Event Replay: Building a Personal Health Operating System

[00:00:00] Chris Nicholson: Hello everyone and thank you for joining us. My name is Chris Nicholson and today, we are going to hear from Amy Deng about how she used AI to navigate a complex health journey and build what she calls a personal health operating system. So Amy started by building these systems for herself as a brain tumor patient using AI to navigate her own health. This made her realize that complex medical care is actually an information management problem, scattered information that needs aggregation in order to be analyzed and understood. That led her now to help other families use AI to navigate their health, which sounds incredibly important. Today she'll share more about how she used AI to navigate her care, and she's going to demo one of the ways she uses AI today to track her symptoms. Amy, thank you for being here.

[00:00:54] Amy Deng: Thank you so much for having me.

Chris Nicholson: Yeah. So much fun. So we're going to start. Maybe you could just introduce yourself and your healthcare journey.

[00:01:01] Amy Deng: Yeah. Hi everyone. My name is Amy and I was a researcher and leader working on AI capabilities evaluations prior to all of this. In July of 2025, after a period of very abnormal fatigue, I was diagnosed with a benign brain tumor called prolactinoma, which is a tumor on the pituitary gland that sits in the middle of your head. I had two surgeries in August and November of 2025, but unfortunately neither was able to fully remove the tumor. Now I'm on lifelong medication to manage tumor growth. I started medication earlier this year. By April of this year, I was hoping to just really get back to normal life. I felt pretty ready, but then I was suddenly hit with a very abnormal wave of fatigue. It's not like anything I'd ever experienced before. I could be having a normal day in the morning, but then just feel so tired in the afternoon that walking a couple of blocks to the grocery store would feel like the hardest thing ever. That period of time ended up changing how I thought about how to be a patient, how to be more agentic, and how normal people can use AI tools to improve their health. It's definitely been a very long journey. I still am a patient. I just learned a couple of weeks ago that I probably sustained some permanent pituitary damage and I would have to receive continuous endocrinology support going forward. So I'm extra motivated to build these systems for myself and for others. I think we're in a very exciting era and the AI capabilities is largely there for us to use these tools to improve our bodies. I think the missing link is rather people's awareness that they can do so and their AI literacy to make this happen. So I'm very excited to be here today and share some of my stories.

[00:02:57] Chris Nicholson: Very cool. Let's zoom out a little bit. So you experienced this deep fatigue, but you were not a tired person in general. So maybe talk to me about the journey into fatigue and how you realized, well, this is something I need to fix.

[00:03:14] Amy Deng: Yeah. So I'm normally a very bubbly person. I fill my weekends with plans and I work a lot during the week. But then when I was trying to get back to work in April, I just couldn't even do 20-30 hours of work a week. I was really trying and I was really not getting there. I would make plans with friends, but then have to cancel very last minute because just when I was about to get ready and go, I would feel the crash coming. I feel like I was living a very small life for a couple of weeks because I couldn't really do all the things that gave me joy.

[00:03:58] Amy Deng: I'm going to have to get on this problem because nobody is going to solve it for me by default.

[00:04:02] Chris Nicholson: Yeah, so you've described fatigue as a complex condition. It sounds like it doesn't have just one cause. How did you go about tracking down the sources of your fatigue?

[00:04:16] Amy Deng: Yeah, so fatigue is, I think, kind of the vaguest symptoms that we can give a physician. If you tell a physician you're fatigued, they can come up with like 25 different causes, and I say should they start with the most basic ones, like are you allergic to anything, or, like, you know, are you in front of your screens too much or something like that. But I think the most important thing when you're fatigued is you want to get out of it as soon as possible because it sucks so much. And so the most important thing I realized at the time was that I need to systematically figure out what mattered and what did not matter out of all the possibilities out there. So one day I just decided to ask ChatGPT, given all of my prior conditions, what are some possibilities? Can you come up with a hypothesis on what could be going on? And then obviously ChatGPT gave me a bunch of hypotheses. So I, like, asked to rank them, prioritize them, and then I asked it help me think about what data can validate and invalidate those hypothesis. I, like, really started to approach this as a research problem, or it's not that complex, it's just how we approach science problems in middle school. Just what is a hypothesis, what data can validate and invalidate it? And then I started thinking about how do I track those data? And some of those are easy, some of those are hard. I can get into more of those if you're interested. And then, once I have the data, I can then analyze and be like, are there correlations between these metrics and my fatigue? And if there isn't, then that's a dead end, we're not gonna pursue that. If there may be something interesting, then we're gonna dig into it, we can try a lifestyle change, or we can talk to doctors about it. And that's kind of the high level how I was approaching this fatigue problem.

[00:05:53] Chris Nicholson: Right, okay, so you're in this, you're in the hole, you're feeling this fatigue, and there's just a ton of potential explanations. And clinicians are offering you some of them. And your problem is, how do I even figure out what's a promising lead to pursue? And so, what I'm hearing is you're using AI and various systems you're building to collect the data that's going to give a heat map for you so you can focus on the top few.

[00:06:20] Amy Deng: Basically, yeah.

[00:06:21] Chris Nicholson: Is that what? So what's an example of something that you had to enter data for and that you surfaced as a lead that you could later kind of discuss with a clinician?

[00:06:32] Amy Deng: Yeah, so I guess I'll give a few examples for what didn't track and give one example for what did work. Because I think when you have 100 possible contributors, ruling things out is just as important as ruling things in. It prevents you from trying desperate measures to help yourself feel better. So one thing that ChatGPT suggested was maybe you have post-exercise, post-exertion fatigue. So for a period of time, I tracked on a spreadsheet. It didn't have to be any complicated tools or anything. I just tracked on a spreadsheet whether I exercised or not, what is my number of steps I walked, which I tracked via my wearable. And I also tracked the variable that I'm trying to change, which is the fatigue, so how I was feeling on a day-to-day. And I basically ranked one to five every hour, how I was feeling on that day. And this showed no correlation between how much exercise I did, how active I was versus how tired I was on average. So that meant that I didn't have to change my exercise. I can continue to exercise. I didn't have to be scared. Similarly, we ruled out blood glucose problem as a possible source. I didn't have to worry about eating a big lunch or eating a big dinner and that causing it. And I didn't have to cut out sweets, it's awesome. And it's very important, yeah. And then something that did show up through this is I tracked roughly my calorie intake and like the macro breakdown of my diet for a bit. And it didn't really show up as something to be concerned about when I looked at it myself or with AI, but just the action of doing so made me realize maybe I should go see a dietician and have a more professional opinion on it. And because I have all the data ready, I can just go to the dietician. I actually sent this dietician a doc with a few examples of what I was eating on the days before and my continuous glucose monitor tracking prior to the appointment. So on the very first appointment she was able to tell me, oh, I think you're under eating, particularly under eating on carbs. And if you maybe just bump up your carb intake, this is a very tractable change that you can track and see if it helps you feel better. And it did indeed help me feel better, like pretty quickly.

[00:08:56] Amy Deng: That ended up contributing to my fatigue, this was one that caused a lot of big magnitude of change and pretty immediately. And I also think that I, I think A, I wouldn't have thought about going to a dietitian if I hadn't really systematically looked at all the factors. I might've wanted to prioritize other things because diet is not like an illness per se, like you might've wanted to dig into whether you have autonomic disorders. Like I would have prioritized my blood work and my like tumor-related stuff. But because I was looking at everything in a systematic manner and like really being level-headed about prioritizing what to look at, what to track, I was able to kind of zoom in on this, which is a correct thing to zoom in on. And then a second thing that I think really helped, this system really helped with, it's just the speed of figuring it out once you've identified something.

[00:09:47] Amy Deng: Because alternatively, I would have gone to this dietitian, we would have spent the first meeting just doing normal intake and question asking, and then she would have told me, okay, go track your diet for two weeks. And then I would have come back in two weeks with that data and she would have arrived the same insight that she arrived at right away. And I would have probably been fatigued for two more weeks.

[00:10:10] Chris Nicholson: For sure two weeks if you're lucky. So what do you think most people are not approaching their clinicians correctly with the right data? You sound like one of the most prepared patients a clinician might meet. And that sounds super useful. But what's the failure mode here? Are most people doing it differently? And could they do it better?

[00:10:25] Amy Deng: Well, I don't know how most people do it. But I think prior to being very systematic about approaching my fatigue, I was a pretty conventional patient. I just scheduled my appointment. I show up, I listened to what the doctors have to say. I do what they tell me. And then we go from there. I think doctors actually really appreciate... I talked to my doctors about this. I'm like, am I overwhelming you? And I think certain doctors really appreciate you showing up prepared. My PCP sister, she loves it when people bring her a chart of some kind. Because when you bring an insight that you formed, it gave them a lot better information to work with. This information that you prepare for them is not at all replacing their judgment. You're in fact giving them better information to exercise their judgment on. So you're empowering your doctor to do a better job for you. I would highly recommend people to ask ChatGPT before a specialist appointment, especially, you know, for normal people, you only get so much time with a specialist. You've taken weeks to months to get that 30 minutes of their time. And I highly recommend folks to ask ChatGPT what information would that doctor want to see at the appointment, what would help them? And a lot of that information might not be things that you can gather yourself. Let's say an MRI needs to be ordered by a clinician.

[00:11:46] Amy Deng: But there are other things that maybe it's just like your day-to-day normal function or things related to your symptom that are qualitative, that isn't something that fits into an EHR record that they would really like to know. And we know that because at an appointment, usually they start the question by asking, how are you feeling? Like, describe how you're feeling. But you're under the pressure to recount everything that you've been going through in like that one minute time that you get to talk. And instead, you can show up with something more structured and very dense. And that gives them like a jumping point to go off and analyze what's going on.

[00:12:25] Chris Nicholson: So one of the lessons I'm really absorbing from you is that everyone who's a patient can actually claim a lot more agency.

[00:12:34] Amy Deng: Yes.

[00:12:34] Chris Nicholson: And one of the effects of that will be accelerating, improving the decisions their clinicians can make in a short amount of time. The 15 minutes they've got with the doctor, so that each appointment matters. And that clinician can say, get this test, or prescribe this, or do this. And all of a sudden, they're feeling better sooner.

[00:12:53] Chris Nicholson: Exactly. So the gap is the agency. So you're doing new things with AI. AI's doing some things, you're doing some things. Is it hard to build those habits? Like how do you keep yourself on to make sure you're bringing everything you think you need to the doctor?

[00:13:07] Amy Deng: I think obviously, your motivation to do such a thing depends on how dire of a situation you're in. I think if you just have a normal cough or whatever, I wouldn't recommend overthinking it or over-engineering anything. But if you are in a situation where every bit of information matters, which I think particularly matters for what I call complex information diseases. So that's anything from chronic illness or undiagnosed things, or things like cancer. Like every bit of information might give you an additional clue in what's happening or might give you additional evidence on the best decision that you can make. I think in those cases it's worth being very systematic. And I think obviously, like all the habits start very small.

[00:13:54] Amy Deng: Like, all the habits start very small. It might just start with you logging something down every other day when you do remember. And I think, like, what I call the health operating system is not anything too complex. It's just kind of a system that you create for yourself that is based on your needs and your preferences and your energy level, how much time you have. Whatever works for you. Whatever works for you. But the gist is the same. The gist is figuring out what you need to track, find the easiest way to track them, find the easiest way to gather the data, and then use AI and user doctor's intelligence to analyze this data.

[00:14:29] Chris Nicholson: So I think a lot of people are probably maybe not feeling great, feeling bad. They might be feeling fatigued and think, asking themselves, is this normal? They might not have a name for it. And so when you say that what's useful is an information management system, I really feel like that's the flip side of this is complex. And information will help us see through the complexity. Yeah.

[00:14:52] Chris Nicholson: It sounds like your care has been complex. You've seen a lot of different people. Yeah. That's the nature of health care. Is it hard to get a unified vision of all, it's like the blind men and the elephant. Is it hard to get a unified vision of their point of view and what they're telling you?

[00:15:08] Amy Deng: I think this is a very interesting thing that you brought up about the American healthcare system. The American healthcare system is structured around specialists and appointments. But a lot of complex illnesses fall in between those cracks. Like my neuroendocrinology team correctly only really looks at the world through the lens of the pituitary. And they cannot order a test or they do not do investigations for blood pressure or glucose or other things. So when you're in the position of navigating a complex disease, you are your only quarterback for your care. And I think that's when AI and the health operating system become very useful because it helps you accomplish the task of gathering information much easier than before. Like I think if we had to go like 10, 20 years back, perhaps you have to hire a software engineer or a team of software engineers to build a dashboard for you or write custom software or something. But now you could just ask Codex to do something for you and that makes it very accessible to everyone.

[00:16:09] Chris Nicholson: Yes, so there's these two things. First of all, it's hard to see into ourselves but you're tracking what's happening in yourself, you're able to quantify it is the first thing. But you're also maybe the only person seeing each doctor one after the other. Exactly. So they're giving you different points of view. You're unifying it along with what you know about yourself. And so it feels like there's a couple of different steps. You're aggregating and then you're analyzing. How does that work? How does... So you're, like concretely, how do you aggregate? So people in the audience know like, what does it mean to aggregate data from a clinician's visit and then create a bigger understanding from it?

[00:16:45] Amy Deng: Yeah, yeah, yeah. Unifying it with everything else. Yeah, so I guess I briefly mentioned already, how do I collect data? I use spreadsheets, wearables, like different tests that I can have access to. And it's a lot of diet tracking apps. And then once I have all of this data, they currently live in like a bajillion different information sources. They're in my Apple Health app. They're in my EHR portal. They're in like my email threads sometimes. And so at the time, I actually just exported all of those data and put them in one folder on my local computer. And then I ran Codex over it. The reason why I chose, so I actually both tried Codex and created a project in ChatGPT. I think there's pros and cons to it. I think the pros to Codex is mostly that some of the data that I have is quite complex. So, and it took a lot. And the coding agent has a better capability of sifting through a lot of complex data. You can, if that's difficult, if you're not technical, you can also just start with uploading all your files on the web, on chat. And I just asked questions that I would like to know. Like what tracks with my fatigue? What is correlated to that? And a big headache was periodically uploading, like periodically renewing, syncing the records because new data might come into my EHR portal or new continuous glucose monitoring might come up. And you would then have to re-export these data, put it back in the folder, redo the analysis with the new data. So I'm actually very glad to see ChatGPT Health doing, providing like Apple health, Function health and EHR connectors now. It would've made my setup a bit easier because then I could, like those data are already synced in the platform and I can just bring in other data that isn't synced.

[00:18:36] Chris Nicholson: Yeah, for folks in the audience, ChatGPT work can do this for you, Codex has worked for you, if you're not a software engineer. I don't wanna plug our stuff a lot, but like ChatGPT work will do that for you. So let.s get back to this journey.

[00:18:52] Chris Nicholson: You're highly energetic, you're accomplishing so much, the tumor happens, the operations happen, you realize you've got fatigue and you begin to get these leads out of it. Where are you at now?

[00:19:06] Amy Deng: I'm feeling maybe like 90, 95% back to normal. I feel very grateful. Most days I can work just as hard, if not better, I guess maybe possibly even better than right before the tumor was diagnosed. Yeah, I feel very grateful for that. But I'm still navigating variations with the rest of the 5% of the time because there's still so much hormonal changes that I'm going through with the additional endocrinology support. So I am very motivated myself to learn how my body works in this new normal, which is why these systems are still useful to me. Yeah, because it gives me insight on just how does my body function now.

[00:19:48] Chris Nicholson: Yeah. So what you've built is incredibly impressive. It also seems very complex in itself. When you're speaking with families dealing with these complex conditions, where do you get them started? Like, what's the one thing they can do that feels doable?

[00:20:04] Amy Deng: That's a really great question. I think it really depends on what is the problem that they're trying to tackle. And I think it largely depends on what kind of situation you're in. So I see people in two large, two buckets. Most of the people fall in the bucket of you're navigating your own care and you, possibly your family's helping you, but you are really navigating your own care. In that case, I would just advocate for the easiest, highest-leverage thing that you can do. And again, it differs. Depends on what condition you're trying to navigate. But plausibly, I would just have a conversation with ChatGPT and be like, this is the thing that I'm trying to figure out. This is my bandwidth. What information do you think I should collect? What do you think I should get started with? Like, this is my background. Am I an engineer? Am I not an engineer? And it can probably help you come up with something pretty reasonable. I think another bucket of situations is family dealing with very complex care that involves many different people. You might involve home health aides. You have multiple family members rotating, taking care of someone. You have multiple doctors. And in that case, everybody's hair is on fire. And you really, the thing that you really want to make sure is that everybody is on the same page, and because everybody is generating information and insights from just everywhere, and you want to pull all that together so you're not missing anything. And in that case, I would probably have a conversation with ChatGPT about this, and see what's a reasonable thing for you to really aggregate all of that data and capturing it.

Chris Nicholson: I have a ton of other questions, but I want to make sure everybody sees your demo. So you brought an example of how you use ChatGPT on this. You're gonna share your screen with us.

[00:21:53] Amy Deng: Yes.

[00:21:54] Chris Nicholson: You're gonna talk us through it.

[00:21:55] Amy Deng: Sure, yeah. So as I mentioned a little bit earlier, a lot of the times the non-clinical data, like what you see, how you feel on a day to day, is very informative, but it's quite hard to track because it requires you to type into your laptop when you're most fatigued. So I thought about what if I could just mumble into my phone when I'm in bed and this data can just be captured on my laptop that I can then use to analyze. So I kind of put together this little workflow with Codex Dispatch, which allows you to control your codex from your phone if your laptop is awake. And then I use ChatGPT to help us analyze it. And now we can watch the demo. So this is me setting up Codex Dispatch, connecting to my MacBook on my phone. And you can obviously look up how you can do this with documentation online. I go into this little forum demo folder and I mumble my symptoms into it. I think I talked about that day. Like, I didn't really feel like eating and I was feeling brain fogged. And I asked some question around, like, connections between tired and having hunger signal. I wonder if those two things are correlated in some way and maybe spiraling with each other.

[00:23:08] Chris Nicholson: Pause for a second.

[00:23:11] Amy Deng: Yeah.

[00:23:12] Chris Nicholson: So when people are too tired to type, which you have been, they can just talk to ChatGPT. So there is voice mode, obviously, and then there is, like, voice transcription. I know voice mode isn't available right now if you're interacting with Codex as such. So I just use the transcription and it's pretty accurate.

[00:23:30] Chris Nicholson: Okay, got it. So you've got the dictation in Codex and in ChatGPT you've got the voice mode.

[00:23:34] Amy Deng: Yeah.

[00:23:34] Chris Nicholson: So basically if you don't have the energy to type, you just have a conversation.

[00:23:38] Amy Deng: Totally.

[00:23:39] Chris Nicholson: Get it in there and you'll basically get the same response. Well, the same response depends on what tool you're using. You'll still be able to apply AI too.

[00:23:46] Amy Deng: Yes, absolutely.

[00:23:48] Chris Nicholson: Okay, awesome, great. That's just, I wanted to clarify that.

[00:23:50] Amy Deng: Totally.

[00:23:50] Amy Deng: Totally. I want to flag, though, that usually in voice mode, voice mode uses a smaller and faster model because it wants to be more instant, so it's maybe less of a frontier model.

[00:23:59] Chris Nicholson: Okay, so that's a great tip. So if you really want the frontier model, you do dictation, and you enter it as text.

[00:24:04] Amy Deng: Yes.

[00:24:05] Chris Nicholson: Okay, awesome.

[00:24:07] Amy Deng: Yes. So back here, basically, I asked Codex to write a skill for me, which I will show you guys in a second. Basically, the skill just says whenever you see Amy rambling her symptoms, you want to log it down on her laptop. It doesn't matter what she was asking. If it's present in the text, write it down on her laptop. So now my Codex is doing work. By the way, this entire time, I had my phone shut down. I'm screen recording, so my phone is alive. But if I didn't have to screen record, I could just put it back in my pocket, go back to what I was doing. Then Codex is just doing the work in the background for me. Now it had written the symptoms to my computer. You can see it's a structured format, where it has the dates, it has the entire raw text that I spoke into, and my symptoms and names and qualifiers for it. Then it also answered my question that I asked above.

[00:25:00] Chris Nicholson: Pausing again real quick here.

[00:25:03] Amy Deng: So this is basically my laptop screen that I later opened my laptop, and it's the same session of Codex that we ran earlier. You can see it's the same output on my screen as well. Here I just wanted to show you guys the skill that I wrote, which basically just says, when the user describes their symptoms, write it down. And by the way, I did not write the skill myself. You don't even need to know what a skill is, to be honest. You just ask Codex, write a skill for me that does this, and then that just, it just didn't work. Here, basically I'm just showing you guys that this is a file that indeed existed on my computer called Symptoms. And this file indeed contains the data that Codex said that I wrote to my laptop. And you can ignore the code editor on the right. You can kind of, if you're not a software engineer, and you don't know how this works, you can just ask Codex to write the logged information to a text file or a docs or something, whatever you have software to view. And then the real magic happens when you actually analyze this data. So I used ChatGPT Health to analyze this data because I wanted to see how it correlates with other records that I have connected to ChatGPT Health, which includes like my wearable data and some of my medical records. And you can see them here. And I went to, yeah, in this chat, I just basically uploaded the file and asked, can you help me visualize some of this data? I like visualizations personally. I think a picture speaks a lot better than words. And you can see very interestingly, four days roughly after I stopped, I have had appetite problems. I started to experience way worse fatigue. So bad fatigue is, I think it's high or it's top. So the fatigue, there's a lag but it follows the appetite disturbance, I guess.

[00:27:00] Amy Deng: Yeah, exactly, and I think it's very good. It's very interesting for me to know how many days until it happened because the next time I stopped eating well for a couple of days, I know I need to get on it. And then another interesting thing that I found is like walking speed is somewhat related to my fatigue, which makes sense, although I don't quite know what to do with this data. But it's an interesting leading indicator if I could just get my fatigue data from my wearable. And then it also showed me, go back a little bit, that my sleep did not track with the fatigue, which is also good to know because it just is, like I said earlier, one less thing to worry about and ruling things out is just as important as ruling things in. And then it gave me a little bit more analysis.

[00:27:39] Chris Nicholson: Yeah, that's the demo. Amazing, okay, so I feel really motivated by your story. You got insights about yourself that sped up the pace at which you could get better care. And now you have not totally recovered, but you've come a long way towards recovering from fatigue. And I hope that a lot of people can do that with these tools. How long did it take you to really feel like it was working? You were starting to get something back that felt promising, where you were like, wow, I think I'm on a different track because I'm using the system.

[00:28:14] Amy Deng: Hm, that's a great question. I think it really depends, for me it may be a couple, like even just one week, one and a half weeks. So, I think that's when you have like 10 days worth of data point and you can roughly start to see what matters and what doesn't matter. I would say though, it is a lot of work and determination to set something up like this. And I also kind of stopped doing the intense version once I felt better, because it's not that useful anymore. So I would highly recommend folks try and play with these tools, try to build something themselves.

[00:28:48] Amy Deng: Be aware of what they really need and not to over-engineer. [00:28:51] Chris Nicholson: Yeah, so people can hook up connectors within Codex. Are there open-source things or prompts that you recommend for people to really bootstrap themselves into a system? [00:29:02] Amy Deng: I think probably the easiest place to start is ChatGPT Health because you just need a few clicks to connect your records. So I personally wrote a skill that I published online, but I think it's very custom to figuring out small things lifestyle-wise, like fatigue-like things. I think for every different research, it will be very different.

[00:29:26] Chris Nicholson: Okay. We will share that link to your skill around fatigue after this session. So I think that we're going to audience Q&A now. We've got questions coming in.

[00:29:40] Chris Nicholson: Let's see here. I have a question from Andrea Germanson who's a member of Deloitte Global Boardroom. Andrea says, what was the most surprising thing that AI helped you discover that you don't think you would have noticed otherwise?

[00:29:57] Amy Deng: The most surprising thing. I think a lot of things in hindsight did not feel surprising. But I think it would have just maybe taken me a long time to get there if I didn't have a systematic way to track these things. I can give an example of the family that I'm helping with right now. And some stuff that got surfaced via the system that I sort of built for them to make sure that all of the people involved in their care is on the same page. I think one thing that got surfaced, for example, was so we were tracking how he's eating on the day to day. And I think folks who are not there in person is not aware of whether his diet has improved. And such is the thing that you don't really think to ask because it's not the highest priority item, but it's very important as part of recovery. And now I have a dashboard showing is he eating well on a day to day? How many booths is he drinking if he is not eating? And that gives a great way for folks to visualize where his caloric intake is. And yeah, I think that would have been something that just wasn't surfaced at all if we didn't have the system.

[00:31:06] Chris Nicholson: Amazing, so aggregating the data helps solve parts of the handoff problem. I think hospitals see it among clinicians, nurses, and doctors. Information doesn't get passed on. It all happens in home health care, right? And it also sounds like it happens between patients and clinicians. So these systems make sure that the right information gets passed on to the right people, especially in these complex situations.

[00:31:27] Amy Deng: Yes, exactly.

[00:31:28] Chris Nicholson: Cool, okay. All right, we've got another question. It's from Jason Deluca, who's at Crossing Point IT Solutions. He says, for someone without your AI research background, you started out with a head start. What is the simplest safe version of a personal health operating system they can build today and what should they avoid?

[00:31:47] Amy Deng: I think I would go back to relay your question to ChatGPT, relay what you're trying to figure out and ask it the simplest thing that you want to make. I think it's really hard for me to answer in the abstract. What is the simplest thing that you can build for yourself because everyone's goal is different. For example, if you're trying to recover from a physical injury and you're doing PT or something, you might want to track your pain level at different days or connect to your wearable to see whether the amount of walking you're doing or calories you're burning tracks if your pain's increase or decrease. But I think it's very interesting because as I'm working with other people on this, I realized that you can be very creative with how you use these tools and sometimes you come up with things or learn things that certain tools can do that you just didn't really realize before. So yeah, I think it's very exciting. I think we're in a very exciting era where the AI capabilities are already largely there, that it can enable everyday people to just go and play with it. I think one thing that I really want to say to all of the audience is that there is pretty much nothing that you cannot learn in this world anymore. I think it's always true, but now more so than ever. So for anything that you don't know how to do, you can choose to do two things. One, ask an agent to do it for you. Just ignore how it did it. Two, you can ask how you can do it yourself and ask it to explain until you have a step-by-step instruction. And then if it doesn't work, you tell it why it doesn't work. And then it'll probably help you through the entire process. So I think it's really exciting that we can, I really want to empower people to build these tools themselves.

[00:33:27] Chris Nicholson: For sure. One thing I'm really learning from you and hearing from you is that you really suffered and you cared more than anybody else on the planet. You cared every second of the day about getting better. And no matter how much your clinicians care, they see you for 15 minutes every so often.

[00:33:46] Chris Nicholson: Extremely motivated and these tools are helping you gain agency to quarterback that and really any patient out there who can talk to a computer could leverage their how much they care about being sick into getting better. Yeah with these tools. That's the amazing thing for us, okay?

[00:34:03] Chris Nicholson: Let's see here. We've got another one Daniel green. Looking back. What do you wish you'd known when you first started this journey? Hmm?

[00:34:12] Amy Deng: I think if I had something to say to myself when I was first diagnosed, I would say that it's not going to be a quick journey. I think I really thought that it's gonna be one and done. I do a surgery when I get out come back to work. I did a version of that but I think I'm still gradually accepting that it's going to be a lifelong process of navigating my new body and there is a lot of grief that comes with it, too.

[00:34:48] Amy Deng: Yeah, yeah, this is nothing with the house operating system but yeah, yeah, that's um...

[00:34:56] Chris Nicholson: Yeah, that makes a lot of sense to me. I think everybody who is lucky enough to live long on this planet faces more and more constraints with their body and some people are unlucky enough to face constraints very very early.

[00:35:12] Chris Nicholson: Right, but everybody runs into those constraints and has to like realize the body there again. It sounds like you've really realized that and you found, I don't know if they're tools of liberation but ways of managing some of the some of the constraints and overcoming constraints is what I hear in your story. Hmm.

[00:35:32] Amy Deng: Yeah.

[00:35:32] Chris Nicholson: Yeah, do you agree?

[00:35:35] Amy Deng: Yeah, I would agree.

Chris Nicholson: One of the things I noticed it's kind of in response to Jason is that he was asking you what advice would you give and you refuse to give universal advice, right? And really the lesson that I'm getting from your relationship to AI is that you're getting great advice by putting a lot of information.

[00:35:51] Amy Deng: Yes, and the quality of the advice correlates directly with how much is known.

Chris Nicholson: Yeah. Yeah. Yeah, right. So the real question is just how do people get into a situation where they can let more be known? Yeah, so they get better advice back.

[00:36:01] Amy Deng: Totally.

Chris Nicholson: Yeah. Yeah. Okay. Yeah.

[00:36:06] Chris Nicholson: I think those are the questions we've we've got Amy, this was mind-blowing and if I could wish one thing for the planet, it would be I wish more people would just get healthier following the path or their own path but learning the lessons from you that you have done. So thank you so much for coming.

[00:36:31] Amy Deng: Oh...

[00:36:34] Chris Nicholson: So here I'm supposed to say something else. If you would like to join us for another forum conversation, you can register for our upcoming K-12 Showcase. That's gonna be lessons from teachers and administrators. It's on August 24th. We're gonna share the registration link for that and additional upcoming Forum events and let me just tell you I've been talking to teachers who are accomplishing so much serving kids with different needs when they're teaching dozens of kids in the classroom in there. They have more than one classroom to teach in a classroom. They're really accomplishing a lot with these tools, so I hope that anyone who cares about education will come to that event.

[00:37:15] Chris Nicholson: You mean that was amazing.

[00:37:18] Amy Deng: Thank you.

Chris Nicholson: I'm so glad we had you.

[00:37:22] Chris Nicholson: Yeah, we're gonna share that link. Thanks again for joining us everybody.