Hey, how's it going? This is Craig Cannon, and you're listening to Y Combinators Podcast. Today's episode is with Matt Cutts. Matt is the administrator of the U.S. Digital Service, and previously, he was the head of the web spam team at Google. You can find him on Twitter at Matt Cutts. All right, here we go. Matt Cutts, welcome to the podcast. Thanks for having me. No problem. So for those who don't know you, you are the administrator of the U.S. Digital Service, and previously, you were at Google, where you were the head of the web spam team, and also the 71st employee in the year 2000.
What was Google like in 2000? Oh, man. So we had three people start that day and that was a new record at the time. Now, you know, you've got like hundreds of people starting each week. But the crazy thing is it was a startup back then, you know, so late nights working crazy hours. I remember one of the first project I worked on was safe search. And so at one point at like 2 a.m., I got something working. I was really happy. So I was going to head home and I was like speeding because I was like super excited. I got this thing working, got a speeding ticket. And so I literally remember like working every weekend until at some point people were like, oh.
three or four years in we're like we don't work on weekends anymore and i was like oh now the culture's changed a little bit but it's super weird to be like the people who were just folks you know amit or lucas or whoever then eventually became entire departments you know sales departments and people who dealt with logs and privacy and but it back then it was just like a small group of people it's crazy how do you because i know that the story with pb creating gmail is just like a one guy goof let's see if we can do this right How did projects get delegated and chosen? How did it all work? Well, it was funny because I started out, I did safe search, and then there was this ski offsite. Like everybody fit on one bus, one 50-person bus back then. So that was a great introduction to the company. And I was skiing and on a lift with a manager. And she was like, hey, Matt, you like doing front-end programming. And I was like, sure, I like front-end programming. And then like, boom, guess what? You're in the ads group now. I'm like, wait, I don't want to be in the ads group. But there were only like five people, and they needed to help. So I was like trying to help out. We did like geolocation. And it took like a year to claw my way back towards ranking. So it was very informal. It was very much like, here's a program.
problem we got to go swarm and tackle it um even writing safe search was because there was a partner that wanted it you know and so we're like okay can we build this in time let's see if we can make it so on like a self-hosted version or like yeah okay to what degree did you feel like the success of google was certain at the time completely uncertain i mean if you go back i think google had raised like 25 million dollars uh from you know like kleiner perkins and sequoia or whatever and so As I recall, the dot-com crash happened, like March of 2000, winter apocalypse. Everybody was like, this is going to be terrible. And so it wasn't at all clear that we were going to be able to make it. I remember when AltaVista, I was worried they were going to crush us. Oh, man. Because like, you know, you have a certain number of ranking signals. And if they had twigged and caught on fast enough, they even like copied our appearance. They had a little front end that you could set it where you could be like, I think they called it like goofy. It was like rainbow colored. So it looked a little like Google. But they didn't get the quality right. So we were OK from that point of view. But it was nonstop for several years. I mean, trying to make sure that.
In those early days, Microsoft didn't realize how much money was coming from search engines, but also AdWords and later AdSense. Can you walk through that product development? Because I'm so curious. What did you start? You're like, oh, this might not be a thing. And then a certain type of ad takes off or you tried something that failed. So it was it was wild because back then people were like, do you have salespeople sell stuff, which was the default? You know, so you go to the most profitable folks and do you sell by CPM cost per thousand or do you sell by cost per click? There was this thing called Overture where people could bid on things. And so there was a whole bunch of shifts in strategy where people are like, let's figure out how to do this. So at one point I was in the ads group and they said, we're going to do this.
prototype of self-service advertising so we're gonna make some little ads on the right hand side and uh oh man i forgot to turn off caching when i ran that experiment and i nearly melted google at that point which was not cool because they were just dynamically serving them constantly to everyone well no no this was this was super super prototype so it was like the ads were like pool tables and playstation and i forget what the third one was for any search for no but in order to show enough you had to have it in the experiment for like 30 of people because not that many people were searching for playstations or whatever uh and so i turned off caching for 30 of google which radically like like racks were melting down all this sort of stuff so i remember we looked at the click-through rate and uh and it was really low um because we just picked some copy you know we hadn't done any ab testing we're like would you like to buy a playstation now kind of thing and as i recall marissa was like this is not good for the user experience and larry page was like Well, maybe but I could imagine click-through going up. So let's let's explore this a little bit more That surprises me because I've heard stories about banner ads for example in the beginning having crazy click-through rate So why was that not working? You know, the only thing I can think of is the copy probably sucked It was off on the right-hand side people probably didn't know what exactly is this thing over here And it was kind of fun because they were like multi-color ads back then you like throwing in pictures and trying to make no pictures But I remember like a one pixel darker color boundary like they were really pretty ads But I think people just didn't even know to click on them and it turns out
having people willing to put in the A-B testing makes a huge difference. And the first self-service ad we got for AdWords was, I think, for a lobster company. You could buy lobsters in Maine and have them packed in dry ice and shipped to you. And that was the point when we were like, oh, there's this whole long tail of people who want to reach people who are looking for the things that they're selling. Pre-social media. Totally.
did you do like a user study you like call the person in main up and like what made you choose to do this i hope that they got in touch with that you'd be like a pizza shop in san jose like not a lot yeah right i really hope they got in touch and we're like did you know you were the very first one yeah and i remember i went to a search conference a few years back and somebody was like hey i was one of the first adwords advertisers and i was like oh cool what did you do and he was like It wasn't family safe. And I was like, oh, interesting. Yeah. And he was like talking about how he tweaked all the keywords and did all the testing. But people were willing to put in that work because you could get, you know, you could find these little cavities where nobody else was searching. And like if you found the right word, you could actually get great click through rates and a great response. I think you see it today with YouTube, right? Like people realize like, oh, there's nothing for this kind of flat earth search. Therefore, they just like fill it up. Yeah. So we used to call that the evil unicorn problem.
Or at least I used to call it that because people come to Google and they'll search for something like flat earth. Right. And there's no good answers because you're looking for the most reputable, useful results to give people. And like there's not that many legitimate folks who are like, oh, yeah, it's totally flat. So but you still have to show 10 results unless you like change the interface to say you're in an untrusted area, which we experimented with that later. And so like the folks who realize there are people searching for evil unicorns, you know, everybody thinking unicorns are perfect, nice, whatever. But you can still search for evil unicorn and then you got to have 10 results for unicorn. And so it's sort of this.
lacuna like a lack of information and so when there's not high quality information you still end up showing something and so the folks who realize they can make flat earth content or whatever we're filling in the gap and so at that point you're just like hey listen we're a common carrier to a certain extent whatever comes through comes through it gets really hard with common carrier and publisher and 230 and all that sort of stuff so there were literally people who sued us because we took action on them because we considered them spammers like they were literally selling page rank like i will link to you and the amount of money is based on the amount of page rank i have uh and then we took action and they were like that's unfair and we're like like wow like we rank the search results and so there was one called uh search king and the result of that lawsuit was that search results are protected by the first amendment so that was that was a useful court precedent and then there was one called kinder start where they were saying page rank is an algorithm and so you have no ability or right to like zero out somebody's page rank or to take action but if you follow that to the natural logical extreme like then you'd never be able to like tweak or adjust the search engine or like manually say oh i haven't you know this one span but we haven't caught it yet our algorithms aren't ready yet so we're not allowed to take action on it and so we won that lawsuit as well but it was super interesting to see how people thought about search whether it was like a newspaper or whether it was like a card catalog in a library or like a magazine and you know people just want high quality relevant results they don't want to delve too deeply into you know i don't want we tried giving people knobs yeah you know we could you know tweak how reputable something would be and nobody nobody ever used it they just click on the first five or whatever typically yeah so where do you fall now with things like youtube
where you just go deep, deep, deep, and you might not even know you're in it. So I think one good thing about Google, having left there several years ago now, is that the people really care about trying to do the right thing. And so trying to return high-quality relevant results. And the same thing for YouTube. It's a different silo within Google, but a lot of the DNA is the same. And so when you see searches for something that doesn't actually exist, and so spammers are ranking for it on YouTube, YouTube wants to take care of that. And so I think some of the recent stuff where people are getting down a rabbit hole, I'm sure that there are engineers thinking very hard about how do we solve this problem and make it work better.
Okay, I could talk to you about Google back in the day for a very long time. I want to be careful. But I was curious, so being early on the WebSpan team and then running WebSpan, you've read these stories about content moderators overseas looking at horrible stuff. Were you exposed to that? Yeah, I mean...
yes more so with safe search where you were trying to detect you know pornography and non-family safe things and for a while so larry and sergey shared an office for a long time because they were on the road and you know we were tight on real estate yeah yeah and so for a long time i had the cubicle right outside of their office and this was right when i was working on safe search and so i was like oh man trying to see if i could find stuff that had slipped through and if i did i would you know try to tweak word weights and stuff like that and so at some point colpreet who was our first lawyer at google came by and he's like hey um hey matt we know you have to like look for pornography. It's part of your job. It's a vital thing that you do, but like, it kind of weirds out when visitors are coming. It looks kind of like you're looking at porn and we know you're doing it for work, but could you put like a whiteboard up to block? I was like, okay, I can, I can do that. So yeah, I did. I did end up seeing a lot of stuff, but, um, but it was a little bit of a different time. Like safe search was just towards pornography and spam was more like buy cheap Viagra and loan consolidation stuff. And so it wasn't nearly as bad as a lot of the content moderators, um, had to deal with.
There is one aspect, though, in which once you've seen all the different ways in which people try to spam and cheat and break the rules, you can't unsee that. Like it's the black hat mindset. Like once you realize, hey, here's the thing where people can recycle the conference badges. My mind immediately goes to what if that's not the conference? And then people have free conference badges that they can then use for their friends on the last two or three days of a conference. Like you literally can't look around the world and not think about how is somebody going to abuse that system? Right. So now do you feel that people are like fundamentally evil? No, no. So even it was funny whenever we were working with there's a lot of publishers and websites that do search engine optimization or SEO. And and there was a little bit of folks early on who were like, oh, that's all evil. That's 100 percent. You're trying to manipulate things, therefore bad, therefore take action. And there was a.
of engineering his name was wars who who really had the right approach he was like look these are small businesses they're trying to do the best they can to make sure that they rank well because they think they have some of the best services on the internet so we shouldn't begrudge them trying to rank well we should give them good things that they can do like make your site better you know make it faster make it easy to navigate and so that was that was really kind of a turning point where a lot of folks who might have been like antagonistic towards seo so it more like this is energy which can be channeled in a positive way which i think is critical because you know folks are just trying to do the best thing for their business there's a few that are Yeah, of course. Bad actors. But for the most part, you know, people just want to know, give me the ground rules. Make sure that everybody's behaving consistently by that. If there's somebody I see violating the ground rules, can I tell you about that? And will you take action on that? And so trying to make sure that people know the right positive things to work on and the right negative things to avoid, I think helped diffuse a lot of the tension where it shouldn't be SEOs or websites versus Google. It should be, or a search engine, it should be working together to give the best results. Yeah, that makes sense. Another thing that's interesting about your time at Google was just how long it was. I think the average tenure now is two years or less at a tech company. I don't know. Yeah, I stayed there for a month short of...
17 years it's a good run yeah yeah so how did you think about your work over that long long period because you came right out of a phd right you didn't finish yeah um maybe someday my dad's like you can still go back i'm like i don't think i'm gonna go back then i'm good um in the beginning like you're you're an ic right uh how long is it before you're managing people what did that whole progression look like yeah i got to code for about five years before they like in 2004 2005 yeah um okay so worked on safe search worked in the ads group and then i was on quality was on quality for the rest of my time at Google. And it was funny because for a while I was like, spam is going to be an issue. And it was it was not a popular opinion within Google. Yeah. For a long time, people thought because they thought the algorithm was so good. Yes. I thought Google couldn't be spammed. And it was because I worked on safe search and I found a loophole that I was like, oh, no. Oh, there's going to be a problem here. Yeah. So you use an example on the history of the Internet podcast where like someone bought.
An expired domain, turn it into a porn site. Yes. Stuff like that. Exactly. And so after that, you know, I remember having an argument with a very early employee of Google who was like, well, that's easy. You just solve expired domains and then you're done. But that doesn't take into account guest books and award programs and fake awards that you'd give just to get links to people and social engineering. And so it was almost like you could see this thing coming down the horizon and everybody else was busy. They were working on other stuff. Google was great at the time, like back in 2000. You'd have a hard time convincing people like there was ever going to be spam. So there was some tension there for a little while. And but eventually started working on spam.
And after a little while longer, again, this one great VP of engineering named Ours was like, he invited me to his office one day. And he was like, Matt, bearing in mind that you can't say no, how would you like to manage the web spam team? And at the time, I was just a lowly engineer. I was like, oh, I guess I can't say no. So I guess I'm a manager now. Is that your managing tactic now for negotiations? Oh, that's such a bad tactic. Because if they realize, they can be like, or I could go work somewhere else, or I could just say no. But it actually worked out because you become less productive, but you enable so many more people to become productive. And so really kind of worked on it for the rest of my time at Google, being a manager and trying to guide and help people to partner with them to figure out how to make the quality of search results better. If you were thrown back into Google right now, how would you be thinking about deep fakes?
oh man so the idea that you can create a video which has somebody you know falsely super posing someone else's picture is a super hard problem because like in theory okay you could do content aware hashes where you could say this video is extremely similar to this other video except for like these pixels but then people can like change the histogram and like add some noise and add a chyron or something so it's fundamentally a hard problem i in general with web spam the philosophy was do as much as you can with an algorithm catch the residual with you know manual spam people who are you know well trained and extremely good at being able to catch things and then use that as the training data for the next wave of algorithms okay um but fundamentally like when you're synthesizing new content as opposed to just republishing old content that's a lot harder to detect okay that's a tricky one so yeah that job is not over yet at google no and i think the main thing is just have policies like if you have a good reason to believe that there's a deep take how do you how do you handle that process because i don't think you'll be able to completely automate the detection but that's a problem already right with like deplatforming demonetizing or monetizing right same issue totally so after that quite
Amazing tenure at Google. What motivated you to join USDS? So it was interesting. I was in Nebraska at the time with my wife because I was working part time and my wife had said, why don't we do something fun? And I was like, OK, you get to pick what's next. Yeah. And she's like, well, my family lives in Nebraska. Seems great. So I'm like, all right, let's try that. And after like eight or nine months, I was like.
Nebraska is wonderful. The people are really nice, but I also want to try something new for a little while. Did you go to the Berkshire meeting? I did. The Berkshire Hathaway meeting, you buy one share of Berkshire Hathaway, which you can buy for like 140 bucks. And then you can go to this meeting and watch them grill Warren Buffett for eight hours. And they also give you free honey bun treats and all the Dairy Queen stuff. So that was a fun April, May kind of thing to do.
But it was interesting because I saw a ton of people that I really respect going to D.C. to try to make a difference in government. And Mikey Dickerson, who was the first guy to run the U.S. Digital Service, had come back to Google. He'd given a talk and talked about the impact that you could have. And so at that point, you know, I was like, all right, something interesting is going on here. I want to see what's going on. And had an amazing six months. The election happened and they needed somebody who could sort of steward the U.S. Digital Service and make sure it would still be in good shape. Because the person in front of me was an appointed official. Mikey was a political appointee. Yeah. And so they needed a sort of interim acting administrator. And so I was willing to fill in for that role.
well and to be clear like a bunch of people stepped up like yes you know some folks left but at this point um something like 75 percent of the people who are in the u.s digital service joined during the trump administration because we've got this sort of two to four year tour of duty model so we're not supposed to stay forever we're supposed to come in and bring in skills from the tech industry make systems work better inside government and then and then head back out again yeah and so a fair number of people like there's about 10% of USDS has been around longer than I have. So at some point I need to find the next person to hand the baton to. So can you just, everyone maybe has heard about it, but doesn't exactly know what you work on. Can you just like break it down? Absolutely. So the US Digital Service started when healthcare.gov went down, caught on fire. The website was spinning shrapnel everywhere. They turn on CNN to see whether the website was back up or not. And it turns out like, you know, a bunch of...
contractors had done good work but the integration points when things were supposed to connect were not really all that well tested and so bringing in best practices like let's get everybody in a room try to do a blameless post-mortem let's add monitor monitoring you know site reliability engineering practices those kinds of things like that was what allowed healthcare.gov to make it through the enrollment period and so after that people said you know we need more technologists in government how many hundred million dollar boondoggles do you have where you spend a bunch of money you spend three years writing the requirements then four more years writing the tech. And by the time you're done, it doesn't work or it doesn't work as well as you expect it. And so the US digital service is sometimes we're a little like a SWAT team. When a system goes down, you know, US Transcom has a database that's down or something like that. We try to work with them and try to figure out how to get it back up. We also do discovery sprints where we'll come in for two weeks and we'll say, we think this is the problem. You might think this is the problem, but it turns out over here is the bigger constraint. And then we'll do anything from months long engagements to years long engagements. So we've worked with USCIS, you know, to try to help immigrants become citizens faster. Veterans Affairs, trying to claim, you know, health benefits and everything like that. And so it's fascinating because you are a
But you come in for a limited tour of duty anywhere from six months to two years to four years. And then a lot of folks head back into industry, but a surprisingly large number of them say, it turns out I'm ruined. for private industry. Like, yeah, once you have helped make a student loan wizard that literally helps your sister, you know, fix her student loan and be able to get a car and have a better life, like, you're like, why am I going to go work on Uber for dog walkers? Or how to deliver weed to people better, which is a fine problem to work on, but you're not saving the world if you're doing that. And so a lot of people are like forming startups in civic tech. They're helping to build state digital services. They're, you know, doing consultancies all the way down to some folks who decide to stay in government, which is really exciting because then you've got, you know.
people with good emotional intelligence hopefully good technical ability who can say that's not the way a computer works oh dude we're gonna get to that like the whole facebook hearing oh my god um but i'm curious about who makes up your team like do you have people like writing cobalt patches like how does it what does it look like we do have one guy who taught himself cobalt for fun really yeah because there's a lot of cobalt yeah like a bunch and in fact i could show you a picture of a um Oh, man, there's I won't name the agency, but there's a room where they file bugs by printing out COBOL on paper and being like the room is sorted by line number. Yeah, yeah, yeah. It's it's super scary. But so we're about one third engineers, one third designers, one third product managers. We also have people who are like procurement experts, you know, who help decide how to buy things better. We've got an amazing talent team. You know, it's always good to have a lawyer. How many folks? We are about 180 people. And and it's fascinating because you've got everything from folks who, you know, have been in government and wanted to.
maximize the amount of bureaucracy that they could hack to, you know, folks who are like Facebook engineers. You know, we had a staff software engineer come from Google. She could only come for the summer, but we were like, come on down, you know, and let's see how much work we can get done. And it's really gratifying to see people do that. That's cool. So yeah, do you have a different kind of onboarding process, given that you know that TorbDuty is limited? Yeah. Most of the time, if you want to join the federal government, you go to a site called usajobs.gov. We, you know, if you're interested in joining the U.S. General Service, go to usds.gov slash apply. And you can literally do it in 90 seconds. Like if your browser has autocomplete, it's like your name.
email address, that kind of stuff, and a resume. So if you've got a resume ready, you just upload that guy. What happens then is if you're a designer, designers look at your resume. If you're a product manager, product managers look at your resume. We actually do one sort of technical interview that's like, okay, how good of an engineer are you? Can you write some code? And then if that goes well, then we'll do a follow-up technical interview and one emotional intelligence interview. We don't want to hire jerks. It goes to a hiring committee that determines whether this person is minimally qualified.
We're always hiring. We have sort of this rolling. It's not like we have batches. We're like the application phase is always open. And and it's fascinating because like, you know, these folks could probably earn a little bit more money, although the government can pay up to one hundred sixty five thousand dollars a year. So not a horrible salary. It's cheaper in D.C. than in San Francisco right now. And so it's we try to streamline the candidate experience to make sure it's not, you know, as governmenty as it as it often is trying to join. Like, yeah, hiring startups and knowing what that experience is like. Yes. Very different. Yes. Let's go back to the Facebook or tech in government.
How are you guys thinking about educating people in government about how the computers actually work? Yeah. So, you know, there's Jen Palka at Code for America, which is an amazing nonprofit, I think has said, and there's a lot of people who use this phrase that government is who shows up. And so there's a, you know, there's amazing amount of passionate people on Capitol Hill and throughout government. But I'll give you one stat. HUD, Housing and Urban Development, has 8,000 plus employees. According to the statistics, the number of IT experts, like there's a job classification, 2210, blah, blah, blah. We'll just say IT people. The number of IT people under 30 at HUD.
is zero so you know if you can get one good technical person to come in it can make a huge difference you can have a huge impact and so um you know a lot of times you might have thousands of people on capitol hill but you know they have to be experts on a bunch of stuff farm bill and you know all kinds of stuff and so technology is not necessarily their core expertise and so getting some technologists who are willing to like come and say you know this particular product is actually snake oil you know or we evaluated this and it looks great you know people are even if it's just the process of selecting a contractor like the dirty little secret is the government doesn't do as much work as the contractors who you know they they you know bring on board to do stuff and so if you get a bad contractor or if you don't know what good looks like because they write agile agile agile all over the request for procurement yeah scrum we i'm a certified scrum master 9 000 or whatever it is which i'm sure there's great yeah i'm not saying it's some of my best friends are scrum masters that's right um but it is i'm gonna start using that line if you don't mind um but it's also the case that like for one procurement we said okay instead of writing a pile of paper what says how good you are
Why don't you submit some code to it and work to us? And we're going to have engineers actually grade the quality of your code. And so if you get better contractors, you get better results whenever, you know, the finished work product is done. And have you figured out a way to give someone a feedback loop that's fast enough that makes them feel like they're having an impact? Because like, all right, let's just say I'm like, I'm amazing. I'm the best software engineer at Google. I'm going to go join HUD. Am I not just going to be pushing a rock up a hill for my entire career? OK, so there are definitely days where you feel like in government you were pushing a rock up a hill. We sometimes use the metaphor of paper cuts. So it took us.
four years to get access to slack for example um get permission to be able to use slack at least within our particular group um but every so often you have those breakthrough days where you're like oh we actually convinced this person that this policy doesn't help people actually hurt people um or Yeah, I could. And the scale of the impact is large. It's huge. Like if you can shift the ship of state by one degree, that's tens of thousands of veterans who are getting benefits or tens of thousands of small businesses who are getting certified faster and more accurately. And so, yeah, there's hard days, but it is super deeply meaningful. And if you absorb a few of the paper cuts, you make it a little bit easier for the next person. So like if by the time they show up, they can get a good laptop on day one that has access to some modern tools. They're like, oh, you know what? Working for government's not that bad. I'll take the next three paper cuts and then the person behind me will have an easier job.
That's great. Yeah, I remember hearing that VA story about, like, you need to, like, downgrade your version of Acrobat to use this product. It's so sad. But so, and I have to say, like, the one thing that I try to avoid, and I've not done it well so far in this interview, is giving more credit to our federal partners. Because, like, what you find is people who know the right answer or who are deeply dedicated, committed, passionate, but for whatever reason don't always have the power to, like, get the right answer to the right level or to push through some regulation or, you know.
overcome some resistance and so if our goal is to come in and find those amazing people who are trying their darndest to try to make sure the right thing happens and give a little bit of extra wind beneath their wings like that's a fantastic model because it's not like we're not the world experts on you know how this part of government we're also 180 people so it's like you couldn't exactly yeah so and that's why you know just being able to find the the leverage points where we can enable good things to happen and work with and collaborate with federal partners who are the true subject matter experts and the real like heroes in the story like that's when things really go well okay so sometimes you kind of like drop in and you're like a pm basically like making that happen okay gotcha all right There are a bunch of questions for you from the internet. So we're going to just knock some out. Okay. So Michael Wang asks, how does USDS decide whether to outsource something to a private company or build a software in-house? Yeah, that's a great question because it kind of goes back to this last thing we talked about. Like fundamentally, that's going to be the federal partner that we work with that's making that decision. And so it's like, do you buy, do you build, you know, can you use something open source or off the shelf, you know, you know, commercial technology. And so it's rarely the case that US digital service is making that precise call. It's more like we might do a discovery sprint.
dig into something for two weeks and be like, you know what, this off the shelf software as a service product will work just fine for the 90% case. And then sometimes it's like, no, you have to build your own grants management software, but we'll help you find a good contractor or help vet them or, or help make sure that the contract is written well, those kinds of things. Uh, and so, you know, if you can just buy something commercially or, or repurpose some open source, like great, you should not reinvent the wheel. Um, but if, if you've got a really unique need, then do us total services there to try to help figure out, okay, how do we, how do we fill that with the minimum amount of work and money? Cool. All right. Next question. Spencer Clark asks,
It would seem that the government is far behind private industry's technology. To what extent is this true and what can be done about it? In addition to that, how should we gauge the progress of institutions like the USDS? That's such a good question. So I sometimes joke, and this is not intended to be a knock against, again, the amazing people who are trying their very best to make things happen in government, but I sometimes joke that government technology is frozen in 1995.
And the reason that I picked that date is because bug bounties, which are just an idea of like if I find a security hole, I'm going to alert a company and the company gives you money and says thank you. Bug bounties were invented in 1995 by Netscape, which was a browser that came before Firefox for those of you who weren't. Yeah, exactly. And then the federal government had never done bug bounties before 2016. So the Defense Digital Service, which is an amazing group of individuals at the Pentagon, ran something called Hack the Pentagon and later hacked the Army, hacked the Air Force, hacked the Marines. They've done a ton of bug bounties. And it increases the security of our country in all kinds of different ways. It's cheaper, faster. You find more security holes. Bug bounties are fantastic. You look at your quiver of tools. It's a great tool. So bug bounties were not put into place until 2016. Now the government is sort of thinking more about vulnerability, disclosure policies, bug bounties, that kind of stuff. Now we're kind of at a phase where I've seen, even just in the three years that I've been in government, a lot of folks are like, OK, how do I move to the cloud?
How do I make sure that it's secure? So if you think Amazon Web Services was introduced in like 2006-ish, if you can, in three-ish years, go from 1995 to 2006, instead of going one year per year, you're doing like three years per year. And again, not to claim that that is all the work of the U.S. Digital Service, because there's amazing groups. There's 18F, which is a group in the General Services Administration. There's a ton of super leaning forward chief information officers, CIOs. So a whole bunch of people all collectively pushing the government means that if we're moving...
through bug bounties through you know on-premises email on to people thinking about how do i move my stuff to the cloud my gauge of progress is hey we're only 13 years behind instead of you know 23 years behind and so so that's pretty good progress from our standpoint that seems awesome and in particular your progress like how do you rank yourself there yeah you know i think the fact that we are the u.s deal service is still here we're still working on projects that matter we're hiring and uh we're able to have an impact for me that's like if we've got good work to do that's the primary you know measure success that i care about okay cool Steven Sturgis asks, with GANs, General Adversarial Networks, getting more and more powerful, is the USDS thinking about the future of data authenticity? That's such a good question. And it makes me put my web spam hat on. So you can use a GAN to make a fake person that doesn't actually exist. And so a picture of someone that looks completely real, but is just...
invented by a computer, which is a huge problem for something like spam because you can AstroTurf comments and be like, I am Bob Smith. And here's a picture that doesn't look like any other. You haven't just stolen someone else's picture. So it makes it harder to figure out, is this comment authentic? And is this data authentic? Luckily, the profit motive to spam government, like there is some, but primarily the sorts of spam that we've seen are things like fake comments, you know, on the FCC or, you know, various other places. The Wall Street Journal had a good article about that. So typically, the US Digital Service is more like we're implementers. You know, if there's a system or process that needs to be examined, we are happy to help.
There's this amazing group called the Office of the Federal CIO, and they think more about policy. So, you know, what should the federal government data strategy look like? What should the federal cloud strategy look like? Those kinds of things. And so, you know, how much open data should people have? And so I think all of those are hugely important. Data authenticity, at least as far as with people spamming or creating fake data, is a little bit outside the scope of the kinds of things that we typically see. You know, as we do see more movement toward like data interoperability. So that might be a way where you could be like, OK, this seems like fake data because it's like two standard deviations.
out from what the typical stuff seems like so you could do those kinds of things but we haven't honestly the kinds of problems we run into at the us digital service are more like here's a paper process can we make it electronic here's an electronic process but it sucks can we make it you know like private industry would do where you can do it on your phone and it's you know no no obvious glitches yeah um and there's so much work to be done just on that sort of before you hit bleeding edge gans totally yeah that makes so much sense uh all right we have another google question so john doherty asks how difficult was it to communicate google's algorithm changes in evolving seo best practices without leaking new spam tactics oh man Good question from John. So it was interesting because it was I would go to search conferences and a lot of Googlers would go to search conferences and we would pick up, you know, what are people talking about as far as black hat tips and tricks or like what are they talking about on search forums and stuff like that. So we got a lot out of that participation. We would learn and then we had to be careful about how we communicated. But at a very high level.
uh my goal and i think a lot of how google communicated the goal was to say look here's where the puck is here's where the puck is going to be like move toward where the puck is going to be like make a site that works on like we were saying you need to have a mobile site you know way early before everybody realized mobile was going to be such a big thing you need your site to be fast you know you need to think about you know do you are you practicing you know good design those sorts of things and so um it usually wasn't that hard because you're like look most sites if you do a site audit there's things where you could just say move toward this mountaintop and you'll be good. Uh, there were definitely a couple of signals or dimensions that I had to be careful about. Um, I talked to a partner at YC on, at demo day and they were sort of saying, I just have to be a hundred percent honest because there's like 5,000 founders. I can't keep them all in mind. And so like, if you're honest, you don't, you don't keep track of what you're saying. Uh, so there might be times when I could, you know, I was, I did my best to always be honest, but I might, you know.
frame things in a way that's like here's the positive way to talk about it and leave out you know and you could probably make some money in the short term doing it this way right but it historically wasn't too bad it wasn't that big of a deal okay um van man 0254 totally a real person exactly totally legit i completely trust this comment uh how can smart tech folks better contribute to regulatory and policy discussions in government it's a great question i mean there's a lot of stuff happening at the federal level but there's also a lot of stuff happening at the state level and even down municipal city county sort of stuff so what i would say is you know, show up, show up to the city council meeting, show up to the, you know, the state legislator and say, I can help answer some policy questions. Uh, there's a guy who literally was like buying DVDs of the Virginia state legislature and their transcripts and like getting them turned into, you know, actual transcripts and then just making that available to more people. And so there's like all these grassroots ways to like encourage people to understand how government works. The other thing I would say is like, smart tech folks you should run for office like we need people who like we don't have that many computer scientists who are elected officials especially at the federal level so like it's not you know it's hard i'm sure but like it's it is a thing that is possible to do right uh and so um especially at the state level or at the federal level like you would be amazed how much of difference it makes to just show up and be like hey i'm a resource if you want to hear about x y or z you know i'm kind of the world expert on this part so like if you have questions about whatever happy to help yeah cool i mean
Also, not for nothing, there are a bunch of government tech startups that have gone through YC based in D.C. and other places. Totally. Yeah. You can do this in a for-profit way. Yeah. And there's a whole new generation of contractors that are like, hey, you know, we see a bunch of opportunities. So, like, yes, you can go the nonprofit route. You can go into government. You can offer your resources to free. You know, you can be a consultant. But you can also form a company. Like, people have started to on health care. And there's so much, you know, redundant waste in there. There's like a decade's worth or a generation's worth of stuff to be done there. There's a bunch of stuff to be done in government as well. Yeah, totally. All right. Ronak Shah asks.
Well, they say, hi, Matt. Nice to hear you on the podcast. What's your best pitch to high performing startups in the Bay Area to adopt more of a human centered design? Says something that the government has been moving towards surprisingly well, but that fast moving startups have neglected. Yeah. You know, there's this there's this myth that like the lore about Steve Jobs was always like, well, if I ask people what they want, you know, they'll tell me they want faster horses instead of cars or something. And so, yes, there is room for the occasional fifth standard deviation genius who's like, I know they think they want this, but they really need this, you know, an iPod or whatever. And.
but most of us are not steve jobs we're just not and so if you talk to users you can only get so far off base like it's amazing to me you know i i went to a someplace recently and i was filling out the you know the register on the ipad and the lobby kind of thing and they ask for an email address and but they don't have like the at sign you gotta go hunt down and press shift shift shift you know to find the at sign and it's just like watching a user doing the journey map seeing what the pain points are like people underestimate how important it is to be beloved, like just goodwill. Like one of the things that people love the most about Google was the logo, right? And it's like, that's not hard to do, but it's like, it's worth putting a few people on coming up with cool, fun Pac-Man logos. You don't think it contributes to your bottom line, but it kind of does. Like whenever Zuckerberg got testified and grilled in front of Congress, the market cap went down to like $129 billion in one day. And I always had a hard time at Google, you know, saying, okay, yes, we should talk to webmasters and publishers and SEOs, search engine optimizers, but how do you know how many people?
should be allocated to that like it should be at least one right the first one helps but then you don't know how far you go until you get to diminishing returns so we always had a hard time quantifying like what is the value of goodwill and i think like losing 129 billion dollars in market cap in in one day is like one really good measure of goodwill about whether people like you or not and so like don't wait until the congressional hearings roll around but this is a dangerous conversation right because i think a lot of tech companies are like okay how do i get out there more without offending the other side right so they're just like wading through very carefully like oh if i make just to make it very simple like if i make the right like me by allowing you know gun videos and whatever gun has whatever the left will hate me so like how do you do that So I would say, yes, there's like 2% of issues that might be hyper-partisan and divide people and polarize people. But there's like 98% of issues that are like, I was literally trying to buy insurance the other day. And I had two websites open. And the first website was like, okay, step one, we're going to need you to make a login. It's going to be this password, six to 30 characters. The second website was like, tell us your personal information. Tell us your credit card number.
Which one do you think I gave my $400 worth of insurance money to? The second one that was like super easy and like no pain points. And you could literally see like one was like 1980s style static websites. And the other one was like hero images, you know, and cool stuff. So but design is not just what is pretty. Of course, it's not. It is thinking about the user and how to make sure that they have a good experience. And I honestly think that is like a secret competitive advantage whenever you talk to, you know, a random company and they think about, you know, what is my net promoter score? And yes, you know, net promoter scores have their own issues. But like if you're not thinking about how much your customers like you, you probably have a competitor who is thinking about that. Totally.
I mean, it's so much that it's a cliche at YC, but we basically shove people out the door to go talk to their users. And we see that with search engine optimization as well. If you talk to five users and say, what would you type to find this page? Or here's your problem. How would you type it? You will be radically surprised by the kind of words they use. Is it a USB drive, thumb drive, USB disk, this kind of stuff. So if you've got a friend who's afraid to insert a USB drive into their computer, you've got to think about why are they afraid? What makes them afraid? What kind of words are they using? All that kind of stuff. Do you have an opinion on the size of your data set? This is like a constant debate among some folks.
I mean, I feel like the first nine or 10 people you talk to get you the biggest amount of value. The team that we have at Veterans Affairs has literally talked to 5,000 veterans. Now, that's over a course of like four years, right? But I mean, we built one feature and it was... So if you've been discharged for like other than honorable reasons, so traumatic brain injury, PTSD, don't ask, don't tell, whatever, it's really hard to get your paper upgraded because you have to like, it depends on the service. You might have to fill out a form. You might have to send it to VA or DOD, Veterans Affairs or Department of Defense. What does paper upgraded mean? So that you can get an honorable discharge. So that you're eligible for health benefits and all sorts of other stuff. And it was crazy because we launched it. People love it. And somebody was like, well, who told you to build this?
You know, where does this fit into the software development lifecycle and the enterprise planning, you know, whatever? And the answer was the veterans told us to build this. And so it took one person. Her name is Natalie, by the way, taking the ball and pushing really hard. And she got this amazing group of folks who helped her. And now that community of veterans has like a tool that they really want. And it is amazing. Like, yep, the first 10 veterans you talk to are the most helpful, but the 5,000th will still help you make your product better. Super interesting. Yeah, I remember one time I was at a post office in Japan and they had a bunch of different grade glasses tied to the stand where you filled it out. I was like, this is so perfect. Wow.
You would never think about it until you see the 400th person like leave their glasses there. Yes. Oh, this is what they want right here. Yes. And it's crazy to me that like.
There are some companies that people love, you know, Vanguard or TiVo or pick your favorite. Right. And typically they love them because they delighted them in some way. Or it can just be like a lot of people like Google because it's just like I show up. It's always up. I get the answers I need. It's fast. It is as relevant as I think humans can reasonably achieve or whatever. And then I leave. And just the sheer like being able to deliver over two decades now, a product that just works and then gets out of your way and doesn't annoy you, doesn't show pop up ads or whatever. Like that is a way to engender a lot of goodwill with people. Yeah. Yeah. I mean, just thoughtful.
is like so it doesn't have to be cute yeah to be thoughtful right whimsy is you know maybe it's optional but yeah but delight or just caring for the user is huge yeah all right adam hoffman asks what are legislators the government and the general populace populace most getting wrong and how they conceptualize the internet oh man that is such a good question i'm not sure i have a great answer i mean most people are not at the level of like ted stevens was where he was like the internet is just a series of tubes you know like people have a more sophisticated uh conceptualization now i think you know the internet is a huge big place and you've got everything from great actors to bad actors um a lot of times the kinds of times when people like want to pass a law or something to forbid something on the internet you can just say like well what if somebody were doing it offline how would you how would you treat it and a lot of the same metaphors apply the other thing is like you don't need to specify the specific mechanism you don't need to say you can fax something because if you if you bake into code or into law that this has to be faxed right that's gonna affect things for the next 30 years until there's a new law that supersedes it and so like baking in the idea of what you want but not
hard coding the specific technologies that are used are a little more likely to make something evergreen so that it's just like the data can be electronically transmitted and then you don't care if it's via fax or chat or whatever you or protocol buffers or json whatever yeah it does you're not you're not hard coding something to a specific technology i think that would that's probably the best i can offer on that one okay all right it makes sense uh next question uh rafael ferrera asks is it possible to live without google i think there are some interesting questions like beneath this though um so he says How do you think Google affected people in searching for answers and content now that everything is just in one click? That's such a good question because, yeah, people lived without Google at least up until 1998, right? Then tens of thousands of years. But now I've been to a restaurant up in Toronto where they literally have a little indentation where you both put your phone in and then you put the wooden thing on the Faraday cage. Yeah, and it's like it tucks your phone away where you have to be present with the person. And it was hilarious because I went to a dinner that was at that restaurant and like three or four times during dinner, I was like, oh, well, I can just look up, you know, when the Eiffel Tower was invented or whatever.
But the food plate is sitting on the little, you know, phone holder. And so you're not able to get your phone. But at the same time, I do think that we're a little more like I feel like my attention span has gone down. You know, you got you don't have time for boredom anymore. Instead, you just hop on Twitter when you have five minutes to waste. Twitter is a great way to waste 35 minutes. I see it oftentimes with friends. I mean, I did myself to like online dating as an example. You get in this like eternal optimization problem. Yeah.
oftentimes you don't think like, oh, I have to do a full loadout if I'm going to load something else in here. But you see people just like, oh, I can get someone who's like 10% more funny or more attractive or something. Same with restaurants, right? You're at this place like, it could be better. It could be a cooler phone holding thing. Well, and it's strange to me that, you know, there's somebody who just wrote a book called How to Break Up With Your Phone in 30 Days. Are you still doing that, by the way? It failed horribly. I tried that as a 30-day challenge. I still have my phone, so I clearly didn't break up that much. But just like, I have been trying to spend a little more time being active on weekends instead of being on my computer all the time. I've actually lost like five pounds doing that. So I'm like,
yeah we could all step away and do a little forest bathing or you know that kind of thing as opposed to just like you spend three hours on the computer and then you're like what did I actually accomplish yep so um so I think this is kind of putting their finger right on the pulse which is like uh maybe the pendulum will swing the other way maybe we'll be a little bit more mindful and like okay I will do this thing with the computer and then I'll put the computer away and talk to a friend or you know visit with somebody hopefully we'll see we'll see yeah tbd meanwhile all the dark patterns in the world and all the infinite scrolls indicate no way yeah yeah i did use like piehole to block most of my time wasting sites okay so like i have to be stuck to it do you like to open the browser on your phone and then i do so i like have to turn off wi-fi to like be able to access twitter now but you know it helps because then you're like you think at least for a second before you get back on there i like grayscale grayscale makes your phone terrible it's so boring you take a picture and you're like i don't know if it's a good picture or not yeah and also just taking time off you like you know i've gone away for like a week but offline and it's shocking how quickly you can like batch process it all when but then when you think about it's like wait i probably spent like
10 hours a week in email. But then I just did a whole week of email in one hour. Yeah. Like, why am I refreshing this constant? Yeah, totally. It feels like, you know, with the latest version of the iPhone and with the latest version of Android, like the pendulum starting to swing the other way, like digital well-being, those sort of features. I think that's super cool. Yeah, it's exciting. Yeah. OK, so Tim Woods asks, which job was more fun and why? Oh, that is not fair. I love all my jobs. OK, so.
at google you know you could get haircut and oil change and do your laundry on site and you know see colin powell and like yeah i mean crazy it was a great place yeah it was a ton of fun the people were phenomenally talented um so on a sheer superficial fun level google's pretty fun um but i gotta say like working at the u.s digital service often hard often difficult often frustrating off the charts meaningful like you know there's a lot of people who say happiness is not this hedonism kind of did you enjoy your day and how much candy did you eat it's like did you work on something that you're gonna feel good about you know on your deathbed kind of stuff yeah so they're radically different and i would not have been able to do the job at the us digital service without my time at google and i'm incredibly grateful for it and uh and a ton of people work hard to make google a fun place and a you know a great place to work um but man the people at the us digital service are folks who are just uh incredibly noble and will sacrifice and will wake up every day and try to sometimes push a rock up a hill and a lot of days the rock just comes back down 90 of the way yeah uh so that kind of perseverance and seeing people willing to do that in order to try to make services work better for the american public is it is super inspiring fundamentally they're just different though like say all else equal in some like crazy alternate reality where uh google salary and usds salary equalizes do you think there would be a swing because i mean purpose is super important right you see like even these people who like go for early retirement like they need to do something you have to do something yeah yeah
Do you think it's really like a salary difference? Yeah, salary is part of it. I've heard people say, why on earth do you have to take a drug test, which you do if you want to join the US jail service? Or why do I have to move to DC? The summer's there, hot and humid, and they suck. All good feedback. I'm like, we can solve some of those problems. We would solve those. I hear from a lot, a lot of people who are looking for more mission and purpose right now. Like if you think about.
you know the me too movement you know some of the stuff affecting the tech industry like people don't always feel good to admit which company they're working for now uh or they don't feel good to say yeah i'm just making a little you know a little bit more money for this particular billionaire or i'm adding infinite scrolling so that people spend more time or you know in my game or my app or something like that and so it is it is super interesting to me if i go to a random conference like xoxo which is a neat design conference and i'm like have you you know considered government service the hit rate is incredibly high it might not be the right time for that person they might want to you know work at a different level of government or you know they might have certain uh political proclivities but like a large fraction of people are like no that's that's on my list at some point maybe i need i was talking to somebody earlier today who was like for right now i need to earn a little more salary but you know in two years or in four years i would love to do this to the point where they're like let me come shadow you for a little while that's cool and yeah so um so it's really inspiring to see that that a lot of folks are like they And if you think about it, there's folks who have student loans, you know, who are a veteran or their mom or dad is a veteran, like almost everybody interacts with the government and almost everybody sees ways that those interactions could be better. Well, I mean, as I told you before we started recording, I just interacted with the USCIS. Yeah, I have some opinions about the product. Right. And OK, so a lot of folks are like, I am not the world's 10x best engineer. Can I still contribute something to government? And as we were saying before you started recording, like.
Um, a lot of the stuff that we do at the US digital service is not rocket science. It's like, Hey, show me the status of my claim online. So I know, do I need to wait two years for my disability claim? Or am I going to get helped in two months? And like, you know, adding a progress bar to see where you are in the process or making a form work on a phone. Like a lot of people have those skills. So if you're, if you're listening right now, like you can do, you know, like a six month tour, you can get a leave if you're at a big company for six months.
And I have to, you know, give up all your stock options or stuff like that. That's what I did. I signed up for a six month tour and that was three years ago. And so we do practice commitment escalation, full disclosure. But it is also 100% the case that like.
you know there is good work to be done and you know two-thirds of trust in government according to mckenzie so maybe take it with a grain of salt and two-thirds i'm always scared when it's such a round number but they say two-thirds of trust in government is driven by the interactions that you have with government so if you want people to trust government more and for it to function more effectively and efficiently and sort of regain trust in an important critical pillar of society you know consider short tour nice okay we got a couple more we're gonna knock it out we good yeah of course um OK, so Snehan Kekri asks, what is math view of the ongoing debate about backdooring encryption for so-called, in quotes, lawful interception? So fundamentally, I'm a technologist. I have a math degree. I have a computer science degree. I have a master's in computer science. I have a PGP, you know, like I have done the public key encryption. And my technical assessment is that, well, let me let me start answering a slightly different way.
A lot of the value that the U.S. Digital Service does is not within a specific silo, but looking at the seams between silos. Because maybe Department of Defense and Veterans Affairs don't talk to each other. And so it's when those service treatment records are transitioning from a service member to a veteran that things might get lost in translation or fall in between the cracks. And it's the same way with security. Like you can have a full frontal assault with a really great protected system.
That's, you know, really locked down. But if you have some little seam over here on the side, like a recovery method, that's not two factor authentication, that's actually just tied to your phone. Then somebody has to do a SIM swap and maybe socially engineer, you know, a customer support rapid pick your favorite carrier to maybe get access to your accounts and then drain your bank account or your blockchain wallet, whatever it is. So like it is often not the case that it's the primary system that gets cracked. The hackers don't care about how elegant it is. They just want to get in. And so it's those seams where two systems join that things often, you know, where there's a problem. So as a technologist, I do not support having a backdoor in encryption at the same time.
That's my personal opinion. That's my personal technical assessment. But I'm also a government employee. And so there's processes in which, you know, people participate in making policy decisions. So if I'm looped in, that's going to be my point of view. You shouldn't have back doors because it represents a vulnerability where bad actors and criminals and all sorts of folks and other nation state governments would totally attack it.
but i also abide by whatever policy processes are run so i say my best i try to convince people of what i think uh but then when the policy decision gets made you know that's the policy decision and that's what party line yeah um has anyone asked you about breaking up tech companies uh only in a personal capacity okay uh no yeah although i will say so unexpected plug. The Federal Trade Commission is looking for a technology sort of coordinator who can basically bridge between two worlds and translate government to technology and back and look at, you know, if a technology company is doing A, B or C, you know, why are the potential reasons for that? You know, what is the business model and why would they structure things that way? So if that's an area of interest to you, the FTC has an open application. I think I've tweeted about it recently. I'll try to retweet about it. And so that's a super interesting.
position where you you could go in and just like i talked about you know having my opinions about encryption how that might or might not affect policy uh you could go in and say here here's my take ftc on you know this small company or this big company or this technology practice. And it's not necessarily an engineering kind of position. It might be like a product manager, you know, because those are the sorts of folks who often translate between the different worlds. So there's a bunch of places in government to slot in. There's also a group called Tech Congress that tries to bring technology people into Congress as staffers so that you can help translate policy and say, here's a good idea, here's a bad idea, you know, be a sounding board for people within Congress. And so there's a bunch of ways to participate in those kinds of discussions. That's awesome. All right. Last time, what's a website if someone wants a job? If you would like a job, please go to usds.gov slash apply. We will have actual people looking at your resume.
we could use engineers, product managers, designers, a lawyer, recruiters. Like if you're a person who can get to yes and you're a good bureaucracy hacker, we would love to talk to you. Yeah. Cool. Thanks, man. Thanks so much for having me. All right. Thanks for listening. So as always, you can find the transcript and the video at blog.ycombinator.com. And if you have a second, it would be awesome to give us a rating and review wherever you find your podcast. See you next time.