Determining your healthy baseline of food through experimentation

Engineer Michael Keller discusses his health journey, wearables, experimentation, and finding what works for your body.


Article highlights

  • Wearing a CGM revealed how even small amounts of foods like raisins and rice spiked Michael's glucose levels higher than expected.
  • Through a process of elimination, Michael discovered culprit ingredients like balsamic vinegar that caused unwanted glucose spikes.
  • Getting baseline nutrition right allows for occasional indulgences without guilt or negative metabolic effects.
  • Real-time CGM data provides personalized feedback for making informed food choices that lead to feeling better.
  • Michael would like to see Levels use data to uncover subtype trends and provide personalized food recommendations.

Michael Keller is an engineer who currently works at Meta. He explored continuous glucose monitoring early on, even considering building his own product with which he could more easily and conveniently monitor his blood glucose. Then he discovered Levels, and is now one of our most tenured members. 

Michael’s Path Toward Optimal Health

Ben Grynol:  You are one of our longest, most tenured Levels members—you’ve been a member for almost two years now. Let’s get into your background. We know you’re based in San Francisco. Were you born and raised in the Bay Area?

Michael Keller: No, I was not. I moved out here in 2016—I guess I’ve been here six years now—when I got recruited by Google. I was in New York before, and was at Google for a while. Then I’ve since switched to Meta, working on the Oculus hardware.

Ben Grynol: Eventually, you discovered Levels. You’ve been a member for two years. And I know it was Dr. Peter Attia who led you down the path of wearing a continuous glucose monitor

Michael Keller: I’ve been listening to Attia for a while now. I became aware of him a few years ago. He often talks about the benefits of CGMs. Before I was aware of Levels, I went to my doctor, and got a prescription for Dexcom. 

I soon realized that it’s a little hard to just be monitoring this on my watch. It actually started the process of building my own app, just for internal use, to do the tracking.Then I became aware of Levels, and I thought, “Oh, my God. They’re doing exactly what I need. I don’t need to do this in my spare time.” I then signed up for the wait list. Just by chance, this was maybe a month after I had got my first sensor.

Ben Grynol: What initially led you to get the prescription from the doctor? When we talk about CGM, we’re still in such an early phase of bringing awareness about it. You’re very much on the forefront of it.

Michael Keller: I was definitely the one who asked. Luckily, my doctor’s very responsive, and we talked through it a bit. He was happy to prescribe one, but it was definitely something that I had brought up.

I’m an engineer. I’m totally nerding out on data. I think it’s just in my nature. A couple of things prompted me. One, I had the tendency to get on a sugar spike and then crash and get cranky and so forth. It’s something that has always been in my nature.

I thought I could probably do a better job of tracking this, just so I’m in a better state overall. And then by listening to Peter, I learned about the pretty serious implications of metabolic health and how it matters and how it’s not obvious what your response is to things. That was definitely a surprise to me, that people are so different, and can have opposite responses to things. I’m also pretty health conscious. That was another big factor. I thought, “Can I do something that’s going to meaningfully impact my health 10, 20, 30 years down the road?”

Ben Grynol: When did you get into health and wellness? When you were growing up as a kid, was it always on your mind, or was there a tipping point?

Michael Keller: It was more of a tipping point. In my 20s, I ate at Wendy’s, and had fast food, and ate pizza, and all that kind of stuff. It wasn’t something I thought about when I was younger. I was probably about 30 when I first moved out to New York. I started feeling maybe more adult, and started being more aware. 

My extended family hasn’t had a history of really taking care of themselves. As they got into their 50s and 60s, I started to see the health repercussions of not being careful. It wasn’t necessarily an abrupt change, but it was a fairly quick change that I made in my 30s, where I really tried to clean up my diet, step up the exercise, and so forth.

Ben Grynol: Was there a single moment you think back to? Sometimes there’ll be a specific podcast—with Dr. Andrew Huberman or Peter Attia, say—and you listen and think, “Never again. I won’t do that one thing.” Matthew Walker, the sleep researcher and author, is a great example of how, even though we know we should sleep more, he puts the information in front of you in such a way where you think, “I have to make changes.” Did you ever have a moment like that?

Michael Keller: Part of it was witnessing some of the health problems some of my family members were running into, especially at the ages they were at. I just looked back and thought, “If you want your outcome to be different, you need to take different actions.” That prodded me in the right direction. From there, it builds on itself. You start eating better, you start feeling better. You go to the gym, you start feeling better. The CGM wasn’t until much later, but somebody like Peter Attia is, obviously, super interesting to me.

I follow him and some of his advice very closely. CGMs provide a meaningful number. It’s not a number for the sake of numbers, which I think you can get a lot of in the health world. It’s something you can change. It is something that matters, and it’s something that’s going to have a long-term impact. In 30 years, my 30-year-old self will be happy I made these changes now. I don’t view it as sacrificing or compromising. It’s a different path, and I’m happy to go down it. I’ll feel better now and I’ll feel better in the future.

CGMs provide a meaningful number. It’s not a number for the sake of numbers, which I think you can get a lot of in the health world. It’s something you can change. It is something that matters, and it’s something that’s going to have a long-term impact.

Ben Grynol: It’s absolutely the long game. When you started to explore some of this, what did your support system look like? Did you have friends who went on this journey with you? Or were you a lone wolf?

Michael Keller: A lot of stuff happened at a similar time. I grew up in Iowa. It’s where I went to school. Then I moved out to New York, and there was definitely a market change. Health and fitness were just much more integral to the lives of my New York friends, compared to those of my earlier friends. It was just something you did. In that way, I didn’t feel like I was standing out. I was in some ways doing the mainstream thing. 

Fewer people are familiar with the topic of a CGM. Some thought, “Why would you do this if you’re not diabetic?” Although, I’ve convinced other people to do this. I’m not a huge outlier in that regard. I probably was slightly nerdier and slightly healthier than the typical person in a friend group, however.

On Navigating Varying Health Attitudes

Ben Grynol: The outlook on fitness and nutrition might be different depending on if you’re from the Midwest, the West Coast, or East Coast. And it’s not to generalize and say anyone outside of the coast isn’t into this. It’s more a matter of the people you surround yourself with, and who support that system. 

Without a doubt, there’s a larger support system of people who are into those sorts of things on the coast. How much do you think geography has to play into it? When you connect with people you would’ve grown up with, or maybe when you go back home, does it feel different? 

Michael Keller: First of all, like you were saying, there’s obviously a variance. There are healthy people everywhere, and also people who don’t try as much. When going back to visit my family, the default is for things to be fried. Diet tends to be more fast food heavy and so forth. In some ways, it’s actually gotten better over the last 10 or 20 years. There is a healthier streak now.

But when you go out, the default you would have does tend to be unhealthy, although there are increasingly lots of healthy options as well. If your peer group is expecting to eat a certain way or exercise in a certain way, in some ways, it makes it easier and makes it part of your normal habits.

Ben Grynol: If you go back home, how do you spread the information? You’ve got a foundation of knowledge that is based on experience. In general, you’ve started to understand metabolic health and you’ve listened to Peter Attia and all these other thought leaders. How do you start to spread that information to your family, or to anyone back home?

The intention is positive, but sometimes people think, “Okay, get off your pedestal.” And you’re trying to help them be healthier and live longer.

Michael Keller: The way that it often manifests for me is trying to, while I’m there, nudge them toward better options. If we were going to go to restaurant A, I might say, “How about we try out this other place you haven’t been before?”

I definitely try to encourage activity, exercise, and so forth. Luckily, my parents are now pretty good about this. It’s much more along the lines of nudging them a little bit and trying to lead by example, rather than pontificating.

Ben Grynol: It all starts with education. You can get as many insights as you want, especially you being a self-proclaimed data nerd. You can get all the data you want, but you’ll have a different understanding because you’re looking at it. You can connect the dots, but the foundation is to read a blog post, say. The foundation for sleep, let’s say, is to read Matthew Walker’s book. That’s your starting point. It is important to make these micro changes so that, behaviorally, people can say, “Hmm, should I think about this versus that?”

The data is what closes the loop and connects the dots for a lot of people. My dad is super healthy. He’s 70 years old, exercises every day, eats healthy. But I think, “Your oatmeal in the morning is causing your blood glucose to spike. Just stop.”

Michael Keller: That is one of the tricky foods that people think is healthy but actually isn’t, necessarily. Choosing to eat oatmeal is a classic example of people thinking they’re making a really good choice, which may not necessarily be as healthy as they think.

Ben Grynol: Exactly. When you start to personalize your diet to what works for you, you realize how crummy you feel. You feel poorly when you eat food that just doesn’t work well for your system, or your schedule. If I’m going to have that thing, like sweet potato or something similar, it’s not about avoiding it altogether, it’s to make sure you understand what’ll happen. But people say, “I’m going to pair that with fat, fiber, protein—maybe I’ll go for a 20-minute walk afterward.” When you do these things, you don’t feel that same sluggishness.

Michael Keller: Exactly. It’s good to know. Those foods are different for everybody. Certainly, getting up and doing some brisk walking afterward can make a big difference. Levels did the Coke study, and that was interesting to quantify the results from that.

Processed food is another one I try to nudge my family away from. I try to encourage them to just try to include more actual raw produce, so that not all of their food is processed. 

Ben Grynol: It makes such a big difference. Especially with processed food, there are so many hidden things—derivatives of sugar that get hidden in ingredient lists.

Michael Keller: It’s definitely much easier in New York or California to acquire healthy foods. In another place, you may have fewer options. I make almost all of my food myself. I really don’t mind it. Then you know exactly what’s in it: lots of fresh ingredients, lots of veggies, and a good protein. I know my baseline food, and I found stuff that works pretty well for me. I don’t have to think about it.

Experimentation and Finding What Works

Ben Grynol: Once you started wearing a CGM, did you discover things you used to eat and then decided, “I’m not doing that now”?

Michael Keller: Yes. I actually struggle with a lot of carbohydrates, although rice, for me, is a pretty bad one. Even, say, a third of a serving is enough to get me to go crazy. I now feel like I can actually sense glucose to some extent. Just with enough data going back and forth, you know what that crash is going to be like, and you’re going to feel queasy or irritable. 

Raisins, even in tiny amounts, are also enough. They are commonly going to be in a big dish. If you go out somewhere, you get those. I try to pick them all out one by one, but I still end up missing some of them, and there’ll be a spike. Ketchup is another one where you think there’s not enough of it to matter. But I mostly don’t consume it anymore. I get a bigger spike from the ketchup than the rest of the meal combined. 

I eat a little bit of fruit as well. I tend to really only eat berries now—blueberries, strawberries, and so forth. It’s most of my fruit supply. Bananas and pineapple are pretty spiky; they’re higher in sugar. That’s probably not a huge surprise. 

I’ve found that cutting carbs, not to the point of keto (I’m not a keto person), but fairly low carb, in general, tends to work much better for me, because a lot of the carbohydrates, even the complex carbohydrates, spike me more than I’m comfortable with. It’s not that it’s terrible, but I don’t ever want to be in the 130s. I just have fewer carbs now. And most of the carbs I have are from vegetables.

Ben Grynol: You really notice that difference. As soon as you start to adjust and you figure out what your healthy baseline is and where you feel good, then you make those adaptations to what you eat. Even the timing—the window—when you eat certain things makes a difference

We’ve got this inaccurate mental model as a society. We say we’re going to have the midnight snack, or a before-bed snack at 10:30 or 11:00 p.m., and we say we’re just fueling our bodies. At the cellular level, metabolically, your body is saying, “What are you doing to me right now? I’m trying to go into a state of sleep. Melatonin’s going up, and now you’re just giving me fuel.” 

Then people have poor-quality sleep. Their glucose levels are all over the map, even if it was a relatively healthy, low-glycemic snack. It doesn’t matter what you’re eating. You’re putting fuel into your body, and it is not wanting to use that fuel right now. What do you think is going to happen? You have a poor night’s sleep, you wake up, you’re groggy, and then you get up and it just oscillates into the next day. You don’t realize that until you see the data behind it.

Michael Keller: Sleep is another one that’s big for me. I’ve traditionally struggled a bit with it, but I do my best. Like you said, it’s impressive how even eating almonds a half-hour before bed can affect the quality of your sleep and show up in the actual numbers. Luckily, I’ve never been a person who’s needed to snack midway through the night. But even if I eat food close to bedtime, it will have a meaningful effect on the quality of my sleep. It’s good to know.

Ben Grynol: Exactly. Because then you become in charge of your own choices and your own health. Anyone can at any time, but you can make very conscious and calculated decisions, like, “I’m not going to eat after 6:00 p.m.” It’s not about saying this is the window for every single person. It’s about knowing what your windows are and when to stop. 

Then you start to feel better. Your energy levels really do change. You go into the new day in a more stable state because of that. You have agency over your own choices in health once you see the feedback. The more wearables you have, the more insights you get.

Michael Keller: As I said before, I’m definitely on the nerdy side. I love the data. I try to always step back and say, “Okay, is this actually useful data for me?” For the CGM, 1000%, it is. Same goes for the sleep tracking. Those are ones I find to be super useful. Then the next step, like you said, is to make decisions that actually make you feel better and make you sleep better.

Ben Grynol: It sounds like you’ve been into wearables for a while. What were you thinking the first time you put on a CGM? 

Michael Keller: I was very excited to get it. I probably did it the first night I had it. There was some apprehension around firing this needle into your body. I thought, “Is it going to be painful? Am I going to put it on in the right place in the right way?” I watched a decent number of YouTube videos before doing it, and everybody said it doesn’t actually hurt at all. I was very excited to do it, and then pretty anxious to start getting feedback from it. 

It takes a while to warm up and get calibrated. I knew it was going to take a couple of weeks before I could start meaningfully going through food and so forth. I was excited to get the data. I was not yet a Levels customer. My first thought was, “How am I actually going to use this?” I knew I could just look at the history, but it’s hard to keep all of that in your head and remember where the peaks were. I thought, “The only way to do this is to go through and build my own app.” I knew I could do this, but it was going to take some time. I started that building process. 

I was very excited at first to see a baseline. You just want to see if the machine thinks you’re in an okay place. I was pretty healthy, so I thought I’d probably be fine. To some extent, you want to try everything, but you also don’t want to be doing everything, and it takes weeks before you start to hit all the different food groups. It was a little hard seeing how even complex carbs were spiking me—maybe not as much as rice would, but it was frustrating to see that. 

I went into it expecting that there was probably some stuff that’s pretty safe to eat on a regular cadence. I didn’t really find that. I discovered that maybe a lower-carb diet was just better for me, which I’m ultimately fine with. I’m happy with the food I eat. But that was one of the frustrations. Everybody’s different, but there’s a food group for everybody. I never really ended up finding those foods. I tried a lot of different things. Quinoa is probably one of the better ones for me, but I still get a spike from it. Even small amounts of rice were enough to make my body go crazy.

Ben Grynol: Sometimes people eat a salad and think there’s no possible way they’re going to get a spike. Let’s say it doesn’t have croutons, or things that people would expect a spike from. But you get the salad that has the raisins, and some dressing with sugar in it. And it feels defeating, because you think you’re being healthy and it’s hard to understand. Then you realize it’s those dried pieces of fruit that have sugar added to them, and their fiber has been stripped away. 

How did you experiment with yourself as far as being able to narrow down things like raisins? 

Michael Keller: People had talked about it before. Raisins are something I don’t normally put in my dishes. They’re not in my pantry. But when you go to other places, it’s something that’s not uncommon to get. Sometimes I’ve either had a certain dish or something pretty analogous to it, and I had no response, but I see that now there are raisins in there. Then all of a sudden I’m at 180.

Sometimes you’re in a place where you just need to eat something. I’d even go through and try to pick them out. I guess I didn’t get all of them or something like that. But anything foreign with raisins for me ends up causing a spike.

Ben Grynol: It happens with things like ketchup, too, or salad dressings. It’s especially poignant if you were having ketchup with something like a fry, where most people suspect the fry to be behind the spike. But let’s say somebody has ketchup on a naked burger or something. They think, “What could it possibly be?” You narrow it down. Maybe you’re not used to reading labels, or maybe you have ketchup that says organic on it. It’s marketing gone wrong because it’s still the same thing.

Michael Keller: This is in my journey. Part of the reason I’ve used a CGM somewhat long term is you get some of the effects pretty quickly. It’s pretty easy to realize, “I added this rice and there was an immediate effect.” There are other foods that take a while before you’ll be able to assess the impact. But you selectively adjust over time, and you realize it was actually this specific ingredient.

I make salads for lunch. It took me a while to make it so that I did not get a huge spike. It turns out the balsamic vinegar I happened to be using had more sugar than the typical balsamic. That one took me a while. I was taking individual things out one at a time. It feels like it’s such a small amount—you’re just drizzling a little bit on there. It’s mostly just acid. That’s one that, over time, you end up seeing. I thought, “Okay, let’s switch to something a little bit better.”

Ben Grynol: In some of those things, it can be highly concentrated, especially if you’re using that ingredient in isolation. But when it starts to get mixed into dishes, it’s more difficult to determine. It’s a process of elimination, or you can do the process of building things up—introduction.

You start with the base of lettuce and nothing else. No spike. Then you add plain chicken—nothing. Then you keep introducing little things and discover the culprit. You’re able to figure it out. One process is not better than the other. They’re both extreme in some senses, but the unpacking can be really, really hard sometimes.

Once you’ve collected the data, it presents it to you as a tradeoff. You can have an opportunity cost: eat this now and it’s yummy, but you know you’re going to pay a little bit of a price for it. It definitely helps me to have that.

Michael Keller: What I ended up doing in that case was pulling an ingredient out one at a time. This is over the course of a couple weeks. This was also just a little bit of day-to-day variance in there. You want to average out a little bit. But in the case of salad, I picked something to not eat for a couple of days. Lettuce wasn’t going to be enough to satisfy me. I needed more than that. But that’s basically the process of elimination, one by one, trying to find out what’s happening.

Ben Grynol: It makes a big difference, having those little swaps. Again, everyone’s going to be different. I don’t do super well with tomatoes. It doesn’t mean I don’t eat them. I’m just super conscious of whether or not I feel like putting them on a salad. I want it, sure, I’ll eat it, or maybe I’ll add a little bit more protein, or a little bit more fat. My wife is different. She can crush tomatoes, and it’s relatively fine for her. 

Michael Keller: Another thing I find a bit confounding is what you’re doing around the process. For me, lunch was not this isolated incident. Sometimes I work out early in the morning, sometimes right before lunch, sometimes right after lunch. That ends up affecting my levels, too. I could take time to try to tease apart everything that’s going on.

Wearables, the Future, and the Importance of Real-Time Data

Ben Grynol: Being an engineer, and the fact that you had the idea to build a product to begin with, what did you hope to get out of a product that didn’t yet exist?

Michael Keller: I thought, “What are the important outcomes here?” It was the peak of the glucose spike and total time above whatever threshold you wanted to establish. It was very much trying to assess those huge factors, which is obviously what Levels does, too. This wasn’t a unique understanding on my side, but that’s basically what I was trying to do.

I’m almost describing Levels here. I would mark when I got the food and then go through and analyze the peaks and how long I was above my threshold. I found that by looking at my phone all the time, it was too hard to aggregate the data, and too hard to remember the actual values.

It’s hard to keep all these things in your brain at once. It’s nice to be able to just go back and review things you see. You get this food you like, and then see you get a spike. Now you know this thing has a number associated with it. With this knowledge, do you really want it right now? Or would you be fine not having it?

Once you’ve collected the data, it presents it to you as a tradeoff. You can have an opportunity cost: eat this now and it’s yummy, but you know you’re going to pay a little bit of a price for it. It definitely helps me to have that. I know that when I’m about to eat something, for the next eight hours, it’s going to show up on that screen. It’s this way to trick myself into thinking, “Hey, the machine’s watching, and it’s going to know if you took something you probably shouldn’t have.” Just knowing that something’s watching is helpful for me to make better choices.

Ben Grynol: You have a good foundation where you understand how sleep, food, exercise—all these things—affect metabolic health. Have you changed how you go about your week productivity-wise? Have you thought about how you structure your lifestyle choices because of those glucose responses you’re able to see?

Michael Keller: My philosophy is to get your baseline really good. I’ve got a baseline of food. I make most of my food at home. I know exactly what it is. You’re rotating between different ingredients. Get the baseline to be really solid. I don’t break 110 with my baseline food. If you really nail that, the occasional cheat day or cheat meal is probably, in the scheme of things, not so bad. 

You have about 100 meals a month, or something like that. It’s fine if you’ve got three or four that are a bit more spiky. Still, there are some things I just still won’t do. But my philosophy is to get the baseline really good, and then worry a little less about what happens at the edges. For me, that always happens on the weekend. It’s pretty rare for it to happen during the week. On a Saturday night, sometimes I’ll do a dessert or something like that, and just enjoy it. I don’t feel super conflicted about that.

Doing the right thing most of the time is where the value comes in, versus if you’re pushing the limits on each meal you’re having, but you don’t ever splurge. You’re probably in a worse place with the latter. That’s the philosophy I’ve taken.

Ben Grynol: It makes such a big difference. The important thing is having flexibility, and agency over your choices. You can make those choices, and that is totally fine. It’s just a matter of shifting when you do. 

If it’s Wednesday and somebody says, “Hey, do you want to have a piece of cheesecake at 1:00 p.m.?” You might say, “I’ll pass.” Without the data, you might say, “Yeah, why not? My friend wants to have cheesecake, let’s do it.” After seeing the data, though, you know you’re not going to feel that well afterward, and that your afternoon is going to suffer.

You’ve seen so many iterations of the product, literally every single one. We continue to evolve the product and the features we’re offering. What are some things you’d like to see, either from a product feature standpoint or even from a membership standpoint, as we start to offer more options for members to be part of Levels in the community?

Michael Keller: Let me first say that I use Levels very differently than I did as a new person. The majority of customers are going to be in this new, exploratory phase. I’m in less of the exploratory phase. The way I actually use Levels now is for what I call my normal meals. I don’t even track them anymore. What I will do is go back and check the day to see if there was anything abnormal. But nothing’s usually abnormal, because I do the same thing.

I use it for new things I’m eating. I used to track every single ingredient and every meal, but because I’ve been using it for a while and have a baseline, that’s not the way I’m using it now. It’s more, here’s a novel situation I’m not quite sure about. And I go and check. 

We’re all very different in how we process foods, in a way I would not have guessed before.
People have different responses. That said, I still believe there’s underlying physiology. I don’t actually think that everybody’s a snowflake. There’s a lot of diversity, but there are underlying metabolic trends. Now you’ve got this bigger collection of people to go off of. I would love to start to group people to get more information.

Even with a small amount of data collection, you can learn if you’re a certain subtype, and which foods are probably good for you, or which foods are bad for you, and so forth—taking that data to the next level. 

My philosophy is to get the baseline really good, and then worry a little less about what happens at the edges.

Levels is doing an awesome job right now of taking pictures of food, viewing it, and having your reports at the end of the day. All that’s excellent. I’m also very interested in where you can get blood tests done. That was really cool to have somebody go out to your house to draw your blood, versus having to visit a lab. I’d love to see that expanded. 

Overall, Levels is doing an amazing job. I 100% recommend it. Maybe, if I was proposing a project, it would be trying to use some of the big data to see if we can ascertain some of the actual effects happening.

Ben Grynol: Are there other markers or analytes you’re interested in, and would like to monitor in real-time?

Michael Keller: I get regular blood testing. The inflammation ones, I think, are super important to be on top of. I know the hsCRP was included there. I also watch the heart stuff closely, because most people are going to die of that eventually. There’s heart disease in my family. 

I’d love to see the testing expand just a little bit, maybe add some additional tests. If I could make a medical device that measured cortisol, I think that would be amazing. I notice it for myself—work can be stressful day in and day out. It’s similar to the glucose, where you maybe don’t realize you’re in it until something happens to make you aware of it. It’s one of those things that actually impacts your overall health and wellbeing quite a bit. 

Insulin’s another one. But I’ve heard that insulin’s next to impossible to do on small scales. But glucose is, certainly, the most important one. But if I had to pick, cortisol would be another great one.

Ben Grynol: Cortisol is very interesting. I’m sure you’ve seen it show up in your CGM data, where, if you feel stressed, you get the fight-or-flight response. You’re sitting in your chair, and your levels shoot up. Then you start to realize how deeply associated cortisol can be with a molecule like glucose. 

There is a benefit to understanding so many different markers in our body and what they’re doing in real time, versus getting your A1C checked once a year, let’s say, and your doctor says you’re doing great. It could be far from your baseline, or it could be looking great, but your baseline is actually not that great. It’s all important to take into account.

Michael Keller: Definitely. The standard deviation matters a lot. And then there’s a lot of variability about how the A1C number actually works out. Some people cycle protein a bit more quickly, for instance. The actual real-time glucose is the real response. The A1C is just an approximation.