C (2:49)

Thank you. It's good to be back. Yeah, there's a blast from the past this week.

D (2:56)

Although I hear for you it's not quite a blast from the past because you've been using your pebble For a while. I want to talk about that. That's awesome. It's right there on your wrist. So let's kick things off though. If somebody were not familiar with pebble and what we're talking about here, maybe we should start there. Although again, I have a feeling that this crowd either has owned one has heard about it.

C (3:21)

It's funny, it's been so long that there are people who, they were alive when pebble came out, but they weren't part of the tech ecosystem and community. And I've been talking to them. It's kind of fun. They're like, what's Pebble? What is that? How old are you? Oh, I'm high school. There was something special about pebble that brought it to a really wide audience and I'm kind of excited that eight years later a new generation might hear about it. So for. For those who haven't had the pleasure, pebble is and was one of the first smartwatches. I think we didn't. I don't know if we coined the term, but I started working on smartwatches in 2008, originally making one for BlackBerry, another blast from the past. So the first watch that I made worked with BlackBerry and then in 2012 we announced pebble and Pebble was the first watch that really went viral. People heard about it. What made pebble special though, is. And I guess we could chat about this more. Also coincidentally, the reason why I think it's time to bring it back. So what made Pebble a Pebble was its paper screen, which is a low power, always on daylight readable reflective screen rather than emissive, which pushes light out of it paper long battery life lasted for a week. Plus fairly simple feature set. Like we didn't try to boil the ocean and do everything that your phone could do. Focused on things like notifications, telling time. You know, who would expect your watch to be really good at telling time? Music control. We had physical buttons. We have physical buttons on the watch so you could press the button to play, pause, skip tracks. You don't have to look at the touchscreen and hit the right tiny little icon. And it was hackable.

D (5:26)

Mm.

C (5:27)

Anyone who wanted could write a watch face or build an app or pebble and that was it. Basically. Like there's, you know, a lot more that people did with it. But that was the core of what pebble is.

D (5:42)

Yeah. And I think that's what ended up being the delight for many people. You talk about in your blog post, the kind of the surprise of all those watch faces. That were available for, that are available for pebble that were not for the Apple Watch. So we, we should not continue to, to kind of beat around the bush here. What's the new news? Because I think people will be very excited about this.

C (6:09)

Yeah, so the big news this week is that pebble is that Google open source Pebble OS. So in 2016, pebble sold its IP to Fitbit, Fitbit got bought by Google. And so, you know, there's always a bigger fish world in tech. Pebble now is under, under Google. They weren't doing anything with it. And so about a year ago I asked very politely if they could open source the Pebble OS code base. Now this is the software that runs on the Watch itself. And they said yes.

D (6:47)

Wow.

C (6:48)

You know, they didn't have to and I'm incredibly thankful that they did. And on Monday they published a GitHub repo that has the entire source code for Pebble's operating system. This was unavailable before. There was no way to build software for pebble or kind of compile and run the, the pebble operating system. It still works perfectly. That's the crazy part. Like this watch is eight years old, nine years old, hasn't had an update in eight years.

D (7:17)

Wow.

C (7:18)

And it still works great. I think that's a testament to the quality of engineering work and product design that the pebble team put in, you know, between 2012 and 2016, in that it still works. Like what other gadgets do you have from 10 years ago that you could turn on, connect to a phone?

D (7:38)

Right.

C (7:40)

The first generation Apple Watch doesn't even work anymore.

D (7:43)

Yeah, that was one of the things that really stood out to me. The folks at iFixit helped with making sure that if your battery had gone bad in your pebble that you could get a replacement for that. And it's funny the pebble, because I do still have mine because I think it's just a, a testament to that time. And also it was for me that moment of excitement of not only a Kickstarter project being a success and being part of that, but also it actually shipping. So all of that is built into this totem that is the pebble for me.

C (8:19)

And in my opinion, like, I think we've lost a little bit of that gadget awe and optimism that maybe we had in spades 10 plus years ago. Like there was an era of gadget blogging that I loved, like the Gizmodo and Gadget, you know, where I was, I was a bit younger and I, you know, I just Loved reading about gadgets. There was something special to me about a gadget because it's, it's like a gadget that just doesn't take itself too seriously. It's not trying to change the world, it's just, yeah, there it exists. And maybe it helps you smile a little bit during the day. It doesn't have to be more than that. And I think we doubled down on that at Pebble. We made it quirky and fun. The OS had these little layers, like when you were charging the watch a coffee cup. The icon was a coffee cup that would fill up to the brim until. And there would be a little bit of steam rising off the top of the cup. That was fun.

D (9:18)

Surprise and delight. Right, so I want to talk about, about what's been going on between the time that pebble was acquired by Fitbit, which was then acquired by Google, and now because there were some people out there who figured out how to make it work. Can you talk about the open source hacky community?

C (9:48)

I took a break starting in 2016. I had spent nine years working on smartwatches and I was like, this is, this is good. We built, we built something nice. You know, go do something else. I went to work at Y Combinator for a bit. Luckily for me and for the entire pebble community, an open source group of hackers called the Rebel Alliance R E B B L E took it upon themselves to reverse engineer the web services that pebble had, the App Store and offer that, as, you know, free software to the community. And so for the last eight years, this group, the Rebel alliance, has been hosting hackathons, sponsoring contests to build new pebble apps and basically keeping the community kind of fresh and enthusiastic. You can see on their discord, there's thousands of people that are chatting about Pebble. And one thing that I love is you can still go to the pebble subreddit R Pebble and every week there's like someone posting about, or actually every day probably there's people posting about pebble and sharing their favorite watch faces or even new apps. Like eight years later, after there are no more watches being built, people were still writing new apps for pebble, which is pretty special.

D (11:16)

Now how many iterations of pebble did the company make before pebble was acquired by Fitbit? I couldn't remember if it was two or three.

C (11:28)

There was pebble, the OG pebble, as we called it in 2012, Pebble Steel in 2014, Pebble Time in 2015, and then Pebble 2 in 2016.

D (11:44)

Got it so there were four total generations. Pebble Steel was one that escaped my memory. So now we're talking about with Pebble's operating system being open sourced and with the I think still robust and excited community that exists there, is it time to bring back Pebble? Are we going to have new hardware? That's, that's kind of wild, that's kind of exciting.

C (12:15)

So I have a problem. I'm running out of Pebbles. I've luckily, I mean, I obviously having worked at Pebble, I had a big stash and so over the last 10 years I've been kind of working my way through. But know even great hardware doesn't last forever, right? Albeit it's lasted extraordinarily long. You can replace the batteries, like you mentioned, iFixit has a bunch of tutorials on how to do it. You go on YouTube and there's like how to replace the bevel, the pebble buttons, how to replace the pebble, everything. It's, it's, it's cool. But you know, I used to tell people that you could buy a Pebble on ebay, but the prices are either skyrocketing like hundreds of dollars or there's just none available anymore. So what's going to happen? I decided to start making new Pebble Watch hardware. Hardware that will run pebble os. I'm going to do it a little bit differently this time. So whereas last time we did it as kind of a prototypical startup, you know, we raised money, built out a big team, had up to 100 plus people working at Pebble. It taught me a lot about how to build a sustainable operation. And this time around, I think the number one goal is how do we build this in a sustainable manner. And I mean that sustainable in that we will continue building hardware long into the future and continue being able to support the operating system and, you know, updated and published new features and that kind of thing. So we're starting small. We're not going to try to do too much all at once. We're going to build a watch and get it out there and see what people think. And then if it's successful and we earn enough money, we'll reinvest that in building, you know, maybe more watches. We'll, we'll kind of cross that bridge when we get to it. But the name of the game is sustainability. So it's going to be a small company, no venture backing, and just building kind of in public and taking it one step at a time.

D (14:27)

Yeah, I feel like that's not common these days. Right. That you say no venture backing and that in and of itself, to me, it's giving an underdog vibe in a good way. I mean that in like a, you know, you're cheering it on, you're saying, you know, what can you do in that very small sustainable way. But you would know better than I when it comes to, I guess it's the mixture of hardware and not going venture backing that is interesting. Do you feel like it's because of your knowledge up to this point, of the things that you made? I even read about the. I think it was a relative of yours who was asking you for help hacking a BlackBerry to display things on the ceiling. Like, you clearly have some, some electrical engineering knowledge there.

C (15:22)

So I think the essence of this is. I wrote a blog post ten years after the first Kickstarter, kind of reminiscing and looking back and kind of analyzing some of the, some of the story. I think what I recognized is that I am, I think I'm more of an inventor than an entrepreneur.

F (15:44)

Okay.

C (15:44)

I love creating products and I love using those products and I very greedily want this product to exist. And so my plan is to like, make sure that the elements that are needed to keep building these type of products exist. And that doesn't necessitate selling tens of thousand, hundreds of thousands, millions of watches. Like, we can build, like I said, a sustainable company without having to make it big. If it's big, great, but it doesn't have to be. I think most of the premise of a traditional startup is that, you know, the definition of a startup based on kind of PG's programs work is startup is a company that's designed to grow quickly. And I think this is not a startup. It's this company designed to build great products, get them out into the world, and if growth happens, great, but it's not a prerequisite.

D (16:48)

Yeah, sounds very much like a passion project, I think. Lastly, I'd ask anything you can share for the people who are curious. You had a little bit of an FAQ that we'll of course link to. But are we imagining a watch that is three times the size of the current one? What are we expecting when it comes.

C (17:14)

To expect literally a pebble? Expect whatever the last pebble you had. Like expect that, you know, this is also recognizing. I think I've recognized that products and software can be done. You can finish something. Like, it doesn't have to continually change and add new features. And like you can be done something. And I'm not saying that pebble is done in that there's literally never going to be a change ever again. But I'm still using the same watch with the same software that was there eight years ago. It's better than the alternative, which is wild to me that like no one in the last eight years has built something that's remotely comparable or usable to the set of features that I'm looking for out of a smart watch. Plenty of people in the last couple days have talked to me and said, oh, is it going to have this feature? Is it going to have that feature? Is it going to be like this? I'm like, no, it's going to be like a Pebble. There's plenty of other options. There's just no options for the people that want this exact set of features. And that's all we're saying.

F (18:29)

Nice.

D (18:30)

Well, I want to thank you so much for taking the time to join me today on the show for getting us excited about the return of Pebble. Of course, the open sourced code that you went and said, hey, look, is it possible you could make this available? That's pretty cool too, all of that work coming together. I know that there are many people out there who are listening to the show who, like I said, probably own Pebbles, pebble or Pebbles and are looking forward to seeing what's next. Thank you again for your time. If people want to stay up to date with everything, is there a place that they should go to follow you or. Yeah, tell us.

C (19:16)

Check out repebble.com Sign up for the wait list. We'll be sharing a lot more information about, you know, new stuff, how you can help if you want to be part of the community and help kind of on the software side as well. Join the fun.

D (19:33)

Awesome. Thanks so much.

F (19:36)

Hey, don't let me interrupt. I know we're having a blast here reliving 2025, but I thought this would be a good time to mention something we do every year around this time that's very important to us and to our ad sales. It's our twit survey. We do it because we don't really. And no podcast does know anything about you. That's, I think, a good thing. We respect your privacy, but we also would like to know a little bit about you to the degree you're willing to help us out. Just some basic information that helps us go to advertisers and say things like, well, 80% of our audience is it, decision makers, that kind of thing. That's why we do this annual survey should only take a few minutes of your time, as I said, is one of the ways you can contribute to keeping TWIT on the air. If you would like to before too long in the next couple of weeks, do it now while you're watching, go to TWiT TV Survey 26. It's our annual 2026 TWiT listener and viewer survey. It's very important to us and I thank you. I really appreciate. And of course, if you don't want to do it or there's questions you don't want to answer, that's fine too. But anyway, you can help us out. We appreciate it. All right, now back to the show.

E (20:49)

Here's to quitting.

A (20:51)

Quitting the couch for a quick run. Quitting the snooze button for a morning workout. Quitting giving up after two weeks.

B (20:58)

You see, staying committed to your fitness goals isn't easy.

E (21:02)

But with Apple Watch, you don't have.

A (21:03)

To do it alone because Apple Watch gives you real time motivation plus advanced.

E (21:09)

Metrics that track all your workouts so.

A (21:11)

You can stick to your New Year's fitness resolutions. And once and for all, quit quitting with Apple Watch. IPhone 11 or later required Shopping is hard, right?

D (21:21)

But I found a better way. Stitch Fix Online Personal styling makes it easy. I just give my stylist my size.

F (21:27)

Style and budget preferences. I order boxes when I want and how I want, no subscription required. And he sends just for me, pieces plus outfit recommendations and styling tips. I keep what works and send back the rest. It's so easy.

D (21:41)

Make style easy. Get started today@stitchfix.com Spotify that's stitchfix.com Spotify.

E (21:49)

Tito's Handmade Vodka is America's favorite vodka for a reason. From the first legal distillery in Texas, Tito's is six times distilled to till it's just right and naturally gluten free, making it a high quality spirit that mixes with just about anything from the smoothest martinis to the best Bloody Marys. Tito's is known for giving back, teaming.

A (22:07)

Up with non profits to serve its.

E (22:09)

Communities and do good for dogs.

A (22:11)

Make your next cocktail with Tito's, distilled and bottled by 5th Generation Inc. Austin, Texas.

E (22:16)

40% alcohol by volume.

A (22:17)

Savor responsibly and I am very excited.

D (22:20)

To say that we are joined by antitrust reporter extraordinaire, Leah Nylan. Welcome back to the show, Leah.

B (22:26)

Thanks for having me.

D (22:28)

Absolutely. So when you were last on the show, I believe we talked about the Justice Department's proposal to force Google to sell Chrome. Could you give us A little update on how that proposal has evolved since then and where things stand now in the remedies phase of the trial. And I have to know too, do you and all of your antitrust friends love the remedy portion? Because to me, that's so interesting.

B (22:53)

I mean, it is fun because it's a little bit more like forward looking and talking about technology as it exists today and how it might exist in the future, as opposed to like, you know, rehashing stuff that happened back in like 2009. But yes. So the Justice Department revised its proposed remedies a little bit in March. They are still seeking a divestiture of the Chrome browser, but they changed. The big change was they eliminated a portion that would have prohibited Google from making investments in AI startups that had been opposed by a number of AI companies, including Anthropic, which already has some investment from Google. They said it would really, you know, harm the AI ecosystem for Google not to be able to invest in companies, in part because a lot of the investment that Google has been giving to companies is in the form of what they call sort of cloud credits. So they've essentially been giving them free access to their cloud infrastructure. And that's a really big part of what AI companies often have to pay for getting all of that computer space. So that was sort of where we started two weeks ago or a week and a half ago when this remedies trial started on April 21. We are now a week and a half in, and the Justice Department just earlier this week, on Tuesday, finished its sort of presentation. So for the first week, the Justice Department presented testimony from a number of different market participants, along with a few groups, Google employees, about how its remedies would sort of change the ecosystem. There are sort of like three big buckets. One of them is obviously the Chrome divestiture. The second big bucket is the Justice Department wants to force Google to share some of the data that underlies its search results. This it would have to give to what are called qualified competitors. So the government won't get to decide who it goes to. It doesn't get to go to just anybody. But if somebody is labeled a qualified competitor, they would get all of the data, the ranking signals and things that underlie Google search. The idea being that other search engines or other companies could use that information to either create their own search engines or improve their own search results. Or in the case of AI companies, it sort of gives them a leg up on trying to create large language models that might compete with Google's. And Then the third big bucket has to do with Google's contracts, which were sort of at the center of the original case. So Google had all of these exclusive contracts with smartphone makers and browser makers to make its search engine the default. The Justice Department wants to bar Google from paying companies for that position. And additionally they want to bar Google from paying to make it a AI app, Gemini, the default on browsers or phone makers. Google says obviously it's not a big fan of that proposal. It says it's not really fair because AI is still a very burgeoning space and it has not been found to be a monopolist in AI. But those are sort of the three big buckets of what the Justice Department has been presenting evidence for the past week and a half.

D (26:12)

Understood. Wow. I really appreciate you because of how well you take what is just these huge topics and make them easy for us to understand. So thank you. Sundar Pichai, of course, has testified at this point calling the government's data sharing remedy, quote, a de facto divestiture of search. Can you talk about the tone of his overall testimony and how Google is positioning its defense at this point?

B (26:41)

Yeah, it was very interesting. This is the third time that Senator Pichai has had to testify in as many years in these antitrust cases. He, of course, testified during the original trial in this case back in 2023. He also testified in another lawsuit over the company's alleged monopolization of the Android system. And then now he's testified again. This was by far the shortest one. He was only on the stand for about an hour and a half, but it was, he took a very serious tone in talking about how he thought this would, like, ultimately damage both Google and the US's sort of stature in innovation and national security. So he said, you know, by forcing us to divest Chrome and give away this data that we have spent years building, it would ultimately undermine Google as a company because, you know, they have invested, you know, he said, millions and billions of dollars over the years in building up the search infrastructure and all of the things that go into the search engine. And the Justice Department would be requiring them to give it away not for free, but for. For very little money. And that, that, you know, is going to hurt the company's ability to invest further in the future because it's not going to be earning as much money. It's going to have to face more competition from these other folks. That in turn, he said, could potentially harm national security because, you know, the us, Google and the US are technological leaders. As of right now, the US has not put any limits on the nationality of qualified competitors. It hasn't said that you have to be a US company in order to access this data, though that could potentially in the future be a, you know, a condition. And then they also said that this would definitely damage their ability to compete in AI because the Justice Department is seeking some of these restrictions on how it distributes its AI model right now. We learned from the trial actually that Google has created an agreement with Samsung in which it is going to be paying billions of dollars to Samsung to put Gemini as the default AI assistant on Samsung's new versions of phones. But Google has stressed repeatedly that these are not exclusive contracts, that these phone makers can also try and install another AI system their own or a different one, and users would have the option to choose between them. You know, some of the other AI companies that testified said, well, yes, you might be able to get on the device, but Google's contracts have actually been blocking them from getting the default. So we heard from Perplexity AI, which had actually reached an agreement with Motorola that on all of the new Motorola phones, Perplexity is going to be installed in there. Both Motorola and Perplexity actually wanted the company to be the default AI assistant, but Google's contract sort of prevented that. So Perplexity will be an option, but not sort of the pre installed option that comes pre working on the phone. So there's been a lot of a little bit in the weeds about whether Google's contracts really prevent this or whether this is a misreading of them. And it probably will end up being up to what the judge decides at the end of the day.

D (29:59)

Understood one kind of, I think, interesting part of, of this aspect because you did, you talked about how, you know, we heard from other executives. OpenAI perplexity, DuckDuckGo all came in to testify. What was the overall point, I guess in terms of the court in understanding how Google's practices affect competitors? And does it seem like those testimonies have had an impact on the remedies portion of the case? Because speaking just from my own, I think it would be difficult to sort of figure out what remedies like to even, to even fathom what it means to find remediation in the situation. So is it often the case that competitors coming in to testify helps to kind of give the court somewhere to go when it comes to the remediation?

B (31:04)

Yeah, so this is a little bit of a new proceeding. So what had happened, you know, this case is very Much based on the Microsoft case that happened 25 years ago. But what happened in Microsoft is the judge didn't have a remedies hearing and the court really ended up dinging the judge for that and said you should have had a separate entire hearing before you decided what to do. So here the judge is like a little bit breaking new ground with how to do this and what to do. So the Justice Department brought these witnesses in, as you mentioned. OpenAI perplexing DarkDarko and also Yahoo. Fun fact. Yahoo still exists. And what they had them testify about was one, how Google's current practices impact their business. So all of the companies talked about how, you know, Google's contracts limit their ability to get distribution through their own contracts with manufacturers, which means that it's more difficult in their view, for consumers to discover their products. And then they talked about how they think the proposals that the Justice Department put forward would help their business. So all of the AI companies and then the two other search engines, you know, sort of described how having either access to Google's data or the ability to, in their view more fairly compete for defaults on search engines and browsers would sort of increase their ability to compete with Google, make, make it easier for consumers to discover them. And what was very, very interesting is all four of those companies said that they would be interested in buying Chrome. So I remember, I heard when, when it first came out that the government was looking to divest Chrome. All of these like analysts were like, nobody would buy it. Like, why would you buy this? It's really dumb. You could make your own browser. But all of these companies got on the stand and said, yes, we would be willing to pay billions of dollars because so many people use Chrome like it is a built in distribution mechanism. And it has been a very successful built in distribution mechanism for Google both in terms of search and now in terms of Gemini, its own AI app. Google has started to integrate Gemini directly into the Chrome app so that sometimes, you know, right up there in the bar when you're typing, you're going to start getting answers from Gemini, not just from Google, Google search.

D (33:27)

Now I wish, I wish I could keep you around longer. I want to round things out by asking you judge. The judge is expected to rule on remedies this summer. And this is the always the final question. What should we be watching for in terms of timing, in terms of next steps and how likely is it that we're going to see everyone's favorite word an appeal?

B (33:51)

Yes. So the judge has, we're having the remedy trial right now, it's supposed to wrap up next Friday. The judge then is waiting a couple more weeks before he's going to have closing statements. So then the government will sort of get to have a last chance to sort of wrap up all of their arguments for the judge along with sort of like their legal reasoning that will happen on May 30. And then the judge has said that he plans to roll by August because that would be one year from when he issued his opinion last year. And he felt it was very important that we have this remedy in place within a year of when he made the original decision. And then of course, Google has already cited plans to appeal. And so that means that we're going to have to go through this entire process where it goes before another court. That can take another nine months to a year. Though I will note the government has been pushing very hard that this is a very like the time is of the essence here. Right. They really don't want this lingering for a very long time. And in some of the other cases involving Google, such as the Epic case, people have argued that in order to expedite the appeal, that is to make it go a little bit faster than it normally would. In the Epic case, they already had their arguments in February and the Ninth Circuit is sort of expected to roll at any time now. So we might, you know, you could see that here in this case, so that we might have a decision, I mean, maybe by the end of the year, but probably early next year.

F (35:19)

Understood.

D (35:20)

Well, Lee and Alan, I want to thank you so much for taking the time to join us today. Always a pleasure to have you on the show. If people would like to keep up with what you're writing about, where do they go to do that?

B (35:30)

Bloomberg.com youm can follow me and it will email you anytime I write a story or you can also follow me on bluesky.

D (35:36)

Wonderful. Thanks so much.

B (35:38)

Thank you.

D (35:40)

Ooh. I am so excited about our next interview. Joining us on the show today, head of research for the Wikimedia foundation, it's Layla Zia. Welcome to the show, Layla.

E (35:51)

Thank you so much, Maika, for having me. Great to be here.

D (35:55)

Absolutely. So the Wikimedia foundation has put out a bit of news about the AI strategy for Wikipedia and I was hoping that you could start by giving us kind of a high level overview of this strategy and how it fits into the long term mission of Wikipedia.

E (36:15)

Happy to do that. Indeed, we have decided to put out a new AI strategy and in this new AI Strategy we are focusing our attention on reinvesting on the humans behind Wikipedia. The community of volunteers behind Wikipedia are the people who are the most unique element of the success for Wikipedia. Wikipedia is a project that for almost 25 years have been developed and cared for by these people and has become a central part of our lives on the Internet and web in a variety of ways, whether we come to Wikipedia directly or use its content or knowledge through other platforms. In the Wikimedia foundation, we decided to invest on more heavily with supporting these editors with AI. More specifically, we decided to invest do a targeted investment in four areas to support the editors. One is to support the work of moderators and patrollers to assure the integrity of knowledge on the projects. The other one is to focus on any task that is basically repetitive and is required and does not require human judgment. It's basically a barrier for editors for doing the work that they're trying to achieve. We also decided to focus on investing in AI for reducing the burden on editors for creating knowledge that already exists on Wikipedia in a given language and perhaps giving them pathways to bring their local perspective and knowledge to the world. And lastly using AI to help the next generation of editors to become editors on Wikipedia through mentorship and assisted mentorship and helping them get onboarded to the projects. These are kind of the four primary areas that we would like to invest in AI in support of the editors on the projects, with again, like a primary focus to be on the human agency and the editors being on the projects and supporting them.

F (38:18)

Absolutely.

D (38:19)

Now, I think we'd be remiss if we didn't talk about the public anxiety that exists around AI replacing in some ways human creators and editors. So I'd love to hear you talk about how this strategy reaffirms the role of Wikipedia's volunteer community in this age of generative AI, that the humans are first, as you've mentioned, to hear about how that actually plays out would be great.

E (38:49)

Yeah. So the primary thesis of the strategy is that we are investing, reinvesting and focusing heavily on editors as humans who are behind knowledge. Really the reason for this is knowledge is socially constructed and humans are needed for creating knowledge. At the same time, we understand that there are a lot of repetitive tasks, tasks that do not require human judgment that are happening on Wikipedia by our editors. So we want to invest in AI in areas that human judgments, deliberation, discussion and consensus building is not needed and leave more time for people to do the things that people are best at, if they choose to do so. So that's basically the primary thesis for us, which is invest in humans, reduce, give them the option to invest less time in areas that their AI can help them with, and then instead give them more time for things that are very uniquely human, which is around deliberation, consensus building and discussion.

D (40:00)

Understood. Now the foundation highlights AI assisted workflows for moderators and patrollers. I'd love to hear kind of some nitty gritty, what kind of tasks are you hoping to automate and how will that support content integrity on the platform?

E (40:18)

Yeah, our moderators and patrollers are key to the success of Wikipedia. These are the people who make sure that the content in Wikipedia is meeting the, the policies that the communities, the content policies that the communities have put together. What we are primarily focusing on in this space is saying that these individuals are handling a lot of repetitive tasks on the projects and we know that AI can assist them in some of these repetitive tasks that don't require human judgment. Again, we're going back to that theme of where human judgment is not needed or much less needed. Can we introduce AI to support these editors? The other thing that I want to bring your audience's attention to is that Wikipedia, while many of us read it in English, Wikipedia is a project that is available in many different languages. In fact, over 300 languages. And these moderators and patrollers operate in all of these different languages. What our teams eventually decide to do in terms of supporting moderators or patrollers in a given language with AI also depends on the specific needs of that language. The needs of a language in which you have, let's say, less than 100 edits per day. In that Wikipedia is going to be different than the needs of a language in which you have a few edits per second in their Wikipedia and the affordances that you have for offering AI. To give you just one concrete example, maybe suppose you're a patroller and you want to update information related to a reference that is being used currently on a Wikipedia article. If you want to find all Wikipedia articles that are currently using this reference. This is a hard task right now, but the task of retrieval and discovery is something that we can support patrollers with and moderators with. So this is one area that we think AI can actually help moderators more effectively pull information that is already available and do their work more effectively on the project.

D (42:28)

Understood. Now, one interesting aspect with you talk about this a little bit. The AI assisted translation to help editors share local perspectives more broadly. I'd love to hear you know the goals of this feature and most importantly, because there's that AI anxiety again of sort of stripping the, the human context out of things. How do you ensure the cultural context is preserved in translation?

E (42:54)

Yeah, excellent question. And this is a topic that, as you may imagine, we care deeply about. So a few things I can share here is that one again, going back to the thesis of this strategy. The focus is on humans and bringing AI in places where humans can do their job, support humans to do their job better on the projects, or more effectively on the projects if they choose to. So when we talk about the translation, it is in the context of that primary thesis, which is we want to support humans. So I think that's really important here. We are not talking about automatically translating content and just putting it in front of people. We're talking about supporting editors for translating that content. At the same time, what we see in smaller Wikipedia languages in terms of article size, is that editors are under tremendous amount of pressure for creating vital knowledge, vital encyclopedic knowledge on their project before getting to the point of being able to bring their local perspective and knowledge to the projects. And this is a tension that we're trying to resolve, which is there is a list of vital articles every Wikipedia should have. Can we, for at least this list of articles, support editors with translation methods and translation power to more effectively translate and more efficiently translate this content in their languages and give them back time so that they can go and bring the knowledge that doesn't exist anywhere on the Internet today, because they are the people who have that knowledge and knowledge perspective. So that is the idea behind the translation component and where we talk about translation in this context.

D (44:42)

Understood. You have you being the foundation of emphasized values like transparency, human agency, multilingual inclusion. Just. Just hearing how those main things have played out in the decisions that you've made with a. I would be great. I think once again, I know we keep pulling it back to. But this is about. You see the reaction that sometimes people will have when we hear, oh, this site or this service is adding AI. Oh no. But it sounds like there's a very thoughtful rollout that is taking place. So, yeah, can you talk about that transparency, that human agency, that multilingual inclusion as these kind of core principles as you roll out these AI tools and maybe how you're continuing to check in on those principles being followed with this?

E (45:36)

Yeah, this is a topic that is dear to our heart and that's why we talk about it in the strategy as well. I'll give you maybe a couple of examples and I'll be led by you if you want me to go through more details. One is on the topic of transparency. Our machine learning team has developed this concept of model cards. So every AI model that we put into production, or even we propose to put in production, needs to have a model card page, which is effectively a page which is publicly accessible. And you can go and check to see what is this model, what is the motivation for creating this model, what kind of use cases we had in mind for creating this model, who is the user that is the target for this model, what are the ethical considerations that we have had for developing these models, what are the caveats that we are seeing, and all the technical details about which models we have used and what data pipelines we have used and all that. All of these are captured in model cards. Model cards are a requirement for any model going to production. And by making them a requirement, effectively, our machine learning team has built in the process for making sure that we have transparency around what model is going out and what do we know about the shortcomings and the great things that the model can do. When it comes to the topic of languages, I think we have developed some good understanding on this front and we are still on a journey to fine tune our understanding. Our general position right now is that we need to look at what are the needs of different Wikipedia languages to decide how are we going to use and if we are going to use AI in all of these languages or not. Again, for some languages that are very small, we may want to use AI in certain ways and we may not want to use AI in some other ways. So thinking about what is actually the need of that language and is it useful to use AI, or maybe there are other pieces of technology that are going to be more useful, given where the project is at that point in time. The other principle that we have around multilingualism is about trying to not lock ourselves in a model that we cannot expand to more languages. So here the philosophy is being, maybe if today we can't use it in 10 more languages because these languages don't have need, let's make sure we don't lock ourselves because our principle is general, we should be able to go to more languages.

D (48:08)

That's wonderful. The model cards, I think in particular, one thing about, you know, Wikipedia in particular is this aspect, that kind of anything that you want to dig into and see how this was built, there's, there's the, I can't think of the word for it right now, but essentially it's all checkable, it's all able to be kind of out there in the open. And so the idea of this model card being an aspect of that is really interesting and I think important. And yeah, the transparency of it is I think going to be somewhat refreshing in comparison to perhaps some of the other ways we've seen AI used. This last question that I have for you, I am very curious to hear what, what your thoughts are. You know, Wikipedia's content is, as far as we can tell, often used to train large language models. How does this strategy of, you know, AI being included in the work that is taking place kind of position Wikipedia and the Wikimedia foundation in the evolving landscape of AI generated knowledge?

E (49:25)

Yeah, for me there are two aspects that I consider when I think about positioning of Wikipedia with regards to AI with the lens of this strategy. One is our focus on the local perspective. I think this is something that we have tried to become really clear about. This is something that we see perhaps less organization, less fewer companies or organizations currently investing in which is bringing in the local perspectives or the local knowledge of the communities from across the globe. And there are communities that are being left, left behind with the speed at which AI companies are operating and moving. So one way that we are thinking about this AI strategy is reaffirming the importance of this local knowledge and perspective and thinking about a Wikipedia that continues to be this model for bringing encyclopedic knowledge to the Internet for all of us, for whichever application we are looking at, whether it's training an AI model or getting an alert on our watch about what we should be, where we should be going next or what is the monument that is around us. The other aspect is really the human centered approach in this. I think this is again what is differentiating the work of Wikipedia from a lot of other content that is available on the Internet and web. In an Internet and web that is being constantly polluted right now with machine generated content. Having this place of human curated and cared for content is going to be critical for all these AI models that need to be constantly retrained and built. And in that way I think Wikipedia is going to play an important role of human curated, generated and cared for content for AI.

D (51:23)

Layla, I want to thank you so much again for taking the time to join us today to extend, explain kind of more about where the Wikimedia foundation is in terms of AI being used. I think that we've learned a lot today. Really excited to see these model cards, for example, and continue to watch this roll out if people want to stay up to date of kind of where the project is, what what they should be looking for. Do you have a place that they should go to do that that we.

E (51:55)

Could include for the model cards? Specifically, if you just search in your search engine of choice. Model Cards Machine Learning Team Wikimedia foundation. You will be landing there and you will see our model cards and from there you can be routed to other places.

D (52:11)

Beautiful. Thank you so much for your time today and we really appreciate it.

E (52:16)

Thank you so much for having me. Have a good rest of your day.

C (52:19)

Micah, you as well.

A (52:21)

At Capella University, learning online doesn't mean learning alone.

E (52:26)

You'll get support from people who care.

A (52:28)

About your success, like your enrollment specialist who gets to know you and the goals you'd like to achieve. You'll also get a designated academic coach who's with you throughout your entire program. Plus, career coaches are available to help you navigate your professional goals. A different future is closer than you think with Capella University.

E (52:49)

Learn more at capella.

D (52:51)

Edu hey Ryan Reynolds here for Mint Mobile.

F (52:53)

You know one of the perks about.

D (52:55)

Having four kids that you know about is actually getting a direct line to.

F (52:59)

The big man up north.

D (53:00)

And this year he wants you to know the best gift that you can.

F (53:03)

Give someone is the gift of Mint.

D (53:05)

Mobile's Unlimited Wireless for $15 a month. Now you don't even need to wrap it.

F (53:09)

Give it a try@mintmobile.com Switch a front.

A (53:13)

Page of $45 for three month plan equivalent to $15 per month required new customer offer for first three months only. Speed slow after 35 gigabytes if network's busy. Taxes and fees extra. See mint mobile.com thank you for being here.

D (53:23)

I know you are being asked by everyone across the world about your impressions and thoughts about the Switch 2. So I am honored that you chose to take some time to join us today. You know to kick things off you in writing this piece talk a lot about Mario Kart World as this was kind of the hands on you got to do spent hours with it hands on time with the Switch Switch 2 itself. For listeners who have yet to catch up. Can you talk about the key takeaways about what is new in this generation, what makes it the Switch two as opposed to the Switch.

A (53:55)

Yeah so they did a lot of under the hood stuff which is different and starting with the chip, it's a new Nvidia chip inside and there's a lot of claims about AI upscaling and improving graphics. It can run 4K and higher refresh rates on a TV. The system itself has a 7.9-inch 1080p screen. It's LCD, but that looks good. That can run higher refresh rates. The Joy Cons are redesigned so they have more rumble and they also magnetically snap on. But a lot of the stuff that's going on with this also it works with cameras in an interesting way, which I'll get to in a sec. It can split your faces up and put it into the game in a way that's pretty clever. And it adds this game chat feature, which is a live onboard 4 to 12 player audio video chat thing. And those are the new things. But you know, a lot of. A lot of those potentials are still kind of unknown. You know, right now we're right at the beginning. Nintendo has a couple of exclusive games, a lot of ports and updates to some games. And it feels a little more like a proier Switch than it does a completely new thing, like kind of by design.

F (55:13)

Understood.

D (55:14)

Now you did mention, and I think this was one of the most interesting things leading up to it, we saw like a new button right on the controller and then we heard about game chat. You describe it as Nintendo's audio video hangout zone. How does it actually work in practice? And I think more importantly, how big a shift is this from Nintendo's previous approach to online multiplayer and online multiplayer conversation?

A (55:40)

Yeah, Nintendo's always been pretty locked down on online stuff. And you know, it can be frustrating for people who are used to a deeply online gaming life. They've gotten a lot better in the Switch years. A lot of that chat stuff was always offloaded to the Nintendo phone app. And I think people kind of found their own. You know, you might have just done Twitch or something else if you wanted to chat. But now this is something that's built in and you have to connect through friend codes and invitations. So you have to still very deliberate thing where you're inviting people in and then inviting them into the chat. And it's pretty instant once you do that. It starts as audio. You can also add video if you have a camera. And it was pretty seamless in the demos we tried with Nintendo. I'm still getting on my feet with it now because we did not have early review units. We just picked them up on Wednesday. And you have to have other people to do game chat. Yeah, so we're just kind of getting our feet with that. But I think it's a good move. And what's interesting is that it is definitely designed to not have to be playing games together. It starts as just, hey, we're all chatting. Then you can do on. You can connect with the games you're playing and play. But it starts as kind of like a party group. And I think that's pretty common. I mean, I know with my kids that's kind of how they're playing all the time. They start with a bunch of chats and they're talking with friends and then they play. And I think it probably is pretty generational because it's not, you know, it's not what I'm doing. But, you know, I don't know how flexible it's going to be and how frustrating that might be for people. But and also was interesting is when I set up game chat, it asked for my phone number to authenticate. So I'm not quite sure what that means because I didn't know that that would be part of it. And then, you know, does that mean that every person who's doing game chat needs a phone number? In which case, what does that mean for kids?

C (57:32)

Yeah.

A (57:33)

Does it mean with the. I mean, there are kid accounts and it might mean that the parent can then authorize it, but I don't know is the answer. But I just thought that was a little bit of a surprise.

D (57:43)

Yeah, that actually, that really surprises me as well, I suppose from the aspect of maybe making it easier to track and ban accounts that are offenders, that could be a good feature. But yeah, if you've got multiple. Because I'm immediately thinking about when I needed to sign up and create accounts on any website and need to make more than one for whatever reason, work stuff or whatever. And the second account, I go to type in the phone number and it goes, well, you've already done that with a different one. So what if you have two kids? Is that going to all work with the same phone number? People going to have to get these. Yeah, that leaves a lot of questions to be answered. And before I ask you the next question, I want to follow up there. You said, I don't know, could you give a little bit more context about the situation? When it comes to the review of the switch to across the entire industry of, you know, journalists who are writing.

A (58:38)

About this thing, it's been very on the fly, you know, I think, you know, we didn't get early review units. Apparently there were some software updates that were needed. You know, we had early demos. We had a demo all the way back in April when they were doing the tour. And then we got another opportunity a week ago to spend more time with some of the key features and kind of get, kind of get explained some of the stuff that we did already know, but in a little more detail. But what we didn't know was how it would feel once you're setting it up and using it. So we're in a hot situation now where we're using it on the fly, just like a lot of people who bought them. I think it further reinforces to me that it's kind of a soft launch, you know, the Switch 2. It's. I mean, it's a big launch and that I think they're going to sell a ton of them. But this is something that Nintendo is going to be getting on its feet over the course of the next year because the Switch is already so popular that the Switch 2 doesn't need to be an instant hit. And a lot of people are probably going to wait a bit. And so as features roll out, as maybe kinks get worked out as more games arrive, I kind of expect it to be a rolling start. And that's how I'm treating it because right now, like, I don't know, for instance, I don't know how fast it is to charge. I was just sharing this on Blue sky, but it seems like it's charging slowly. And someone at the Verge chatted back and said, yeah, they saw the same thing. So all those questions, things that are important, we're learning. So it's going to take a good week. Usually I take a good week or more to review something, ideally, and, you know, I'm on day, day one and a half at the moment, while also.

D (60:18)

Coming on shows like this to talk about it. But the game time, one of the things you talked about, camera connected features, something kind of interesting, floating live video faces in Mario Kart world. I mean, that's, that's fun. But it does seem technically ambitious. What did it actually feel like using that, that feature in real gameplay?

A (60:40)

So it's kind of impressed upon me that the camera, which I was, I was feeling pretty dismissive of with the first Nintendo Switch event because they were showing it mainly with Mario Party Jamboree, with some fun but gimmicky things that felt like, you know, memories of PlayStation, I, Microsoft Kinect, all these things from the past that you might remember. And I thought, okay, you know, they're kind of doing that, but I think they're going to be doing a lot more with it. First of all, it works with the game chat and it can actually do some interesting kind of green Screeny stuff of your face and zooming in. But the camera enabled modes can separate out four different people at once from the same camera, which I am not, I'm not, I'm not studying camera tech all the time, but that felt like a magic trick even to me having seen a lot of this tech that felt pretty new. And, and I think it's a trick that the processor is able to pull off too, so you know, it will. You'll pre identify which faces to, to keep in frame. I'm not sure how good the tracking actually is, but at least it keeps it in frame. You may have to move your head in that zone, but the point is it will then cut out and live feed your faces into the game. So as you're driving the Mario Kart, you can look up ahead and I can see my colleague Ahmad all the way up in first place like I did yesterday. And it's live. It's like slow refresh rate a little bit. But you can taunt someone, make a face.

D (62:09)

That's super cool.

A (62:10)

It is cool. It also lets you know where they are too. Because there's a lot of times when you're playing these games like Smash Brothers or, or Mario Kart, where you're kind of like, where are they again?

D (62:20)

And yeah, who is who? Yeah, exactly.

A (62:24)

Now you're like, I'm gonna get him or no, I'm not, I'm gonna lose. But I think it added a lot to the fun and I think it, it, it really turned us into a like, oh, maybe I need to get the camera. So at 55 for their camera, you can also plug in your own USB C cameras apparently. But as someone was asking me, they wanted my thoughts on that and I haven't gotten there yet, so I don't have a lot of cameras lying around. That is an interesting question. How? Well, you could just plug in your own. But I think it's a really interesting feature.

D (62:55)

Yeah, absolutely. And what I love about Nintendo is the novel ideas that the company seems to come up with that are quirky. But then you go, you know what? The delight makes the quirkiness just work.

A (63:10)

You did.

D (63:12)

Speaking a little bit more about Mario Kart world, you did talk about playing with 24 people. That's pretty wild. What makes that scale of multiplayer so different and engaging versus playing against Princess Peach Robot?

A (63:30)

It feels like Mad Max. I felt like when we were playing with the 24 gaming journalists and tech journalists in that demo and you're playing in like the tracks feel pretty wide open too, to Accommodate those number, because that gets pretty. You get pretty bottlenecked. It just reminded me of like, Fury Road with like, all the cars going at once. And you're like, things are firing. And you're like, how do I get out of this? I think it changes the strategy at that scale. But also there's a new mode that I think really makes great use of it. That Knockout Tour, which a lot of people are saying is kind of like. Like, could have been the name. Someone said this online. They're like, that could have been the name for Mario Kart World. I think that's. That's true. Where kind of like a 99 player, you know, kind of survival mode. This is like every. Every race, every. Every one of the courses, they'll cut you off. You know, you have to make it to 18th place or you're out. You have to make it to whatever interest. Yeah. And it adds an extra pressure, but not like, oh, I have to get in first place right now, but I better not fall too far behind. And I think it works really well for that large, chaotic group. I think that's a really good fit because you lose track of. And also getting into first place feels almost impossible with 24 people. But you could try to be into, like a top group and then see how luck plays out. Once you get down to like eight, you know, maybe. Maybe you catch fire. So I think that's. I think it's great use of it. And I don't want to go back, you know, I. It hooked me on the large player size if I can find other people online.

D (65:14)

Yeah. Especially this early on. Right, Right. So let's. Let's round things out here. After two extended demo sessions, you still weren't fully convinced. You mentioned that the Switch 2 is a must buy right here at launch, that, you know, people need to run and go grab it as quickly as possible. What do you think would. To change or what software would need to arrive to tip the balance into the go grab it now situation? Or is it just a matter of let's just wait and see?

A (65:42)

I think it's like, also, it depends on the audience too. You know, I know a lot of people, like, I know a lot of dads that are running out and getting it anecdotally, you know, And I think, like, if you. If you have the money, if you have a large Switch library, if you like the idea of, again, the Pro appeal, where it's kind of like the PS5 Pro, where you're like, oh, I have a huge library and I think upgrading all these games in my library matters to me, but for somebody else who's thinking about spending money on this, it's significant amount of money. There are more expensive things in the world, but $450 is not cheap. And I think it needs to develop a bigger, unique library of games to justify that. They're leaning on the Switch library, but right now you just have Mario Kart World and this welcome tour little thing, which is cute, but should have been free, you know. But then again, it's only $10 if you. It's like a tour of the Switch 2 and its features and multiplayer. I think if you're really starved for extra things to show off, the mouse functions like, it's probably worth throwing $10 at it. But it feels like Nintendo was trying to just, like, get a few extra bucks out of you. And I think. I think that some of the ports, the new ports, like Cyberpunk 2077, which I'm still going to be getting into playing, they show a lot of potential for where it could be like a Steam deck and play more things like PC and PS5 and Xbox. But we have to kind of see how that plays out, see how really capable it is. And I want more Nintendo whimsy, actually. So, speaking of, I was a little let down by the whimsy approach. I think I was expecting. Even I wrote a piece saying I was expecting even more wild cards. I know they come out of left field with things those may still come. Sometimes Nintendo gets burned on those. Like Labo, the cardboard folding thing they did a number of years ago, which I still think was genius, but was probably too ambitious. And then they have some other interesting accessories and things. And while the mouse feature is fun, it's a mouse. Like, you can use the joy cons, it's okay. And the camera thing is fine. But again, it doesn't work when you're in handheld mode, so that's like you have to be docked in front of a tv. I'm mostly a handheld player, so I really like Switch in that mode, so I think it needs. I'm looking forward to a little more. Also, indie game from Nintendo Whimsy. Nintendo often would take shots in different directions with games. You go, what the heck is this? And right now they're leaning on some pretty safe franchise moves. You know, Donkey Kong, Mario Kart, more of the things you like. Like, I. I think they need to take some swings in, like, what is this? What is this, you know, crazy frog thing that I don't know, like, you surprises like Pikmin, when Pikmin first came out around Animal Crossing. You know, I think that'll come. But it does seem like right now in 2025, Nintendo is looking for smooth continuity with what they've already got.

D (68:40)

Understood. Well, we will continue to watch what, what, what makes its way out there as everybody at the same time is getting their hands on this. And as you mentioned, you know, kind of figuring out what people, what issues people are running into. If people want to keep up with what you're doing, where should they go to do that?

A (68:59)

You can follow me on Blue Sky. I'm on there a ton. You can also check out all my stuff on CNET, which I am always on and on CNET's YouTube channel. And those are pretty good places to start.

D (69:12)

Beautiful. Thank you so much, Scott, for joining us today. And I'm sure we'll see you again soon.

A (69:16)

Thanks a lot. Yeah, I'll be playing.

D (69:21)

All right, we are back from the break and I have to tell you, I am very excited to be talking to our next guest about a topic that I think is incredibly important and fascinating and perhaps a little bit frustrating as well. Joining us today is Derek Kravitz of Consumer Reports. Welcome to the show, Derek.

F (69:40)

Yeah, thanks for having me.

D (69:41)

Yeah. So this is a story that I came across about Instacart. And I have to tell you, I am an Instacart user, have been for some time. And seeing this and seeing the undertaking that you all went under is pretty wild. I wanted to ask, as we get into kind of the differences in pricing that you saw in this study, first and foremost, I would just like to know, what was it that actually led the team across Consumer Reports and the rest of the group who worked on this, what was it that led them and you to investigate Instacart's pricing practices in the first place? Because I don't think I would have necessarily thought about what you all found. What made you suspect that shoppers might be seeing different prices for the same item items?

F (70:34)

Yeah. So it all came about in about six months ago. We had just published an investigation about Kroger, a big grocery chain. Most U.S. states have, you know, King Supers or the Kroger eponymous grocery chain in most US States and territories. We had just published an investigation about pricing practices there. Right, right. And so paper labels being out of date or expired and them reaping the profits from that, that's a long standing thing that happens at grocery stores. Thousands of labels to switch out paper and plastic labels and it can take a long time to do that. And sometimes they get out of date. And because you can make money from that practice, a lot of companies are loathe to, to fix it or correct it. We did a second investigation looking at how Kroger also collects a lot of personal data about you, so who you are and through their loyalty program. And they drive most purchases through the free loyalty program. And so the data that they collect was, you know, a lot. Name, address, phone number, email, all the, you know, device id, all the things that companies collect, collect. But then they make inferences about you, Right. So how likely you are to be in the market for a new or used car, how likely you are to be ready for a domestic or international vacation, how likely you are to, you know, have a pet dog or a cat. Very basic things. And, you know, it was surprising because the inferences that we saw were largely incorrect. Correct. So we share them back with the people that requested them for us to see what they had on them. And the inferences were wrong in many cases. And we did a story about that. So then fast forward, we do this big policy roundtable in D.C. where we bring in experts and lawyers and advocates and all these other folks that know a lot about this issue. And one of them, Kitty Wells over at Groundwork Collaborative, came up to us and said, you know what? We should look at Instacart. The same stuff you're seeing at Kroger, the same things that you're trying to highlight. It's happening there and we can prove it, but we need a lot of people to do it. We need hundreds, and we need it very tightly controlled, all organized, all sort of well thought out and planned. But we could, we could figure it out, but we need, and we also need time, we need months to do this.

D (73:12)

So we were like, yeah, this, this is the part that is, I think, really cool to me that you. And this is my next question. I, I would love to, I would love for the audience to hear about the process of figuring out how to go about testing this, how to then conduct the study, the mechanics of getting so many shoppers to do what you're, what you have them do. I'm kind of trying not to spoil it because I just think it's so cool. Yeah, tell us about what all was involved in what feels like a bit of a logistical nightmare.

F (73:47)

Yeah, I mean, it's sort of like a cool participatory, volunteer led project where you. We asked all of our subscribers and members, would you be interested in joining us for live video experiments and if so, respond and join us during one the of our calls. So we did, and, you know, we developed a run a show, sort of a very tightly controlled, again, plan in. In which they would open up Instacart or download the app for the first time and then start putting items, specific items, the same items in their carts, all from one new location that we gave them, setting their home address for one location and doing it at a specific store. And all of us were doing it together at the same time. So the idea behind all of that is to try to remove all the other factors that might present, you know, or make the data a little bit more noisy.

C (74:44)

Right.

F (74:45)

We wanted something that was clear and definitive. And so we did that with 437 people over four sessions. Different areas of the country we targeted. So we did an experiment in Ohio. We did one in Washington, D.C. we did one in Seattle, Washington, St. Paul, Minnesota, and found same things. Right. So we, we looked at Target, we looked at Safeway, then we did a confirmation test and looked at 15 different other retailers. And we saw signals where people were being sorted into specific price buckets and very, very specific down to the cent. And for whatever reason, these folks were being placed into, you know, groups that all got the same price for the same exact items at the same exact time. And it was very, almost even distributions among some buckets. And for us, that raised a lot of questions or alarm bells. Why are we seeing this? Why isn't it more diffuse? Why isn't everyone getting a little bit, you know, either. Different prices are all the same, right?

D (75:56)

Yeah.

F (75:57)

And when we looked at one place, Snook Markets, everyone got the same price. No noise, no. No variance, no nothing else. But then when we looked at and same thing for Sam's Club and a few others, but then we looked at Target, Costco, Kroger, Safeway, Albertsons, all tons of noise, tons of signals everywhere. And, you know, for. From a data perspective, from. From like an investigative journalism perspective, that's exciting because then you're like, oh, okay, well, this. We have something in hand. Now we got to find the answer. Now we got to find out what this all actually means.

D (76:34)

Yeah. And that, I think, is a fascinating point that I want to kind of focus in on, which is it's important for people to understand that you had them all put in a specific address, which then, as you said, kind of took out that factor, made that, that factor a control. And so instead it was based on some other set of data points and trying to Figure out what those are. The study found 74% of grocery items had different prices for different shoppers, priced up to 23% higher than others. Can you talk a little bit more about the price variations and like what those actually looked like for specific products? If there's sort of a category or a set of categories where you saw this more than I think that's especially helpful for people to kind of latch onto. Oh, so you're charging me this for eggs and them that for eggs. What were the differences that you saw in that way?

F (77:35)

Yeah, so if you check out our website, this is actually on the homepage, so consumerreports.org if you go a little bit down the page, you'll see the story. About halfway through the story we have this, what they call a scrolly telling graphic where you can you scroll down and it sort of text appears and new visuals appear. And it really lays it out in a way that sometimes text or even audio or video can't really do clearly. But we basically highlight a particular test, right? In this case it's Seattle and we're looking at a Safeway. And this is September and we're all shopping for the same 20 products. And that's apples and baby carrots and Cheerios and corn flakes and eggs, 12 pack of eggs, Wheat thins, Heinz ketchup, Skippy peanut butter, Clif bars, you know, Ruffles, potato chips. All these things that Americans really like to buy. Both name brands and then some also some in house store brands. So the, the Safeway select, you know, brands which are supposed to be a little bit cheaper, right. And in that test we saw people with five different basket totals, right, ranging from about $114 all the way up to $124. Might not seem like a lot $10. But if you extrapolate that over the course of a year, if you're buying this every week, that's real money. We found maybe $1,200 for a family of four. And so that's about a 8% difference, right. When you do the percentages. So when you break out the, the actual products, right. So you're looking at each individual item in that cart. Some products had no variants, right. So for whatever reason, Instacart and, and the retailer Safeway have determined in this particular case, let's not experiment with the prices, let's leave them be, right? Premium saltines, Heinz ketchup, Barilla for Folly Pasta, no variant variants, right. So then you go to the middle tier. So suddenly you have, you know, two Price buckets. For whatever reason, these products have two different prices. And some people, a lot of people get the higher price and a few select people get the lower price. Cheerios, ruffles, potato chips, Lucerne eggs, all had about an 8 or 9% difference. Then there were products that had a ton of variants and multiple different price points. Just using an example, Skippy peanut butter, really popular, 23.4% variance. The high, some of the highest variances we saw a lot of people at 299, smaller group of people at 323, a slightly larger group of people at 349, similar group at 369. Right. And no pattern necessarily as to whether they're loyalty members, particular area of the country, whether they're wealthy, whether they're, you know, race, gender, nothing else mattered. They want to test the prices to see how people would react. Right. And it's all a big price experiment. And that data informs then future price. Right. And Safeway can take that data and set their own in store prices using these numbers. And it's incredibly valuable data. It means hundreds of millions of dollars in additional revenue for these companies. So that's why they do it.

D (81:02)

Wow. So yeah, you in the study and the reports talk about how Instacart acquired a company called Eversight in 2022. It enabled this kind of AI powered price experimentation as part of what the app could do. The team talked to the company what, or you know, at least tried to understand what the, what the companies, what it was hoping to do with this. Can you talk about what they say the technology is designed to do? And then how does that compare to what your research actually uncovered?

F (81:37)

Yeah, so, you know, Instacart's pretty upfront about this. You know, they say to online and to business clients and participating grocery retailers, hey, we have this tool, tool, it's called Eversight. It is a product, software product that we offer to you retailers or, or consumer, packaged good companies, brands, craft, Heinz, those places, if you take this product, it'll help you price goods. Right? And we'll be able to learn from these experiments and, and help you make more money. And you know, they also, on the flip side, they told us, hey, look, this also helps with affordability in some cases, you know, if a retailer knows an ideal price point, the willingness to pay for someone, they might lower a price in order to sell more. Right. Supply and demand. And so, you know, it can work out to your benefit, but at the end of the day, it's a service for retailers and brands to Figure out these ideal perfect high price points on individual products. And Instacart told us, yes, this is happening. Yes, your tests reflect what we're doing. You know, some difference in, in terms of, you know, who's doing it. They told us 10 retailers are doing it. We want to know which 10 retailers, of course, because that's like a, a follow up question you would ask, yeah, where is this happening? They wouldn't tell us the 10. So of course, as journalists and researchers, we want to know that. So we did our own confirmation test with as many retailers as we could, trying to figure out who the 10 were. Sort of like a game of Battleship. And we did, we found out, you know, six of them. Kroger, Albertsons, Safeway, Costco, Target. And we went to all of them and we asked them what's going on and they said, hey, yeah, we're sure we're doing this in, in some cases, or we don't really want to, you know, comment about it or, or talk more about it. But you know, we're, we're partners. In one case, Target, they said, no, we don't have a business relationship at all with Instacart. So what you're seeing, we think is them scraping our data from publicly available websites and then putting an extra markup on it in order to offset their technological and operational costs is how they describe it. Basically a markup, right. A little bit of extra on top of the target prices. And we went back to Instacart and said, hey, you say only 10 partners do this. Target saying they do it and they're not a partner. So what say you? And they said, oh, okay, yeah, they do do it and we do do it there and we do have these price experiments there. But they then ended, that's over with. So that, no, that's not happening anymore. And we do, we, we scrape the website and we, we offset it with an additional charge. So for us, that's of course, as journalists and researchers, that's okay, that's interesting. We want to know more. So that's where we sort of landed when the story published. And, but we're still investigating, we're still trying to figure out, you know, unanswered unknowns, you know, things that, that we see in the data that, you know, don't really have an explanation.

D (85:05)

Yeah, that's the, the big thing, right, is that as you point out, there doesn't seem to be any. Because one would think, when I first read this report, I was thinking myself and how especially around the holidays, the different holidays that involve making things, that's when I really like to use it because then it means I don't have to go into the very crowded store. And so then I was thinking, thinking I've got some big bills on Instagram or Instagram Instacart. And so then I'm going because I have had these instances where I spent a lot of money versus what you would maybe normally do if you're just doing your weekend or your week shopping. I wonder if then they're up charging me because of that. But yeah, as you point out, it seems like there's no tie to it and they're not being very clear about that. I guess then, because I want everybody to go and read this report so we won't get too much more into it. But I would love to hear your take in terms of then impact on. Let's talk about the real world impact of what this does and especially given that there's such a randomness to it as far as what you've seen thus far. What do people do? Is there anything that they can do when it comes to this? Or is it just a matter of, of the awareness that if you're using this service, you may be getting a different price from what someone else gets.

F (86:36)

So two things worth talking about on this one. One is that one thing we can sort of tell from the data is that shopping history and buyer behavior is really important to these companies. Every economist we speak speak, we spoke to every grocery industry consultant we spoke to. They all say the same thing. That your shopping history and your behavior, you're more likely to do an impulse purchase for, you know, an item at Friday at 5pm and also maybe get a, you know, some alcohol and maybe some, you know, chips and various other things all at the same time, you know, on that particular day of the week. That behavior really informs sort of how these companies view you as a customer. And we went through patents, patents. In the US you have to file a patent whenever you have new tech or a product that you are offering. And you want to protect that intellectual property, right? That's yours and you don't want anyone else to steal it. So we pulled the patents from Instacart and Eversight, the company. They acquired about 302 patents I believe exactly over the course of the last decade. And in a lot of those patents it's very clear what they're doing. They're grouping people in what they call sub populations in order to figure out, you know, this group of shoppers responds to this type of promotion or discount in this particular way. So let's group them together and offer them this price. And this group of shoppers responds to this particular price in this way. So that is the takeaway, right? So it's invisible to us. We can't really tell which group we're in. But we do know that's why. And we do know that again, shopping history, buyer behavior, those are the best leading indicators. It's not demographics, it's not your wealth, it's not, you know, race or ethnicity. It's really, you know, how you spend your money and the factors that go into a particular type of purchase. And so the big takeaway for people to maybe learn from this is even if you don't use Instacart, their data, they sell it to everyone. They sell it to a lot of big grocery retailers and also food companies, brands that sell, that create the products and then sell them in the grocery store. And this type of data is unique because they're essentially creating the market in economic terms. They're sort of creating the data out of thin air. And then they're selling that data back to companies. Then those companies can use that data to, you know, set their prices and offer promotions and discounts. So it's, it's incredibly sophisticated. It's really novel. When we talk, spoke to regulators, Lina Khan and others, they're like, you know, we need to have a first order conversation is what she's told us. Do we even want this as a thing? Should we allow this type of price experimentation or dynamic pricing or optimization, whatever phrase you want to call it, should we allow this at all to happen? And should we regulate at the federal level or this whole hodgepodge of state laws that might come up around it? And so that's the open question. Regulators are playing catch up, so they don't know how best to deal with this. But that's what consumers are now faced with.

D (90:14)

Wow. I'm just always in awe of Consumer Reports and the work that you all do on our behalf. Thank you so much for taking the time to join us today, Derek, to talk about this. And yeah, everybody will of course have it linked in the show Notes. Go check out this report and many more over on the site. If people would like to stay up to date with the work that you're doing, where are the places online they should go to do so?

F (90:41)

Yeah. ConsumerReports.org Great, great spot for ratings and reviews and also investigative journalism. We do a lot of research, things that we can't quite figure out through a rating or review. We need a lot more time, a lot more patience and resources behind that. All of our investigative journalism is free, not paywalled. So check it out on our website.

D (91:05)

Awesome. Thank you.

F (91:06)

We appreciate it. Thank you.

D (91:08)

I just want to quickly say thank you so much for a great year. I really appreciate the support. Your downloads, your subscriptions, your club twit joins all of that means the world to us here on the network work. And thank you so much for what you've done to make this year's Tech News Weekly possible. Looking forward to bringing you even more even better in the new year. Catch you then.

A (91:33)

Bye bye.

D (91:45)

Limu Gate and Doug, Here we have the Limu Imu in its natural habitat, helping people customize their car insurance and save hundreds with Liberty Mutual.

F (91:58)

Fascinating.

D (91:59)

It's accompanied by his natural ally, Doug.

A (92:03)

Limu is that guy with the binoculars watching us.

F (92:05)

Cut the camera.

D (92:06)

They see us. Only pay for what you need@libertymutual.com Liberty Liberty Liberty Liberty Savings vary underwritten by Liberty Mutual Insurance Company and affiliates. Excludes Massachusetts.