Is AI More Sustainable if You Generate It Underwater?

AI data centers are so hot right now. Each time generative AI services churn through their large language models to make a chatbot answer one of your questions, it takes a great deal of processing power to sift through all that data. Doing so can use massive amounts of energy, which means the proliferation of AI is raising questions about how sustainable this tech actually is and how it affects the ecosystems around it. Some companies think they have a solution: running those data centers underwater, where they can use the surrounding seawater to cool and better control the temperature of the hard working GPUs inside. But it turns out just plopping something into the ocean isn't always a foolproof plan for reducing its environmental impact.

This week on Gadget Lab, WIRED writers Paresh Dave and Reece Rogers join the show to talk about their reporting on underwater data centers and how the race to power AI systems is taking its toll on the environment.

Show Notes

Read Paresh and Reece’s story about the plan to put an underwater data center in the San Francisco Bay. Read Reece’s stories about how this is AI’s hyper-consumption era and how to wade through all the AI hype. Read Lauren’s story about the social network inhabited only by bots. Read Karen Hao’s story in The Atlantic about how companies like Microsoft are taking water from the desert to use for cooling down AI data centers. Here’s the Black Cat substack article about the character Harper from Industry. Follow all of WIRED’s AI and climate coverage.

Recommendations

Paresh recommends checking out cookbooks from your local library. Reece recommends the soundtrack of the first Twilight movie for all your Fall feels. Lauren recommends the HBO show Industry. Mike recommends Anna Weiner’s profile of bicycle designer Grant Peterson in The New Yorker.

Reece Rogers can be found on social media @thiccreese. Paresh Dave is @peard33. Lauren Goode is @LaurenGoode. Michael Calore is @[email protected]. Bling the main hotline at @GadgetLab. The show is produced by Boone Ashworth (@booneashworth). Our theme music is by Solar Keys.

How to Listen

You can always listen to this week's podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here's how:

If you're on an iPhone or iPad, open the app called Podcasts, or just tap this link. You can also download an app like Overcast or Pocket Casts, and search for Gadget Lab. If you use Android, you can find us in the Google Podcasts app just by tapping here. We’re on Spotify too. And in case you really need it, here's the RSS feed.

Transcript

Note: This is an automated transcript, which may contain errors.

Michael Calore: Lauren.

Lauren Goode: Mike.

Michael Calore: How often are you thinking about your environmental footprint?

Lauren Goode: All the time.

Michael Calore: Really?

Lauren Goode: Yeah, like all the time. Really, everything I do, even if I just get takeout and there's a plastic bag, I think, "How many times should I responsibly reuse this? And why don't we have another solution?" And every time I get in my car instead of biking somewhere, which you encourage me to bike and sometimes I don't because I get lazy, I'm like, "This is terrible." Yes, I do. Some of my thinking is behavior-changing and some is not.

Michael Calore: OK. Would you ever stop using the internet for environmental reasons?

Lauren Goode: I'm going to go with probably not. Why? Should I worry about that now too? Are you thinking about that?

Michael Calore: Well, I do think about generative AI, because it has an abnormally large environmental footprint.

Lauren Goode: Yeah, it does. What we're hearing is that these chatbots are incredibly resource-intensive; well, the data centers are.

Michael Calore: Yes. And I do think about that all the time because generative AI stuff like ChatGPT and Gemini is everywhere, and I think twice about using it.

Lauren Goode: And you are already very good about keeping your footprint low.

Michael Calore: I am, yes.

Lauren Goode: You don't eat meat, you don't have a car. You're a thoughtful person in the world about this.

Michael Calore: I try to be.

Lauren Goode: Yeah. So does this mean you're going to lecture me this episode?

Michael Calore: No, I'm not going to lecture you, but it's true that most people don't think about the environmental impacts of generative AI. The tools are on their phone. They like them, they're helpful. They have productivity stuff that they can't live without, so they don't even really think about it.

Lauren Goode: We should talk about that.

Michael Calore: We absolutely should.

Lauren Goode: And not lecture.

Michael Calore: Not lecture.

Lauren Goode: Let's do it.

[Gadget Lab intro theme music plays]

Michael Calore: Hi, everyone. Welcome to Gadget Lab. I'm Michael Calore, WIRED's Director of Consumer Tech and Culture.

Lauren Goode: And I'm Lauren Goode. I'm a senior writer at WIRED.

Michael Calore: We are also joined this week by WIRED senior writer, Paresh Dave, and WIRED staff writer Reece Rogers. Welcome back to the show, both of you.

Paresh Dave: Excited to be here. Happy fall.

Michael Calore: Happy fall.

Reece Rogers: Happy fall. Howdy. Howdy.

Lauren Goode: Happy fall. It's so dark now.

Michael Calore: All the time?

Lauren Goode: Yes. The sun's setting before 7:00 now, and it's dark when I get up.

Michael Calore: Bring it. I just unpacked all my sweaters.

Lauren Goode: OK. Happy fall. OK, great. All right.

Michael Calore: I just unpacked all of my sweaters. Some of them even have zippers. Well, look, Reece and Paresh, we brought you guys on the show because you both write a lot about artificial intelligence for WIRED. And in those stories, you talk a lot about the risks and pitfalls associated with AI. But one thing that maybe doesn't get as much attention is the energy side of generative AI, about how much energy it takes to run the chatbot on your phone or write a poem or use the tool that erases tourists from your beach photos.

These things require massive amounts of computing power, and the companies that run these systems need massive server farms to process all of the requests. So not only does it take a lot of power to run all of those computers, it also takes a lot of power and water to cool those computers. You both recently co-wrote a story about a startup that is looking at a novel way of cooling these AI servers. And we're going to talk more broadly about the energy and water needs of AI in the second half of the show. But first, I want to begin with this story about the startup that wants to dunk a data center in the San Francisco Bay, and why that is problematic in its own way. Who would like to go first?

Reece Rogers: Yeah, I'll take this. As soon as I saw the startup NetworkOcean planned to test an underwater data center off the coast of San Francisco, I was curious. I had so many questions. I just needed to know more. So we invited the co-founders, Sam Mandel and Eric Kim, here to the WIRED office for chat. And according to the founders, a key benefit to this approach is that the data centers get really hot and often evaporate huge amounts of water to stay cool. But by putting a capsule with GPUs in the ocean, then the surrounding water can cool the capsule down without any of this evaporation. And this vast water evaporation from AI data centers is a major environmental issue and definitely not talked about enough. So I think trying to find ways to reduce this water usage is critical. But many of the experts Paresh and I spoke with were quite concerned about the heat emanating from this underwater data center and how that could impact nearby plants and creatures in the Bay. And by the way, they didn't even get a permit for that test.

Michael Calore: How big was the thing that they wanted to dunk in the Bay?

Reece Rogers: It was a capsule that I would imagine it closer to a submarine size, and full of GPUs. And to be clear, they were just wanting to test it in the Bay. And they wouldn't exactly tell us where they wanted to do full-time deployment, but they were looking at different areas in the United States and internationally.

Lauren Goode: So these folks did not have a permit, but they thought they could just put a little submarine in the Bay. What did they do, go to Crissy Field and waddle it out to the shore and drop it in? I know that we're going to be talking about AI and energy, but I just want to know the logistics of this. How do you plop it in the water? How do you dunk it in San Francisco Bay without a permit?

Reece Rogers: That's a fantastic question. And originally we were invited to come along and watch them do the test, but after the article came out, I haven't heard from them. So I don't know. I don't think we'll be able to see the test.

Lauren Goode: Also, I would not get in a submarine that's created by a startup. I'm just going to ... OK, please continue.

Paresh Dave: But what we can look at is Microsoft about a decade ago now at this point, and then for a few years after that, did a test off the coast of Central California in San Luis Obispo Bay, where they used a pier of a local university that the university has a research pier, and they sort of attached a data center in the water sort of next to that pier and then studied it for a little bit. And then they did similar tests in Scotland off of sort of a research pier out there.

And in both cases, Microsoft sort of found that underwater data centers are a practical idea, but for whatever reason, they haven't actually deployed any. So it really raises concerns. And then we found out that Microsoft also had not gotten a permit for its test in California and got dinged for it. But fortunately, there were not sort of serious effects that the regulators could find when they found out about it. So Microsoft didn't have to pay a fine or anything.

Michael Calore: So for the people who are administering the waterways of California or Seattle or wherever, what are the things that are giving them pause about sinking computers into the water in order to cool the computers?

Paresh Dave: The number one thing is temperature. So in the Bay, for example, things that you stick into the Bay or things that you do in the Bay are not supposed to increase the water temperature by any more than four degrees at any specific point in the Bay. And out near Sacramento in another part of the Bay, there's actually a floating data center that uses water from the Bay to help cooling. And part of the sort of agreement there was that you can use the water as long as the water that you put back into the Bay, the sort of water that helps cool those servers, doesn't increase the temperature by more than four degrees.

So that's the number one concern, because higher temperatures in the water lead to sort of breeding toxic algae. It can sort of harm fish. Then the other big issues are just displacement. You might hurt sort of an eelgrass bed that's an important place for fish to live, or you might leak some sort of toxic metal into the Bay, which obviously is problematic. So there's all these other sort of concerns, navigation of ships, but the main one is water temperature.

Lauren Goode: So draw a line for me here, and maybe for some of the folks who are listening, who are trying to understand how this helps with emissions. Is the idea just that by dunking a data center into the water, you are cooling it and therefore there's a reduction in carbon emissions? Or is the idea that by putting it underwater, there aren't emissions in the way that we typically think about them?

Paresh Dave: There could still be emissions, but there hope is that they would be using wind power, at least in the case of NetworkOcean, that they'd be using wind power to power the energy. And on top of that, you would be using the water to cool. And because you wouldn't need other forms of cooling necessarily, you'd actually be reducing the amount of energy that you're using as well, not just relying on the wind energy as well. So all of that combined reduces the potential emissions of an underwater data center.

Lauren Goode: Got it. So what you're doing is you're creating a low to no electricity option for the data center, thereby reducing the emissions.

Paresh Dave: Exactly.

Reece Rogers: And the water usage is a major factor, because when we say water consumption, it's not just like you or I using water to take a shower or anything. They're evaporating this water into the atmosphere, which can have an effect on nearby environments. So by putting it underwater, the idea is that you're not actually evaporating the water, you're just heating the nearby ocean water.

Lauren Goode: Got it. OK. So with all these permitting issues, what's the future of a company like NetworkOcean?

Paresh Dave: Yeah, there's another company that's set up some data centers out in Southeast Asia. They wouldn't tell us anything about the exact locations, but they said that they've gone through permitting processes, and they haven't seen sort of a negative effect on the environment. And they're working on more projects as we speak, and they're particularly eyeing sort of locations that are already messed up. So even if there was a potential environmental impact, it wouldn't be the worst area to have something go wrong.

And there's a company out of China that's created some underwater data centers. They didn't respond to our request for comment, so hard to know exactly what they're up to. But certainly, there are companies that are pursuing this and see an opportunity here. I would expect as we sort of see this energy crisis in the data center world, we're going to see more creative ideas come up. And look forward to keeping the scrutiny on them.

Michael Calore: All right, well, Paresh and Reece, we look forward to reading all of your future stories about underwater data centers. For now, let's take a quick break, and we'll come right back.

[Break]

Michael Calore: So NetworkOcean is just one of several startups or entities trying to minimize the impact of generative AI on the environment, as we just discussed. But I guess my big question here is, how bad exactly is generative AI for the planet? We hear a lot, there are a lot of stories about it, but if you could just paint it for us in a big picture, how quickly are we marching towards devastation by embracing all of these AI tools?

Reece Rogers: Fairly quickly, I would say, potentially.

Michael Calore: Oh, no.

Reece Rogers: A couple months ago, I wrote this piece about how we're in this hyper consumption era of the internet because AI data centers and any generative AI processes use a lot more computation than maybe a traditional Google search. But AI data centers are still a small portion of our overall usage compared to buildings, manufacturing, oil refineries. So even though it's still small, it's on an upward trend.

And that's concerning for a few reasons. Even if the numbers stay small in the aggregate, they can have a big impact on localities, on where these data centers are. Utility bills can go up. There can be a lot of strain on these power grids. And also, the water consumption; a lot of millions and millions of gallons of water are evaporated from local places. And it does eventually return, but this water goes up into the atmosphere and can stick around there for a while. But maybe one of the larger things to take a look at with these is that they're often sold as helping to fix the climate crisis. And I'm very skeptical of that. There was this great piece by Karen Hao in The Atlantic recently about how Microsoft was selling fossil fuel companies on AI tools to supercharge their extraction. And that doesn't sound like a very green innovation to me.

Michael Calore: No.

Lauren Goode: Now, part of this equation is not just how many queries we have for Google Gemini or how many chats we have with ChatGPT; it goes all the way down the chain. So it goes from our usage to the data centers that we've been talking about, but then even further down the chain to the GPUs and other physical hardware devices that are powering these giant data centers. And some tech companies factor in those suppliers and some don't. Can you guys talk a little bit about that?

Paresh Dave: Yeah, I would say part of the challenge in addition to that is there's so few suppliers. So the market is kind of concentrated. There's only a couple of companies that make servers. There's only a couple of companies that make chips. There's only a few companies that make network equipment. So there's not a lot of competition, even. These companies that build up the data centers and assemble all of this equipment together have been trying to put pressure, from what I've learned, on the manufacturers of the gear and to say, "Try to figure out how to build this with less carbon. Let's try to make our supply chains more efficient." But the problem is demand from consumers is just skyrocketing in a way that no one can keep up with. There's not enough energy, there's not enough gear, there's not enough data centers. And in this sort of mad rush, people are just trying to meet the need that's out there. They're not even having enough time to take a step back and improve that supply chain that you're referring to, Lauren.

Michael Calore: And I'm curious to learn what the big companies have in place as far as a mitigation plan for how much energy their new AI tools that they're pushing on us; and to be fair, that we are demanding. Everybody is excited to try the new version of ChatGPT or the new version of Gemini when it comes out. I say everybody; I don't mean everybody, but I mean most people that we talk to are excited about these things. We're excited about them. So if all of a sudden we're all flooding the Google servers with Gemini requests, how is Google equipping itself to handle that surge? From what I remember, the company had pretty aggressive carbon neutrality goals just a few years ago before this AI era started.

Paresh Dave: One of the things we're seeing them do, maybe not Google specifically, but a lot of these companies like Microsoft and OpenAI, is because they don't have enough energy to meet their data center and AI needs, they're actually turning to forms of non-renewable or sort of non-clean energy, whether it be nuclear energy or coal-fired plants, that are coming back online to serve these needs. So that's concerning on the one hand.

The two things that I'm curious to see are how companies nudge their users, especially their businesses that use their cloud tools, how they nudge them to maybe schedule certain processes. Like if you're trying to generate a movie, a Hollywood big movie in the future, are they going to sort of nudge you to run that process at a time when the data center is using a lot of solar power, and so that way you don't have to burn as much coal or rely on nuclear power or something? Or even consumers, if they sort of empower consumers with more information, maybe I don't make that Gemini search right now. Maybe I wait a few hours to when I know that my request is going to be powered by solar power, if it's not super urgent. I mean, I spent 45 minutes trying to generate an image last week, trying to make edits and stuff. I don't know, I mean, it was no rush. I could have been doing it at a really stupid time and putting out a lot of emissions into the planet as a result.

Michael Calore: And Paresh-

Lauren Goode: Wait, wait, hold on.

Michael Calore: What?

Lauren Goode: We have to ask Paresh what the picture was. Who spends 45 minutes on a photo?

Michael Calore: Paresh, what was the photo that you spent 45 minutes generating?

Paresh Dave: It was sort of like an event announcement. At the recent Pixel event, I think Google showed off how you can use their image generator on the Pixel to create an event invite. So I tried doing it using ChatGPT, something pretty similar. It does not know how to spell, let me tell you that. And what else did it keep doing? I would say, "Get rid of this," and then it would add something else in its place even though I just said, "Get rid of it." Don't add something in. And yeah, it still is terrible In the end. I'll show it to you someday.

Michael Calore: It's a good observation to make, because the new phones are coming out right now, right? All the new Pixel phones, all the new iPhones are coming out. They're all loaded up with these AI features and these image generators, right? Image Playground is what Apple calls theirs. And there's going to be millions of people doing this. So we are wholeheartedly embracing these things as part of our lives now, and probably not thinking as much about ... like you, Paresh, you, being very thoughtful, are going to wait until you know that your party invite is being generated by solar power. But I'm sure 98% of the people out there doing this are not going to be as mindful.

Paresh Dave: That's the big concern.

Reece Rogers: And it's kind of pushed on you. You don't have to download the ChatGPT app. You don't have to look for these tools. You just open Facebook and they're like, "Please, please use Meta AI." So almost feels inescapable.

Lauren Goode: If there's any upside to this, which there doesn't seem to be much of, maybe it's just that it's made us that much more aware of our overall energy consumption when we use the internet. I don't think I've really thought that hard about the amount of energy that these data centers require or how many natural resources they may be taking up, I don't know until ... within the past decade for sure, maybe five or six years now. And now with gen AI taking up an increasing sliver of the percentage of energy that data centers use, but that's sliver getting bigger and bigger, now it's in the conversation. We're all talking about it. It is making me take pause a little bit.

Michael Calore: I think the last time that I thought about it was there was an announcement about the new iMac about 15 years ago where it had this very remarkably, for the time, low power draw when it was in standby mode. And I was like, "Oh, that's exciting. There's something that's going to save me 60 cents a month on my electric bill." And that's literally the last time I've thought about how much energy and power computing takes up.

Lauren Goode: Right, or we thought about it in tangible form, right, to repair: can I use this gadget for longer than I originally intended to? Can I make it last so it doesn't end up in a landfill?

Paresh Dave: I mean, this is getting really in the weeds, what you just described, Mike, consumers don't feel that pain anymore, because increasingly, because of solar and how energy works in places like California, you're almost paying nothing for the actual energy. You're paying for the transmission of that energy to you, which may be sort of a flat fee. And so you sort of making all these searches and stuff, it's not hitting your pocketbook really, but it is hitting the virtual environmental pocketbook that we all share. And so it makes it harder to even think about these issues.

Michael Calore: OK, so how do we all feel now after talking about it for a stretch of time? Are we all going to stop making Genmoji?

Lauren Goode: I mean, I have a confession, which is that even as a WIRED senior tech writer, I don't use these tools all that often. I don't instinctively open ChatGPT throughout the day or use Google Gemini all that often. I did recently test a new generative AI app called SocialAI, and I was thinking about ... All you do is interact with bots on the app. It's all you do, literally. You don't interact with another human being. And I was like, "This seems sort of pointless and very extractive from an energy perspective."

I think what concerns me a little bit more is this idea that it doesn't have to be overt generative AI that is using this energy. MIT Tech Review did a story a couple of years ago looking at the energy usage of generative AI, and talked about how even something like a recommendation engine in an app that we use every day is using intensive computing and therefore using more energy. So I think it's more like my day-to-day stuff that I use that already uses AI that isn't necessarily "Make me a party invite or make me an image of a cat with six legs," or something. We are extractive beings on the Earth using the internet.

Reece Rogers: Yeah. And it kind of, to me, feels like this isn't a personal responsibility point. If I use 100 generative AI tools, I actually would feel OK with it, even if it does have a little bit of impact on the environment, because a lot of these are being pushed out by the companies and made at a larger scale. So I feel like even if I tried to not use a single AI tool ever, to save the environment, it really is on these companies and people building generative AI data centers to examine their output. And we should put pressure on them, but I think me not using meta AI ever isn't going to save the planet.

Lauren Goode: Well said.

Michael Calore: Well said.

Lauren Goode: Reece just said what I said, but a lot more succinctly and clearly.

Michael Calore: All right, well, everybody send them links to this episode so we can ... I promise I wouldn't lecture anybody, but I did not promise that I wouldn't guilt people into not using these tools responsibly.

Lauren Goode: Paresh, what are your thoughts?

Paresh Dave: I mean, I do think there's a little bit of personal responsibility involved, but I think the thing that maybe gives people hope is that you mentioned recommendation advertising algorithms, for example, there are startups that have emerged that are trying to make those systems more energy efficient. So there are people who are cognizant of the problems that they have created maybe with their last startups that were very extractive. And now we're still building technologies that are extractive, but trying to make them less extractive, and are cognizant of this issue. And like I said, the tech companies are putting pressure on their suppliers to make their data centers more efficient.

I think what someone like me who's in the media wants is more transparency about all of this. I wrote about one AI data center company, CoreWeave, a few months ago, that internally has been promising to sort of do an environmental disclosures report and be transparent about how much energy that they are using, but still waiting for that report as of a couple months ago. So we just want to understand the problem better and so that we can sort of discuss potential solutions.

Michael Calore: All right, well, I guess everybody gets to make their own decisions about how they should shrink their carbon footprint and by how much. So now you have all the information and you can do what you want with it. How's that?

Lauren Goode: Reader service.

Michael Calore: Reader service.

Lauren Goode: Listener service.

Michael Calore: Let's take another break, and we'll come back with our recommendations.

[Break]

Michael Calore: All right. Paresh, we'd like you to go first. Please tell us, what is your recommendation?

Paresh Dave: I would say it's a very simple one. Check out cookbooks from your library, whether in person or online. I hadn't actually thought about doing that until last week for the first time. I will say I was a little disappointed in the recipe, but it was a recipe that I saw online. It said that it was in a cookbook. And it was like, "Oh, rather than paying" ... not that you shouldn't pay for subscriptions online to services, but it's on a recipe website that I hadn't really heard about before. I didn't know if I was ever going to use it again. So it's like rather than paying for this sort of subscription, why don't I just get this book? And maybe there's other recipes in the book that are interesting that I plan to make later this week, so I'll let you know how those go. But yeah, I mean, cookbooks from your library.

Michael Calore: Was it glue pizza?

Paresh Dave: It was not. It was not a ChatGPT-generated book.

Michael Calore: I get cookbooks from my local library and I love it, because it lets me try the cookbook and see if it's something that I want to buy.

Paresh Dave: I will say it is kind of concerning that if you get the cookbook online, you can take screenshots from it and save those recipes. I don't know, should these apps allow screenshots or not is an interesting question.

Michael Calore: All right, well, that's a question for another day, but thank you, that's a very good recommendation. Reece, what is your recommendation?

Reece Rogers: Yes, my obsession of the week has been the soundtrack to the original Twilight movie. Say what you will about the Twilight franchise, but this soundtrack is absolute bangers, banger after banger. You have Supermassive Black Hole by Muse, Decode by Paramore, Eyes on Fire by Blue Foundation. The seasons are changing and this soundtrack will remain in rotation for me until autumn's over.

Lauren Goode: Does this also have the Christina Perry song?

Reece Rogers: I think so.

Lauren Goode: Oh, my God. I mean, come on.

Reece Rogers: It's so good. I am transported to fall. The movie's on my watch list. I know it's campy, I know it's bad, but I have a soft spot for it.

Lauren Goode: That's fair.

Michael Calore: The sparkly vampires, they always win. OK, moving on as quickly as possible from that one. Lauren, what is your recommendation?

Lauren Goode: My recommendation is a thank you to Mike for your recommendation last week. You recommended Industry on HBO Max. I started it this weekend, and I'm hooked.

Michael Calore: Nice. Say more.

Lauren Goode: I'm just going to say everyone should go back and listen to the entire episode from last week, leave us a review, and also hear Mike's description of the show. But it's great. It's about a bunch of young bankers, traders in London, at a bank called Pierpoint, and it's very cutthroat. And the first six months of them working at the bank is basically a giant trial for them. And then they're going to find out at the end of that trial period if they've got a real job there or if they've been unceremoniously dismissed. And so the first season is great, a lot of high stakes, high tensions, a lot of sexual tension, a lot of partying, a lot of drugs. It's a journey. And now I'm in the second season. And I don't know what happens next, but I'm into it.

Michael Calore: It keeps going.

Lauren Goode: Yes, you can only binge so much and still maintain our wonderful jobs and lives.

Michael Calore: So Paresh has recommended autumnal recipes, Reece has recommended autumnal bangers. You have recommended autumnal bankers.

Lauren Goode: Or bankers. That was a really bad joke, but yes.

Paresh Dave: Very on point, very on point. I'll just throw it in there, shout out to the Substack The Black Cat, by a former tech journalist, that did an interesting analysis of Harper, one of the characters on the show last week.

Lauren Goode: Oh, I'm totally going to check that out.

Michael Calore: Nice.

Lauren Goode: Yes, right. There are a lot of themes. I think if you read WIRED or listen to this podcast because you are fascinated by the Silicon Valley idea of meritocracy, there are a lot of themes in Industry about that too.

Michael Calore: For sure.

Lauren Goode: Yeah. Mike, what's your recommendation?

Michael Calore: I'm going to recommend a piece of long-form journalism from our colleagues/sister publication at Condé Nast. It appeared in The New Yorker. It's in the current issue of The New Yorker. And it is by the writer Anna Weiner, and it is called “The Art of Taking It Slow.” And it is a profile of Grant Petersen, who is the founder of Rivendell Bicycle company. He makes these very beautiful, kind of expensive but very covetable steel-framed bicycles here in the Bay Area. And he has a lot of sort of ... I don't want to say antiquated, but let's say charming views about the bicycle industry and the current state of the bicycle industry. So if you've shopped for a bicycle recently or if you know somebody who's an avid cyclist, you know that cycling right now is all about performance and going as fast as possible and putting on the Lycra and going out on a Saturday and doing 60 miles and Strava. And it's very competitive and it's very aggressive.

His philosophy is maybe don't do that. Maybe just ride a bike like you were a kid. Use it for your errands. Use it to have fun. Sit upright. Wear whatever you want, wear clothing that is functional. And he has gathered this fascination and cult-like status among people who feel this way about cycling. And he's sort of their figurehead, their guru. So this is a profile of this guy. And it tells you his story. It tells you how he developed this philosophy, how he built his little tiny bike company to something that is now actually sustainable after 30, 40 years in business. And it's also just a really great piece about what it's like to ride a bike around the hills of Northern California.

Lauren Goode: So autumnal biking?

Michael Calore: Autumnal biking, yes, although I think the story was written in the summer, because it's all triple-digit temperatures, which the same here. It's often triple-digit temperatures in inland Northern California until mid-October.

Lauren Goode: And we bring it all back to climate.

Michael Calore: Thank you, all, for listening. If you have feedback, you can find all of us on social media. Just check the show notes. Our producer is Boone Ashworth. We'll be back in two weeks with a new show. And until then, goodbye.

[Gadget Lab outro theme music plays]