44:39.140 --> 00:44:40.640 Yeah, I'm smiling over here. 00:44:41.060 --> 00:44:41.240 Yeah, 00:44:41.600 --> 00:44:41.820 TY. 00:44:42.360 --> 00:44:42.640 Hi, 00:44:42.900 --> 00:44:43.020 TY. 00:44:43.360 --> 00:44:44.700 I don't know what people prefer. 00:44:45.000 --> 00:44:46.480 I asked Charlie and he said ty. 00:44:46.840 --> 00:44:47.000 Okay, 00:44:47.210 --> 00:44:47.340 great. 00:44:47.620 --> 00:44:48.120 Because I'm like, 00:44:48.360 --> 00:44:48.480 Charlie, 00:44:49.020 --> 00:44:49.780 we need some help here. 00:44:49.780 --> 00:44:50.980 You can't just leave this hanging. 00:44:50.990 --> 00:44:52.020 You got to tell us how to say it. 00:44:52.020 --> 00:44:52.420 He said ty. 00:44:52.680 --> 00:44:53.940 TY is excellent. 00:44:54.180 --> 00:44:57.600 I obviously am a big fan of the tools from the Astral team. 00:44:58.020 --> 00:45:00.600 Not surprising given my tenure at Pydantic, 00:45:00.860 --> 00:45:04.280 but I'm very big on like typed code when possible. 00:45:04.390 --> 00:45:10.520 I think the Python core team has also done a ton of work to make expressing types clearly 00:45:11.000 --> 00:45:12.300 and well in Python, 00:45:12.670 --> 00:45:14.440 like much easier even without Pydantic. 00:45:14.460 --> 00:45:18.740 And it's super exciting now with like ty that type checking can be so fast. 00:45:19.200 --> 00:45:20.000 And I think, 00:45:20.360 --> 00:45:26.460 you know, where developers used to view typing things as like maybe a pain or like more boilerplate E agents are like, 00:45:26.780 --> 00:45:27.880 take a lot of that load off. 00:45:27.920 --> 00:45:29.740 So I think the more we can do this, the better. 00:45:30.200 --> 00:45:30.280 Yeah, 00:45:30.600 --> 00:45:31.760 no surprising opinions for me. 00:45:31.840 --> 00:45:32.860 No, I kind of expected that. 00:45:32.940 --> 00:45:33.220 That's great. 00:45:33.400 --> 00:45:33.520 Yeah, 00:45:33.700 --> 00:45:34.980 I'm probably on the other side of that. 00:45:35.000 --> 00:45:40.500 I mean, I recognize the value of type hints and typing for the LLMs and the harnesses, etc. 00:45:40.920 --> 00:45:41.440 and IDEs. 00:45:42.300 --> 00:45:46.340 But one of the things that brought me to Python the language was the stay out of your way, 00:45:46.740 --> 00:45:48.200 get stuff done quickly, 00:45:48.620 --> 00:45:49.600 efficiency of the language. 00:45:50.100 --> 00:45:56.460 I feel like that came with the duck typing and the use of white space as a meaningful piece where I wasn't writing. 00:45:57.000 --> 00:46:00.100 One of the things that always bugged me about other languages was a lot of the ASCII art, 00:46:00.380 --> 00:46:05.820 the hash rockets and yada yada stuff that Python never brought to the table and kept clean. 00:46:06.260 --> 00:46:07.780 And so when I code, 00:46:07.940 --> 00:46:09.740 if I ever hand code anything, 00:46:09.960 --> 00:46:11.800 I'm not using type hints typically. 00:46:12.420 --> 00:46:25.240 Although if I'm using the AIs and the harnesses, I want them to do that because I want to have that determinism when the code is running through tests and catching those kinds of those classes of bugs before they even run. 00:46:25.720 --> 00:46:26.740 So I'm for it. 00:46:26.950 --> 00:46:29.040 I just that's not why I came to Python. 00:46:29.280 --> 00:46:30.040 You'll stay anyway, 00:46:30.300 --> 00:46:31.180 but that's not why you're here. 00:46:31.190 --> 00:46:32.000 I will stay anyway. 00:46:32.280 --> 00:46:32.620 I'm here. 00:46:32.860 --> 00:46:33.180 I'm here. 00:46:33.340 --> 00:46:33.940 I'm not going anywhere. 00:46:34.250 --> 00:46:34.960 I love the language. 00:46:35.560 --> 00:46:38.160 And I'm glad it's still an option. 00:46:38.410 --> 00:46:40.180 Like it's, I'm not forced to use it. 00:46:40.430 --> 00:46:41.220 Like there are other languages 00:46:41.310 --> 00:46:42.960 where you can't even code in the language 00:46:43.070 --> 00:46:45.960 without touching that aspect of the syntax. 00:46:46.270 --> 00:46:48.200 And I'm glad Python still allows me 00:46:48.280 --> 00:46:50.040 to write what I love about it. 00:46:50.100 --> 00:46:51.160 Pat Rubin, what do you think? 00:46:51.380 --> 00:46:51.920 Yeah, with DevOps, 00:46:52.600 --> 00:46:54.240 it's mostly single file scripts. 00:46:54.480 --> 00:46:58.020 So I don't type in any of those side projects. 00:46:58.520 --> 00:47:00.080 There's heavy AI involved there. 00:47:00.360 --> 00:47:01.260 So usually I do. 00:47:01.400 --> 00:47:03.600 And some of those may end up being public at some point. 00:47:03.780 --> 00:47:05.460 So it makes sense to type in those. 00:47:05.660 --> 00:47:05.740 Yeah. 00:47:05.750 --> 00:47:09.020 And even if you don't ask it, a lot of times AI just does it anyway. 00:47:09.240 --> 00:47:10.160 I'm kind of like Calvin, 00:47:10.520 --> 00:47:14.220 where I grew up as like a dynamic language kind of guy, 00:47:14.880 --> 00:47:14.980 sort 00:47:14.980 --> 00:47:15.260 of saying, 00:47:15.460 --> 00:47:15.540 oh, 00:47:15.780 --> 00:47:16.980 come on, all these types, 00:47:17.290 --> 00:47:18.900 like, really, 00:47:19.130 --> 00:47:20.180 do we need it? 00:47:20.460 --> 00:47:23.340 And then like, you're getting to the, well, I guess if you're not a good programmer, then 00:47:23.390 --> 00:47:23.780 you need it. 00:47:24.000 --> 00:47:25.600 But if you're like a cool person, 00:47:25.780 --> 00:47:25.880 like, 00:47:26.040 --> 00:47:27.500 you know, smart person like me, then you don't. 00:47:27.760 --> 00:47:29.840 And then I started to discover a few different things. 00:47:30.200 --> 00:47:31.060 Like, first of all, 00:47:31.220 --> 00:47:32.040 if you have a team, 00:47:32.580 --> 00:47:36.220 not everyone on the team is necessarily schooled in everything, 00:47:36.720 --> 00:47:37.720 reading all the documentation, 00:47:38.040 --> 00:47:39.620 reading all the API stuff, on, on, on. 00:47:39.720 --> 00:47:40.480 They can make mistakes. 00:47:41.260 --> 00:47:45.740 And also the people coming from Staggly type languages feel like there's a safety net. 00:47:45.940 --> 00:47:48.000 Like they like the fact that there's typing there. 00:47:48.460 --> 00:47:50.920 And so all this together makes me feel like, 00:47:51.120 --> 00:47:51.220 yeah, 00:47:51.600 --> 00:47:52.740 I'm all in now. 00:47:52.960 --> 00:47:54.920 Maybe not in necessarily writing my own code directly, 00:47:55.440 --> 00:47:56.940 but very much so in LLMs. 00:47:56.960 --> 00:47:59.100 I think it's a great way to keep them on a short leash. 00:47:59.640 --> 00:48:02.420 And I recently and embarrassingly recently realized, 00:48:02.580 --> 00:48:02.720 okay, 00:48:02.880 --> 00:48:03.620 type-ins are great, 00:48:04.000 --> 00:48:04.520 but Pydantic 00:48:04.700 --> 00:48:05.120 is amazing. 00:48:05.480 --> 00:48:10.520 It's like the difference between having the Postgres columns being typed and having the 00:48:10.760 --> 00:48:13.240 check sort of stuff on them saying, 00:48:13.420 --> 00:48:14.860 well, it must be greater than zero, it must be less 00:48:14.900 --> 00:48:15.200 than this. 00:48:15.580 --> 00:48:18.940 You can really, really tighten the belt there on what's acceptable. 00:48:19.500 --> 00:48:20.440 I think it's a great thing. 00:48:20.620 --> 00:48:26.020 And kudos also to the core developers who have managed to do this so elegantly and allow 00:48:26.020 --> 00:48:28.840 you to sort of slide into it at whatever pace is good for you. 00:48:28.940 --> 00:48:29.540 I'm a big fan. 00:48:29.880 --> 00:48:30.300 as well. 00:48:30.570 --> 00:48:32.160 I think I grew up different. 00:48:32.680 --> 00:48:34.500 I learned my first language was C++, 00:48:34.710 --> 00:48:35.400 then I did a bunch of 00:48:35.490 --> 00:48:35.880 C#. 00:48:36.380 --> 00:48:40.780 So even type Python looks clean to my eyes of what I have been doing. 00:48:41.460 --> 00:48:41.940 But that said, 00:48:41.970 --> 00:48:44.540 I think there's a time and place for these kinds of things. 00:48:44.720 --> 00:48:45.760 So maybe on some sort of, 00:48:45.900 --> 00:48:46.680 if you have a package, 00:48:46.960 --> 00:48:50.800 all the public functions should be typed so that things using it can verify 00:48:51.000 --> 00:48:53.900 they're doing it correctly. But that doesn't necessarily mean the internals have to be 00:48:54.400 --> 00:48:56.000 completely blown up with typends everywhere, 00:48:56.260 --> 00:48:56.360 right? 00:48:56.580 --> 00:48:57.580 It's there to communicate. 00:48:57.940 --> 00:48:58.839 I originally 00:48:58.860 --> 00:49:01.840 really love the idea because it makes editors work better. 00:49:02.200 --> 00:49:03.540 You see a squiggly, that probably 00:49:03.660 --> 00:49:06.740 means you're doing something wrong and it can figure that out where otherwise it wouldn't. 00:49:07.020 --> 00:49:10.900 But these days it's also keeping AIs in line as some of y'all have mentioned. 00:49:11.180 --> 00:49:11.700 So yeah, 00:49:12.060 --> 00:49:12.600 for the, 00:49:13.000 --> 00:49:14.880 when I have any of my agents do anything, 00:49:15.140 --> 00:49:15.420 I'm like, 00:49:15.700 --> 00:49:16.360 you're going to run ty, 00:49:16.900 --> 00:49:17.700 you're going to run ruff, 00:49:18.080 --> 00:49:22.120 you're going to run the tests and you don't call yourself done or you don't consider it correct 00:49:22.400 --> 00:49:24.380 until all those things pass, you know, and just keep them, 00:49:24.800 --> 00:49:26.060 keep it all zero warnings. 00:49:26.840 --> 00:49:26.920 Like, 00:49:27.140 --> 00:49:27.360 you know, 00:49:27.720 --> 00:49:28.500 I think that's the way. 00:49:28.680 --> 00:49:41.300 I sometimes say to people that the things we can tell agents and LLMs are the sorts of things that would get us like written up for being of abusive behavior as a manager if we were to do all these things to the software developers on our team. 00:49:41.840 --> 00:49:42.640 But if like, 00:49:42.880 --> 00:49:48.020 you know, they can't complain or they'll call me a genius even after I forced to do this stuff, 00:49:48.360 --> 00:49:49.120 then I'm all in. 00:49:49.120 --> 00:49:56.000 I think some people would have considered themselves to be pedantic about these kinds of errors and not clean them up. 00:49:56.260 --> 00:49:57.560 And so I feel like this is a good thing. 00:49:57.810 --> 00:50:00.540 We're not going to get through a whole list, but I think an important section. 00:50:01.100 --> 00:50:04.960 Well, OK, let's just go go round really quick and ran out of this type checking one. 00:50:05.460 --> 00:50:06.600 What type checker do you use? 00:50:06.900 --> 00:50:08.220 Is it all mypy or what are we doing? 00:50:08.430 --> 00:50:11.960 I use mypy because I'm embarrassingly old school with this. 00:50:12.030 --> 00:50:14.660 But when you sent out the email for this for this episode, 00:50:15.050 --> 00:50:15.940 I was like, oh, let me look at 00:50:16.020 --> 00:50:16.520 these other things. 00:50:16.990 --> 00:50:17.080 Oh, 00:50:17.400 --> 00:50:18.400 I should try something else. 00:50:18.580 --> 00:50:19.260 They're much faster. 00:50:19.819 --> 00:50:22.300 So this episode alone might have changed my opinion on that. 00:50:22.490 --> 00:50:23.960 TY for me used to be mypy. 00:50:24.820 --> 00:50:26.460 mypy is the only thing that ends up in mine, 00:50:26.900 --> 00:50:27.880 just due to the LLMs. 00:50:28.060 --> 00:50:28.480 Yeah, they do. 00:50:28.600 --> 00:50:31.120 They do bias towards it, but I tell it it's wrong. 00:50:31.380 --> 00:50:33.780 I tell it it has to use ty as well. 00:50:33.960 --> 00:50:36.260 I was looking at what our current opinions on this are. 00:50:36.360 --> 00:50:37.820 My guess is we are on TIE. 00:50:38.360 --> 00:50:41.540 I'm not writing any of the production code for any of our customers at this moment. 00:50:41.740 --> 00:50:43.580 So I'm not the best person to ask about that. 00:50:43.600 --> 00:50:45.940 I know we were on mypy long ago. 00:50:46.320 --> 00:50:51.440 So my guess is because of the way the team bubbles up new technologies into our standards, 00:50:51.740 --> 00:50:56.540 best practices like in scaf our little scaffolding tool that's the place where i'd go look to see 00:50:56.540 --> 00:51:00.860 what we're using i'd really love ty but also gotta give a little shout i got my pyrefly shirt on from 00:51:00.960 --> 00:51:06.960 having the pyrefly team on so i wore that for for today i think pyrefly is really neat as well and 00:51:07.220 --> 00:51:12.100 it's also rust also super fast got some good extensions and plugins and stuff as well okay 00:51:12.360 --> 00:51:16.280 a couple more bits we got to talk about databases i kind of felt like this was going to be 00:51:16.440 --> 00:51:20.519 straightforward but not as straightforward as uv and uv was not straightforward so let's go love 00:51:20.540 --> 00:51:20.900 Postgres, 00:51:21.180 --> 00:51:21.520 period. 00:51:21.960 --> 00:51:24.380 Yeah, I think I start off with Postgres for a lot of things. 00:51:24.600 --> 00:51:25.240 But again, 00:51:25.520 --> 00:51:25.820 here's my 00:51:25.950 --> 00:51:26.060 like, 00:51:26.560 --> 00:51:27.660 Sydney's talking about AI again, 00:51:27.860 --> 00:51:31.580 take a lot of the artifacts that agents end up producing are 00:51:32.080 --> 00:51:36.040 really large. And so like some mix of like Postgres and then blob storage, 00:51:36.280 --> 00:51:36.560 I think, 00:51:37.090 --> 00:51:39.920 often works well for the applications that I'm thinking about. 00:51:40.170 --> 00:51:41.880 I saw you had on the list DuckDB. 00:51:42.500 --> 00:51:44.420 I think DuckDB is very entertaining. 00:51:44.970 --> 00:51:47.299 I think a lot of what I've been thinking about with databases 00:51:47.320 --> 00:51:50.580 recently is like how data models are like back in vogue, right? 00:51:50.760 --> 00:51:50.880 Like, 00:51:51.080 --> 00:51:51.960 are we actually structuring 00:51:52.180 --> 00:51:56.640 data correctly so that like agents or other things can appropriately like query at scale? 00:51:56.820 --> 00:51:57.460 Postgres over here, 00:51:57.730 --> 00:51:58.060 for sure. 00:51:58.430 --> 00:52:00.000 Although I'll give a shout out to the ZODB. 00:52:00.200 --> 00:52:01.040 It served me well for 00:52:01.070 --> 00:52:01.500 many decades. 00:52:01.860 --> 00:52:03.260 Yeah, production is all Postgres, 00:52:03.580 --> 00:52:04.440 almost all. 00:52:04.730 --> 00:52:06.560 I do use SQLite and some DuckDB. 00:52:07.160 --> 00:52:10.180 One of my side projects use lightning memory mapped database. 00:52:10.880 --> 00:52:12.700 Thought it sounded cool and it serves 00:52:12.730 --> 00:52:13.100 the purpose. 00:52:13.580 --> 00:52:15.200 So it was something different to try. 00:52:15.360 --> 00:52:16.660 I'm going to give a shout out to Team Mongo. 00:52:16.720 --> 00:52:18.940 I'm a big fan of MongoDB operationally. 00:52:19.000 --> 00:52:20.300 It's just so easy. 00:52:20.560 --> 00:52:22.380 Almost all the migrations we run are like, 00:52:22.380 --> 00:52:24.040 we need to add this column or something. 00:52:24.260 --> 00:52:28.340 And it's just the way that your code just in the database kind of just grow together. 00:52:28.520 --> 00:52:29.640 I really like that. 00:52:30.100 --> 00:52:30.260 Though, 00:52:30.780 --> 00:52:34.940 I am on an absolute kick for small embedded databases. 00:52:35.440 --> 00:52:36.140 So like Sydney, 00:52:36.540 --> 00:52:36.880 DuckDB, 00:52:37.460 --> 00:52:37.980 SQLite, 00:52:38.040 --> 00:52:40.820 I think there's actually some really interesting things you can do. 00:52:41.000 --> 00:52:45.299 SQLite has gotten a lot of work to make it more production ready 00:52:45.320 --> 00:52:46.320 with the write-ahead log 00:52:46.350 --> 00:52:47.440 and certain connection modes. 00:52:47.600 --> 00:52:48.620 It's got durability. 00:52:49.359 --> 00:52:51.060 Failure's got cross-process locking. 00:52:51.520 --> 00:52:53.400 If what you're doing lives on a single machine, 00:52:53.820 --> 00:52:56.160 it's hard to think like an embedded database 00:52:56.350 --> 00:52:57.100 works a lot of the time. 00:52:57.260 --> 00:52:58.440 I would agree with that on DuckDB, 00:52:58.680 --> 00:53:01.060 especially the JSON features that it has. 00:53:01.290 --> 00:53:02.460 If you're dealing with a lot of that 00:53:02.990 --> 00:53:03.700 type of data structure, 00:53:04.090 --> 00:53:05.440 you can build quick little apps 00:53:05.620 --> 00:53:06.900 to like little CLI tools 00:53:07.120 --> 00:53:08.260 for messing around with that. 00:53:08.510 --> 00:53:09.240 And it's super fast. 00:53:09.380 --> 00:53:09.660 It's funny. 00:53:09.930 --> 00:53:12.360 I have not used DuckDB as a database. 00:53:12.880 --> 00:53:15.580 I've used it as a query system on pandas, 00:53:15.960 --> 00:53:20.100 where it is stockingly faster than pandas itself on pandas data. 00:53:20.500 --> 00:53:21.520 I don't know how they pull that trick, 00:53:21.800 --> 00:53:22.720 but kudos to them. 00:53:22.860 --> 00:53:23.440 It's really impressive. 00:53:23.550 --> 00:53:25.900 I just had the Duck Lake folks on, 00:53:26.280 --> 00:53:27.980 and this is also super interesting, 00:53:28.220 --> 00:53:28.340 basically, 00:53:28.890 --> 00:53:33.260 like ultra scalable stuff by just putting a bunch of Parquet files on S3 00:53:33.820 --> 00:53:37.040 with a DuckDB or even something like SQLite or Postgres 00:53:37.260 --> 00:53:38.960 as a sort of metadata front end. 00:53:39.250 --> 00:53:40.300 It's very interesting, these things. 00:53:40.840 --> 00:53:41.120 All right. 00:53:41.420 --> 00:53:42.720 Well, what have we got left here? 00:53:42.780 --> 00:53:44.960 Let's jump down to docs. 00:53:45.480 --> 00:53:46.180 Where are we with docs? 00:53:46.800 --> 00:53:49.220 And I'm talking not wikis for people, 00:53:49.460 --> 00:53:50.440 but I've got a package, 00:53:50.670 --> 00:53:51.920 a library framework, 00:53:52.370 --> 00:53:53.140 and I want people to use it. 00:53:53.460 --> 00:53:55.160 So I went to FastAPI. 00:53:55.750 --> 00:53:56.340 I pulled that up. 00:53:56.500 --> 00:53:57.960 What is powering their docs there? 00:53:58.260 --> 00:54:01.240 This is actually quite a bit going on in this space these days. 00:54:01.740 --> 00:54:02.620 There's been MK docs. 00:54:02.760 --> 00:54:04.900 I always said it, but I think make docs. 00:54:04.970 --> 00:54:05.280 I don't know. 00:54:05.490 --> 00:54:06.580 I used to call it MK docs. 00:54:06.700 --> 00:54:07.480 I say MK docs. 00:54:07.640 --> 00:54:08.260 That's what we use. 00:54:08.460 --> 00:54:09.260 We use MK docs. 00:54:09.540 --> 00:54:09.940 I love it. 00:54:10.040 --> 00:54:12.280 There's actually been some stuff going on with MK docs, 00:54:12.420 --> 00:54:15.540 And I can't remember, there was like going to be a big rewrite, but a lot of the plugins 00:54:15.600 --> 00:54:16.480 are kind of working. 00:54:16.880 --> 00:54:20.560 So one of the two things that I put for this list is Zensical, 00:54:21.160 --> 00:54:22.140 which is a really interesting 00:54:22.720 --> 00:54:26.960 next version of docs from the material for make docs folks. 00:54:27.400 --> 00:54:30.000 And I just looked and that's what they've got for FastAPI. 00:54:30.140 --> 00:54:31.100 That's what Sebastian has. 00:54:31.560 --> 00:54:32.480 So that's very, 00:54:33.080 --> 00:54:34.760 very interesting as a project. 00:54:34.880 --> 00:54:37.600 I think I believe it's Rust based if I remember correctly. 00:54:37.800 --> 00:54:39.480 So obviously got that speed, 00:54:39.880 --> 00:54:40.300 speed element. 00:54:40.620 --> 00:54:42.220 Another one that I thought was interesting, 00:54:42.520 --> 00:54:44.580 I also had these folks on from Posit, 00:54:44.800 --> 00:54:45.800 is Great Docs. 00:54:46.220 --> 00:54:48.680 Have you all heard of Great Docs? I feel like this doesn't get nearly enough love. 00:54:48.820 --> 00:54:51.020 I saw that on the show and I thought it looked awesome, 00:54:51.400 --> 00:54:54.640 although I don't have much use for it because I'm not publishing packages, 00:54:54.810 --> 00:54:55.640 but it does look awesome. 00:54:55.860 --> 00:55:00.660 I still love these things for my own personal experience. I want beautiful rendered versions of my docs. 00:55:00.940 --> 00:55:05.420 So that's why even for a personal project, I still may use something like MK Docs or Great Docs. 00:55:05.480 --> 00:55:09.180 I had all these little projects and I finally broke down. I'm like, they need something more than a readme. 00:55:09.400 --> 00:55:11.040 And I migrated them all to great docs. 00:55:11.420 --> 00:55:12.920 And I'll show you one more thing in a second 00:55:13.440 --> 00:55:14.840 to tie back to Sydney's comment earlier. 00:55:15.020 --> 00:55:15.740 But go ahead, Ruben. 00:55:15.820 --> 00:55:16.440 No, I was just going to say, 00:55:16.640 --> 00:55:18.240 like so many of the things I've never heard about, 00:55:18.420 --> 00:55:19.380 like great docs, I was thinking, 00:55:19.540 --> 00:55:21.280 oh, I wonder if it's related to great tables. 00:55:21.400 --> 00:55:22.780 And I see it's also from people at Posit. 00:55:22.940 --> 00:55:25.500 It's the same people who did great tables as well. 00:55:25.620 --> 00:55:25.980 Yeah, yeah. 00:55:26.120 --> 00:55:27.060 And they've thought a lot about it. 00:55:27.420 --> 00:55:28.940 This one is kind of in your world, right? 00:55:29.040 --> 00:55:30.220 This is a data science company. 00:55:30.400 --> 00:55:32.300 This is Posit and folks. 00:55:32.740 --> 00:55:34.600 Mintlify's done some pretty good work 00:55:34.720 --> 00:55:35.720 on like baking docs, 00:55:36.240 --> 00:55:36.640 agent native, 00:55:36.940 --> 00:55:37.400 which is cool. 00:55:37.980 --> 00:55:40.080 And so like on our LangChain docs, for example, 00:55:40.520 --> 00:55:42.320 we have like copy, 00:55:43.040 --> 00:55:43.200 you know, 00:55:43.360 --> 00:55:44.680 MCP URL, 00:55:45.050 --> 00:55:45.240 like 00:55:45.560 --> 00:55:48.600 copy this as markdown for an LLM, 00:55:49.000 --> 00:55:49.800 like on all the pages. 00:55:50.050 --> 00:55:54.800 And they also have components like copy this prompt that you can give to your LLM to use 00:55:54.800 --> 00:55:55.400 this page. 00:55:55.570 --> 00:55:59.020 I think Pydantic has also moved over to Mintlify now. 00:55:59.190 --> 00:56:00.160 We used to be on MkDocs. 00:56:00.300 --> 00:56:00.980 It's inconclusive. 00:56:01.340 --> 00:56:02.200 The tests are inconclusive. 00:56:02.380 --> 00:56:03.260 But LangChain docs, 00:56:03.660 --> 00:56:05.040 yeah, definitely on Mintlify. 00:56:05.500 --> 00:56:07.440 Love the agent's focus. 00:56:07.900 --> 00:56:09.700 One of the things I like about great docs, 00:56:10.100 --> 00:56:11.720 which I did not know about Mitlefy. 00:56:11.960 --> 00:56:15.260 So this may well be right in the same area. 00:56:15.840 --> 00:56:19.320 But I'll go over here to my website because I have concrete examples. 00:56:19.580 --> 00:56:21.600 So all my little libraries and they each, 00:56:21.880 --> 00:56:23.720 we'll just go to the analytics one. 00:56:24.200 --> 00:56:26.520 So you go over here and for each page, 00:56:27.100 --> 00:56:27.580 let's say, 00:56:27.960 --> 00:56:28.320 I don't know, 00:56:28.710 --> 00:56:31.240 it has a skills section. 00:56:31.630 --> 00:56:35.620 So it has a dedicated skill and you have multiple skills and it shows you how to install the skill. 00:56:35.700 --> 00:56:40.040 just you know this is just great docs making this right like it sounds like mentlify does the same 00:56:40.160 --> 00:56:44.260 thing you can write this but if you don't write anything it'll automatically do it and if you're 00:56:44.260 --> 00:56:50.560 on any page like login async you can go up to the top and put instead of html md and it gives you a 00:56:50.700 --> 00:56:56.120 markdown version so then all the ai agents that want to read your docs can just do.md you can 00:56:56.250 --> 00:57:02.499 copy the page which copies it as markdown it's got you know llm.txt and txt full that it generates 00:57:02.520 --> 00:57:03.720 I don't know. I just think this is super neat. 00:57:04.160 --> 00:57:05.960 But it sounds like you're already doing something 00:57:06.180 --> 00:57:07.260 similar with Mitlify, 00:57:07.440 --> 00:57:07.640 which I, 00:57:08.000 --> 00:57:09.740 like I said, this is the first I'm hearing of it. So it's cool. 00:57:09.940 --> 00:57:09.980 Yeah. 00:57:10.160 --> 00:57:11.120 All right. Let's round it out. 00:57:11.280 --> 00:57:14.260 Last area before we are pretty much out of time, 00:57:14.400 --> 00:57:16.020 but let's talk about hosting. 00:57:16.400 --> 00:57:17.860 Talking hyperscalers, 00:57:18.020 --> 00:57:18.200 AWS, 00:57:18.780 --> 00:57:19.160 other things, 00:57:19.580 --> 00:57:20.200 something smaller. 00:57:20.660 --> 00:57:21.120 What do we 00:57:21.200 --> 00:57:22.480 got? I just use DigitalOcean, 00:57:23.040 --> 00:57:23.640 like super, 00:57:23.820 --> 00:57:24.640 super simple droplets, 00:57:24.980 --> 00:57:25.980 but I'm doing simple stuff. 00:57:26.260 --> 00:57:28.440 I'm not a good example unless you have very small, 00:57:28.660 --> 00:57:29.080 simple things. 00:57:29.340 --> 00:57:30.380 I wouldn't undersell it. I think 00:57:30.420 --> 00:57:32.500 DigitalOcean is really powerful. 00:57:32.780 --> 00:57:36.320 And if you run your stuff in, say, Docker or whatever, 00:57:36.460 --> 00:57:37.740 and you get a big server, 00:57:37.960 --> 00:57:39.600 you can go a really long ways 00:57:40.320 --> 00:57:42.080 by bringing your own stuff to the stack 00:57:42.240 --> 00:57:44.160 instead of tying into a bunch of cloud services. 00:57:44.700 --> 00:57:44.960 Certainly, 00:57:45.160 --> 00:57:47.360 there's a lot less going on in terms of the console 00:57:47.580 --> 00:57:49.060 compared to AWS or Azure or whatever. 00:57:49.220 --> 00:57:50.180 We're all in on containers. 00:57:50.640 --> 00:57:52.200 So if there's a way to run a container, 00:57:52.460 --> 00:57:54.620 which DigitalOcean has their app platform, 00:57:55.060 --> 00:57:57.020 which is kind of nice for throwing a container up 00:57:57.180 --> 00:57:58.399 and hooking it to a database 00:57:58.420 --> 00:58:00.660 and be able to run some quick little app. 00:58:00.890 --> 00:58:02.200 I've got kind of two modes of operation. 00:58:02.380 --> 00:58:04.840 I'll either be running like full-blown Kubernetes, 00:58:05.310 --> 00:58:08.040 whether we're using like Talos Linux giant cluster 00:58:08.380 --> 00:58:10.200 or using K3S in a single node 00:58:10.250 --> 00:58:12.860 to be able to orchestrate the container deployments. 00:58:13.260 --> 00:58:15.620 That's the real killer feature for me for Kubernetes 00:58:15.710 --> 00:58:18.840 is the control plane URL for releasing code 00:58:19.110 --> 00:58:20.860 without having to like shell into a box 00:58:21.040 --> 00:58:22.800 and clone or pull anything. 00:58:23.640 --> 00:58:24.920 And then on the other side, 00:58:25.130 --> 00:58:27.379 I really like, I've been using Cloudflare workers 00:58:27.840 --> 00:58:31.460 and Cloudflare API gateways to throw quick code up there. 00:58:31.570 --> 00:58:35.380 And now Python is a first-class citizen for Cloudflare workers. 00:58:36.030 --> 00:58:40.100 So I've been using it to slowly replace Zapier zaps 00:58:40.640 --> 00:58:45.140 for my specific use cases where Zapier had some oddities 00:58:45.420 --> 00:58:46.900 or hoops I had to jump through. 00:58:46.990 --> 00:58:49.120 I just coded something up and released it into Cloudflare. 00:58:49.260 --> 00:58:49.840 That's very cool. 00:58:50.060 --> 00:58:51.760 And foreshadowing, Alvin, 00:58:51.980 --> 00:58:57.040 I'm having the Python Cloudflare worker team on podcasts in a few weeks. 00:58:57.220 --> 00:58:57.640 That's amazing. 00:58:57.780 --> 00:58:59.620 Who else wants to chime in on this thing? 00:58:59.860 --> 00:59:02.720 We use like 90 or more percent AWS. 00:59:03.480 --> 00:59:07.720 And we've got multiple EKS clusters and an ECS cluster. 00:59:08.320 --> 00:59:09.800 So lots of containerization. 00:59:10.200 --> 00:59:14.880 We do have some workloads and authentication over on the Azure platform. 00:59:15.380 --> 00:59:21.340 And then we sort of backed into Oracle Cloud because they purchased Zenedge several years ago. 00:59:21.460 --> 00:59:25.760 And we use them as a kind of a CDN and they run a WAF for us. 00:59:26.020 --> 00:59:28.140 You already talked about DigitalOcean. 00:59:28.360 --> 00:59:30.800 That's right. I did DigitalOcean for a long time. I really like them. 00:59:30.980 --> 00:59:34.940 I actually switched stuff over to Hetzner when they first came to the U.S. 00:59:35.160 --> 00:59:37.760 and still running on them. And I think they're really interesting. 00:59:38.200 --> 00:59:38.600 However, 00:59:39.340 --> 00:59:41.760 this AI stuff has, I don't know if anyone else has noticed, 00:59:41.960 --> 00:59:45.360 it's changed the pricing of hardware and certain things just a little bit. 00:59:45.640 --> 00:59:53.200 So I have a, I have a eight core 16 gig server at Hetzner that I pay $35 a month for. 00:59:53.680 --> 00:59:55.500 It's new price is $130. 00:59:56.140 --> 00:59:58.340 So that's not just, oh, it's a little bit more. 00:59:58.760 --> 00:59:59.200 So actually, 00:59:59.600 --> 01:00:01.840 even for all the talk about Hetzner being cheap, 01:00:02.080 --> 01:00:03.940 DigitalOcean now is a better dollar 01:00:04.010 --> 01:00:05.100 for dollar deal. 01:00:05.370 --> 01:00:08.560 You get a better deal on DigitalOcean than you do even on places like Hetzner. 01:00:08.800 --> 01:00:12.260 So Hetzner, so shout out to DigitalOcean for staying true. 01:00:12.270 --> 01:00:12.800 I think that's cool. 01:00:12.960 --> 01:00:15.200 So Hensner is a dedicated box for you though. 01:00:15.300 --> 01:00:15.780 Is that right? 01:00:16.319 --> 01:00:17.340 Digitalization has a little bit more. 01:00:17.600 --> 01:00:19.580 Like I think they have RDS and a few things, 01:00:19.780 --> 01:00:21.220 but I didn't really care because I didn't use them. 01:00:21.280 --> 01:00:22.140 I just have some of my own things, 01:00:22.540 --> 01:00:23.320 which like I said, 01:00:23.580 --> 01:00:25.560 maybe more embedded databases in the future. 01:00:25.660 --> 01:00:26.240 We'll see about that. 01:00:26.320 --> 01:00:27.000 It's going to be super interesting. 01:00:27.160 --> 01:00:28.380 All right, let's close this out really quick. 01:00:28.780 --> 01:00:30.020 We talked about AI a bunch, 01:00:30.140 --> 01:00:31.480 but just go down the list, 01:00:31.600 --> 01:00:33.300 go down the list of people, 01:00:33.480 --> 01:00:34.180 not the list of tools, 01:00:34.520 --> 01:00:37.040 and just give us your AI tooling thoughts real quick. 01:00:37.280 --> 01:00:37.480 Calvin, 01:00:37.800 --> 01:00:38.180 you go first. 01:00:38.340 --> 01:00:38.640 You're at the top. 01:00:38.720 --> 01:00:40.780 Definitely Pi and Claude Code, 01:00:41.020 --> 01:00:41.500 mostly Pi. 01:00:41.840 --> 01:00:47.780 using typically the soul like whatever the one that just came out 6.1 I think it's the latest 01:00:48.260 --> 01:00:55.880 on OpenAI or I like Sonnet 5.5 for most things I find I get really annoyed when they default me to 01:00:55.880 --> 01:01:00.940 the latest opus and it starts just eating through tokens and I'm like what happened oh a new release 01:01:01.040 --> 01:01:05.060 came out so that's what I'm at I like I like Pi for that stuff with superpowers just to skills. 01:01:05.320 --> 01:01:11.159 Sydney. There's so many cool tools I think generally just thinking more broadly about AI usage like 01:01:11.400 --> 01:01:13.080 I think having a coding agent makes sense. 01:01:13.680 --> 01:01:14.880 There's so many good ones. 01:01:15.190 --> 01:01:16.940 I think having an agent in Slack 01:01:17.300 --> 01:01:18.120 that you use, 01:01:18.280 --> 01:01:21.880 having an agent in wherever you keep track of documents. 01:01:22.160 --> 01:01:22.660 So for me, 01:01:22.820 --> 01:01:23.440 that's Notion. 01:01:23.880 --> 01:01:27.700 And then generally, I'm pretty bullish on MCPs and skills, 01:01:27.980 --> 01:01:29.460 just because that's where you basically 01:01:29.870 --> 01:01:34.900 for MCPs can connect your data and use cases, and then for skills can distill domain knowledge. 01:01:35.300 --> 01:01:37.840 And then maybe the thing that I'm most excited about with agents, 01:01:38.250 --> 01:01:40.240 both all of these categories, 01:01:40.320 --> 01:01:42.640 is like improving them over time with memory. 01:01:42.920 --> 01:01:44.940 So like making sure my coding agents 01:01:45.700 --> 01:01:46.880 remember my specific preferences 01:01:47.200 --> 01:01:48.800 and making sure my agents 01:01:49.300 --> 01:01:51.180 that are helping me with like blog edits 01:01:52.040 --> 01:01:54.580 distill information from previous like human reviewers, 01:01:54.880 --> 01:01:55.400 things like that. 01:01:55.720 --> 01:01:58.760 But a lot of my job now is sharing those ideas. 01:01:59.000 --> 01:02:00.060 So if you want to know more, 01:02:00.480 --> 01:02:02.540 the LangChain blog is a good place to start. 01:02:02.700 --> 01:02:04.980 We've kind of settled on using DevNetwork 01:02:05.320 --> 01:02:07.519 and what you said earlier 01:02:07.540 --> 01:02:11.320 about changing all the LLMs on the backend, 01:02:12.060 --> 01:02:12.720 that can be scary. 01:02:12.780 --> 01:02:14.720 I usually let it decide automatically 01:02:15.040 --> 01:02:15.980 which ones to use, 01:02:16.180 --> 01:02:18.760 but we're trying to keep an eye on token spend. 01:02:19.140 --> 01:02:20.280 So we'll have to take a look 01:02:20.280 --> 01:02:22.480 at your analytics for that too. 01:02:22.880 --> 01:02:24.040 I tend to use Kiro too. 01:02:24.360 --> 01:02:25.880 I like to use VS Code locally, 01:02:26.460 --> 01:02:30.040 but we're also spoiled with the public LLMs 01:02:30.140 --> 01:02:30.780 that are so capable 01:02:30.980 --> 01:02:32.580 that when I try to run something locally, 01:02:33.040 --> 01:02:34.200 it's a little underwhelming. 01:02:34.920 --> 01:02:35.320 Yeah, 01:02:35.520 --> 01:02:36.400 as I mentioned earlier, 01:02:36.800 --> 01:02:40.300 I've been using Claude Code for like 95% of what I do, 01:02:40.710 --> 01:02:41.820 and I'm very happy with it. 01:02:42.070 --> 01:02:44.400 I have played quite a bit with OpenCode as well. 01:02:44.780 --> 01:02:48.360 I sometimes tell people I feel like Claude Code is like the Mac and OpenCode is like Linux, 01:02:48.660 --> 01:02:50.040 where it's like instantly extensible, 01:02:50.300 --> 01:02:52.500 but you have to spend a little more time with it and 01:02:52.780 --> 01:02:53.680 configuring and so forth. 01:02:54.000 --> 01:02:57.140 And Pi is one of these things that I keep saying, I really have to play with it because 01:02:57.360 --> 01:02:58.520 everyone's saying great things about it. 01:02:58.720 --> 01:03:01.360 So I really have to play with it because everyone's saying great things about it. 01:03:01.420 --> 01:03:01.640 Awesome. 01:03:02.030 --> 01:03:02.780 Well, for me also, 01:03:03.240 --> 01:03:06.760 like Ruben, Claude Code, usually inside VS Code or something along 01:03:06.780 --> 01:03:13.340 lines for my little personal automations i've been using a lot of glm 5 3 flash which i think is 01:03:13.460 --> 01:03:19.160 actually super neat as well still not self-hostable but you know it's been running in other places that 01:03:19.160 --> 01:03:23.240 are not not too expensive right you know 5 10 bucks a month something like super cool though 01:03:23.440 --> 01:03:31.340 well for the self-hosting models quin 3 8 26 b is incredibly capable and runs on most you know 32 01:03:31.940 --> 01:03:33.440 gig plus machines. 01:03:33.940 --> 01:03:34.020 Awesome. 01:03:34.180 --> 01:03:35.300 A GLM 5.3 flash, 01:03:35.520 --> 01:03:35.680 I think, 01:03:35.800 --> 01:03:37.740 is 750 gigs of RAM you need. 01:03:37.880 --> 01:03:39.920 And the non-flash is like 1.5 terabytes 01:03:40.020 --> 01:03:40.300 of RAM. 01:03:40.460 --> 01:03:41.820 So not hosting those. I'll check out 01:03:41.920 --> 01:03:43.680 the Quinn one. That's awesome. All right, everyone. 01:03:44.340 --> 01:03:45.860 Thank you so much for being here. This has been 01:03:46.339 --> 01:03:46.980 really interesting. 01:03:47.240 --> 01:03:47.940 We didn't even 01:03:48.180 --> 01:03:49.180 touch on a bunch of topics. 01:03:49.640 --> 01:03:49.960 There's 01:03:50.120 --> 01:03:51.580 so much to talk about, so much happening. 01:03:52.119 --> 01:03:53.600 And you all have great 01:03:54.020 --> 01:03:54.120 insight. 01:03:54.460 --> 01:03:55.600 So thanks for being here. Thanks, Michael. 01:03:55.760 --> 01:03:56.080 This was great. 01:03:56.340 --> 01:03:56.680 Thanks so much. 01:03:56.940 --> 01:03:57.260 Thank you. 01:03:58.880 --> 01:03:59.760 This has been another 01:03:59.760 --> 01:04:01.020 episode of Talk Python To Me. 01:04:01.290 --> 01:04:02.120 Thank you to our sponsors. 01:04:02.330 --> 01:04:03.320 Be sure to check out what they're 01:04:03.480 --> 01:04:05.180 offering. It really helps support the show. 01:04:05.940 --> 01:04:07.740 If you or your team needs to learn Python, 01:04:07.970 --> 01:04:13.820 we have over 270 hours of beginner and advanced courses on topics ranging from complete beginners 01:04:14.090 --> 01:04:15.500 to async code, Flask, 01:04:15.720 --> 01:04:15.920 Django, 01:04:16.340 --> 01:04:16.900 HTMX, 01:04:17.060 --> 01:04:18.720 and even LLMs. Best of all, 01:04:18.940 --> 01:04:20.680 there's no subscription in sight. 01:04:21.080 --> 01:04:22.840 Browse the catalog at talkpython.fm. 01:04:23.640 --> 01:04:25.840 And if you're not already subscribed to the show on 01:04:25.860 --> 01:04:26.900 your favorite podcast player, 01:04:27.140 --> 01:04:28.160 what are you waiting for? 01:04:28.980 --> 01:04:30.620 Just search for Python in your podcast player. 01:04:30.730 --> 01:04:31.600 We should be right at the top. 01:04:32.040 --> 01:04:34.800 If you enjoyed that geeky rap song, you can download the full track. 01:04:35.030 --> 01:04:36.920 The link is actually in your podcast blur show notes. 01:04:37.760 --> 01:04:38.380 This is your host, 01:04:38.600 --> 01:04:39.080 Michael Kennedy. 01:04:39.480 --> 01:04:40.540 Thank you so much for listening. 01:04:40.750 --> 01:04:41.540 I really appreciate it. 01:04:41.980 --> 01:04:42.680 I'll see you next time. 01:04:54.320 --> 01:04:57.120 I'm out.