WEBVTT 00:00:00.020 --> 00:00:03.360 You ask an AI a question, and it answers with total confidence. 00:00:03.880 --> 00:00:07.540 Most of the time, a confidently wrong answer is just an annoyance. 00:00:08.080 --> 00:00:11.620 But what if the question is medical, and there's a real patient on the other end? 00:00:12.260 --> 00:00:16.139 In that world, a hallucination isn't a bug. It's a patient safety event. 00:00:16.740 --> 00:00:22.980 Sumit Gundawar is a London-based software engineer who builds a clinical platform for a UK longevity medical clinic. 00:00:23.560 --> 00:00:27.200 And his whole argument is that in high-stakes AI, the model is the easy part. 00:00:27.660 --> 00:00:29.420 Earning trust is the real engineering. 00:00:30.220 --> 00:00:36.320 We dig into grounding, refusal logic, human-in-the-loop design, and the messy frontier of longevity and biohacking. 00:00:36.880 --> 00:00:42.060 Plus, a live demo of an assistant that refuses to answer when it can't back up the claim. 00:00:42.460 --> 00:00:49.200 Let's get into it. This is Talk Python To Me, episode 554, recorded June 25th, 2026. 00:01:07.500 --> 00:01:11.820 Welcome to Talk Python To Me, the number one Python podcast for developers and data scientists. 00:01:12.340 --> 00:01:17.660 This is your host, Michael Kennedy. I'm a PSF fellow who's been coding for over 25 years. 00:01:18.080 --> 00:01:19.360 Let's connect on social media. 00:01:19.840 --> 00:01:22.820 You'll find me and Talk Python on Mastodon, Bluesky, and X. 00:01:23.180 --> 00:01:25.000 The social links are all in your show notes. 00:01:25.780 --> 00:01:29.280 You can find over 10 years of past episodes at talkpython.fm. 00:01:29.460 --> 00:01:32.660 And if you want to be part of the show, you can join our recording live streams. 00:01:32.940 --> 00:01:33.500 That's right. 00:01:33.740 --> 00:01:36.940 We live stream the raw uncut version of each episode on YouTube. 00:01:37.540 --> 00:01:41.960 Just visit talkpython.fm/youtube to see the schedule of upcoming events. 00:01:42.220 --> 00:01:45.840 Be sure to subscribe there and press the bell so you'll get notified anytime we're recording. 00:01:46.300 --> 00:01:52.120 This episode is brought to you by Six Feet Up, the Python and AI experts who solve hard software problems. 00:01:52.840 --> 00:01:58.040 Whether it's scaling an application, deriving insights from data, or getting results from AI, 00:01:58.520 --> 00:02:01.140 Six Feet Up helps you move forward faster. 00:02:01.860 --> 00:02:06.400 See what's possible with Six Feet Up. Visit talkpython.fm/sixfeetup. 00:02:07.580 --> 00:02:10.740 Hello, welcome to Talk Python, Amish Simi. It's great to have you here. 00:02:11.120 --> 00:02:12.140 Thank you so much for having me. 00:02:13.060 --> 00:02:17.820 I'm really excited to talk about some high stakes use case of AI. 00:02:18.380 --> 00:02:25.520 And to be clear for everyone listening, I imagine the area we're going to focus on is not we're using AI to write code. 00:02:25.800 --> 00:02:34.700 We're not using AI to write code, but instead we're actually integrating an LLM into the process, kind of like an API would, right? 00:02:34.920 --> 00:02:36.000 Kind of like, yes. 00:02:36.800 --> 00:02:38.620 In general, it could be something similar. 00:02:39.060 --> 00:02:39.480 Right, exactly. 00:02:39.840 --> 00:02:44.820 So having the AI be part of the actual execution of the code, not the creation of the code. 00:02:45.160 --> 00:02:47.740 Not that you couldn't potentially use AI to create code as well. 00:02:47.880 --> 00:02:48.740 That works that way. 00:02:48.990 --> 00:02:53.200 But the focus is really about using AI in the execution of your apps. 00:02:53.210 --> 00:02:53.320 Yeah. 00:02:53.640 --> 00:02:56.680 And how if you can trust the answer that it gives. 00:02:57.860 --> 00:02:58.840 What do you mean you can't trust it? 00:02:59.080 --> 00:03:01.680 You can, of course, but depends on the situation. 00:03:02.220 --> 00:03:02.780 I'm just kidding. 00:03:03.320 --> 00:03:03.940 I'm just kidding. 00:03:04.200 --> 00:03:05.320 You know, the meme. 00:03:05.540 --> 00:03:06.240 You're absolutely right. 00:03:06.400 --> 00:03:08.800 So this is a medical field area. 00:03:08.840 --> 00:03:10.820 So let me give you a medical joke for AI. 00:03:11.010 --> 00:03:14.460 And I might get it switched, but the point of the joke will be just the same. 00:03:14.580 --> 00:03:15.960 It'd be fun to kick the show off like this. 00:03:16.220 --> 00:03:21.700 So I saw this cartoon and there was a surgeon coming out of a post-op situation to speak 00:03:21.820 --> 00:03:25.120 to a patient who had just had a surgery to remove their appendix. 00:03:25.780 --> 00:03:31.840 But the surgeon had like a head that had one of the main AI companies as their logo instead 00:03:31.910 --> 00:03:33.200 of like an actual human head. 00:03:33.620 --> 00:03:38.680 The patient says to the AI surgeon, says, doctor, if the appendix is on the left side, 00:03:39.060 --> 00:03:44.420 Why is the scar on the right side? You know what? You're absolutely right. Let me try that again. 00:03:46.600 --> 00:03:47.480 Yeah, we don't want this, do we? 00:03:48.540 --> 00:03:50.240 Yes, that's what we have time to avoid. 00:03:52.180 --> 00:03:53.940 You're absolutely right. Let's do that again. 00:03:55.200 --> 00:03:58.200 Simeet, before we jump into that, maybe we'll have more jokes, who knows. 00:03:58.600 --> 00:04:02.880 But before we jump into that, give people a bit of background on who you are. Give them an introduction. 00:04:03.180 --> 00:04:21.720 Question would be, I'm a software engineer. I work at a wellness clinic in London. My background is full stack development. But before this, I used to work in data engineering and data analyst role where I worked on like demand forecasting and other similar domains like in data analytics perspective. 00:04:21.799 --> 00:04:29.660 Here now I do a little bit of R&D, plus I do front end to sometimes back end and like basically full stack development. 00:04:30.060 --> 00:04:32.880 So this is one project I recently got on healthcare. 00:04:33.440 --> 00:04:36.920 And yeah, so I've been working on this for some time now, for a year now. 00:04:37.180 --> 00:04:39.900 Okay. And how'd you get into the AI side of things? 00:04:40.180 --> 00:04:50.400 Like that's a pretty big shift to go from just data analysis and pandas and those kind of things in front end to now I'm using AI as execution, as I said at the start. 00:04:50.560 --> 00:05:08.920 Yeah, I think it started during the COVID times. We had gotten like GPT-3 and GPT-4, it just came out and all these new models like constantly upgrading themselves. And that is when I thought that, okay, that analyst draw is not going to last longer. 00:05:09.320 --> 00:05:10.440 I realized that back then. 00:05:10.750 --> 00:05:12.940 And I was like, okay, I have to prepare for something better. 00:05:13.050 --> 00:05:16.640 I have to prepare for something that AI won't replace me. 00:05:16.900 --> 00:05:22.440 Like, you know, we hear of these layoffs happening, you know, 10,000 people laid off, 12,000 people laid off. 00:05:22.940 --> 00:05:27.440 And I didn't want to be a part of that, which is why I started studying again. 00:05:27.920 --> 00:05:33.240 And then I was introduced to this clinic later on after my studies got completed. 00:05:33.620 --> 00:05:34.900 And then I joined here. 00:05:35.200 --> 00:05:35.420 Awesome. 00:05:35.710 --> 00:05:38.440 I think that's a great little message for people out there. 00:05:38.620 --> 00:05:39.820 Just you've got to keep learning. 00:05:40.050 --> 00:05:40.820 You've got to keep studying. 00:05:41.120 --> 00:05:44.580 And some of the things that come down, you could be really excited about. 00:05:44.820 --> 00:05:46.640 Some of them, you know, I know a lot of people are really like, 00:05:46.760 --> 00:05:48.560 I don't want anything to do with this AI stuff. 00:05:48.740 --> 00:05:52.200 It's like, you could have said, I don't want anything to do with the web as well, 00:05:52.410 --> 00:05:53.240 you know, 20 years ago. 00:05:53.420 --> 00:05:54.060 But guess what? 00:05:54.300 --> 00:05:57.680 Like, it's the main part of software development these days, you know? 00:05:57.780 --> 00:05:58.020 Yeah. 00:05:58.460 --> 00:06:01.780 We keep hearing our stories for coming out from meta nowadays. 00:06:02.240 --> 00:06:02.340 Yeah. 00:06:02.600 --> 00:06:03.620 Have you heard of the news? 00:06:03.940 --> 00:06:04.960 Like what's happening in meta? 00:06:05.220 --> 00:06:05.920 Oh my gosh. 00:06:06.540 --> 00:06:07.120 So many things. 00:06:07.260 --> 00:06:12.100 I don't remember the details enough to recount them with enough accuracy, but yeah, it's a little crazy. 00:06:12.280 --> 00:06:13.380 It is a little crazy. 00:06:13.660 --> 00:06:19.160 The engineers, they're calling it a gulag, you know, like the Russian gulags, Soviet Union gulags. 00:06:19.860 --> 00:06:21.020 Yeah, that doesn't sound great. 00:06:21.220 --> 00:06:24.900 And they're saying that they have been moved into a separate entity and a separate team. 00:06:25.020 --> 00:06:27.840 And all they do is weekly two tasks. 00:06:28.840 --> 00:06:32.840 And they have to code very, very high level language so that AI can learn. 00:06:33.280 --> 00:06:37.020 Basically, they are prompting AI to learn from them, to replace them. 00:06:37.760 --> 00:06:45.400 Yeah, it's a little bit morbid. I do think that a lot of these layoffs that you pointed out, I think a lot of these companies are using AI as an excuse. 00:06:45.980 --> 00:06:46.780 Yeah, that's true. 00:06:46.880 --> 00:06:55.820 Because if you say we're laying off people because of AI, your stock goes up. If you're saying we're laying off people because we hired way too many people during COVID, your stock goes down. 00:06:56.020 --> 00:07:00.620 So, well, you can just say AI. What difference does it make? It lawsably could be because of AI. 00:07:01.300 --> 00:07:07.740 I do think AI will have this effect, but I think a lot of the layouts we're seeing are opportunistic to some degree. 00:07:08.100 --> 00:07:08.760 So I don't know. 00:07:08.770 --> 00:07:12.560 That doesn't really change things for people who got laid off, but it's a weird time. 00:07:12.840 --> 00:07:14.080 It is a weird time, yeah. 00:07:14.300 --> 00:07:15.720 It's such a developing time. 00:07:16.100 --> 00:07:17.240 Every day there is something new. 00:07:18.640 --> 00:07:20.460 So everyone listening, I will put it out there. 00:07:20.650 --> 00:07:24.160 Like there's, you know, all the many thousands of people who are listening. 00:07:24.500 --> 00:07:27.740 I'm going to go out there and rely on them and say every single one of them uses AI for something, 00:07:28.040 --> 00:07:29.480 or at least has experimented with AI. 00:07:29.960 --> 00:07:32.140 asking a question to ChatGPT or something. 00:07:32.370 --> 00:07:36.580 But not many people have had high stakes experience, right? 00:07:36.800 --> 00:07:40.200 Like my silly joke of the AI surgeon, you're absolutely right. 00:07:40.470 --> 00:07:47.800 But we've seen AI used for radiology, for x-rays, cancer detection, that kind of stuff. 00:07:48.040 --> 00:07:52.000 We've seen it for patient advice, classification, all kinds of things. 00:07:52.010 --> 00:07:57.040 I guess also probably mortgage, insurance, other things that are opaque and really affect your life. 00:07:57.220 --> 00:08:01.720 But give us some sense, at least in your world, some of the different types of high stakes 00:08:02.060 --> 00:08:06.120 AI that you might have seen, like maybe set the stage for this conversation. 00:08:07.500 --> 00:08:13.860 I think in a high stakes environment, what would happen is the answer is not a problem. 00:08:14.220 --> 00:08:15.700 The answer is usually right. 00:08:16.020 --> 00:08:18.300 But the keyword here is the usually part. 00:08:18.620 --> 00:08:23.980 It's not always right that, you know, the 1% or the 2% that it gets wrong very confidently 00:08:24.060 --> 00:08:25.540 is like, yes, this is the way to do. 00:08:25.800 --> 00:08:28.000 That is the part I try to avoid. 00:08:28.420 --> 00:08:30.320 You know, I try to block that part. 00:08:30.420 --> 00:08:33.380 So this is the part where when you use it in a high-stakes environment, 00:08:33.740 --> 00:08:36.940 like, for example, if you use it in a medical line, 00:08:37.200 --> 00:08:40.479 what would happen is that 1% or 2% will have a real impact on a human being 00:08:40.640 --> 00:08:41.539 and on their life. 00:08:41.849 --> 00:08:43.680 Like, you know, it can be a life-and-death situation. 00:08:43.979 --> 00:08:44.860 We never know. 00:08:44.980 --> 00:08:47.740 So which is why this is the kind of scale that we talk about. 00:08:47.840 --> 00:08:51.060 Yeah, and if you're working at a national health care level, 00:08:51.560 --> 00:08:53.400 1% is actually a lot of people. 00:08:53.840 --> 00:08:56.620 There's millions of people that gets affected by it. 00:08:57.000 --> 00:09:01.180 Even in a finance industry, I think if you apply this, it can cost millions. 00:09:01.690 --> 00:09:05.560 Even if like 1% error, it could go up to millions and billions of dollars. 00:09:06.780 --> 00:09:06.960 Yeah. 00:09:08.300 --> 00:09:10.840 We've had these kinds of situations before. 00:09:11.760 --> 00:09:20.340 It's kind of fun to go out and just do a little research and look through like the worst Excel, Microsoft Excel failures of all time. 00:09:20.460 --> 00:09:27.020 And there's like billion dollars of mistakes that have been made because there was a problem in some complex spreadsheet. 00:09:27.400 --> 00:09:29.940 That's also true for software, right? 00:09:30.080 --> 00:09:38.760 We've had the Mars Lander, you know, probably the most comical, ridiculous version is like the U.S. NASA Mars Lander that just plunged. 00:09:38.760 --> 00:09:44.940 I think it plunged into Mars, not the Earth, but it certainly plunged into a planet because there were two contractors. 00:09:45.120 --> 00:09:46.500 I don't know. Let's say Boeing and Lockheed Martin. 00:09:46.520 --> 00:09:47.740 I don't remember exactly what it was. 00:09:47.840 --> 00:09:54.080 But yeah, one of the teams use metric and one of them use, imperial units. 00:09:54.880 --> 00:09:58.660 And it's just like, well, first of all, in science, you should be using metric period. 00:09:58.880 --> 00:10:03.380 Like I love myself some Fahrenheit and so on, but if you're doing science, it should 00:10:03.410 --> 00:10:03.960 be in metric. 00:10:04.320 --> 00:10:08.120 So there's a clear blame to go around somewhere on this, but I'm just saying there's like, 00:10:08.620 --> 00:10:10.220 it's not unique to software. 00:10:10.400 --> 00:10:15.760 It's not unique to AI that, that is, these problems are serious and expensive or affect 00:10:15.900 --> 00:10:16.280 many people. 00:10:16.480 --> 00:10:20.500 But I think why this feels so different is it's, it's not known. 00:10:20.570 --> 00:10:22.880 You can't predict deterministically. 00:10:22.990 --> 00:10:25.020 You can't say, yeah, we've debugged the heck out of it. 00:10:25.080 --> 00:10:26.280 We've curated it, right? 00:10:26.540 --> 00:10:30.460 OpenAI could release a new model that you don't even, it doesn't even get renamed, but like 00:10:30.550 --> 00:10:31.900 a little tweak behind the scenes. 00:10:32.320 --> 00:10:33.760 All of a sudden it starts answering different. 00:10:34.160 --> 00:10:35.620 And now people have cancer when they didn't. 00:10:35.710 --> 00:10:38.080 I mean, they don't necessarily really do, but they're being told they do. 00:10:38.150 --> 00:10:38.600 You know what I mean? 00:10:38.780 --> 00:10:38.980 Yeah. 00:10:39.110 --> 00:10:39.220 Yeah. 00:10:39.330 --> 00:10:42.580 I would compare Opus 4.7 and 4.8. 00:10:43.760 --> 00:10:43.940 Okay. 00:10:44.300 --> 00:10:45.020 Was that a big change? 00:10:45.880 --> 00:10:48.660 Well, it was a change. It degraded, I think. 00:10:48.920 --> 00:10:49.440 Oh, interesting. 00:10:49.920 --> 00:10:52.680 Did you guys ever experiment with Fable while it was out, Fable and Mythos? 00:10:53.020 --> 00:10:55.280 That week it was out before it got yanked? 00:10:55.540 --> 00:10:57.660 I used Fable, not the Mythos. 00:10:57.860 --> 00:11:01.660 Mythos is, I think, it's only a result for a certain amount of companies. 00:11:01.800 --> 00:11:03.820 I'm pretty sure that Fable and Mythos are identical. 00:11:04.080 --> 00:11:06.620 The only difference is that Fable has guardrails. 00:11:07.000 --> 00:11:09.060 Like, you can't talk about the sensitive stuff. 00:11:09.240 --> 00:11:11.580 Of course, those got broken and it got taken away because of it. 00:11:12.020 --> 00:11:18.220 Yeah, I think it's just what they do is they put some safeguarding around prompt injection. 00:11:18.590 --> 00:11:23.840 What they do is they try to give that, like for example, if I talk to Fable, right, 00:11:24.460 --> 00:11:30.500 they give it to Opus first and then let it decide if it has any prompt engineering or something 00:11:30.740 --> 00:11:33.700 which would hack it and then they pass it to Fable if it is safe. 00:11:33.840 --> 00:11:34.020 Yeah. 00:11:34.280 --> 00:11:36.620 That would be one of the ideas of doing that. 00:11:36.840 --> 00:11:40.760 Yeah, that seems like a good idea to do something like that. 00:11:40.800 --> 00:11:43.920 I have heard some rumors that it might be coming back. 00:11:44.020 --> 00:11:44.940 I read something about yesterday. 00:11:45.280 --> 00:11:49.600 There's some changes in the API address and some other things like, hey, this might be coming back. 00:11:49.800 --> 00:11:50.440 So I don't know. 00:11:51.440 --> 00:11:53.900 Let's take this as a more abstract conversation. 00:11:54.160 --> 00:12:00.880 So as you've seen the models improve, clearly GPT-3 and Opus are a different kind of smart, right? 00:12:01.020 --> 00:12:04.800 Like over the last three or four years, things have changed quite dramatically, right? 00:12:05.100 --> 00:12:09.580 So how has that affected thinking about these AI systems, getting these smart models? 00:12:09.660 --> 00:12:12.120 Does it make a meaningful difference or not really? 00:12:12.400 --> 00:12:14.220 Well, it depends on the use case, I think. 00:12:14.520 --> 00:12:22.580 Like if you're using it for general coding and like, for example, wipe coding, making some, for example, the demo I will show today, it is a wipe coding, right? 00:12:22.760 --> 00:12:30.220 So it was a weekend project just to show something, you know, like to have some synthetic data to create some front end UI. 00:12:30.540 --> 00:12:31.880 I think it's great for that part. 00:12:32.260 --> 00:12:38.440 But when you try to apply it to very, very complex systems, I don't think it will work out. 00:12:38.500 --> 00:12:46.300 Like, for example, if you give it a very huge repository and it does not have the context window of keeping it track, there are different scales. 00:12:46.420 --> 00:12:48.660 There are different tools you can use, like graph images. 00:12:49.100 --> 00:12:52.080 And there are, like, files you can create, scale files you can create. 00:12:52.220 --> 00:12:53.080 But the result would be... 00:12:53.080 --> 00:12:58.880 In subagents, you can split up all the stuff it has to work on across subagents, but not always or not entirely. 00:12:59.120 --> 00:13:02.200 It just won't be the same, you know, like when you get the raw output. 00:13:02.480 --> 00:13:04.280 But I think it has been improving a lot. 00:13:04.580 --> 00:13:10.520 But yeah, I think the new problem that these AI companies are going to face would be data, data deficiency. 00:13:11.020 --> 00:13:12.260 They have used all of internet. 00:13:12.880 --> 00:13:20.980 And what nowadays they have only a like a wipe coded articles and wipe coded blog post. 00:13:21.300 --> 00:13:26.020 And I think that those are the ones that will be indexed next into the systems. 00:13:26.220 --> 00:13:31.120 And then they will have a real engineering challenge of getting the right output out. 00:13:31.400 --> 00:13:45.180 It's a little bit like a mad cow disease or something where it's just like it's eating its own outputs and it's just going to be like eventually, I don't know, some iterative, weird, discreet version of itself, you know? 00:13:45.440 --> 00:13:46.000 Yeah, I think. 00:13:47.020 --> 00:13:47.940 Yeah, crazy, crazy. 00:13:48.350 --> 00:13:50.220 Okay, well, looking forward to the demo. 00:13:50.680 --> 00:13:54.840 But let's bring it over to clinical AI in 2026. 00:13:55.080 --> 00:14:02.520 So give us a sense that to the extent that you can, how you guys are using it or how you're seeing these more broadly in the industry, some of the challenges. 00:14:03.140 --> 00:14:12.860 Yeah. Well, I have had a couple of calls and some research usually I do on other competitors on their applications and how they use AI. 00:14:13.240 --> 00:14:24.160 And the most I've seen is like summarization and like when a clinician or a doctor puts in their details into the platform for a specific record, 00:14:24.600 --> 00:14:30.500 or what AI does, it just summarizes and gives it like a pointer, like a bullet point system. 00:14:30.980 --> 00:14:37.700 Or it does a note taker system like during a consultation, it will have a note taker online, which will listen to the whole conversation. 00:14:38.000 --> 00:14:41.440 It will have a transcript as well as the summary of what was discussed. 00:14:41.720 --> 00:14:50.240 So through a legal point, you know, if something wrong happens, a patient cannot like claim that it was not told to me or something. 00:14:50.270 --> 00:14:52.200 It was not discussed upon during an interview. 00:14:52.250 --> 00:14:53.620 So that it helps on. 00:14:53.920 --> 00:14:59.500 But what it doesn't help on nowadays, it doesn't help on the systems is how to make decisions out of it. 00:14:59.700 --> 00:15:01.480 So it is really, really complex. 00:15:01.670 --> 00:15:05.600 We haven't used like even we haven't used it widely. 00:15:05.910 --> 00:15:10.460 It is like, you know, still being stress tested and still being updated every day. 00:15:10.720 --> 00:15:15.460 work on it. Like I do this testing, that testing, and that when I try to simulate it with your 00:15:15.700 --> 00:15:20.080 patient and, you know, it fails, then I try to go back and see what failed, where it failed, 00:15:20.420 --> 00:15:24.760 like log tracing of everything. So yeah, I think that is where clinical AI is right now. 00:15:26.920 --> 00:15:31.160 This portion of Talk Python To Me is brought to you by Six Feet Up. Let me ask you a question. 00:15:31.600 --> 00:15:36.819 What's stopping you? Maybe it's an application that won't scale or an AI initiative that just 00:15:36.840 --> 00:15:43.760 isn't delivering, that's where Six Feet Up comes in. With deep expertise in Python and AI, they solve 00:15:43.940 --> 00:15:49.400 hard software problems, modernize platforms, and get teams to market faster. These folks have been 00:15:49.480 --> 00:15:54.040 doing Python since version one. They know the frameworks and ecosystems like the back of their 00:15:54.140 --> 00:15:59.900 hands. Six Feet Up's impact speaks for itself. Automated healthcare pipelines for hospitals, 00:16:00.420 --> 00:16:05.959 helping NASA explore Pluto, building severe weather prediction tools, and applying AI to connect 00:16:05.980 --> 00:16:11.440 farmers with vital crop data. When the stakes are high and the problems are hard, Six Feet Up is the 00:16:11.680 --> 00:16:17.140 partner that delivers. See what's possible with Six Feet Up. Visit talkpython.fm/sixfeetup. 00:16:17.560 --> 00:16:21.580 The link is on the episode page and in your podcast player's show notes. Thanks to Six Feet 00:16:21.660 --> 00:16:27.240 Up for sponsoring the show. It's still basic, pretty basic. Like even there are no ticker 00:16:27.360 --> 00:16:31.759 applications that does the same thing, you know, like it's not an innovative idea that they have 00:16:31.780 --> 00:16:35.440 they've applied. They've just taken some basic ideas from different companies and 00:16:35.640 --> 00:16:39.520 different tools and then just combined it all, vaulted on together into one system 00:16:39.600 --> 00:16:43.580 and they're just like giving it on a subscription. Yeah, yeah. You could probably 00:16:43.660 --> 00:16:47.180 almost just piece this together if it weren't for rules like HIPAA. 00:16:47.500 --> 00:16:51.460 In the US, we have HIPAA rules, which is about you're not allowed to share your medical data 00:16:51.540 --> 00:16:55.480 in most circumstances. So like, for example, 00:16:55.620 --> 00:16:59.000 one of my favorite note-taking apps that I've come across lately is Granola. 00:16:59.540 --> 00:17:01.120 sponsored, just, just a shout out. 00:17:01.120 --> 00:17:02.120 I really like this thing. 00:17:02.240 --> 00:17:04.579 It's like a new taking up marketing now. 00:17:05.040 --> 00:17:05.260 Okay. 00:17:05.459 --> 00:17:07.120 They really are, but it really works well. 00:17:07.260 --> 00:17:12.880 Like it will listen to basically any, it would do this video stream that we're doing. 00:17:13.000 --> 00:17:14.020 It would listen to zoom. 00:17:14.020 --> 00:17:15.120 It will listen to FaceTime, whatever. 00:17:15.459 --> 00:17:19.579 And, but you take notes and then it does the transcript and then AI like considers your 00:17:19.720 --> 00:17:21.939 notes and the transcript and then like enhances it. 00:17:22.160 --> 00:17:23.800 Things like that exist and they're awesome. 00:17:24.100 --> 00:17:25.459 They're not perfect, but they're pretty awesome. 00:17:25.699 --> 00:17:28.560 But you just can't use them in a medical situation, right? 00:17:28.660 --> 00:17:33.520 Because there's probably a whole chain of untrustworthy things happening to that data, 00:17:34.150 --> 00:17:38.000 that transcript and so on that is not allowed, right? 00:17:38.640 --> 00:17:39.400 Yeah, that's true. 00:17:39.580 --> 00:17:41.520 But I imagine that works pretty well, right? 00:17:41.600 --> 00:17:42.420 Like I'm a doctor. 00:17:42.470 --> 00:17:43.300 I just saw a patient. 00:17:43.330 --> 00:17:44.080 I could walk in. 00:17:44.190 --> 00:17:47.800 Like ideally, maybe I could just like hold a button and speak to my watch. 00:17:48.000 --> 00:17:53.220 Like I just saw this patient, so-and-so, and here's what I think, like a three sentence 00:17:53.360 --> 00:17:54.140 and just keep walking. 00:17:54.420 --> 00:17:55.680 You know, that would actually be pretty sweet. 00:17:55.840 --> 00:18:00.040 I think there was one device, some variable, you know, around your neck. 00:18:00.440 --> 00:18:01.940 I remember, I just don't remember the name. 00:18:02.300 --> 00:18:05.000 I see, like a smart stethoscope you can speak to. 00:18:05.820 --> 00:18:09.640 No, it was just a small device which would just stay here, and it had a camera. 00:18:10.300 --> 00:18:11.120 Oh, interesting, okay. 00:18:11.500 --> 00:18:13.300 Human, human AI, human pin. 00:18:13.460 --> 00:18:15.020 Yeah, yeah, yeah, interesting, okay. 00:18:15.380 --> 00:18:16.220 It had a microphone. 00:18:16.360 --> 00:18:20.660 It would listen to everything that you do when it's on you. 00:18:20.980 --> 00:18:23.120 In principle, I like that if it weren't being shared. 00:18:23.340 --> 00:18:28.520 I would like to be able to recall stuff, but it just doesn't make sense for the medical situation. 00:18:29.080 --> 00:18:35.100 So what you're saying, like trying to summarize is basically that kind of stuff is pretty common in the medical space now, 00:18:35.520 --> 00:18:40.820 but maybe a specific provider that has all the right protections and brings a couple of these together. 00:18:41.280 --> 00:18:47.200 I imagine and probably not quite of a good way, as good of a way as these really specialized VC-backed tools, but who knows? 00:18:47.380 --> 00:18:52.340 Yeah, I mean, I've come across tools that are very heavily focused on AI. 00:18:52.750 --> 00:18:54.940 And what they do is like, they had a very good platform. 00:18:55.150 --> 00:18:57.580 Like, you know, it was working well, no problem at all. 00:18:57.860 --> 00:18:59.520 New features kept adding on. 00:18:59.610 --> 00:19:04.280 Like, then AI era started and they like pivoted really hard into AI. 00:19:04.590 --> 00:19:08.980 And they started marketing themselves as an AI first clinical platform. 00:19:09.250 --> 00:19:11.300 And they are still developing it, by the way. 00:19:11.300 --> 00:19:12.520 It's not like they have finished it. 00:19:12.540 --> 00:19:24.880 But yeah, I see that they don't, even though they have a big team of developers working on it, and every time they do something like that, like there is a new feature rolling out and there's always downtime on the production application. 00:19:25.050 --> 00:19:28.120 So it's like, it's a pain to get through it. 00:19:28.320 --> 00:19:28.920 Wow. Interesting. 00:19:29.360 --> 00:19:36.780 I do feel like that a lot of the challenges of software are moving a little bit to the DevOps operational side of things. 00:19:37.220 --> 00:19:42.240 I don't know if you've noticed that, but, you know, it's really pretty easy to ask AI to build you something. 00:19:42.820 --> 00:19:47.160 And either you work with it in an engineering way or it's like you pointed out like a vibe code thing. 00:19:47.540 --> 00:19:51.560 But at some point you're like, I need to put this on the internet for people with a database and stuff. 00:19:51.920 --> 00:19:54.300 And there's a lot of people out there who have never done that. 00:19:55.080 --> 00:19:59.220 And it's just, I think it's, that's kind of a wall or a big step. 00:19:59.360 --> 00:20:03.220 People are being beginning to hit more and more because now they have all these apps they didn't used to have. 00:20:03.360 --> 00:20:05.780 But then how do they get them on the, out to the world? 00:20:05.920 --> 00:20:06.760 Yes. 00:20:07.280 --> 00:20:10.280 Yeah, it sounds a little bit like this company might be struggling with that a bit. 00:20:10.580 --> 00:20:11.720 I think so. 00:20:13.340 --> 00:20:20.860 I'm not really sure how they're doing financially, but I think they are not really doing great because of all this pivot and constant updates. 00:20:21.340 --> 00:20:27.220 But I think this downtime is not just their problem, but I think it's a general problem that is being caused. 00:20:27.540 --> 00:20:33.180 Like even about, I think, three days ago, Facebook crashed in one of the regions of the world. 00:20:33.340 --> 00:20:34.740 And there was an error. 00:20:35.020 --> 00:20:39.760 It's the first time ever that Facebook application showed an error that JSON is not parsable. 00:20:40.420 --> 00:20:41.800 Oh, I remember that. 00:20:41.950 --> 00:20:43.240 Right, right, right, right, right. 00:20:43.360 --> 00:20:48.260 They vibe-coded something that went out and generated truly malformed JSON. 00:20:49.420 --> 00:20:49.520 Yes. 00:20:50.020 --> 00:20:51.680 And somehow it didn't even get tested, yeah. 00:20:51.940 --> 00:20:54.240 Yeah, it ended up on the front end. 00:20:54.270 --> 00:20:55.780 It ended up in the production application. 00:20:56.060 --> 00:20:57.040 How can that happen? 00:20:57.040 --> 00:21:00.040 You have thousands of engineers, so many AI tools. 00:21:00.560 --> 00:21:06.700 And they remarked, like, there are so many companies that market themselves as QA2s. 00:21:06.740 --> 00:21:09.000 You know, there is like a RabbitMQ, I think there is one. 00:21:09.270 --> 00:21:11.220 They market themselves as code quality experts. 00:21:12.020 --> 00:21:15.980 And it's still this kind of things get ended up on the production application. 00:21:16.280 --> 00:21:16.660 Yeah, yeah. 00:21:17.140 --> 00:21:17.480 CodeRabbit. 00:21:17.480 --> 00:21:18.920 I think it's CodeRabbit you're thinking of. 00:21:18.980 --> 00:21:19.960 It's CodeRabbit, yeah. 00:21:20.280 --> 00:21:20.540 Yeah, yeah. 00:21:20.880 --> 00:21:22.360 Rabbit and Q is like more Redis. 00:21:22.520 --> 00:21:23.400 But yeah, CodeRabbit. 00:21:23.580 --> 00:21:25.860 There is a rabbit that knows about code quality. 00:21:26.060 --> 00:21:26.720 It's out there somewhere. 00:21:27.470 --> 00:21:28.840 Yeah, that's pretty wild. 00:21:29.640 --> 00:21:32.080 And recently I've been seeing this GitHub issue. 00:21:32.500 --> 00:21:33.960 Have you seen the status pages? 00:21:34.780 --> 00:21:35.600 Oh my gosh. 00:21:36.060 --> 00:21:36.140 Yeah. 00:21:36.560 --> 00:21:38.000 And how accurate are they, right? 00:21:38.180 --> 00:21:39.860 Like they're like, oh, we got 97% uptime. 00:21:39.880 --> 00:21:41.880 People are like, it's more like 86%. 00:21:41.880 --> 00:21:42.620 86, yeah. 00:21:43.640 --> 00:21:45.440 I think there is a separate page for GitHub, yeah. 00:21:46.140 --> 00:21:47.020 The missing GitHub page. 00:21:47.120 --> 00:21:47.240 Yeah. 00:21:47.480 --> 00:21:48.120 Check this out. 00:21:48.260 --> 00:21:52.880 I think it's starting to, I thought GitHub would just, GitHub would be Google. 00:21:53.060 --> 00:21:56.260 It would just never, never change, never go away. 00:21:56.360 --> 00:21:59.160 It would just constantly be where the magic was, you know what I mean? 00:21:59.380 --> 00:22:02.120 But let me show you, let me pull this up real quick. 00:22:02.360 --> 00:22:06.260 I just heard about this thing from Cursor. 00:22:06.560 --> 00:22:10.540 And by the way, Cursor has a little bit of extra money compared to what they had. 00:22:11.059 --> 00:22:11.740 60 million. 00:22:12.600 --> 00:22:13.420 Oh my gosh. 00:22:13.640 --> 00:22:14.020 You know what? 00:22:14.240 --> 00:22:14.880 Here's the crazy thing. 00:22:15.000 --> 00:22:17.140 People are like, I use Cursor for a couple of years. 00:22:17.140 --> 00:22:17.980 I love Cursor. 00:22:18.200 --> 00:22:20.900 I'd stop using it just purely on a pricing perspective. 00:22:21.140 --> 00:22:22.420 Otherwise, I would still be using Cursor. 00:22:22.460 --> 00:22:22.860 I love it. 00:22:23.160 --> 00:22:29.300 But you would constantly see these people saying, there's no way that this company is going to be around in a year. 00:22:29.360 --> 00:22:35.140 there's no they're losing money xyz there's just no way they were bought for 60 billion dollars 00:22:35.440 --> 00:22:40.180 they turned out okay net net okay so what am i even bringing this up for they are coming up this 00:22:40.260 --> 00:22:46.100 follow-up on your comment here they are coming up with this um project called origin which is a 00:22:46.480 --> 00:22:51.820 git forge aka like GitHub built for the agentic era and apparently it can do something like handle 00:22:51.940 --> 00:22:57.079 like her repost i'm like what do you push this per second or something i don't know some insane 00:22:57.340 --> 00:22:59.140 metric you shouldn't really have to think about. 00:22:59.510 --> 00:23:01.260 But there's a lot. There's going to be some interesting stuff 00:23:01.380 --> 00:23:02.680 like this built, like this origin. 00:23:03.150 --> 00:23:05.200 I don't know. I'll check it out. Maybe I should sign up 00:23:05.300 --> 00:23:07.380 for its wait list. There we go. 00:23:07.700 --> 00:23:08.740 Signed up. They'll reach out to me. 00:23:08.990 --> 00:23:11.200 But I think it's an interesting idea that things 00:23:11.420 --> 00:23:13.000 could evolve and get better and be 00:23:13.190 --> 00:23:15.020 maybe more focused on the way people use tools. 00:23:15.380 --> 00:23:17.240 Yeah. I think recently about 00:23:17.760 --> 00:23:19.560 I think yesterday, I think it was 00:23:19.900 --> 00:23:21.240 Mick Cherney announced something 00:23:21.460 --> 00:23:23.100 that they are moving into medical 00:23:23.480 --> 00:23:25.199 health AI. Oh yeah. How much 00:23:25.220 --> 00:23:26.380 How relevant is that? 00:23:26.650 --> 00:23:27.360 Tell people about this. 00:23:27.630 --> 00:23:28.000 Do you know? 00:23:28.300 --> 00:23:28.760 It's crazy. 00:23:28.810 --> 00:23:29.840 I saw a whole video on it. 00:23:30.020 --> 00:23:30.540 Yeah, I did. 00:23:30.620 --> 00:23:30.760 I did. 00:23:30.830 --> 00:23:33.940 I checked on the blog post and I also read the whole video. 00:23:35.020 --> 00:23:37.120 But the concept is very good. 00:23:37.260 --> 00:23:43.040