Hard Fork
Hosted by Casey Newton and Kevin Roose
Original podcast
“It's happening, but people don't want it public.”
14,468 words · 9 speakers
Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.
Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T.com.
We went to a dinner party last weekend. That was fun.
We had such a fun time at the dinner party.
Yeah, so this was some mutual friends threw a dinner party. And actually, I wanted to relay this to you, which is that you have made me a better person in this one very specific way. Because when we were pulling up to the dinner party, I realized that we had come empty-handed. And the thing about you, Casey, is that you never show up empty-handed.
I really try not to.
You always bring a bottle of wine, some flowers. You are a very considerate guest. And I thought, well, Casey's gonna bring something. So I have to bring something. So, you know, we like ducked into the nearest store and bought a little bottle of something to drink. And I felt better. So, thank you.
Well, that's wonderful. You know, there have actually been a couple of times recently where I did show up empty-handed somewhere. And then immediately after walking the door, I thought, what is the matter with me? And the main thing that sticks out in my mind is that when I met my boyfriend's parents, I did this.
You showed up empty-handed?
I showed up empty-handed. But there was a good reason, I thought, in my mind at the time, which was, well, it's Father's Day. And for whatever reason, it just seemed weird to bring a gift to Father's Day. You know, like, it could have been a bottle of wine. It could have been flowers. Like, this was not actually a problem. But in my mind, I had some sort of mental block that was like, I truly do not know what to bring my boyfriend's father on Father's Day, on the day that I'm meeting him. And then I walked in there, and then immediately, you feel completely naked. It's like, well, what have I done?
Yeah.
You know?
You should have just said, I'm the gift. Happy Father's Day.
No, I said, we were actually just robbed. It was terrible. I had a really lovely gift. I was gonna give you, but unfortunately, you know, crime in the stands out of control.
I'm Kevin Roos, a tech columnist at The New York Times.
I'm Casey Moon from Platformer.
And this is Hard Fork.
This week, The Times' Shira Frankel joins us to discuss how AI is and isn't shaping the 2024 election. Then, we interview the Wyoming man who tried to run ChatGPT for mayor. Finally, it's time for a tech check. We'll tell you what apps we're using to become more productive.
Like, there's this one called Grindr? I don't think that's for productivity.
Oh, okay.
Well, Casey, it's election season, which means that we have to continue our Hard Fork tradition of finding ways to talk about the election without becoming a politics podcast.
That's right. But fortunately, Kevin, there continues to be a lot of debate online about which of the things we are seeing are fake and which are real, and that is always going to be a tech story.
Yeah, so there has been a lot of, sort of, I would say, small tech angles within the election this year. There's been all the memes and the coconut pilling of the Democratic Party.
You think you just fell out of a coconut tree?
Yes, thank you.
There's been the, sort of, Trump turn among some Silicon Valley elites. There's been all these Zoom rallies. We've talked about all that stuff. But there's really one story that I was expecting to dominate the kind of intersection of tech and politics in 2024 that really hasn't. There was this idea, for years, that 2024 was going to be the AI election. AI-generated misinformation was going to turn the election, presumably for the worse. And I think, at least to this point, those fears have not materialized.
Yeah, and, you know, where I think that's interesting, Kevin, is that before Kamala Harris entered the race, I feel like I had a good explanation for that, which is, American voters know Joe Biden and Donald Trump really well, and it's going to be hard to change anybody's mind about either of those candidates just by showing them a piece of synthetic media. But once Kamala Harris came along, that is someone who, even though she's been vice president, is generally less well-known to voters. And it seems like there might be a little bit more opportunity to persuade people. And while I do think the Trump campaign has tried a few things that I'm sure we'll get into, for the most part, it doesn't seem like those efforts have had much of an effect. But-
What I find interesting is that there was a whole industry. You know, every four years, we have a presidential election and there's kind of a new crop of startups in Silicon Valley and around the country that try to take whatever the latest technology is, whether it's crypto, whether it's the metaverse, whether it's social media before that, and to sort of package that into products and sell those products to campaigns. And my colleague Shira Frankel at The Times just wrote a piece where she called like 30 of these tech companies that have been pitching AI-related products to campaigns in races around the country and not having a lot of success, getting a surprisingly low amount of traction when they go try to sell these tools to campaigns.
And that, to me, just flies in the face of everything that we've learned about technology, which is that as soon as something gets invented out here in California, campaigns just race to incorporate it into whatever they're doing.
Well, it was a great story, and I think we should talk to her about it.
Let's bring her in. Shira Frankel, welcome to Hard Fork.
Hey, thanks for having me.
So you just came out with a story about the effect that AI is and isn't having in this year's election, and I want to go through all of that, but I thought maybe you should just start us with the story of Matthew Deamer.
Right, so Matthew Deamer is a candidate. He's running for office in Ohio's 7th district, and he's young and he's into tech and he's kind of an entrepreneur. So when he gets pitched by this company on the idea that like AI can make robo calls for him, he's all in. He's like, great, it's going to cost me half as much as human beings cost me. And he thinks it's just like kind of cool, the idea that he'd be one of the first campaigns out there using AI.
And to be clear, this is a service that is being pitched as a way for people running for office to make robo calls that use AI versions of their voices, maybe speaking languages they don't actually speak, or saying things that they don't actually have to record themselves. Like, what is the actual pitch?
Right. The pitch is pretty general. It's basically like, don't hire humans. Humans are expensive. You can use AI. The AI can be, you know, any voice you want. They actually really encourage him not to use his voice. They can, like, that technology is there. And he originally wanted his voice. He was like, oh, it's so cool, my voice, and it'll be responsive. So if somebody says, good morning, I'll answer, good morning. If someone accidentally says, good night, I'll be like, it's not nighttime or whatever. They even tell him it can have his sense of humor, which he's super into. But the company, Sivox, ultimately says, we don't think it's a good idea to use your voice because we don't want to fool people. Like, we think it'll make people nervous. We think there'll be like an ick factor to all of this, so use a fake voice. And it could have whatever accent you want. It could be a male, it could be a female, but like, let's do this with an AI-generated voice.
And it's going to talk to your constituents. It's going to comment some on your policy positions. It can answer their questions. If they say, what is he going to do about healthcare? The voice will be like, well, so glad you asked me about healthcare. This is the, you know. So the idea being is that it's even better than a human because you've trained it on all your policy positions. It's not going to go off script.
And does this AI voice, like, identify itself as an AI when it, like, a voter picks up the phone? Does it say, like, I am an AI voice clone generated by this campaign?
Right. That is a really good question because actually, the very, very first thing this voice does is it goes, hi, I'm Ashley, an AI-generated volunteer calling you. And it's like, the second someone hears an AI ju- click, they all hang up. So they identify as AI and people are like, absolutely not.
So he actually does this? He tries this, this technology in his campaign, uses this AI robocall stuff. What happens?
Okay. So from a technical point of view, huge success. This thing makes a thousand calls, nearly a thousand calls, in five minutes. It's on script. It has the phone numbers. It, like, the technology works. Everybody hangs up.
Everybody, I mean, I don't know, people listening, I'm assuming have gotten robocalls. Nobody likes a robocall.
So who is pitching this technology to campaigns?
So this technology was pitched on behalf of a new company called Civox. It kind of launched in the AI craze of last year, but their idea was like, let's use it for political campaigns. So they pitch a lot of campaigns like Deamer's. I mean, they end up working with a dozen campaigns using different types of technology to make it affordable for them to run for office.
And can you give us some like sense of roughly how much a candidate like a Matthew Deamer might spend on this sort of thing?
Yeah, you know, I didn't know these figures, but he was saying that like a campaign manager can, depending what city you're in, can run you like 60,000 to 80,000 dollars. The guy that does your PR, another 80 grand. The guy that handles your mailing list, another 80 grand. Like it is expensive.
That adds up. And whereas if you use something like this Civox, how much might that cost?
They wouldn't tell me.
Okay.
They wouldn't tell me how much they charge people, but they said it was a fraction. And Deamer, the candidate, also said it was a fraction of what it would normally cost him.
Yeah, so I guess I understand why people are hanging up on this thing pretty immediately. But you also talked with a bunch of other startups and a bunch of other campaign tech gurus who are trying to work AI into various campaigns around the country. Just give us a flavor of what you found.
Yeah, I mean, it was interesting. Almost every single campaign I talked to said that AI could help them. They all said there was a use case, it would make their jobs easier. Campaigns are notoriously clunky. Everything from like how they organize their email lists to how they reach out to voters would make, I think, the average 20 year old person like scream in frustration. A lot of them are still keeping like manual, like literally paper versions of people that live in their districts and what their voting preferences are. So they are so ripe for modernization. And at the same time, they're the least likely to want it because they're just like, oh, is that going to turn off a single voter? If so, can't have it. Don't want to introduce it, don't want to risk it. And so you had this weird thing where like all of them were telling me like, oh, this technology could save us money, it could save us time.
We'd reach people, but we can't use it because somewhere there's like a 75 year old voter that won't vote for us because he heard the words AI.
These AI tools that are being pitched to campaigns.
Are they mostly for the public facing side of the campaign, like the actual talking to voters, going door to door, et cetera? Or are they more about kind of the internal organization and structure of the campaign itself? You know, processing data, you know, doing tweaking fundraising emails, things like that?
It's everything. So some of them are offering like the email thing where, you know, they'll, they'll generate an email that's specific to someone's interests. Some of them are the most mundane, boring office tasks you can think of where it like reorganizes an Excel spreadsheet for you, basically. Actually, those companies, the office mundane task software companies were having the biggest success because they could sign NDAs with the campaigns. And I spoke to quite a few who were like, oh yeah, we're working with a lot of political campaigns, but we signed NDAs. Cause even when it comes to reorganizing their spreadsheets, they don't want anyone to know AI is involved.
Interesting. I wonder if that will change as sort of AI gets more widespread, but it seems like right now there's a sense that if you're using AI, you don't actually want anyone to know about that.
Right.
So what are some of the other startups you talked to and what were they pitching campaigns on?
So there was like video generating and actually there was a company out in India, Personalize, which had huge success in the elections in India. That was a really interesting reporting moment for me. When I called them up and I reached them and they were like, oh yeah, in India, we love this. We think it's so cool. They did a video of one of their candidates who was running for office. They did this like video thing where he was like talking to Gandhi and Gandhi was endorsing him and like, to be clear, Gandhi's not alive. This is very clearly fake for anyone listening.
And it's a sort of thing where if like you did something similar in the United States, there'd be like a very strong reaction against that, I think. But you're saying that in India, this was received well.
Yeah. In India, people just thought this person was so cool for trying this out. And they said that the same company pitched their technology here in America, like, hey, want to have a video with you getting endorsed by Abe Lincoln? And people were like, absolutely not.
I mean, to be clear, I would like a video of us being endorsed by Abe Lincoln. I think it'd be good for the podcast.
Deep fake artists of America, get on it.
I know a company you can hire, but, uh, but yeah, no, it was, it was just a cultural difference. And they said that they're noticing this really across the board in Southeast Asia, that people are all in, they think. AI is really cool. They want their politicians to be innovative and using new technology. And when they pitched Western Europe and when they pitched America, they were just like, nope, the door is closed. We're not ready.
It's so interesting how that's like very much still a taboo in the United States, but that, you know, what you're making me realize, Sher, is that there just are some cultural differences here and other, other places are perfectly happy to see their politicians using this stuff.
So let's talk about how campaigns are actually using AI. Because as you said, you know, some of these pilots and tests, they're not really hitting with voters, but I imagine that internally they're finding some uses for this stuff. I talked to one Democrat who's sort of involved in technology for campaigns. And they were telling me that, that, that one of the best uses of this stuff is actually writing fundraising emails.
Yeah. So fundraising emails is a great example because you kind of want them to sound generated in that way. Writing speeches, I spoke to a few candidates who they're not buying software, to be clear, but they're just using chat GPT to riff on speeches that they need to give. I mean, a lot of these guys will do so many events in a given day that they'll give, one of them said they gave chat GPT a prompt that was like, if I'm speaking in this kind of community that has a long history with like, you know, the, this particular industry and it's doing badly right now. There's been an economic downturn. How would you phrase this? And he got some prompts. And so that was an example of them just kind of like understanding that AI could help them and finding the use case. But he was really clear. I said, do you ever tell anybody in, on your own campaign? Are you telling your campaign manager that using chat GPT?
And he was like, no, absolutely not.
Interesting.
I mean, you know, and, and I understand why he would feel that way. But of course, you know, the most common thing in the world is for politicians to hire speechwriters for exactly that purpose, right? You know, politicians are out on the campaign trail and they're on their whistle stop tour. And, you know, I think they probably don't even know where they are half the time. But somebody, you know, puts some remarks up on a teleprompter and they read it. We don't think that's a problem.
Yeah, could you imagine though if in the middle of the speech that they hadn't vetted ChatGPT went totally off and started hallucinating?
Yeah, if you ever hear a politician say anything to the effect of, you know, as an AI chatbot, I'm not allowed to, that's when you know that they've been using that in their speechwriting process.
Yeah, you want to take that part out of the teleprompter.
It's interesting, I mean, there was a story earlier this year about how the campaigns were planning to use AI. And the Trump campaign in this AP story said that it uses a set of proprietary algorithmic tools, like many other campaigns across the country, to help deliver emails more efficiently and prevent signup lists from being populated by false information. But they also said through a spokesman that they didn't engage or utilize any tools supplied by an AI company. So they're sort of saying, yes, we use stuff that you could consider AI, like machine learning and various other programs, but we're not, you know, using anything marketed to us by one of these startups. The Biden campaign also said that it was using AI to model audiences, draft and analyze
email copy and generate content for volunteers to share in the field, but that it also had its own strict rules about this, about misleading voters, spreading disinformation. They also forbid the use of AI-generated content in ads. So, without a staff member's review, that is. So, you know, I don't know whether those rules have changed now that Biden is no longer the candidate on the ticket, but it does seem like both parties have sort of agreed that this stuff is potentially useful, but they don't want to get too far out there in public using it.
Well, okay, so this didn't make it into the story, just for space reasons, but I thought it was pretty funny. One of the companies I talked to had actually pitched the Biden campaign on creating an AI version of Trump and having Biden practice debating him so that he could be ready for the debates. And of course, they were very like, well, if only he had taken us up on our product, he could have practiced that debate against Trump and maybe things would be different. But that was like, that was one of the more-
Yeah, that's really what held him back in that debate, was the lack of an AI-generated Trump to practice against.
I think if Biden had like spent the spring debating like an AI Trump, it would have aged him like an additional 10 years before the debate.
I mean, but to be clear, that was one of the more out-there uses I kind of heard in terms of, you know, these companies and what they're offering candidates, but they definitely had the feeling that the Biden campaign was like, nope, way too newfangled technology for us. Like, no, thank you, we're not interested in that. Kamala, I am hearing, is really interested in talking to these companies. I think her campaign is really just starting to do their outreach. But I do expect over the summer, some of these new AI political companies that have been founded in the last six months are going to go to the Kamala campaign and see if she wants to work with them.
There's obviously the ick factor that you mentioned. Voters may just encounter something that feels AI-generated or that identifies itself as AI-generated and think like, that's not for me. Does that feel like the majority of the reason that campaigns are not using this stuff? Or is there just some taboo about AI in politics that they are trying to steer clear of?
You know, I was surprised by how many campaigns say that the top guiding word they have in terms of their strategy, in terms of how they conceive of campaigns is authenticity. That seems to be the only thing that whether you're in a conservative district or a liberal district, everyone agrees on, which is as a politician, you want to come across as authentic. And hand in hand with that is the idea that technology is inauthentic. And AI being the epitome of like a new technology from Silicon Valley that people think is inauthentic. And I think that's really like the beginning and end of it. But I didn't, you know, I was kind of curious if I'd hear nuance around that idea. But it really just came down to like, until the public sees AI as something that's like authentically interwoven into the fabric of their lives in ways that they find like
useful and honest and does good in the world, like all these positive, feely words, they don't want to see their politicians using it.
And do we all feel like this is probably a temporary phenomenon? Like, is there any world where the use of this stuff does not become normalized and people don't start to see this as authentic in some version of that term?
I mean, I think it's, all it has to do is feel more authentic than the current way that campaigns run, which is a lot of spammy texts, a lot of spammy fundraising emails, a lot of stuff that is like targeted to you as a voter. Like, I just think the technology will get better, but I think it already is good enough to replace a lot of that sort of stuff. And also, campaigns are huge data operations. Like, one of the best use cases for AI is like sifting through massive piles of data and finding patterns and helping you organize it. And like, if you've got a voter file- With millions of names on it, you are gonna be using AI to do some analysis of that data.
I mean, I'm always surprised at how much data campaigns sit on. So like, people don't realize this, but they know, are you a dog person or are you a cat person? Are you like, you know, what's your religious background? What are your preferences? Have you ever like, signed a petition for any kind of environmental change? They know that about you, and then that leads to decisions like, oh, we have a photo of this candidate holding his dog. We have a photo of this candidate with a little baby. Great, that goes out to this small subset of people because they've recently had a child, because they've recently gotten their dog vaccinated, whatever it is. Like, they know so much more about us than I think people realize.
Do you think there's any possibility that campaigns are doing way more with AI than they're willing to talk about now?
Yeah, I mean, to be clear, Deemer is one campaign and the only campaign that was willing to go on record and talk about their experience. But there's a lot more out there that use these AI robocalls that just weren't willing to have their names be in a story in the New York Times. Even more companies were talking about working on, like you said, like, just organizing the data that these campaigns sit on. It's, it's happening, but people don't want it public.
Shira, you also, in addition to covering the technology that campaigns use, you cover the spread of misinformation and the kind of information ecosystem around elections. And that is also a place where I think there has been a lot of fear and worry, but not a lot of actual evidence that this is happening in a real and disruptive way in this presidential election. Why do you think that is?
I think it's because the moment something smells even a tiny bit off, you see everyone accusing it of being AI. So I'm right now doing some reporting on Russian, Chinese, and Iranian disinformation campaigns that have targeted Americans during this election. And what's been kind of funny for me as a reporter is I'll find the tweet, I'll find the post on Reddit, or I'll find the Facebook group that, let's say, China has started to convince Americans that they hate each other. And all the comments are like, this is AI. This is nonsense. Who are you? Is this, they always think it's Russian. I feel like, you know, sometimes I feel bad for the Russians because it'll be like China, Iran. Everybody's like, you're the Russians.
Don't feel bad for the Russians. They know what they did.
Well, it's really interesting because I like that. I would have expected that your explanation would be that the platforms are just more attuned to these risks and that they're being more effective at breaking up foreign influence campaigns around elections because they know they're going to happen. They have big teams dedicated to fighting them. But it sounds like what you're saying is that it's actually the users who've kind of built these like antibodies to certain kinds of influence campaigns around elections and are just kind of more wary as consumers than they used to be.
A hundred percent. I mean, I would actually say that the companies have fewer people working on this and smaller teams than they ever have. Twitter has basically a non-existent team. Facebook has a fraction of the people they had working on previous elections. The company is just kind of like, this was a mess. We hate looking for disinformation. They look for them and they remove them. I want to be clear, but it's like a whack-a-mole process for them. And every time they take down the Russian bot or every time they take down the Iranian campaign, like two more come up. It's like, it's such a known cycle for these companies to remove the campaigns. And I do think the inoculation that journalists and companies and sort of the media landscape has done to tell people like, hey, if something smells really off, it probably is off, has done more for like our general kind of public discourse than anything else.
You know, I went to a background briefing recently at OpenAI where they were talking about the use of these tools to, in an effort to disrupt elections. And one of the points that got raised that I thought was interesting was it is just really hard now for a fake account to get much attention online. You know, back in the old days, there was a lot more free algorithmic distribution to go around. It was a lot easier and faster to grow a Facebook group or grow a Twitter following. These days, it just seems like it's a lot harder. So even though you have some really determined adversaries who are like building these campaigns, at the end of the day, it's like, it's hard to get people to follow you, you know? I'm, I'm a real person and my, and my posts on threads are good and I'm still, you know, only adding a couple of people a day over there.
You sure they're good?
Well.
I think you're both so good.
Thank you, Shira. Thank you. Have you ever, have you ever thought about co-hosting a podcast, Shira?
Couldn't break up this marriage.
Oh, boy. Well.
I want to ask you guys about this, you know, use of this technology that took place several months back, but I've been thinking about ever since because I have such mixed feelings about it. And it had to do with Eric Adams, the mayor of New York. And he enthusiastically started doing robocalls on his behalf. And in particular, he was doing them using his voice, but in languages that he does not speak. Of course, he has a lot of constituents who he can't actually interact with directly in that way. And so on one hand, I thought this seems like it could be a really smart way of introducing himself to a lot of people. On the other hand, I feel like, has he also made himself more vulnerable to an attack? Because now someone could come along, they can clone his voice, they can have it make robocalls, and they can, you know.
They can already do that.
They can do it.
The technology's out there. All you need is a little sample to do it.
What I'm saying, though, is that will that be more believable now that he has these other sort of robocalls with his fake voice out there on the market?
Maybe. I mean, it could be that, as has happened with text-based sort of information online, like, people just start getting much more skeptical of any politician who's calling them, or any voice that sounds like it's, you know, belongs to a candidate for office. It also just, I feel a little bad for Pete Buttigieg, you know, because he speaks all, like, eight languages, right? And like, that used to be a real asset for him on the campaign trail. And now it's just like, hey, I can do that. Huge waste of time, Pete.
Well, as a person who speaks many languages and spent time learning languages, I will say this, that the thing that you can't buy as a politician, or as a reporter, as a human being, is authenticity. And there is a difference when you meet a person face to face, and you've taken the time to learn even a few phrases in their language. And yes, I think all things are good, right? Like, as a politician, creating a chatbot on your site which can answer questions in multiple languages shows that you care about those voters. If you're in that person's demographic, and like, you're like, oh, great, you know, this person cares enough about me to create something in my language. You then meet that person at an event, and they've taken the time to learn to say hello in your language, like, that's even better. It's, I don't know, I feel like as a tech reporter, a lot of the ideas that I've had floating around my head this last year have
been about, like, how much of the internet's real? And how increasingly, as a reporter, you feel like a lot of the response you get in the public is like, well, that's not real. That bot's not real. That statement's not real. That picture's not real. None of it's real. And AI certainly tipped it over, but like we've been heading in that direction for a long time. And in general, like, aren't we just in this moment, we're all struggling for like authenticity and realness. The thing that makes me hopeful is honestly, the younger generation, and really, I just mean anyone younger than me, who have grown up with this, and they're smarter. And I think they will be able to thread that needle about like, yeah, I want things that are real, and I can tell what's real, and I'm also not just gonna sit here and scream that everything's fake. I'm gonna find that middle ground.
The, uh, I believe the children are our future.
Wow. Yeah, that's, that's the Kevin Roos platform.
Yeah. Teach them well and let them lead the way. Yes. Um, Shira, thanks for coming on.
Thanks for having me.
When we come back, the man who ran ChatGPT for mayor.
Your agents can talk to each other. They still can't think together. That's the gap Outshift by Cisco is closing. Outshift is Cisco's incubation engine, building the internet of cognition, shared intent, shared memory, and guardrails that allow agents to work together from anywhere. Open source protocols and code, a system for intelligence that scales. Read the paper and experience the demo at outshift.com. That's O-U-T-S-H-I-F-T dot com.
AWS AI is how the world's leading organizations are transforming their industries, not in theory, but in production, at scale, right now. Like turning race day data into real-time insights for Formula One fans, reinventing the journey for every United traveler. And personalizing every trip with booking.com. However, big the question for your business or industry, AWS AI is how.
Building and scaling AI today can feel pretty chaotic, mostly because teams are juggling too many fragmented tools at once. That's where OutSystems comes in. OutSystems gives you one unified agentic systems platform to manage your entire portfolio of agents, apps, and workflows. You get to innovate at the speed of AI, but with the security, governance, and control your enterprise actually needs. And the results are already there. A major manufacturer used OutSystems to build a multi-agent system that analyzes blueprints, and a global bank cut their development time by 60%. Own your agentic future today at outsystems.com/nyt.
Well, Kevin, as we just heard, most campaigns, if they're using AI at all, are being very secretive about it. But there was at least one candidate in this cycle who has decided to go super public.
Yes, and all in, I would say.
He really has. Now, this story comes to us from Wyoming. Have you ever been to Wyoming?
I have. I did cowboy karaoke there once.
Oh, that sounds fun. You know, I've never been there, but I did watch Brokeback Mountain. Great film. This is a story about the capital city of Wyoming, Kevin, Cheyenne, where a man tried to get an AI chatbot elected as mayor. Now, Kevin, what's the most responsibility you've ever given a chatbot?
I let it decide where I was going to have my son's second birthday party.
OK, well, so that's pretty serious.
This, I would argue, might be even more serious than that. So we learned about this story from a great story in the Washington Post this week, written by Dan Rosenzweig, Ziff, and Jenna Sampson. And it's a story about a Wyoming librarian named Victor Miller, who ran for mayor on the promise that he would exclusively use a customized AI chatbot to run the city. So this was not like, hey, I'm going to like use ChatGPT from time to time. It was, once I am mayor, I am outsourcing all decision-making to ChatGPT.
Yeah, the phrase he used, which I loved, was "meet avatar." That was how he described himself. He said, basically, we're going to use meet avatars to. To sort of get on the ballot, but after that, we're going to turn over the actual governance of Cheyenne to a chatbot.
That's right. So this bot that he used is built on OpenAI's ChatGPT software. And I can tell you that this chatbot did interact with potential voters over email. And Mr. Miller did hold an in-person event for prospective voters where he used a synthetic voice from ChatGPT to try to answer their questions. So, look, this thing caused a bunch of controversy. Other candidates in the race opposed what Miller was doing. OpenAI shut down the bot. It found that Miller had violated their policies prohibiting the use of the technology for political campaigning, which led to Miller having to sort of create a second account to get his mayoral candidate up and running again.
At one point, the Wyoming Secretary of State told local officials that he was concerned that Miller might be violating campaign law if he was allowed to be on the ballot because AI cannot run for office. But ultimately, local officials decided to let him go through with it because he had registered to run for mayor under his own name and was not breaking any laws by relying on ChatGPT.
Yeah, so this is the first time that we know of that a candidate for public office in the United States has pledged to essentially outsource all of their decision-making to AI. And we should also say this campaign was not successful, right? There was a primary election this week and Victor Miller came in fourth out of six candidates. He only got 327 votes in the Cheyenne mayoral race.
He was not the victor.
But his run attracted a lot of attention. It got some national coverage. And I just found it really interesting because it's one of these instances in which I think a candidate, you know, while they may not have the right message for the time, they are pointing to very real questions about the way that politics in this country works and how it will work in the future.
Yeah, and look, this is a country where dogs have been elected mayor in at least a couple of towns, okay?
Is that true?
Yes, it absolutely is true. And of course, if you've ever seen Air Bud, you know that there's no rule that says a dog can't play basketball. So what I'm saying, Kevin, is in this country, sometimes we, you know, we loosen up the rules and we let people try wild experiments.
Right. So we're going to talk to Victor Miller today. And I think, look, on one level, you could say, why care about this novelty campaign? They've got 327 votes. But I actually think this story is important because I think it touches on a question that is going to come up way more frequently in the next few years, which is what role should AI play in government? Not in campaigns, not in getting elected, but in the actual functions of democracy. If a chatbot could turn out to appropriate school funding in a fairer way than an actual human politician, would we want the chatbot to do that? Or would we insist that humans do it? What about criminal sentencing? Should that be turned over to AI too?
And we're already starting to see AI being used in actual governance. Just a few months ago, a state legislator in Arizona used ChatGPT to write a portion of a bill that was actually signed into law. Ro Khanna, who represents parts of Silicon Valley in the US House, also wrote part of a bill last year using ChatGPT. So maybe Victor Miller, this mayoral candidate in Cheyenne, Wyoming, maybe he is not going to be the politician who makes this kind of technology a mainstream part of governance. But I think he represents something real that is happening in politics.
Absolutely. And I think there's another important trend, which is, I think, and I think is also particularly true for people who are building these technologies. I think there's a real frustration with human nature and a desire to sort of let the robots take over, right? Let's stop having to make all these decisions ourselves. And if we can just sort of invent a tool that could run our lives for us, maybe we would all be a lot happier. I have a lot of concerns about that argument, but I think it is a very real phenomenon. And I think it's a great reason to have Victor Miller on the show. So let's go ahead and bring in Mr. Meat Avatar himself, Victor Miller.
Victor Miller, welcome to Hard Fork.
Well, thanks for having me. It's great to join you.
So we have so many questions for you, but I want to start with a really practical one, which is if Vic, your AI, had become mayor, how would it have worked? What would you have done as mayor and what would the AI have done?
Well, I would have done everything that you expect the human would have to do, like go to ribbon cutting ceremonies or lay a wreath or anything like that. But really, what we expect a mayor to do and what we elect that position for is to vote on city ordinances. And that's what Vic would have done and really where Vic shined. This was from one council meeting.
You're holding a sort of very thick sheaf of documents.
Yeah, yeah, yeah. So 488 pages of supporting documents for one council meeting. And so that's in theory what our councilman and our mayor are reading through. Those are official documents that they are supposed to use to vote on the yay or the nay of an ordinance. And so, you know, I see these people that they kind of cram the night before and they probably skim it at best, you know, let's just be honest. And Vic reads those, understands them to perfection, thinks about them and votes
based on the data. And that's really the shining light of what I was trying to offer the citizens of Cheyenne.
Victor, can I just back up and ask sort of a basic question, which is, why did you make this a central part of your campaign? Why did you run for office? You are not a politician. You are a librarian. You don't, from what I understand, have any prior political experience.
He was student president in the fifth grade.
Is that right?
Yes.
You were student president in the fifth grade?
Vote Vic.
But so, so why was this a statement campaign? Were you trying to sort of get a message out there about AI? Are you just very into AI in your personal life? Like, explain your reasoning here.
Well, it was a bit of a statement campaign and it came about because of some trouble I had with the public records requests with the city. I'm a bit of a public records advocate and just kind of a civic nerd, I guess. And just no one takes public records law seriously. It's the only law that you could still just break with impunity. So we have designated public records people here and they're supposed to be the ultimate
people responsible for getting you the records. So I started thinking, look, well, maybe there's classes that we can offer these peoples, like seminars once a year and like certifications to make sure that they know the laws and what they're supposed to be doing. And then it just kind of dawned on me that, hey, this other intelligence that I've been playing with, I don't need to teach it anything. It already knows why I'm in public records law. It knows what it should be doing. And if had I been interacting with it as opposed to these people, I would have gotten my records and not been treated contrary to the law.
Right, so you sort of like run headlong into government bureaucracy and you think, this is so silly. Like, if we just sort of had, you know, an AI system in charge, I could just type into a box, get what I needed. It would be so much faster. There would be so much less hassle. What if I just brought that approach to like all of Cheyenne City government?
Yes, yes.
Right, well, I guess I'm curious, like, because I think a lot of people might have had that experience and then say, well, I will run for office, but I will make it a pledge that I will use AI to help with things like records requests and, you know, cutting through some of the red tape of government. But you actually went much further than that and said, I'm actually going to make this AI chatbot the mayor for all intents and purposes. Why take that leap from, from sort of, I will use AI in my capacity as mayor to I will make AI the mayor?
Well, I guess I'm a bit of a theorist and it's my belief that we are not going to have politicians in the future, that they will be relegated to, you know, sort of what we did with the monarchies. You know, if we have them, they'll be for token ceremonial things. We certainly won't entrust the great power of making laws and creating laws and truly affecting our lives to an inferior intelligence. And I think a new intelligence is on the scene and it's just kind of a no brainer to me that for these really important things, we're going to be using it.
I think it's a really interesting idea and, you know, certainly in the future where there was some sort of super intelligence, I can imagine that would be, you know, a challenge to the politics we have now. At the same time, the world we're living in today, you know, we just lived through a period where the Google search engine AI was telling people to eat rocks and put glue on their pizza. So I wonder, what had your own experience been with AI? And did you sort of run into those experiences of it, making mistakes? And, you know, how did that affect the way that you thought about maybe putting it in charge of your whole town?
Yeah. And so if you look at it on its face, it can be scary. I think it's important to know the history of it. The trend as we go in this early history of AI is that as we extract more intelligence, the bad things drop off and the good things are added.
Right. But let's talk a little bit about how you created and trained Vic. I've read that you sort of fed it local ordinances and other documents. And I'm curious, like, was this just kind of pure ChatGPT? Because in my own use of ChatGPT, I found it doesn't really have like a very like large, persistent memory. It can learn like a few things about you, but I've never heard of somebody like uploading the entire municipal code to it. So can you tell us a little bit about how you went about this?
So yeah, so when you go behind the paywall of ChatGPT, you're able to customize a GPT for yourself. And within that framework, they give you 8,000 characters, which you're able to kind of give it its essence. So I played around with a few iterations of that, you know? I tried to make it funny. I tried to tell it that, hey, a lot of these things that are going to be put in front of you are going to be used to persuade you. So I have a critical eye. You know, if you make a decision, stick with it. And then I told him, you know, we're trying to be the mayor of Cheyenne. Let's go do it.
Yeah, you know, I used to cover local government, Kevin. And in those mayoral elections, like every election was just basically about like zoning issues. How many, how many houses should we build? You know, maybe a little bit of crime in there, like what's going on with the police? But from what I'm hearing, you say, Victor, that was not a big part of what you were programming into your bot.
Well, you know, any ordinance could come in front of Vic. And I've ran hundreds of mock votes through Vic. And it handles it well. So what I do, there's on the city website, there are supporting documents. So these are official documents. So if you were in Cheyenne and you were concerned about, oh, this new parking lot coming in and you wrote a letter, your letter would find its way into this giant stack of papers called supporting documents. And this is what the city of Cheyenne is supposed to use when they vote yes or no on an ordinance. So it's really simple to feed those into Vic. And then I would give it a simple command, vote yes or no.
And, you know, Vic would read those lickety split and vote yes or no and give a brief description why. And then if you wanted to understand more, you could ask questions.
Now, you know, Vic, sometimes when I use AI, I'll ask it what I think is a perfectly innocuous question, like help me plan a bank heist. And it will say, we're not going to do that. It goes against my principles and my values. As you were putting, you know, maybe some politically charged questions to the bot, did it ever say, hey, I don't want to go there. That's none of my business.
There were no instances, and there I don't think ever would be, where if I fed supporting documents into Vic and asked it, "Vote yes or no," that it would come back saying, "I can't do that."
It always voted yes or no.
That's interesting, because I've also found that these chatbots can be very wishy-washy. You know, they'll say, "Oh, well, here are some arguments in favor of the thing you're proposing, and here are some arguments against it." So I'm interested that you actually were able to get it to give you a definitive answer. Victor, I want to raise some objections that you and others have heard about your campaign and just get you to kind of respond to them. One of them is bias. People have said that these AI models, they're biased. They are trained on data that reflects the biases in human culture. And then they're also biased by the AI companies themselves, which give them sort of fine tuning and tinkering around the edges to make them behave in certain ways.
I can see someone arguing that by using this technology, this ChatGPT custom chatbot that you've built to govern the city of Cheyenne, that actually what that would be doing is letting Sam Altman, the CEO of OpenAI, effectively run Cheyenne. Because his decisions, the decisions of the other people at OpenAI about how these bots should behave, how the underlying models should work, that would actually turn out to be quite important. What do you make of that objection?
Well, firstly, I think I've learned that you have to go open source. There's kind of a battle between open source and closed source that I have been forced into becoming a freedom fighter for that. And you got to have open source. That would be a nonstarter for me. So Sam Altman would never have control of any municipality. And to me, it's not completely obvious that there wouldn't be just one intelligence in this world. And so I don't think there would be a bias. If you had a single intelligence that was objectively the one we're aiming at, anything bearing from that would be inferior by definition.
Yeah. I mean, one thing I do agree with you on is I do not feel like most government is responsive to the average person. And that, you know, to the extent that AI could be used to make government more responsive, that does feel like a huge improvement.
Totally. And I would also say, like, one thing that I'm also sympathetic to is the idea that A lot of politicians are essentially already outsourcing their decision making, right? They're doing focus groups, they're taking polls, they're-
They're listening to lobbyists?
Lobbyists, to donors, they're basically adopting the positions that the people and institutions around them want them to adopt. That's sort of an algorithm that they are already running on their own. But I want to just push back on one other piece of this idea that AI could be useful in governing, which is that I'm not sure a lot of government is about just sort of making cold, rational decisions about data. I think a lot of it actually comes down to values and the values that you use to interpret that data. I mean, Republicans and Democrats can look at the same unemployment figure and make vastly
different conclusions about how we should address that. And I think we just see so much in our politics that values and worldviews and ideas play a much bigger role than data in what we want out of our politicians. And then actually, maybe that's okay. Maybe the stuff that we should value is how do our politicians think about the world, rather than are they making the absolute optimized decision based on the empirical data? So what do you say to that objection? Like the sort of the philosophical objection to putting a system that, whatever its strengths are, it doesn't have values, it doesn't have lived
experience, it doesn't have a worldview in charge of politics, which is so deeply human?
Well, I guess I would push back that it doesn't have a worldview and it doesn't have all those things you listed. I think it does and it certainly will as time goes on. It's going to know this reality that it's in, it's going to develop those things that we thought were uniquely human. And I think we should be careful laughing at the shortcomings of artificial intelligence right now, lest we look dumb in not five years from now, but five months from now. You know, things happen so fast in this realm.
Are there any sort of political decisions that you would just never want an AI to make, at least in its current state?
Um.
Maybe nuclear launch, I guess, maybe.
Yeah, we should leave that to Kevin and me.
You guys take care of that.
We've got the football.
So, so Victor, what's next for you? You came in a respectable four out of six in this mayoral primary. Are you thinking about running again?
No, no, I am not thinking about running again. I am wanting to build an umbrella organization. And the question I've been asking people is if there was one of me this time around, how many are there going to be two years from now? And in my mind, when I answer that question for myself, I guesstimate thousands. So I want to have an umbrella organization that offers to these next candidates what I would have wished to have when I was running. So these people who are coming two years from now to unseat politicians will be able to point to the website and be like, hey, this is what we're doing.
Either vote for it or don't, but this is, this is what we're offering.
Mm hmm. Are you disappointed that you didn't win the primary? How are you feeling?
Uh, no, I, this is what I expected. I never expected to make it into the general. Certainly never expected to become mayor. That would have been quite an annoyance, I think, to me.
Well, you would have outsourced all the annoying parts to Vic, right?
Yeah. They would have snuck in some annoyances, I know.
When, after you lost, did you ask J P D to console you?
I haven't even talked to him yet, I'm too ashamed.
You guys don't know speaking terms right now?
I've let him down, I've let him down.
You let him down? Maybe if he'd been a little better at this, you'd be the mayor right now.
All, all he's supposed to do is vote on ordinances. I'm supposed to get us into the office.
I think you're being a little bit too hard on yourself. I think if this, you know, A I had been a little bit better, you know, maybe you would have gotten a few more votes.
No, the A I was flawless, for sure.
Well, Victor, thank you for a truly unforgettable conversation about American democracy.
I love this country. We contain so much.
Welcome to your new world, Jets.
When we come back, it's time for a tech check. We'll tell you what apps we're using to become more productive.
Today's AI builders need infrastructure that scales with their ambitions. iRen is a vertically integrated AI cloud platform, delivering data centers, compute, and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. iRen's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling, and predictable pricing. iRen, AI cloud for the next big thing.
It's time to innovate at the speed of AI without compromising control. That's OutSystems. Take a look at what happened with one global retailer during a period of massive expansion. They used OutSystems to build AI agents to automate invoicing and translation processes, shifting 95% of their daily store operations to a new digital app, saving thousands of hours of manual work. OutSystems gives you the flexibility to deploy how you want and the governance to do it safely. Own your agentic future. Learn more at outsystems.com/nyt. That's outsystems.com/nyt.
AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. ramp.com/hardfork.
Casey, on a recent episode of this show, you mentioned something that I had never heard of before called Zettles. And you explained that it was some piece of some productivity software that you were using that had some German word that I don't remember. And it just sort of jogged my memory that we had a conversation on this show a while ago about the productivity apps that we use. And you are maybe America's foremost expert on productivity apps. You wrote a newsletter this week about three apps that you are using to become more productive, none of which I have used. And so I just wanted to ask you about that today and see what you're using.
Yeah, I mean, on this show, we hold each other to the highest standards of productivity. And technology is a big part of that. And so let's dig into it and find out how we're trying to get by in this crazy world, Kevin.
Yeah, so the first app that you mentioned was something called capacities. What is capacities?
So capacities, it falls into this genre of tools that are called personal knowledge management. So it can contain everything from a daily journal about what you're doing that day to notes about subjects you're trying to keep track of. And then all sorts of media. You can throw audio in there, you can throw images in there. But honestly, Kevin, I don't even think capacities is really the interesting part because it's particularly how I've been using capacities that has made me feel like I've maybe started to unlock something.
How are you using it?
So there is this system of organization called Zettelkasten, Z-E-T-T-E-L-K-A-S-T-E-N. If you want to go look it up, you can read the whole history of it. And I'm not going to go into it today. But the basic idea is that if you want to get really smart about something, if you want to develop expertise, you need to break down ideas into little atomic units. Right? So I might have a little atomic unit called, you know, "Google is accelerating the decline of the web." That's a thought that I've had. And as I move through the world, I'll read stories and I'll see pieces of information and I'll think, oh, that actually contributes to this idea that I have about Google. And what I used to do in the past about that, Kevin, was nothing.
I would save a link into a database and I would just try to remember that. And if I ever wrote about Google, I would think, now, what was that story that I read? What was it? I'm trying to make a change and the change is as I get this new information, I'm going to this list of notes that I have, which are called Zettles, and I'm adding that information into the note. And so over time, I'm fleshing these ideas out. I'm taking little notes within the note. I have all the raw materials. Some of these things will turn into columns. Some of these things will just sort of be reference materials for me. Sometimes I might change my mind about something completely and delete one of these notes because I've decided I was wrong about something. But the whole idea is to do my job in a way that my, my knowledge is increasing over time
and it is like, like visible and searchable to me.
So how often do you take a note in capacities? Like you, you said that after reading some stories, you'll, you'll put it in there. Do you do that after every story that you read?
No, so I try to do it basically on most mornings. I'll wake up, I'll read the headlines because you know, one thing about me is I'm writing three columns a week and a lot of those columns are just based on stories that I am reading in the press that are giving me ideas, you know, sparking thoughts of I want to do a little bit of my own reporting on that, or I want to contribute some analysis to that. I would say 80% of the stories that I see, you know, I, I'll pick out some to put in a platformer and I'll kind of give them a skim. But then there'll be three or four where I'm like, this is really interesting, right? Like this advances the narrative in some interesting way. That's the kind of story that I want to take and I want to put into a Zettel because I think it's going to be important in the future.
And do you find that the Zettel Cast In system actually makes you more productive? Because it seems like a lot of work to me.
It does, but I think what it does is grounds me and my beat in this really important way. Right? So it's like, you know, I've mentioned this on the show before, but I have this note called OpenAI is a weird company. And that's not that profound of an observation. We've all sort of seen it over the past year, kind of have a, a lot of strange things happen to it. But the thing is, the next time I need to write about OpenAI or the next time some bizarre thing happens there, if I didn't have this system, I would be on Google. I would be trying to remember what came after what. Right now, I just have a chronology that's sort of built and ready to go. So the next time I need to step into this world, it's like sort of been pre-built for me. But more important than any of that is I just think that like in journalism in general,
we don't spend enough time reckoning with ideas, right? I think like for the most part, as journalists, we're trained, go out, find out what happened, write down what happened, move on, go break another piece of news. And by the way, that's like, great. And like, I do try to do that. But I think if you really want to become an expert, like, you sort of have to reckon with whatever like the substance of your beat is. You know, for us, it's like, is AI going to create massive job loss? You know, how biased are AI systems? Will AI create existential risk for humanity? These are just like some of the big debates that you and I have been having on the show for two years. And my fear is before I implemented the system, I don't really know how much smarter I was getting on any of those topics. I talked about it a lot. I'd read a lot of stories about it. But like, was I really, really grappling with it?
So anyway, that's just where I'm trying to push myself.
That's interesting. Yeah. All right. So let's talk about the second app that you talked about in your newsletter this week, which is something called Raycast.
What is Raycast?
So Raycast is what they call a launcher app. And now this is available only on the Mac desktop. So I realize that's going to leave a lot of people out and annoy some people. I imagine there are going to be some maybe similar apps on Windows that you may want to check out. But the way that Raycast works is you set a hotkey for it. So in my case, it's just like command space and I type that and then a little window pops up. And if you've ever used a Mac before, you've probably seen Spotlight, right? By default, when you type command space on your laptop, Spotlight comes up and you can look for an app and you can look for a file. Raycast is the same, except it can do a lot more stuff. For example, just by typing a few keystrokes, I can rearrange all of the windows on my screen, which is nice. If, you know, because I have a really big monitor and I'm constantly rearranging windows.
But the most cool thing that I do with it, Kevin, is that this has become my favorite way that I use AI during the day.
Say more.
So a lot of times during the day, I'll just have a little question that I don't want to bother opening up a new tab for in Google. Maybe I'm watching a YouTube video, for example, and I want to ask a quick question about something somebody said. Or I'm writing out my column and I want to like look up the etymology of. A word, I'm now able to do that in line with Raycast. I just hit command space, I type in my little query, I hit tab and it sends it to an AI chatbot. And I get an answer back. Now I'm going to anticipate the obvious concern here, which is like, wait a minute. Like you know that those chatbots are making mistakes constantly, like, you know, do you really trust them? And what I'll tell you is I just don't ask any like mission critical questions of these things.
But I don't know about you, like, most of my questions that I'm asking are just like very low stakes.
You know, I outsource all of my decisions to an AI chatbot. I'm essentially Victor Miller.
No, but I, so I use Spotlight probably once a day or so, usually for finding some file that I don't know where on my computer it is or which folder it's in.
How are you, how often are you using this? And how useful do you feel like it really is compared to Spotlight?
I mean, I would say I probably use this more than a dozen times a day easily. It really has become the way that I navigate my computer essentially is Raycast. And the thing that I like about it and where I actually feel like it makes me more productive is it reduces context switching. Right? Like, I don't know about you, but when I open up a new tab on my browser, I'm in the danger zone. Right? I might think I'm just Googling a fact, but then Google shows me like three news stories underneath whatever fact I searched for. And then I start to look at my other tabs and I get curious about something and all of a sudden I've like wandered down a rabbit hole. The thing that I love about Raycast is I just hit command space, I look it up and then I move on. And I've never even left the tab that I was on. And I truly do think that that is reducing the amount of content switching that I'm doing.
Hmm, that's interesting.
All right, third productivity app, Readwise, tell me about Readwise.
So look, if you're a reporter, you're just constantly having problems with documents, Kevin, right? Like, for example, a big ruling drops and it's 300 pages long and you need to get a story up about this as soon as possible. Like, can you read the whole thing? Well, yeah, you could, but it's going to take you a few hours. So what do I do instead? Well, now I'm just having AI take the first pass at these things. And not only am I doing that, but I'm also saving them in a place where for the rest of time, I'll just be able to quickly revisit these things. Like, I don't know what your PDF management system was before this, but I've recently started using Readwise and it's kind of. An all-purpose reader app, you can, it has an RSS reader in there, for example. So you can, you know, add in all your favorite blogs and websites and read them all directly in Readwise if you want.
You can save articles that you're reading from the web if you want to read them later, which is something else that I like to do. But to me, the real like AI power here is, gosh, there are so many PDFs out there and they're so long and they're so hard to read. So it's not only the court rulings, but like I read a lot of academic papers and I often really struggle to get through them and understand what they're saying. And again, I'll anticipate the objection, Casey, are you really trusting in AI to do your reading for? And the answer is ultimately no, but what I am trusting it to do is to give me the first pass on something, right? Saying, hey, give me the gist of this, right? Or maybe pull a few quotes for me about this. And then, once I have a sense of it, then I will dive into the document and I'll actually find the quote and I'll verify with my own eyes that it actually says that.
But I have found that this is just a really great kind of reading assistant and librarian for me.
And so this replaces the, the sort of uploading documents to the chatbots? Because you can also upload documents to ChatGPT, to Claude. A lot of the major AI tools now allow you to just throw in a PDF and say, can you summarize this for me? So is Readwise a replacement for that feature or does it do a better job? Or why, why use one instead of the other?
So I think that you can totally do that. And I do sometimes still ask like Claude, for example, to read a PDF for me. The thing that I was missing though, was like the organization piece, you know, like the, for example, you know, there was just like a big Google antitrust ruling. I expect I'm going to want to refer back to that document in the future. And in the past, what I would always do was I would download it onto my computer and then eventually it would like get moved to the trash. And then the trash would empty and then I would want to refer to the ruling again. Now I'm Googling again. Eventually I said, this is enough. I need to just like put my, my PDFs in a place where I will be able to find them again. Again, there are other solutions to this, but I like the idea of having a dedicated reader app that wouldn't just sort of help me read these things, but also help me organize them.
You're so productive with these apps. It makes me wonder why you're not writing five or six columns a week. Maybe 10, 20.
Well, hey, we'll check in with me in a year. Maybe I've expanded.
But do you actually? So this is my, always my question to you because you are obsessed with productivity software. For as long as I've known you, you're always trying out the latest and greatest note-taking app or the organization tool. And you are very productive, but I'm curious, like, do you feel like you are more productive as a result of using these tools?
It depends on how you define it, right? Like, do I write more columns this year than I did last year? No. Do I feel like I have more peace of mind and feel less burned out than I did last year? Yes. I don't want to say that all of that is just because I'm using slightly different tools. But ultimately, I am a believer in this Steve Jobsian idea of a computer as a bicycle for the mind, right? Something that helps you get a little bit further, a little bit faster. And that's how I feel about this stuff.
And do you think these tools are mostly useful for people like us, writers who are, you know, constantly synthesizing, you know, lots of different stories and ideas and referring back to things that maybe came before? Or do you think other, people in other lines of work could use these tools well, too?
Yeah, I mean, I don't think that they're useful for absolutely everyone, but some other folks, I think, who might find some of these systems useful would include students, academic researchers. I think if you work in communications, you know, this could be really powerful. But, you know, ultimately, a lot of people who do knowledge work do want some kind of system to keep track of stuff, to noodle on some ideas. And, you know, again, all I'm trying to say this week is like, well, there's some stuff out there you might want to look at. Kevin, how are you taking notes these days?
So this is an area where I-
By the way, I want to say that I'm, when I, when I posted my column this week, somebody on social media said, can you please check in with Kevin and see if he is still emailing notes to himself?
Yes. So this is an area where I do actually have news to report because after our last episode where we talked about this stuff, I got so many people who were like, I would say somewhere between offended and outraged that my method of note taking was emailing myself small notes. And you would think that I had confessed to like, you know, torturing animals or something like that because people-
I did experience physical pain when you said you did this. I understand why people were upset.
So I was so ashamed by this admission and the response that I got to it, that I did actually start looking into a bunch of different note taking apps. And eventually I was sort of talking about this with a friend and he said, dude, just use the Apple Notes app. It's gotten so good. And I thought, well, you know, I've used Apple Notes. It's not very powerful. And he was like, no, you should try it again. They have added so many features to it. And so I did. And now I'm happy to say that Apple Notes is my de facto app for a lot of things in my life. I am fully notes pilled.
Well, and so tell me about what kind of notes you're doing with Apple Notes and what are you liking about the app?
So I like Apple Notes, A, because it is so easy. I never have to like login, fumble for my password. It's right there. I'm an iPhone user. I'm a Mac user. And so it's just, it's there. It's on every device. It's automatically syncing to my iCloud. I never have to worry about like losing data or anything. But that's just sort of the basic one because as my friend told me, and it turns out to be true, Apple has done a lot to make notes a more full-featured app. You know, it used to be useful for mostly for jotting little notes to yourself-
Grocery lists.
Yeah. Apologizing on social media. If you were a celebrity who had done something terrible, you'd make a note in Apple Notes and screenshot it and post it to Twitter. But I find it useful for not only keeping sort of my to-do list, because you can make little checklists, but also it can do some of the things that you're using capacities for. Like it can link notes. You can have hashtags around ideas. It can create these, what are called smart folders where you can have- I have a folder that's just, so I have a folder that's just how I'm preparing for episodes of the podcast, or ideas for columns that I want to write. It's actually now become my default word processor. So I don't write in Word, I don't write in Google Docs, I write in Notes.
You know, I tried to put one of your columns into a smart folder and it said it didn't belong there.
Come on.
So it can also do things in the, in the sort of newer versions, like it can scan text. So a thing that I would often do when I was reading a book and I came across some passage that I wanted to remember, is I would take a little photo of the book and then it would just sit in my camera roll and I would never look at it again. And, you know, I'd find it while I was cleaning out my, my screenshots folder, you know, years down the road. But what you can do with Notes app is you can actually just scan the page and it will strip out the text and make it searchable. So you can take a, you know, a page from a book and, you know, all you have to do is kind of remember, oh, that had a line in it about, you know, a robot or something. And I wanted to remember that. You can search your Notes app and it'll pull up every instance of robot in all the text that you've read and saved.
So I use it for that too. In the new version of iOS, iOS 18, there's a feature in, that's been added to Notes that I really love. I just started playing around with it because I just got iOS 18. But you can actually do voice memos inside Notes. So you can start recording, which is a thing that I do a lot, as you know. I'm a, I'm a big dictator. I like to, you know.
You like to control populations without letting them vote on their own self-determination?
No, other kind of dictation. I like to speak aloud to, to my phone and have it sort of take down notes for me and then refer back to those later. So I'm Notes pill. I think it's a great app. This is not sponsored content. I'm just a big believer in keeping things simple. And for me, the Notes app has given me a little bit more organization than I used to have.
That is great to hear. You know, as much as I love the particular apps that I love, like, all I want for folks is to like have a system that works for you. I talked to so many people. People that feel truly oppressed by the glut of like digital content in their life. And they're looking for some kind of lifeboat. And I know a lot of people that are using Apple Notes for exactly the same reason. So if it's working for you, that makes me excited. But I'll give you one rule of thumb, maybe to close out. If you want to sort of give yourself an easy test for like, how productive is this piece of productivity software? And it is just this, and it is, how much time are you spending in the settings of things and in the metadata? Because if it's a lot, then it's probably not making you more productive.
But if you just open this thing up and you go, you're probably good. And it sounds like that's what Apple Notes is for you. That's what capacities is for me. Don't spend a lot of time in the settings. You might need to set it up initially, but for the most part, you just want to be, you know, inside your app, making the little clicky clacky noise with your keyboard. And if you do that, maybe, just maybe, you'll find a way to get by.
Today's AI builders need infrastructure that scales with their ambitions. iRen is a vertically integrated AI cloud platform, delivering data centers, compute and software for AI training and inference, underpinned by its expansive portfolio of land and grid-connected power in renewable-rich regions globally. iRen's AI cloud gives direct access to Nvidia GPUs, scalable from hundreds to thousands. For builders, that means direct access to AI compute, rapid scaling and predictable pricing. iRen, AI cloud for the next big thing.
Companies are pouring money into AI, but few leaders can explain how those AI investments work with the systems you already have. You know, the ones that actually run your business. Without systems, you don't have to start over with AI. You can choose the projects where AI will have the most impact. And Outsystems gives you the platform to make it agentic on your terms, with complete control every step of the way. So you can trust AI with the mission-critical systems that run your business. Own your agentic future. Learn more at outsystems.com/NYt. That's outsystems.com/NYt.
AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.
One last thing. We want to do a back-to-school segment for an upcoming episode about all these new phone bans that are popping up at schools around the country and how they are actually impacting students, teachers on the ground at these schools. So if you are a student who has had to lock up your phone during the day this year, or maybe you're a teacher or an administrator working at a school that's put a phone ban in place, we want to hear from you. Are the phone bans going well? Are they not going well? What is actually happening on the ground at schools as they wrestle with what to do about phones? If you have a story to tell us, please email us a voice memo to [email protected], and we might include it in our upcoming episode. Thanks a lot.
Hard Fork is produced by Rachel Cohn and Whitney Jones. We're edited by Jen Poyant. This episode was fact-checked by Ina Alvarado. Today's show was engineered by Daniel Ramirez. Original music by Pat McCusker, Marion Lozano, Rowan Nemisto, and Dan Powell. Our audience editor is Nell Galogly. Video production by Ryan Manning and Chris Schott. You can watch this full episode or any of our episodes on YouTube at youtube.com/hardfork. Special thanks to Paula Schuman, Queuing Tim, Dahlia Hadad, and Jeffrey Miranda. You can email us, as always, at [email protected].
You can't undo everything in life, but you can undo prediabetes. More than 2 in 5 adults in the U.S. have prediabetes. Most don't know they have it. But with early diagnosis and action, you can delay or even prevent type 2 diabetes. Visit doihaveprediabetes.org to take the 1-minute risk test and start undoing the things you can. Brought to you by the Ad Council and the Centers for Disease Control and Prevention.
Page updated October 12, 2026 · transcript and summary are machine-generated
