Inside DuckDuckGo

DuckDuckGo

Behind the scenes with the DuckDuckGo team — sharing insights on product, engineering, leadership, and AI. insideduckduckgo.substack.com

  1. Duck Tales: DuckDuckGo donations — why we’ve donated $8M+ to organizations that align with our vision (Episode 13)

    DEC 17

    Duck Tales: DuckDuckGo donations — why we’ve donated $8M+ to organizations that align with our vision (Episode 13)

    In this episode, Gabriel (Founder) and Joe (Policy team) discuss why we donate, the types of organizations we donate to, and some examples of impact. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Show notes: See our full list of donations here. Gabriel: Hello everybody, welcome back to DuckTales. I am Gabriel, the founder of DuckDuckGo. Today I have with me Joe, and we’re gonna talk about donations, but you want to start introducing Joe. Joe: Thanks Gabriel. ⁓ So I am the senior public policy manager for DuckDuckGo. ⁓ What does that mean? ⁓ It sort of means that ⁓ I’m sort of the person in DC who sort of tilts at windmills, talks to Congress, works with lawmakers, and otherwise tries to keep the rest of the company abreast of how the US government and states are ⁓ pushing different types of regulations, solving online privacy. ⁓ and trying to make the internet a better place for everybody. Gabriel: Cool, yeah, and there’s probably a lot of good feature episodes in all that, which is the main part of your A small of your job is you took over really running the process in which we make our donations every year. I realize it’s a small part of your job, but it’s important one, and I think important to our users, so you want to explain just kind of what that is? Joe: Yeah, no, and look, I guess I shouldn’t oversell it, but I actually think it’s a really impactful part of the job. ⁓ It’s really, I think, like, it’s both, and we can talk about this, humbling and satisfying to be able to sort of allocate money to causes that are out there to try and improve trust online. ⁓ I’m our, I guess our DRI, our directly responsible individual for our corporate donations. ⁓ And this has actually become a pretty elaborate internal process to look at a whole bunch of different organizations. Now, I already mentioned I’m in DC, so I think about civil society groups, the Electronic Frontier Foundations of the world. ⁓ But we actually, you know, we give to a whole lot of different entities and organizations, ⁓ you know, sort of open source technology groups, ⁓ online technology reporters. and other organizations that do really impactful journalism on technology and data privacy. ⁓ And so, you know, I think there’s always sort of a push and pull to try and convince you to give us more money to give out each year. ⁓ But, you know, we give out, you know, over a million dollars to, I think this year it was something like 29 different organizations. ⁓ And it’s a, you know, it’s a detailed process. We spend a lot of months just arguing amongst ourselves about how we should allocate that money. And we’re, you know, we have a bunch of different criteria for what goes into this. ⁓ But, you know, we’re trying to both provide impactful donations. And so that means we give to a lot of small groups. ⁓ We’re also trying to sort of ensure that, you know, these groups are not just like aligned with us on one or two things, but are really out there trying to make the internet a better place. I mean, you know, if anybody’s been watching DuckTales, I think everybody would realize that DuckTales Go is a pretty mission driven company. And we’ve got this goal of expanding trust online. And we can’t do that ourselves. And there are a lot of different entities out there in the United States and globally ⁓ that are playing a really important role making the internet more trustworthy. And so we’re constantly trying to find ⁓ new voices to elevate and new projects to support. Gabriel: Yeah, as completely slightly a tangent, but you mentioned it because this is inside dark echo. At dark echo, we have this concept of the DRI directly responsible individual. What that means is someone who owns something and. We have, as you might imagine, tons of processes internally, one of them being this art under nation’s process and every one of them has an owner. Um, and so, yeah, that’s just a little insight in baseball for us, but hundreds of those and Joe owns this process and it often gets handed, you know, over time that changes ownership, but, um, that really means taking ownership of it and kind of, um, driving it forward and seeing it to completion. Um, Joe: Thank you. Gabriel: But yeah, donations, I also see it as extremely key to our vision. So we actually started doing this a long time ago. ⁓ I think I have the, for anyone who wants to look at everything we’ve donated to, we have a page, ducktogo.com slash donations. Joe: Good, it’s good that you got the microsite out there, that’s important. Gabriel: I just went to it. Yeah. Yeah. Joe: Yeah Gabriel: lists, it literally lists everything we’ve donated to, you know, including this year, 2020, all the way back to, guess, the first year we did this was 2011, which is a long time ago. What I was going to say is directly related to what you said is that was right around the time when we, 2011 was the first year we had our first employee. It was me just before that. know, Cade came on and when we did that, that’s when we started laying out our vision, like explicitly our mission and vision. Um, and we, the vision in particular, rates of standard of trust online hasn’t changed since, and it’s not going to change, but it was, it was kind of tied to that. We always said, you know, We’re one company. I think it’s really like what you said. We can only do so much on our own. We have a much broader vision. How can we push that forward? Well, we can donate to other places that could really help. Yeah, go ahead. Sorry. Yeah. Yeah, I think it was in service of, to another one of your points, of making a bigger impact. Joe: Okay. Gabriel: And so like my original thought was the dollar amounts at the very beginning were very small because we were very small and had no real revenues or profits. So there wasn’t ⁓ much to give, but we figured out, you know, we still put aside money to try to do so. ⁓ And so when we were doing that, and I think some of this is carried through to today, we were really trying to find organizations that could really do a lot with a small donation, you know, like, ⁓ and that might’ve been. project where an extra few thousand dollars could really help. And when it came to now, like maybe we give 25,000 or something and that enables somebody to ⁓ spin up a project or create a, someone part-time to on something. I think those are the kinds of things that can really make an impact. ⁓ I think the reason for doing the thematic ones early on is, you know, just concentrating the small amount of money we have into one thing and hoping to make an actual on that thing. ⁓ Now the dollar amounts are greater but we’re still picking themes it’s just probably three or four themes broadly in donations. I mean that’s the way I look at it. Yeah. Joe: Yeah, yeah. mean, I think now we pretty much are settled on, I would say, open technology or improving internet technology efforts, ⁓ data privacy. And then over the past few years, as we’ve been increasingly concerned about fair competition in digital markets, that’s become a bigger chuck. ⁓ But those are sort of the three buckets that think we’ve sort of narrowed in on, at least since I’ve been in charge of this. Gabriel: Yeah, and the process has evolved. Why don’t you take us through kind of the current bones of the process as it stands today. Joe: Yeah, that’s a good question. And I think actually a good opportunity to encourage folks to reach out if they want. you know, we like, there’s a couple of us internally that try to do a, some time over the year just sort of monitoring what groups are working on, ⁓ what they’re doing, whether it’s, ⁓ you know, basically sort of a subjective assessment of how impactful they are. ⁓ And then we have, I think it’s usually about a three month process that we kick off in you know, early summer to just sort of figure out like generally what are we interested in supporting over the year. And I think things that I think it’s worth highlighting that, you know, A, we’re a pretty community driven company. We’re also a team member driven company. I think it’s been really rewarding to have just ⁓ colleagues suggest, you know, organizations I had never heard of and that gets in there. So, you know, We are very much open to new ideas ⁓ and we support organizations globally, so I’m sure there’s plenty of things we have not heard of. So we create like a giant list of organizations and then we put it into a little bit of a rubric ⁓ and basically ask a number of questions about what we think the organization can do in the next year, ⁓ what actually would be impactful. Your point is totally valid. We’ve given money to support things like organizations being able to train up an employee. you know, think as you well know, like going from one to two employees can be hugely impactful for certain organizations. We’ve also, ⁓ you know, our support has allowed certain organizations to offer like health insurance to employees, which is not the, you know, really sort of makes you realize how, you know, some money can go a long way in things that are not just related to core, ⁓ you know, tech. projects, but literally people’s day-to-day well-being. ⁓ So we have this sort of rubric involved, and then we narrow it down. So part of the issue is we’d like to have a diverse, both geographic and ideologically diverse group of organizations. ⁓ We also sort of, at this point, have standardized our donation buckets a little bit. ⁓ we also like to keep a similar number of organizations. ⁓ You know, we started with, I think, six maybe, and we’re up to 29. And it becomes, and I think I’m one of these people,

    20 min
  2. Duck Tales: Hiring at DuckDuckGo, and why we have a 95% employee retention rate (Episode 12)

    DEC 10

    Duck Tales: Hiring at DuckDuckGo, and why we have a 95% employee retention rate (Episode 12)

    In this episode, Beah (SVP, Product) and Zbig (Director, Talent) discuss our approach to hiring, and how it’s designed to reflect our unique, cross-functional and mostly async ways of working. Show notes: Check out our careers page and open positions here. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Beah: Hello. Hi, everyone. Welcome to DuckTales, ⁓ where we go behind the scenes at DuckDuckGo and discuss the stories, technology, and people that help build privacy tools for everyone. In each episode, you’ll hear from employees about our vision, our product updates, our approach to AI, or how we operate as a company, which is the zone that today’s episode will fall in. ⁓ Zbig: Okay. Beah: introduce myself briefly. I’m Beah Burger-Lenahan. I lead the product team here at DuckDuckGo. And we have with me Zbig. Hey Zbig, do want to introduce yourself? Zbig: Yeah, of course. Hi, Beah. Hi, everybody. I’m Zbig. I’m on the Talent Acquisition team. I’ve been around for almost seven years now at DuckDuckGo, and I feel very privileged to have seen this amazing journey of us scaling the team from about 50 to over 350, where we are right now and I’m super excited to be here and talk about hiring. Beah: Awesome, thanks, the big. Yeah, so today we’re gonna talk about how we hire, what that process looks like, why, and so forth. Hopefully it’ll be interesting to everyone. ⁓ So maybe just to get started, can you kind of lay out at like the highest levels of our approach to hiring, what that process looks like? Zbig: course. Yeah. So I think at a, at a highest level, um, there is, you know, lot of intentionality and discipline to how we hire. So we would do that only when we have like a really clear and well articulated need. So when we’re scoping roles, right? Like we would want to deeply understand what the actual tangible problems are that we want that particular role to solve for us. And then we designed the hiring process to test for the skills, the competencies that are critical to being able to do that. So in practice, that means that we base that hiring process ⁓ largely on test projects ⁓ rather than interviews. Well, there are usually a couple of interviews in every process, but we definitely attach way more weight to test projects. We also tried to design that hiring process in way that ⁓ it’s truly like a two-way street and allows Canada to discover how we work. So, you know, they can make a well-informed decision at the end about whether or not this is good place for them. And I think one other thing worth highlighting ⁓ that comes out of this intentionality and discipline is the fact that, you know, never in the company history, we were forced to do any group layoffs. And when you look at the tech market these days, that’s very rare. Beah: Yeah, because we only hire people that we know we really need. Zbig: that will really advance the goals of the company, right? Beah: Yeah, got it. Okay, thanks. So let’s talk about the projects since they are, as you said, kind of the core of our process. Are these projects, do we expect people to do them for free? Zbig: Sure. Yeah. No, we offer payment, which is kind of like average bubbly, but it’s usually like anything from 50 to 100 US dollars per an estimated hour of work required to complete a project. These are, and this is just like recognize the time candidates put into those. These projects are always role specific and designed to simulate the actual work one would be doing on the job. You can complete them async on your own time. And I truly believe that allows candidates to demonstrate the best of their abilities without the typical pressure and stress of interviews. And also like when you think of the async format, that much more closely reflects how we work on a daily basis at DuckDuck. Because I think that probably like, I don’t know, 70, 80 % of collaboration here happens asynchronously. There are a few meetings throughout the week. And ⁓ we’ve picked this format for the hiring process very intentionally. There’s actual research behind that that proves that work samples, test projects that are role specific, they’re much less prone to bias and error than interviews. And what we found through that ⁓ is that they are great predictors of future success on the job. And when you look at our retention rate, that’s 95%. You know, our engagement scores are also very high, like I think 86 % way above market benchmarks for companies of our size and at our stage of growth. like there’s actual, you know, quantitative validation that this approach works very well for us. Beah: Yeah, is it challenging to get people to commit to doing these projects? It can be a bit of time. Zbig: It is one of the bigger challenges of our hiring process because these test projects can take anything from like three, four hours up to 15 and in some cases maybe 20. And we typically expect folks to complete two of those, two sometimes three. So that’s like a time investment of about 25, 30 hours in test projects. I think we do our best to really to adjust to Canada’s availability. There’s like from our side, there’s no pressure on like when exactly they should be starting on those. They can plan ahead, set that time aside. If they need an extension, they can just let us know. And we usually have no problem to grant that because we understand that different things can come up in life that could derail their plans. So So we really try to adjust our pacing to candidates, be very flexible with those timeframes to ⁓ make it feasible for them. Beah: Got it. I’m shifting gears a little bit. So where do we hire in the world and why? Zbig: Hmm. So we hire globally though, not everywhere. We have like a selection of countries where that we’ve decided scale to based on like ⁓ a pretty nuanced estimate on the available town pools, how those relate to our needs, ⁓ how easy it is also to like scale within them from the administrative financial logistics perspective, perspectives, but we are able to hire across tens of countries around the the globe. And I think that’s awesome because that gives us access to amazing talent ⁓ that often when you think of some of these countries, there aren’t that many opportunities to work on something of a comparable scale that Go offers. So we’re way more competitive there, right? And we’re truly... I’m actually in Krakow in Poland and we have a pretty pretty big contingent here, like I think one of the fourth, fifth biggest representations geographically across the company with over 20 people. yeah, there’s definitely a great advantage of this remote setup that enables that. ⁓ And I think also what makes it possible is the fact that as a company, we’re designed from ground up for full remote collaboration, right? When you think of the companies that were switching to remote during the pandemic, that often didn’t work out and many of them are now calling people to go back to the office because they didn’t really have the right processes and culture to enable effective collaboration. I think that’s definitely not the case here. Beah: Where are you located today? Zbig: Beah: Yeah. Yeah. I will say, mean, the remote ⁓ employee base has challenges for sure, and it has a lot of upsides as well. I think for me, like one kind of just fun perk of it is that it’s cool to know people from all over the world. ⁓ Zbig: yeah, for sure. Beah: you know, when I think about like traveling I think, ooh, like who can I go see? Zbig: Yeah. And also when you think of like, we’re building a global product or a set of global products, right? And I think it’s invaluable to have people from different geographies representing different cultures and perspective, because that helps us inform like how we can be building those products for a broader audience that really ⁓ meets their needs and solves jobs for them. Beah: Yeah, makes sense. So you mentioned ⁓ one of the challenges of our hiring process is getting people to commit the time to the projects. Are there any other significant challenges? Zbig: Mm. Yeah. I think the probably the biggest one is related to how we work. Cause as you know, we’re pretty uniquely organized over here. There’s no middle management. There’s no like separate project management function. And in practice, that means that everybody really is expected to be able to scope, propose and execute projects. Sometimes that means, you know, managing a cross-functional project team and that skillset is not that easy to get on top of the functional expertise, Like functional competencies. So ⁓ we often end up hiring, you know, we do most of our hiring in engineering and at a senior individual contributor level, like senior engineer, and we often end up hiring folks, you know, performing these like more senior leadership related responsibilities elsewhere, sometimes even like holding more senior titles than what, we have on the job description. Disclaimer though, don’t use job titles internally. We don’t want them to get out of picking the best solution or going in the most optimal route. We try not to make decisions based on authority. And that’s worked well for us, I think, over the years. Beah: . . I’m curious. I’ve never asked you this question or I don’t know the answer. Do candidates find it appealing that we don’t use titles internally or does it put some people off? Zbig: Thank I think many of them, I think many of them do, and they do highlight, well, we get a lot of that feedback from candidates in the hiring process, but some of them do raise it as an objection. And I get it. you know, I think typically on the market, there’s a lot of weight attached to job titles, right? Bec

    20 min
  3. Duck Tales: How DuckDuckGo makes the sites you visit less annoying and more private (Episode 11)

    DEC 3

    Duck Tales: How DuckDuckGo makes the sites you visit less annoying and more private (Episode 11)

    In this episode, Beah (Product) and Max (Frontend) discuss cookie pop-up protection, why our solution is uniquely private, and the feedback loops we use to help us reject cookies across more of the sites you visit. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Beah: Hello and welcome to DuckTales where we go behind the scenes with DuckDuckGo and discuss the stories, technology and people that help build privacy tools for everyone. In each episode, you’ll hear from employees ⁓ about our vision, product updates and our approach to AI or how we operate as a company. In this case, today we’re going to be talking about a feature that I dearly love. ⁓ cookie pop-up protection with ⁓ Max here. So let me just do some quick introductions, I guess, before I’m kind of getting a little ahead of myself. I’m Bea Berger-Lenahan. I lead the product team here at Tech Tech Go. And I’m going to be asking Max a few questions. Max, would you like to introduce yourself? Max: Yeah, sure. Hey, ⁓ my name is Max. I am an engineer in the front-end team at DuckDuckGo. Been here for about three years, a little more. Yeah, I’m excited to talk about cookie pop-up protection. Beah: Awesome. Thank you, Max. We’re glad you’re here. I’m glad you’re here. ⁓ So first, just tell me, tell all of us a little bit about what cookie pop-up protection is, how it works. Max: Yeah, so this is the ⁓ feature in our browsers that handles cookie pop-up for you. ⁓ And in a nutshell, it... ⁓ that’s a good question. ⁓ I mean, I think most people have seen a cookie pop-up, but yeah, the definitions vary, but we’re talking about these... Beah: What’s a cookie pop-up first? Hahaha Max: dialogues that websites show you on the first visit that typically tell you something about their data sharing practices and the use of cookies and similar technologies. And sometimes they give you a way to opt out of some optional tracking ⁓ or cookies. And that’s what we’re actually doing. We’re automating, ⁓ basically clicking reject buttons for you or whatever it takes to...toggle all these little checkboxes and saving the settings. ⁓ I could demo it if that’s okay. ⁓ So let me share my screen. ⁓ Beah: That’d be great. Max: So for the sake of the demo, I’ve disabled the feature in the settings right now. It’s enabled by default, ⁓ but I’m just going to show you. ⁓ So if we go to Sky Scanner, for example, and I’m in the Netherlands, so you see a Dutch version, but there is this huge cookie pop-up ⁓ when you load the page. And if I enable the feature, cookie pop-up protection and reload the page, you’re not gonna see this pop up anymore. And what happened, and then there will be a ⁓ little notification in the address bar. And if you drill down, you’ll see the explanation Beah: Okay. Okay. Max: what happened. But basically what happened behind the scenes is we clicked on the reject button rejecting the cookies automatically. And that’s why we call it cookie pop-up protection. ⁓ So for us, this is a privacy protection feature because it actually ⁓ chooses the most private option for you, which is not always easy. Let’s see. Beah: Mm-hmm. Yeah, I mean, I don’t know if we have data on this, but I imagine very few people are willing to go into, you know, click the option to actually adjust ⁓ settings and start toggling things on and off on the regular. Max: Yeah, ⁓ that’s for sure. So ⁓ some pop-ups can be really tricky to opt out. ⁓ You would need to go to click, Settings and then toggle a bunch of check boxes and then click Save. This can become... Like most people, think they just click Accept button. ⁓ And ⁓ yeah, this is of course not good for your privacy. ⁓ So we help... Beah: Yeah. Max: getting through these dark patterns. Beah: Yeah, makes sense. why did we build this? What’s the origin story? Max: Yeah, so ⁓ like many other features that go, it started as like a hack project, which is when someone goes in and tries to tackle the problem in a couple of days. And ⁓ of course, cookie pop-ups are universally annoying and wanted to do something about it. ⁓ And we built some prototypes. And then eventually we built a feature on top of a ⁓ prior work of my colleague, Sam Macbeth, who... ⁓ So we have an open source library that does most of the ⁓ things that we... ⁓ And we ⁓ use it and it powers all our... ⁓ This feature in all our browsers. Beah: Nice. ⁓ Max, did I cut off your demo? Did you want to show anything else there? Max: No, I’m trying to stop presenting it just doesn’t work. I’m clicking the button Beah: Oh, okay. Alright, I was just worried I cut you off. Alright, we’ll see if it responds at some point. So, okay, so just to recap, ooh, there it goes, okay. Just to recap, we are a, removing the annoyance of you’re like trying to go to Skyscanner, I don’t know what that is, you’re trying to go to Skyscanner and instead of getting whatever it is that’s on Skyscanner, you’re getting this big like notification in your face, we’re making that go away and we’re going in and we’re changing the settings to be more privacy respecting. That sounds great. What’s the downside? Max: Correct. And that’s, so like ⁓ this ⁓ approach actually is actually quite intentional, right? So as I mentioned this, we’re trying to maximize user privacy and ⁓ because there are other solutions on the market that do like ranging from clicking accept button, which is not acceptable for us. But also ⁓ there’s another approach of like preventing the interaction. And for us, this was very important to do it this way, to actually actively opt out because, well, first of all, ⁓ this is like the only way to opt out of ⁓ server side tracking we know of. ⁓ the second, it gives a clear signal to the website through the official channels. Max: And then finally, in some legislations, it’s actually the only way to opt out. So for example, in California, they can sell your data by default unless you click on the button. So ⁓ yeah, we think that as long as the site is compliant with the law, this approach is better for privacy. ⁓ And if it’s not compliant, we still have our tracker blocking and other privacy protections to fall back to. And so this is of course, so speaking of challenges, ⁓ this is a bit more involved than just, you know, blocking some requests to or blocking the pop-ups from loading. ⁓And so it needs a bit more effort because we actually need to automate each and every pop-up vendor. So it takes a bit more effort. But yeah, this is something we chose to do. I think we, for a while now, we’ve covered most of the, all of the major pop-up vendors, which is like 80, 90 % of top sites in Europe and the US. Beah: So that’s roughly the percent of cookie pop-ups that we think we’re successfully blocking at this point. Max: Yes, so that is 80 or 90 % of all pop-ups that you see on the top sites are handled. And one of the biggest challenges is this long tail of sites, because of course, no one visits just the top sites. And like, each of us has this one site that no one else visits. Beah: Mm-hmm. Okay. Max: And yeah, this is something we’ve been focusing on lately. We’re trying to ⁓ experiment in with automated approaches and using AI as well. And we’ve had some good success in the past months with it. So I think we’re gonna ramp up the this long tail coverage in the coming weeks and months. Yeah, and Beah: And how are you finding those? Do you want to talk about like how your finding those sites, which includes internal reporting, right? Max: Yeah, so we have a few different ⁓ feedback loops, as I say. of course, we have ⁓ our own crawling. So we ⁓ regularly crawl top sites ⁓ and trying to detect new pop-ups and handle them. ⁓ Then we have user reports, ⁓ breakage reports, and just user feedback reports. that we have special systems that filter out and surface the reports related to cookie pop-ups. And we also have very active internal reporting, which is DuckDuckGo employees who go above and beyond and just report new sites to us. is a very important source of feedback because we can get back to those people and verify. Beah: Who’s the number one reporter of cookie pop us. Max: the number one is Gabe. ⁓ So our CEO, he’s like, I think it’s fair to say that half of all the internal reports come from him. I have no idea how he does it. Beah: Hahaha I know. Yeah, sometimes I think maybe I can catch him, but I don’t know. I don’t know that I can. ⁓ So if a user watching this encounters a cookie pop-up, what should they do? How should they report it? Max: Yeah. ⁓ So it depends on what kind of user there are. Like the easiest thing would be to send the feedback through the app. We have this ⁓ feature. Or if something actually doesn’t work, then feel free to send the breakage report, site breakage report at this. But if you’re actually a developer, then This whole thing is open source. And we welcome external contributions. You can go to GitHub, ⁓ find this library, called AutoConsent, and file some issues or even pull requests. This is always welcome. And we’ve had some external contributions before ⁓ from also other companies who are using this library. It’s not only ⁓ used in DuckDuckGo apps. ⁓ So yeah, if you’re that person, we’ll be happy to. Beah: Nice. So to recap, have to be, you only get this feature if you’re using our browser. If you’re using search and you click in another browser and you click on a search result and you land on a page with a cookie pop-up, we can’t really do anything to help you there, much as we’d love to. So you got to install our browser. But if you are using

    15 min
  4. Duck Tales: The DuckDuckGo Subscription — more protection & peace of mind (Episode 10)

    NOV 26

    Duck Tales: The DuckDuckGo Subscription — more protection & peace of mind (Episode 10)

    In this episode, Gabriel (Founder) and Chris (Subscription team) discuss why we built the DuckDuckGo Subscription, its four features, and how it protects more of what you do online. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Show notes: Learn more about the DuckDuckGo Subscription here. Gabriel: Hello everybody. Welcome back to DuckTales, everything DuckTale go. I’m Gabriel, founder, CEO. I got with me Chris. Do want to introduce yourself? Chris: Yeah, hi, I’m Chris Calvi, ⁓ long time listener, first time caller. Had to say that. ⁓ But ⁓ I’ve been at DuckDuckGo for a little over four years. I’m on the partnerships team here. And ⁓ I guess I’ll take a second to talk about the partnerships team, what we are. ⁓ We are primarily the team that handles all of the relationships with third party companies that DuckDuckGo works with. Chris: We’re a relatively small team, about 10 people. you know, classic examples of companies that we’d be working with are any of the live information that you’re seeing when you search DuckDuckGo. So that would be like sports scores or stock quotes, weather, that sort of information, also like flights and lyrics, all of those relationships, we get that information, we license it from...from other companies. that would be an example of work we would be doing. And then also, we will even work on things like infrastructure partnerships with cloud partners and AI companies and all that stuff. So that’s what we do over here at Partnerships more recently. Gabriel: Interestingly though, the way we work, and this gets more inside deck to go since the name of our blog on this, we don’t really work functionally though. We have objectives inside the company to get a particular thing done, and that thing usually involves many different functional teams. What that also means is those objectives have owners and they can be from any functional team. ⁓ You tell me how you think of that, but I think if you do what doing recently actually was not very partnershipy. You’ve been helping launch different parts of our Dr. Go subscription, which involves some partnerships, but like your day-to-day job is not always partnership related. Chris: Yeah, mean, absolutely. And I think you as people, if they listen to a bunch of these episodes, they’ll see that. People might be leading an objective and they’re on the design team or in this case, the partnership team. A lot of times it is somebody from the product team. And in this case, I’m working a little closer on the product side. So you’re right. Gabriel: And yeah, so we’re going to talk today a little bit more deeply because you’ve been working on it about the DuckDuckGo subscription. We’ve mentioned it on some of these episodes, but haven’t really given a big overview. So that’s what this is going to be. The subscription is a bundle. So it’s got a bunch of different things in it, which maybe you can get an overview in a second. But just to say that we might have separate episodes about going deep into some ⁓ One of those things are a technical aspect to them, but this is we can give more of an overview of kind of where it came from and what it exists today and kind of where we’re headed with it. You want to start and just tell us kind of like what’s in it and how it works. Chris: Yeah. Yeah, that sounds good. The DuckDuckGo subscription is a relatively new thing for us. We launched this in April of 2024, so it’s really only been around for a year and a half compared to the search product, for instance, that’s been around for a very long time. And what it does is really, I mean, in a nutshell, it supercharges some of the existing functionality you have. It gives you extra privacy features within your DuckDuckGo ⁓ app and experience. ⁓ It does a little beyond that too, which we’ll talk about, but that’s the key thing. ⁓ The other thing I will note is that if for long time subscribers, you may have originally remembered it being called Privacy Pro. That was the original name for it. We recently just changed it to DuckDuckGo subscription for to keep things simple. So we’ll call it the subscription for the rest of the episode. ⁓ But before I like... Gabriel: Yeah, give us the basics. yeah, what’s in it? Yeah. Chris: Yeah, the basic things, the four basic things in there are the, you got the VPN, which we’ll talk about what that is in a second. You get access to advanced AI models within Duck AI. You ⁓ get personal information removal, which I’ll talk about what that is, as well as identity theft restoration. But before I go down all that, I did want to ask back to you about ⁓ why we came up with a subscription. and then I’ll kind of unpack each of those core pieces. Gabriel: Yeah. Yeah, I mean, the way I think about it more broadly is we want to be your clean internet experience. Like if you want a more protected way to operate online where you’re not getting followed by ads and you’re not giving up your personal information, you kind of adopt the DuckDuckGo ecosystem. And so that’s our browser, it’s our search engine, it’s our AI product. And generally we want to offer those essentially for free. mean, there’s advertising model in the search engine that pays for this, but we can, we want to give it for free in the sense that mo because more people can use it around the world. However, there are certain protections that costs us a lot of money to do that we hadn’t been able to offer because of that. So VPN is the obvious, a canonical example there, because when you turn on the VPN, now all your bandwidth is running through that VPN and a lot of that’s video. ⁓ And it’s just not something that can be advertising supported. doesn’t, the numbers don’t add up. And so we’ve wanted to offer that because it is a key part of being fully protected for people who want that kind of fuller protection, but we couldn’t offer it for free. And so we needed a subscription. At that point, I thought, hey, we could just offer a VPN, but... I think there are other things in this category that have real marginal cost in business terms. ⁓ And it would be great if we could bundle them all together at one low price and allow DuckDuckGo users who really want more protection ⁓ to be able to get it in one place. You don’t have to sign up for multiple things. And that’s where the subscription came from. ⁓ I think we’re living up to that. hope over time we add more and more things and make it more and more valuable without increasing the price or much at all. ⁓ But yeah, that’s where it came from. Chris: Yeah, I mean, that makes sense to me. to dive into the four pieces of functionality, I’ll start with VPN, because you mentioned that. We probably should give a little bit of a high level of what a VPN is. So it stands for virtual private network. But what that means is that we basically have this anonymous private secure no logs VPN server that’s sitting out there, right? And we have these in 40 locations around the world. You likely are connecting to the one closest to you, but you can change it to go to another country. But what it does is all of your internet traffic, the requests that you’re making to, for instance, as you browse the web, are gonna be securely essentially tunneled through that server and then sent out to the resource where you’re requesting that information. So the website that you’re accessing, they’re not going to see, under normal conditions, without a VPN, they would see your IP address. In this case, they’re going to see the DuckDuckGo server’s IP address, which is meaningless, essentially, to them from a privacy perspective. So it preserves your privacy on that front. then they’re sending that information back, and then it’s coming back to you. So then the second place where it’s adding privacy is on your own, sort of like your internet connection. So if you’re... ⁓ you’re making that request in a lot of cases your internet provider or if you’re using like a public Wi-Fi, they can see the sites that you are accessing and sometimes can see the information that’s coming back and forth. And in this case where it’s that since it’s all being encrypted and coming through the duck, going to DuckDuckGo server first, they’re only seeing that you’re connecting to this one server and that they’re not seeing any of the data in between. So it gives you sort of that two different protections when you use a VPN. I don’t know if I’ve done it justice. Chris: Yeah, I think you did a great job there. The only thing I would add to that is, yeah, so it really shields your IP address and other information from leaking to both your internet provider and the sites that you’re visiting, as well as all the places in between too. People may not realize like when you route across the internet, you connect to something and then you connect to like a bunch of different hops in between lots of other internet providers. And it shields not only from the end points, but everything in between. ⁓ Additionally, the IP address is often used to get your location. ⁓ And so by effectively shielding your IP address, you’re effectively shielding your location. And the final thing I would add is like, it may seem esoteric, like do I need to shield this stuff from these people? ⁓ But it’s been well documented that ISPs, internet providers are selling this data all the time. ⁓ And so... If you want to be maxing protected, you do want to have the VPN. ⁓ And you know, I leave it on all the time, our VPN. The other thing, the final thing I would add, which you know, doesn’t relate to the privacy necessarily, but it kind of relates to it in the sense t

    26 min
  5. Duck Tales: Marketing at DuckDuckGo — how we practice the privacy we preach (Episode 9)

    NOV 19

    Duck Tales: Marketing at DuckDuckGo — how we practice the privacy we preach (Episode 9)

    In this episode, Cristina (SVP, Marketing) and Chuck (Front‑end) discuss private marketing at DuckDuckGo, from making decisions with less data to the role of privacy engineers in marketing projects. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Cristina: Hi, and welcome to DuckTales, where we go behind the scenes at DuckDuckGo and discuss the stories, technology, and people that help build privacy tools for everyone. In each episode, you’ll hear from employees about our vision, product updates, engineering approach to AI, or how we operate as a company. Today, we’re going to chat about how most companies collect a ton of information through their marketing activities and how DuckDuckGo, given our privacy policy of we don’t track you, do things like attribution very differently. I’m Cristina, I’m on the marketing team, and today I’ll be interviewing Chuck. Chuck, you wear a lot of different hats. Can you introduce yourself and some of what you work on? Chuck: Sure. ⁓ I am technically on the front end team and work on the front end of our search projects, our products, and our subscription products. ⁓ But I ultimately do whatever I need to do to get the job done, which is kind of our DuckDuckGo ethos. I do some product management, some data science, back end engineering. I work with the marketers. It’s fun. ⁓ And I need access. So getting to where lots of them is. Cristina: Hahaha. Fair enough. Well, thank you. So much like our product philosophy, privacy is core to the ethos of our marketing. There are so many common practices we don’t do, identifying and targeting individual users, retargeting, using behavioral data, using third party cookies and pixels, the list goes on. And we’ve also declined working with a lot of vendors because they don’t meet our privacy standards. As a consumer, that’s something I really appreciate. But frankly, as a marketer, it makes the job very hard. But it’s getting a bit easier thanks to work from people like Chuck, which is why I was so excited to talk to you today. So Chuck, when you first started working with the marketing team, what was your reaction to our limitations and what we were hoping to achieve? Chuck: Honestly, I was a little shocked. ⁓ There’s a pretty well-understood playbook for how marketing in a space like this should look. A playbook of tactics and tools that are well-understood. And every company will do it differently, and every brand and product will have their own personality. But we pretty much use none of those tools. Cristina: Yeah, can you help people understand what the industry norms are for marketing attribution and data and how we do it differently? Chuck: So when you visit your favorite social media site and it’s trying to decide how to fill the ad slot in your feed, the ad platform will take what it knows about you as a person, your search history, who you follow, and what it knows about your situation, like where you are and who you’re with, and line it up with their ad inventory. They’ll do some very complex math to determine the perfect ad to show you that will maximize profits for the platform and the advertiser. So the more better data they have about you, the better they can target the ads and the more money they can make. I know that’s something you’ve talked about with Peter on a previous episode, that the financial incentive for the trackers that are ubiquitous online is data that feeds the machine that helps them make more money off of your ad space. That entire ecosystem just flies in the face of our privacy principles. In fact, some of our apps will block those trackers to keep your browsing private. So when we advertise, we refuse to use those tools like you just listed that are common in digital marketing, like retargeting or reporting different types of conversions after the ad click. ⁓ just to protect the privacy of our users. Instead, we’ll collect limited data only when there’s a very clear and urgent rationale for it. And when we do, we’re transparent about what we collect and how we use it. And we’re possibly most important. We’re really careful never to let those logs link two different events to the same person. That’s really difficult to do. ⁓ We have a really fantastic privacy engineering team that reviews every project and their implementation to make sure that the work we’re doing is aligning with our principles. I’ve also gotten really comfortable making decisions with just the imperfect or incomplete data, trying to identify the solutions that meet 80 % of the business needs without, with 20 % of like the potential input. Cristina: Yeah, it feels like a lot less than 20 % of what’s actually available to us. Well, yeah. So thank you for unpacking that. That’s a helpful foundation. Can you go a bit deeper and talk about what that looks like in practice at DuckDuckGo? Chuck: Yeah. That’s probably fair. Yeah, so we largely ⁓ don’t work with other vendors ⁓ in the marketing space and rely on the tools we own and build ourselves instead. That makes sure that we aren’t incidentally feeding the machine with our own users’ data, which is really easy to do if you’re not careful. ⁓ We have a couple of tools in our toolbox, too. We’ll do as much summarization and analysis of data locally before we ever send it back. So rather than saying that a user of our browser searches 15 times in a day and ⁓ sending 15 different events for those searches, we’ll send a periodic report that will say they searched 15 times during that day. We’ll reduce the precision of those signals even further. So instead of saying that that person made 15 searches, we’ll say they’re a medium volume search user. And then when we do our analysis on an ad campaign, we’ll look at the summaries of the data rather than the raw data ⁓ so that we’re looking across our users rather than the individual humans. And if it comes down to it, we are willing to redact data that might be too identifying for a person, whether it might contain PII or if it looks too unique and may be able to be traceable back to a person, we’d rather delete it and not use it than jeopardize that person’s privacy. Cristina: Well, thank you for ⁓ sharing how our ethos really comes to life there. And I’d love for you to touch on one of your claims to fame at DuckDuckGo, which is creating a better, more privacy-respecting system that we call Origin. Can you talk about how you got the idea and how you brought it to life? Chuck: Yeah, so we were struggling to run small scale campaigns that test new ad platforms or creatives. ⁓ With the tools that we have, the only way that we could do that without jeopardizing user privacy is to run big, broad, expensive, scaled campaigns. But we’re a small company. We want to move nimbly. And that made it really difficult for us to quickly validate our direction and make sure that we were dedicating our resources in the right time or in the right place. So I spent some time with our marketing leaders, including you, Cristina. ⁓ trying to understand the norms and the challenges they were facing, the tools that weren’t in their toolbox. And I brought that to the privacy team. ⁓ We worked backwards, starting with user privacy as a first principle to the business goals and landed on a solution that kind of looks like this. ⁓ You see an ad and you click on it for DuckDuckGo and you install our app from it. When that app first runs, we will send one signal that says that you installed the app from that ad in that location. And then once a day, we’ll build a summary of those signals that give us pretty coarse insights that say, you know, we had 10 users install our app from that ad on that ad platform on that day. Then we’ll line that data up with other information that the ad platform gives us, like how many impressions there were of the ad and how many times it was clicked and how much that cost us. And that’ll give us some high level insights we can use to start making decisions, like how much it costs us to ⁓ per install from that ad. There’s nothing groundbreaking here technologically. It’s actually intentionally very simple and that helps us maintain the privacy properties because we have a high elevation view of everything that’s happening. We never share data outside of DuckDuckGo, so we aren’t feeding that machine. There are never person level insights. We’re looking at broad signals across our audiences. There’s no risk of PII and we’re only collecting the data that we need to make those decisions, nothing more. But it still lets our marketing team make informed decisions while working quickly and doing their jobs well. Cristina: Well, thank you. ⁓ More importantly, thank you for the months and months of work you did on that. ⁓ You say it’s nothing revolutionary, but actually, I think it’s a pretty novel approach. We don’t know of any other companies using technology like this. Typically, they use the entire suite of tools available to them. ⁓ But hopefully, one day, it won’t feel like such a novel approach, and this will become more of the industry standard. At least my naive perspective can hope for that. Chuck: Of course. I hope so. We’ll see if capitalism agrees with this. Cristina: So any parting thoughts you’d like to leave on the future of privacy respecting marketing? Chuck: One of the things that I really love about and appreciate about DuckDuckGo is the example that we set for other companies. ⁓ On the search engine side, could we collect data at massive scale and hyper-target ads to our users? Absolutely, but we don’t need to. And we love being an example of a sustainable business that respects user privacy in their search

    10 min
  6. Duck Tales: Improving AI chat organization, and feature decisions at DuckDuckGo (Episode 8)

    NOV 12

    Duck Tales: Improving AI chat organization, and feature decisions at DuckDuckGo (Episode 8)

    In this episode, Gabriel (Founder) and Esteban (Design) discuss AI chat organization, from automatic chat naming to ‘pinning’ your most used chats. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Gabriel: Hello, welcome to DuckTales. Inside DuckDuckGo features people, et cetera. You got me as the host again, this time I’m the founder and CEO of DuckDuckGo. And I got Esteban with me today. Esteban, wanna introduce yourself? Esteban: Sure. Hi everyone. I’m Esteban. Hi Gabriel. Esteban, I’m a designer in the team. Gabriel: Sweet. And that’s what we’re talking about today, Duck AI, and in particular, kind of new features we’ve been introducing around conversation organization that Estevan has been designing and working on. And we’ve been releasing kind of a few of these over the last couple of months. You want to just jump in, you want to share your screen and walk us through it. I realize we have some of this on audio now only. Esteban: Sure. Gabriel: Let’s also try to describe what we’re seeing while we do it. I’ll do the same thing. Esteban: Yes, for sure. Let me just share my screen and I’ll walk you through a few of the things that we recently dumped. Yes, so if you’re like me, you have tons of like conversations going on at the same time. What we have seen from users is that people who are very engaged with our platform, they end up having tons of conversations and then sometimes it’s harder to get back to them and to find the one that you needed. So we launched three things. The first one, I’ll show you an example. I was trying to see if I should ask you for a promotion during this podcast. Clearly no. ⁓ But the cool thing was that we, yeah, I will wait for a better time. ⁓ The first thing that we shipped was that we had ⁓ a title, the title of our, Gabriel: Good answer. Esteban: Chats were basically the same first prompt. It was just duplicated. That would make it harder to read. So the first thing that we launched was this automatic titling. We’re using the LLM to come up with a better title at the top. So it’s easier to find. ⁓ Gabriel: So this is similar to how ChatGPT and some other competitors are doing it, right? So like we used to do just to repeat, so I understand fully, I think I do, ⁓ having reviewed this when it launched, but just to make sure. ⁓ Yeah, we used to just literally repeat the prompt as your title, like the beginning of the prompt. And so that it was kind of weird sometimes and duplicated sometimes, but now we run in the background something that... ⁓ Esteban: Definitely Right.Yeah. Gabriel: Summarizes it for you and then automatically does this. You don’t have to do anything, right? Esteban: Yes. Exactly. And for example, I’m trying a new one. What day is today? That’s my prompt. And then immediately what I get as a title is day to day. The one thing is that sometimes the title is not exactly what you will call it. Like maybe day to day is not super descriptive. So we added this minor option, no? So we can say, I don’t know, day to day, it’s Thursday. Simple but useful. Gabriel: So if you want to do whatever you want, yeah, like if you don’t like our casing or something, you can put everything in lowercase or whatever, you can just rename it at will, yeah. Esteban: Yeah. Exactly. Yes, so after you have you want to name stuff, it’s very helpful when you have like several chats about something similar, but you want to have it personalized so it’s easier to find. So super simple, nothing that I don’t know, it’s blowing innovation in a way, but very useful. The second one is we noticed how ⁓ lots of our users were asking us for ways to save conversations and finding the conversations faster. We also have a limit of how many conversations you can have right now. There’s just 30 chats after the 30th chat. The next one gets deleted. We’re working on that, but we wanted to offer a way for you to say what’s important for you and why are the things that you want to come back to it. So I don’t know, this one. Let’s say it’s something that I was working on today, but I want to come back to it. add it to the top of the list by pinning it. So now it’s pinned at the top and it will always be there. That also means that if DocAI needs to delete one of your chats, it will go to the last one on the list and this one will not get deleted. You can pin up to five chats right now and yeah, it’s always accessible, always at the top, easy to reach. Gabriel: Got it, so it’s kind of the equivalent of like a favorite or a star or a pin in this case. All of these things are accessible from this three dot hover menu next to the ⁓ chat, chat to the individual chat. And then also you’re saying, is good, that ⁓ because the reason why we have the 30 limit at the moment is because all this is stored locally on your browser. It’s not actually stored on our servers. ⁓ Esteban: Exactly. Gabriel: we’re working on an encrypted storage that we won’t have the keys to where you can get a much higher limits on it. But at the moment, or if you just want to keep it local, there is a local limit because your browser has storage limits. ⁓ But what this will allow you to do is keep ones around that you really want around, right? So you’re saying if you favorite these or pin them, in our words here, you have a pin section at the top. But if you start making a lot of chats, the non-pin ones will get. Esteban: second. Gabriel: kind of blown away first. Esteban: Exactly. Exactly. ⁓ We have plenty of requests about chat organization and some people are going as far as like, want to create projects, want to create groups, I want to organize my stuff, which it all makes total sense. And I guess up to a point you need those sort of organization tools, but the simple ones will cover most of the needs for most of our users, I would say, because maybe you have a few chats where you keep coming back to them, but then you have a bunch of quick requests, quick checkups with the LLM. And something like this is super simple. It doesn’t require a lot of effort. And yeah, we hope it helps a lot of our users. Gabriel: And you mentioned in there a few times like we did this and we’re working on this, these features in particular because we get a lot of user requests for them. ⁓ So speaking to that a little bit, like my understanding is, you know, we get lots of feedback. ⁓ We’re looking through it all and kind of organizing it, like which are the most like important issues to work on. But then also when we launch features, cause to your point, people ask for all sorts of different things. And then we, and like you in particular, design. And you’re like, well, I think this is gonna be a satisfaction of a lot of people’s requests, even if that’s not exactly the thing they asked for. And then we put it out and then how do we know whether it worked or not? Like, what are we looking at? Esteban: Yes, so to answer your question, we see a usage in particular. We don’t have, of course, data about a particular users, but we know that roughly X amount of people are using this feature. We also know from social media, we got a little bit of love ⁓ after posting that we launched this. And then the other thing that’s a really interesting metric is we see the feedback coming in. ⁓ and comparing with what kind of feedback we were getting a month ago, and we see a big decrease on charge organization requests. As said, there are still things that we can do and we’re working on them, but we see a lot of people, a lot less people requesting for things like this, which is also good news. Gabriel: I love that metric. mean, because it really is, I guess we are lucky that we have enough users and enough sample size where we can be like, wow, chat organization is a category we can ascribe feedback to. And now it’s like halved or whatever after these features come out, right? Esteban: Right. Yeah, so something we did for this project that is super interesting, it was fun for me to do was that we have this category, like chat organization, and then we get feedback directly from our users asking for a specific feature. I went and looked into all the feedback that was related to chat organization, and I tried to see why we’re asking for a specific feature. Some of them actually mentioned it, no, I want this so I can do that. And then those needs, I matched them with what type of features will work, even just by naming different types of features. You said that this was sort of favorites or it could be pinning or it could be saved or it could be bookmarked. We also have bookmarks in the browser. So how do we name the feature related to the benefit that I will have and the need that it will solve, but also how much effort will require from the users to actually get the benefit that they were asking for? ⁓ Create projects, it’s a lot of effort. And they will get a lot of benefit, but not many people are willing to spend time organizing on their chats. But pinning is one tap away So that’s a sort of ⁓ prioritization exercise that we did to define exactly what are the new features that we’re going to launch. Gabriel: Yeah, that’s a super interesting point because like all different features have different levels of complexity for us to build. And then also for the user to actually use to your point, like I’ve seen, yeah, I’ve seen lots of requests for people wanting really complex things, but then very small percentage of people would actually go through the effort to use those things. Whereas to your point here, pinning, especially the chat title is just automatic. We get that for free to everybody. But then the pinning is just literally just a one click. You don

    12 min
  7. Duck Tales: How DuckDuckGo protects users from different types of scams (Episode 7)

    NOV 5

    Duck Tales: How DuckDuckGo protects users from different types of scams (Episode 7)

    In this episode, Gabriel (Founder) and Thom (Security Team) discuss Scam Blocker. How it works, the types of scams it protects against, and why our ‘bad pages’ list is updated so often. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. Show notes: See the full blog post on Scam Blocker. Gabriel: Hello, welcome to DuckTales. I’m Gabriel, CEO and founder of DuckDuckGo. DuckTales is everything kind of inside DuckDuckGo. Today we have a new topic. I don’t think we have discussed much about security in our browser. I got Thom here. Thom, you want to introduce yourself? Thom: Yeah, sure. Hi, I’m Thom. I’m one of the security engineers here at DuckDuckGo. I spend most of my time kind of in and out of browser security, product security, that kind of stuff. Yeah, that’s the kind of stuff I love. Gabriel: Sweet. And I think we’re here today to talk about our Scam Blocker. If you follow our blog, we actually had a pretty big article about this when it launched a few months ago. And so you can always check that out too, but we’re going to tell you all about it here and some of the inside info on how it came together. Yeah, Thom, you want to just explain generally what it is? What is Scam Blocker exactly? Thom: Yeah, sure. So I guess Scam Blocker is what we call our in-browser phishing and scam protection. It kind of runs in the background and checks websites as you visit them all locally in the browser. And we kind of have a pretty big data set here that we get from Netcraft. So we can protect against all sorts of scams — this isn’t your standard phishing protection. We try and protect against cloned e-commerce sites, fake crypto exchanges, scareware like fake virus pages, and advertising of fake products and stuff. So we have quite a lot that we’re trying to protect against, but this feature as a whole is that warning page that you get when you’re about to visit something that could be scammy or phishing related. Gabriel: So let’s talk about that distinction a little bit. I guess backing up a little, how did this come together? How did we end up building this and then building it kind of differently than other companies? Thom: Yeah, so it came from a long way back. Originally, we had this idea that we wanted to improve our tracking protection and all of this stuff — trying to make our browsers as safe as possible for our users. We knew that we wanted to do something in this space, but the challenge was that it’s quite easy to build a feature like this where it ends up looking like you need to check people’s browsing activity — and we can’t do that from a privacy perspective. So we knew that we had to do this in a privacy-preserving way, and we didn’t like the idea of sending any data to Google or Microsoft because they pretty much own this space in terms of browser protections. We weren’t comfortable with that, so that kind of led us down the path of building it ourselves. Gabriel: Interesting. So like at a high level, our browser has a privacy protection list instead of blocking that we built ourselves because we didn’t believe anyone else was doing it up to the standard that we think it should be. But that’s all kind of behind the scenes on pages that you visit, assuming that was a page you actually wanted to visit. Privacy and security overlap, but as I understand it, some pages you visit are actually bad for you — not because there’s hidden trackers, but because the page itself has malware or scams. Those are the pages we wanted to cover. And in doing that, you need to have a list of bad pages. Thom: Yeah. Gabriel: Everyone else seems to be using Google or Microsoft, and all the other browsers are just kind of riding on Google Safe Browsing. But we wanted to go somewhere different. So we found this vendor Netcraft, who maintains a big list, and it turns out they have an even bigger list than Google’s because they cover these other categories, right? Thom: Yeah, exactly. Gabriel: Like some of these scam categories that you mentioned are not traditional malware phishing. They’re theoretically legitimate businesses that are scamming you. So for whatever reason, they’re not on Google’s list. Is that kind of how to think about it? Thom: Yeah. That’s a good way of saying it. Some of these are quite unique. One of the interesting cases I like to refer to is that sometimes even a blog post could be a scam. If this is a blog post advertising a fake product that’s going to steal your money, that’s a problem. A lot of these scam sites start somewhere trusted, like a Medium article or GitHub page, and then send you down fishy paths until you end up somewhere meant to steal your money. That’s the kind of thing we’re looking at here with Netcraft. We get data that lets us look at the source of it rather than waiting for you to click through multiple times to get there. Gabriel: So we license this data set from Netcraft who’s aggregating all of these scams from different signals. And then what do we do with it exactly? How does it work to be embedded in the browser? Thom: Basically, we pull this data — it’s constantly evolving, which is one of the challenges. We have to update it pretty much every five minutes on the backend. We pull it, process it, filter out some of the lower-risk things, and then compress it. Gabriel: Five minutes is so quick. So it’s really happening in real time. I didn’t realize we were doing it that real time. Thom: Yeah, it’s rapid. If you take a random phishing link now and look again in five minutes, chances are it’s gone. Gabriel: And that’s because all these people are reporting these things, right? It’s an arms race — things get blocked quick, they switch domains, and all sorts of crazy stuff. Thom: Exactly. It’s this constant cat-and-mouse game. Gabriel: Cool. Sorry to interrupt. Every five minutes, we’re updating this list on our backend. Thom: Yeah, and then we compress this into a small format. Our browsers pull this data every 10 to 20 minutes depending on platform. That’s how the update mechanism works. Gabriel: Got it. So once it’s sitting in the browser, the browser checks against the list. If you’re going somewhere that’s on the list, that’s when you see the warning page. Are we similar to others where you get a big warning page but can accept the risk? And do all these warning pages look the same or are there different types? Thom: Yeah, pretty much the same. You get a warning page explaining the case. We have three types of warning pages — they vary slightly in iconography and copy. They’re for malware, phishing, and scam. Malware means you might download something malicious, phishing is about credentials or credit cards, and scam is broader — like a dodgy e-commerce site. Gabriel: Got it. So any surprises in building this or challenges that arose getting it live to production? Thom: Yeah, a few. The first one is that we have four browsers — four different platforms. The core part of the feature is constantly updating, but the other challenge is intercepting navigation requests. Every browser does this differently. So we had to map out how each does it and figure out ways to do it efficiently. We pride ourselves on our browsers being quick — we don’t want to affect load times. So we had to make sure the check runs quickly, just before a page loads. There’s a lot to consider. That was one of the biggest challenges. Gabriel: Yeah, that makes sense. It basically seems like one project, but it’s four big projects — MacOS, Windows, Android, and iOS. Cool. So how has it gone? Any good response? I know we put out a blog post and got some press when it launched. It seemed positive from my view, but from your point of view, what did you think? Thom: I think we had good positive feedback. One unique thing about this feature is that it’s in the background — its success hinges on people not really seeing it. If loads of people are seeing the error page, then we’ve probably done something wrong. But overall, it’s gone well. Gabriel: Yeah, that’s a good point. It’s like our other privacy protections — always on, not breaking sites, contributing to peace of mind. It’s protection that’s there, not in your face. Thom: Precisely. People who’ve come across it said it works well and gives them peace of mind. Gabriel: Cool. So it sounds like it kind of went off without a hitch. Is there anything left to do now? Are we kind of in maintenance mode with it? Thom: Yeah, pretty much in maintenance mode. We have about three or four people monitoring metrics. But we’re exploring ways to enhance the data, maybe adding new or better data sets. We might tailor data sets by platform — for example, malware is more prevalent on Windows, scams more on mobile. I’ve also been reading about using small language models fine-tuned to detect scammy websites locally. It’s promising research — local-only, privacy-preserving — though I don’t see it in the browser anytime soon. Gabriel: That sounds fun. A good hack day project — and who knows, lots of those end up in the product. I definitely think we should ship local models or get access to local ones on the device. The problem’s been that either local models aren’t very good or the downloadable ones are too big, like three gigs. But I think it’s coming. I think there’ll be a future where we have local models in the browser, shipped by default or opt-in, maybe with extra protection. That would be an interesting incentive to download a local model if it gives extra security protection. Thom: Yeah, exactly — extra security protections. I’d love that. Gabriel: All right. Well, we’ll end here. Thanks, Thom, for coming on. Hope everyone enjoyed hearing about

    15 min
  8. Duck Tales: The internet’s privacy problem, and how DuckDuckGo is solving it (Episode 6)

    OCT 29

    Duck Tales: The internet’s privacy problem, and how DuckDuckGo is solving it (Episode 6)

    In this episode, Cristina (SVP, Marketing) and Peter (Director, Product) discuss digital fingerprinting, privacy washing, and how hidden trackers appear in the majority of popular websites. Plus, the steps you can take to protect yourself online. Disclaimers: (1) The audio, video (above), and transcript (below) are unedited and may contain minor inaccuracies or transcription errors. (2) This website is operated by Substack. This is their privacy policy. CristinaHi, and welcome to DuckTales, where we go behind the scenes at DuckDuckGo and discuss the stories, technology, and people that help provide privacy tools for everyone. In each episode, you’ll hear from employees about our vision, product updates, approach to AI, or how we operate as a company. Today, we’re going to chat about the online privacy problem and DuckDuckGo’s web protections. I’m Cristina. I’m on the marketing team. And today, I’ll be interviewing Peter. Peter, would you like to introduce yourself, maybe what team you’re on and where you spend a lot of your time? ⁓ Peter Absolutely. Hi, Cristina. I’m Peter. I’m on the product team at DuckDuckGo, which I typically work on our browsers and our privacy protection. So happy and excited to talk about the mystifying world of online tracking and privacy today. CristinaAwesome, likewise, well let’s jump in. So I think a lot of people would be surprised to hear just how much information about them is being tracked online. Some seemingly irrelevant to what they’re doing and some pretty creepy in how detailed it is and how all the dots are being connected. Can you give some examples of the pervasiveness of this tracking? Peter Absolutely. know, anyone I talk to about online privacy, the first thing they’ll tell me, and I’m sure you’ve heard the same, is microphones must be listening to them. ⁓ Yeah, everyone can give an example of a conversation in their household where not too long thereafter, they’re seeing advertisements, creepy advertisements, following them around online based on, you know, what it is they were talking about. ⁓ And the reality is the amount of surveillance that happens is like microphones are listening to you everywhere, but the methods are not actually microphones. The methods are actual trackers on websites, on search engines and browsers and apps, which we’ll talk about that are always collecting information about you. ⁓ So just to break those down a little bit, most people, if you think about someone in their daily life, they’re going to go do a search online, whether it’s on their smartphone or on their computer. The search engine that most people use is, of course, Google, most dominant search engine in the world. They collect basically anything and everything about you. ⁓ And so that search engine is one source of this data collection. And then ⁓ the browser you use to actually do those searches, often owned by some of the same companies like Google, ⁓ like Google Chrome specifically, these browsers also directly collect information about you. So if you’re not using a private search or a private browser, a lot of information is directly collected about you. But then, of course, after you do a search and you get onto a website, the websites themselves have trackers embedded in them. And specifically, we’ve done actually a lot of analysis on this. 85 % of the top websites on the web have Google trackers included in them, and about 36 % have Meta or Facebook trackers overall. And these trackers are pieces of code that run on the websites that send information about you, what you’re doing on the site, what products you’re looking at, what’s in your shopping cart, and so on to companies that are not the owners of the websites. The same is true of your mobile apps. So just as it happens, the surveillance on websites, it happens in your mobile apps. ⁓ In fact, 96 % of the popular top free Android Apps send data to third-party companies. And of those, 87 % send data to Google, 68 % send data to Guest It, Meta, and Facebook. Top two trackers overall. And then, of course, there’s other sources too. When you use emails, emails contain trackers. When you open them, little code fires. It tells the email sender when you open their email, where you were when they opened the email. And then there’s a lot of other scenarios too. Like if you go to the store, What do they ask you when you make a purchase at the store? Can we have your email address? And they say, oh, it’s for a loyalty program. You can get points or whatever it is. But the reality is they’re actually usually taking that email address and then directly uploading it to Facebook, to Instagram, so that they can buy advertisements targeting you later. And so you combine all this. And you have this pervasive tracking and then targeting that’s happening. that makes it feel like ultimately there must be microphones listening to you, but it’s just happening throughout your day overall. CristinaIt’s pretty chilling that I could be on almost any site or Android app or reading email or at the mall buying a new shirt and companies like Google are tracking me. So what type of information are they collecting? Peter So they’re typically after two sets of things. And when I say they, I use Google and Meta, Facebook as examples, but there’s thousands of other ad tech companies that are often in the mix trying to collect something about you as well. ⁓ They’re looking first for an identifier. So they want something that’s gonna be able to tie what you’re doing to an identity so they know who it is, or even if they might not know who exactly it is, they wanna know it’s the same person. So of course, email address could be an identifier, your name could be an identifier, phone number could be an identifier. Those are the obvious ones that they would want. And by the way, this is why so many websites try to get you to log in on those websites, often with your Google login, because then they can tie all this, whatever you’re doing on that website to your identity. And then of course, I think most people have heard of cookies, and seen cookie banners come up when they visit websites. Cookies are another form of identifier, might not be your name or your email address, but it is a unique code. And so that when these trackers that are across all these websites see the same cookie identifiers across those websites, they all, this is the same person. And so whatever you did on this site, we can link it to whatever you did on this other site. And then there’s a couple other identifiers such as ⁓ digital fingerprints, which really use information about your device, like your screen resolution and your battery, literally the state of your headphone jack on your smartphones, they piece this together into a digital fingerprint that is unique. And so if they see the same set of attributes about your device on a different website or different app, again, they can infer this is the same person overall. So that’s the first thing they want, identifiers. And then the second thing they want is something about you, behavior, interests, actions. ⁓ And so it might be as high level as Cristina’s into snowboarding. ⁓ But it could be as low level as the specific things that you had in your shopping cart, what you purchased in real life in Home Depot last week. ⁓ Whatever it is, they basically want to collect it, put it together into a behavioral profile that they can then turn around to advertisers and offer very hyper-targeting to these individuals overall. And just to give you a sort of creepy example, we’ve done a lot of studies on this with websites and apps. And we looked at health websites and health applications, ones where you may look up health conditions or prescription drugs. And we literally observe these trackers included in these apps or websites sending information about your health conditions, your sexual orientation, and even prescription drug information to third-party companies overall, things that people would be absolutely shocked to hear overall. CristinaThat’s definitely not information I want shared without my permission. ⁓ And while historically I might have thought something like, ⁓ battery life or headphone jacks, whatever, don’t care, when you start piecing it together to make this fingerprint like you’re talking about, yeah, it gets super scary. You know, I’ve heard some people say, ugh, it’s impossible to do anything when it comes to these giant companies and all these clever ways they’re collecting information. Anything I could do would just be a drop in the ocean. How is DuckDuckGo thinking about a user-led approach to solving the privacy problem? Peter DuckDuckGo, obviously, most people know us through our private search engine. And of course, our private search doesn’t collect information about users. That’s what sets it apart. And even our advertisements themselves on DuckDuckGo search are just based on what you’re searching for. But ⁓ we realized that protecting people in their searches is not enough. We needed to protect people’s privacy more broadly. And so that’s why DuckDuckGo introduced you some years back. ⁓ browsers as well. And so you could use our search and our browser to more broadly protect you. ⁓ Let me share my screen a little bit here just to show you a sort of comparison we put together. So we put together a comparison for people. I won’t go over all the details. feel free to take a look at this later, duckduckgo.com slash compare dash privacy. But ⁓ basically, when you’re trying to protect CristinaThat’d be great. Peter privacy broadly through all these threats I step through. You really need protections for each one of those threats and the methods of data collection. And so that’s what we try to incorporate into our browser overall. And so you’ll see our browser has

    16 min
See All (13)

About

Behind the scenes with the DuckDuckGo team — sharing insights on product, engineering, leadership, and AI. insideduckduckgo.substack.com

Information

  • Creator
    DuckDuckGo
  • Years Active
    2K
  • Episodes
    13
  • Rating
    Clean
  • Copyright
    © DuckDuckGo
  • Show Website
    Inside DuckDuckGo

You Might Also Like