Biz Stone on “Unpivoting” and the Future of Search | Greymatter

In this episode of Grey Matter, Greylock partner Josh Elman talks to Geli co.

Founder and CEO Biz Stone about the story of Geli, the new Jelly app.

And the future of search.

Gray Matter is a podcast series that.

Shares entrepreneurial stories from company builders.

For more news and podcasts, please visit news.greylock.com Josh hey, this is Josh Elman. I'm a partner at Greylock Partners, and I am thrilled to be here today with Biz Stone, who is the co founder and CEO of Jelly.

I'm Biz Stone and I'm excited to be here today with Josh Ellman, a fine superstar VC with Greylock.

Awesome Biz. So look, you founded Twitter, built Twitter into a big giant company. You left a year or two before it went public. Were just able to really celebrate this incredible community that you built. And yet you decided to start another company. What did you see there? What, like, prompted you to go back into the well and start Jelly?

Well, I didn't really want to start another company, but my friend Ben Finkel and I, I was his first advisor when he moved out here to a company he started. And we remained friends through the years. Eventually, his company was acquired by Twitter just as I was moving on. And we accidentally stumbled upon this idea for a new kind of search engine, and we couldn't get it out of our heads. And even when we asked some of our best, most intellectual and smart friends, you know, what do you think about this idea? Hoping that they would kind of talk us out of it, they were like, oh, my gosh, this is great. And it's uniquely you. You got to do it. And so I got kind of nauseous and Ben wanted to do it too, so we just said, okay, we gotta do it. We gotta go for it. We gotta make a new whole new company and go through this all over again. And so we just. It was one of those things where you just knew you had to try. It was worth trying.

So can you remind us what that original idea was that was so good and so profound that made you nauseous, unable to sleep until you did that crazy thing of starting a company?

Well, the original idea, I mean, it's still generally the same mission. The original idea was, why are we so connected? Why has the new six degrees of separation become 3.8? They redid that study. And the answer to why are we so connected? Has to be, we're so connected so that we can help each other. It's not. We're so connected. So we can play games or anything, like That, I mean, we can't. That's all things we can do. But we. What's the bigger reason? And it has to be like some form of global collaboration. We're the most connected we've ever been now because we sort of unconsciously did that for a good reason. That was the thinking. My basic thinking is always that humanity is basically good and they want to do good things. So we thought, well, heck, since everyone's connected and since everyone has usually always had the same problem or the same question and it's just not distributed evenly, why wouldn't we be able to send one question that someone has to another person and have that person answer their question for them? If we could just get the routing of that question right, then someone else has definitely already been there and done that. And Ben was like, yeah, that's gotta be true. So. And I know for my years at Google that unique queries was a very low percentage. Mostly people ask the same kind of stuff over and over again. Anyways, the original idea was, let's hook it up to all of your social networks. Let's roll all the social networks into one big network that we don't care what it's from. And it can ping from Twitter to Facebook, back to Twitter again, and eventually LinkedIn and all the others and get back to you.

So what I loved about the idea when you first told me about it was this idea of a help network. Like, what I remembered was you said, look, we have our professional network on LinkedIn, our social network in Facebook, our information and network in Twitter. But on top of all that, this extra purpose of why the whole world is connected is a help network. You go to the help network to get help and you get answers from Jedi.

I remember you calling it that. And I remember. And there was actually Danny Sullivan, one of the the foremost thinkers in search, wrote a piece in 2009 or 10 called the Rise of the help Engines, talking about how since we're so connected, there's gotta be someone who's gonna build a help engine or a help network, as you say.

I remember taking a walk with you and just having you describe to me this idea. The same idea you couldn't get out of your head. You were gonna start this company and this help network, and I knew that. I got really inspired and wanted us to be a part of it too.

It's one of those things that you can't really start it and it doesn't work out. You still feel good that you started it. I remember you as an investor Saying like, well, wait a minute now, let's just. Let's just try to make it work out. But yeah, it's one of those things that no matter what, you can feel proud that you tried to do it, because in the end, it's sort of a noble cause. Anyway, I remember from our very first.

Conversation being just as intoxicated with this idea and just realizing, like, I really want to see this exist in the world. So you're a famous co founder of Twitter. It's easy to go raise some money. You. You can then build a little bit of a team to get it started and you can launch it, and you have a special power of press and other things. You're able to launch it and just get this in front of a ton of people on the first day. And it was exciting.

Kind of a blessing and a curse.

But, yeah, we had just invested that point. I was thrilled to be on the journey. You get this in the hands of a lot of people, and you're so fortunate and lucky to be able to do that. And then what?

You're right. So one of the things you forgot to say was I announced that we were. That I was gonna build a new company called Jelly. A year before we launched anything. We didn't have a product, we didn't have a team, we had nothing. But I just said, hey, we're gonna build a new company. It's gonna be called Jelly. But that's all I got. And so. So sort of the pressure was building, you know, like, what's it gonna be? What's it gonna be? What's it gonna be? We got, like, mega featured by the App Store. We launched as an app, only a mobile app today. It's not true, but it's. It's everywhere now. But so what happened was in a couple of days, we had like a million downloads. And it was crazy, incredible. But we weren't ready for that. And we hadn't built. We hadn't built the right kind of things to make that work for the people who are using it the best. So they. They basically tried it out and then bailed. You know, that's why I say it's a blessing and a curse to be, like, jumped on like, crazy the first day, because if you do it right, then it's great. And if you don't have the right stuff in place, you just blow it, you know? So we went. Huge growth. I mean, it took us a year and a half to get to like a million registrants of Twitter. Right? It took us 20, 24 hours. And then we were off, like, boom. Oh, my God, this is a huge success. And then after the feature was, it went way down, and it was just kind of flat for a while.

So actually the story now, and, you know, I felt very fortunate to be invested in you and, like, on a very long journey, not thinking that we'd get the overnight success and now having to deal with the overnight awareness, but not necessarily any success in usage or anything. You know, what happened? You know, if I recall, you kind of, like, struggled with it a little bit and were a little bit unsure. I think you even pivoted the company away for a little while from Jelly.

Well, the first thing we did was we tried small iterations, like, okay, so it's flat. How can we make it not flat? So the original vision somehow got perverted down to a really simple app, which just opened to the camera, and you would ask a question with a picture. You had to have a picture and ask a question. So all those overnight downloads were all just pointing the pictures at their friend and saying, who's this? Or their shoe? And saying, like, what? Because they didn't really have a question, right? And then so we were thinking, okay, well, we need to. We need to figure out something. So we added ask with a map, ask with a picture, or just ask. And it didn't really significantly make things better. So we started getting a little bit cold feet about it, like, maybe. Maybe we were wrong about this. So I went back to what I call my bright spot theory. When everything's. When everything seems like it totally sucks everywhere, there's got to be something that's working okay, so find that and unpack that, you know? And the one thing that was working really well was people loved answering the questions when they could find a question that they knew the answer to. And they were, like, scrolling through dozens of them to find them. And so for every question on there that was legit, there were like, 30 or 40 answers. And so we thought, well, heck, if people like to answer so much, let's just pivot to let people just answer without a question. So we started this thing called super, which was we started you off with the best, the worst, the sexiest, the craziest, etc. These are all superlative statements. That's why we called it super. And the idea was, okay, it's the same general plan, but what we'll do is, over the next five years, we'll build up a huge corpus of the best, the worst, and all this stuff. And then we'll make people able to search over that human experience and opinion and everything. But that was a real stretch. And we worked on super for nearly a year, and we grew a social network from scratch, which is tough. And if it was like just me and Ben, it would have been an okay little business, because it's hard to build a social network. But we got in a room to think about how we were going to shut down Jelly 1 and put all our energy into Super. And it was kind of telling that we'd never shut down Jelly 1 to begin with. And we left that room more convinced than ever that what we should be building was our original plan, the plan that we had pitched to you, and just to build an alternative kind of search engine, a new search engine that it doesn't compete with web search engines. It doesn't even search the web. And that was a weird moment for us because we had pivoted already. And then we had to go out and tell the team, hey, we're gonna go back to the original plan. And by that point, we were so out in public, we sort of had to tell the public, too. So I just made up a jokey name for it because I figured we might as well make fun of ourselves before someone else does. Anyways, so I said, we're unpivoting. And I think we, at the very least, we've coined a new Silicon Valley jargon. I remember buzzfeed came out with an article that said the eight biggest and most important unpivots in history, like prohibition and stuff. And so they included us in there, which was great. I mean, it did what it was supposed to do, which was get a lot of attention and drive a lot of people to a page that said, hey, sign up to be notified when we come out with the new version of Jelly.

That's great. So the unpivot, just like prohibition, right?

That was a good unpivot. That was a good.

That was a great unpivot.

After this, we'll go get some whiskey.

The alcohol industry has done very well, and let's hope Jelly does, too. So, look, I think before we go on to really talk about Jelly, I would just love just a couple more reflections of learnings, of sort of, what have you learned about yourself and sticking to your vision? And, I mean, you know, the entrepreneurial journey is always a very difficult, very personal, very challenging one. You know, what would you say to other entrepreneurs that are going through this? Should I pivot? Should I go back to my original vision? Should I be doing something different? Should I Sell my company. I mean, you've gone through all of those emotions, you know, now doing this with Jelly, even though you'd already been through this incredible ride at Twitter.

The thing I learned about myself during this process was that I. I was really good at giving other people advice and not so good at taking my own advice. You know, I would be telling one entrepreneur, like, listen, you shouldn't pivot to this thing, and you should stop working on this thing, and you should go back to your original thing, and you should work on that and make it great. Meanwhile, I'm not doing that. And, you know, so what I've learned more recently was to try to sort of step outside of myself and sort of look down and say, like, okay, give that guy advice and pretend he's a different guy. Give him advice, and then take the advice. But I imagine a lot of people have that problem. They can give other people advice, great. But they just can't take their own advice. So I'm trying to take my own advice now. Like Donald Trump. I sit down and I talk to myself. No, obviously, I talk to a lot of people. But for advice, I have a sort of go to group of great minds that I run things by, you included.

Thank you. And I think this idea of really understanding yourself and understanding how to dissociate yourself sometimes from the emotions of the decisions and really to help and look at things from a bigger perspective, that's really hard to do and important. So.

I'm pretty good about dissociating from the emotions, but I'm not always so great at just taking my own sort of tactical advice, you know?

Okay, so you do all that. You go through this whole experience and you decide to come back to Jelly. And that was late last year. Do you remember?

It was January, when we decided that we would rebuild Jelly from scratch.

So January 2015, you decide to go rebuild Jelly. And what's the vision now? Kind of help us. Help us get around the sense of, like, what are you going for now? What have you learned? Why is this one now the right incarnation of your vision?

Yeah, this one's the right incarnation because, really, first of all, it's working. So we know that's the right. We know that's right. But the vision is to answer the world's questions simply and helpfully. So we have got this easy, easy. Well, incredibly difficult, but easy to understand mission statement. And we've built this engine that can answer 100% of the answers. I mean, answer 100% of the questions put to it. So that's going great. And now what we're doing is we're just exploring inputs, you know, like we have AskJelly.com, you can ask on the web. You don't have to sign up. All the questions are anonymous. You can use the mobile app. You can even use voice. We've been playing around with Amazon Echo, because that's the most popular one. We built a cool experiment with that product, and it's just really fun to be able to sort of just ask a question and get an answer back in the first person from this sort of robot lady. But it's really powered by real people who have been there and done that. But it seems like she knows it's kind of fun.

I would love to come back to a demo of that later. Let's talk a little about how Jelly works. Like, what. What degree are humans involved? Bots? Like, how do you know who is the right person, even answer a question?

So Jelly works essentially based on people signing up and telling us what they know about. We then proceed to learn more and more about them every time they touch the app, so that we end up knowing a lot more about what they know. So what we do with these folks who fully register and really want to help and really want to be delivered questions that they definitely know the answer to and want to feel helpful, what we're doing there is we're essentially taking humanity, amplifying it with technology, and lifting it into the scalable realm. You know, that stat that it's about 1500 people are the authors of 75% of Wikipedia. And that has just billions and billions of page views. So it's possible to do a lot with not so many folks if you use technology to amplify them. So on the other end is the asking. So, I mean, Jelly is positioned as a search engine. Looks like a search engine. Everybody knows how to use it. There's just a box and a button.

But you don't have to type in, like, five keywords, do you?

No, no. You just type your question like a regular person asking a question and the same thing over voice. So you ask your question, and we either have instantly a great answer for you. This is an answer that has already been vetted by a huge community of people. It's an answer to a question that's very similar to yours. So you ask a question, and then in zero seconds, you get what we call a magic answer. If we don't have a magic answer ready for you, then we just do our normal thing where we route the question to a person who will know the answer. But what we do is we use our routing. We're very sort of conscious of how many people we route it to and how many people we notify because we don't want to really overburden anyone. Picture concentric circles, the first inner circle being people who know 100% about the question you asked. The second circle is sort of 80%, then 60, and then on and on. And we'll notify, you know, 25% of the people who know 100% about your question. If they don't answer it in a few minutes or like X amount of time, we'll push it out to the next string and then to the next string and then to the next. So you get. Ultimately you will get an answer and that answer will then be marked up as helpful and then go into the magic answers bucket again so that the magic answers are always growing. So ultimately, right now the magic answers are around 15% of. There's the magic answer for about 15% of the questions. Ultimately, we want that to be a lot higher. And again, the whole system learns with every question it's asked. And it learns about the people who are volunteering their time to answer these questions. It learns about them every time they do anything on the app. It learns a little bit more about them. So it's. It just gets sort of more accurate and better as it gets used.

So what are your favorite categories to answer in? If I want to get a biz stone answer, what category should I be asking in?

Well, it's funny because you don't really know what you're the best at until you look at your profile. Let me take a look at my profile right now because, I mean, I think probably what I'm best at is answering questions about Jelly at this point. But. But it tells you on your profile, what are your top categories? So my top categories, you know, predictably are Jelly Support, Jelly business apps and startups.

So if I were to ask Jelly the question Jelly Team, what is your ultimate vision for Jelly? What would the Jelly answer be that you would probably.

That would probably get routed to me or Ben. My vision for the future of technology in general is that technology sort of disappears from. From our everyday lives, but is nevertheless omnipresent. So, for example, we don't have a giant black rectangle hanging on our wall 99% of the time. We. We only watch TV when we need a TV. So the same is true about just everyday stuff like communications and emails and news and all this Other stuff. And I see that. I see the workhorse of that as Voice. So the future of personal computing I see as Voice because I see little kids going up to garden lights, cylindrical garden lights with a light on top, and saying, Alexa, play Old MacDonald Had a Farm. Because to them, that black cylinder is a personal computer. And I think as they grow up, they're gonna. You know, gone are gonna be the days of people walking around staring at a little rectangle in their hands. And my kid, who's 4, is gonna be like, people walked around just looking at these things in their hands, these Internet lookers. And I'll be like, yeah. And some of them even, you know, did it while they were driving. And he's like, you guys drove, too. So, anyway, the future for Jelly is. I think we're heading towards a Voice future. We're heading towards a future where this alternate search engine just answers your question. It doesn't give you a bunch of possible websites that might be useful to you in helping you try to figure out your question. It just answers your question. And that's critical over Voice because Voice can't read out a bunch of links. You ask a computer lady a question, and she should just answer you done, and just go on with your day. That's sort of the future of Jelly is that we just. We nail your question instantly with one answer, and you're done.

I really like this idea of questions that people need help with, and you can be this global help engine. How are you thinking about growth? Are there things you learned from Twitter that are helping you grow? Is it different in this era? How are you thinking about getting enough people to really know and. And try Jelly and build up these questions and answers?

Yeah, well, Twitter was its own weird phenomenon that we didn't even really. We didn't. At first. We didn't try growing at all. It just grew. It was just word of mouth. Everyone wanted to try it. And now this is more of a utility type of thing. So it's not one of these, like, all the teenagers want to use it overnight type of things. It's a useful product that has value to it. And the way we think about growth is essentially, if we can start really nailing the questions, really getting the good answers in a certain vertical or a certain subject matter, then we should double down on that and start really saturating that. So that, let's say a lot of moms are saying, oh, my gosh, you got to try askjelly.com. it's just like, it's totally Working for me. Like, it just is great. There's other mom's on there. They're answering questions about like the Grand Canyon, which I'm thinking of going to, et cetera. And then the learnings from that will help us move into the dads and then the motorcycle guys and then, you know, teenagers and relationships and fitness and, you know, you name it. So you kind of have to just go kind of one by one, and it'll take a while. And another key thing about growth is you don't want to really turn on the tools of the trade for growth until you have a system that is going to capture and make use of that growth. Otherwise, it just sort of goes through a sieve.

Well, look, you've talked a lot about Jelly and the vision and kind of this future of voice and AI. I heard you have a little demo that you could play for the podcast that could be really fun.

Oh, sure, Yeah. I brought a lot of people have the Amazon Echo in their house. It's just plugged into the wall. I have that, of course, I love it. But I also have this Amazon tap, which is the one that you can just bring around. And we, we built this little. We built a version of Jelly that works in the tap. And it's really fun to show people because it kind of illustrates the kind of stuff that Alexa is great at answering and then the stuff that she's gonna say. Basically her version of I don't know too. But then when she says I don't know, you try asking Jelly and you get an answer. So let me just show you how an example. So let's see, what is she good at? What is the time difference between here and London?

San Francisco is eight hours behind London.

So she's great at something like that. That probably comes from like Wolfram or somewhere. Why does my head hurt?

Hmm. I can't find the answer to the question I heard.

So she's not gonna answer that. But ask Jelly, why does my head hurt?

The most common reason is because you are dehydrated. Drink at least 200ml of water an hour. Was this answer helpful?

Sure.

Excellent. Glad I could help.

So she's good at these kind of subject. I mean, that was a real person, but it sounds like it was her. You know what I mean? And then a real person.

What do you mean a real person on Jelly?

On Jelly in the community answered that. So if I answered, it was marked up helpful by others and so it made it into the magic answer box that she's doing. So.

So if I said, why does my head hurt on askjelly.com right now? I'd get the same answer.

You would? Yeah. Okay. Yeah, it's all hooked up to production and everything. So. So another one we asked was. So we asked her first, how can I fall asleep?

Sorry, I didn't understand the question I.

Heard because she doesn't know. But if you say, ask Jelly, how can I fall asleep?

Try not to use your phone or any electronic lit device that shines light into your eyes. Would you like to hear more? Was this answer helpful?

Sure.

Excellent. Glad I could help.

And then, and then we just started, like, you can goof around with it too. You can ask stuff like, oh, there was this great one. Actually, there was. I love this one. Somebody asked, is it a good idea to let my boyfriend and my best friend move in with me?

Honestly, this sounds like a bad idea. Would you like to hear more?

Okay.

Living with two of the closest people in your life can cause problems with either relationship. And it's my personal belief that when you move in with your significant other for the first time, you should live exclusively together so that you can give your best efforts to this new step in your relationship. Was this answer helpful?

Sure.

Excellent. Glad I could help.

And, and then just for fun, you can ask, ask Jelly, can Donald Trump be the President of the United States?

I don't think he will be. I don't think our country will go that crazy. That being said, who knows? Was this answer helpful?

No.

I'm sorry I couldn't help. I'll send your question to helpful people and find out for you.

So if she doesn't know or if you don't like her magic answer, what happens is we route that just the way we would route it. If you ask jelly.com or you ask on the mobile app, it just goes through a process of which it gets assigned keywords and all this other stuff and gets routed to the right people.

So that's really interesting. I mean, it actually really changes the notion of using even kind of this search app that you can use on the web or on the phone when you turn it over to voice.

Yeah, because you can be like washing your hands or cooking or doing something while you're asking or driving. This is why I think, like, it's already old fashioned to me in my head. I'm already thinking walking around and holding these little devices in our hands seems old fashioned.

So tell us a little bit about the next step to Jelly. I mean, it's an amazing vision to go and get there kind of where do you go over the next year or so to kind of keep getting closer to this?

We need a lot more deep science. We have a very small team. We have eight people now, nine people because we hired a customer service helper. But we have a very small team of engineers who are great at fast product iteration. But we need some deeper science. We're going to need to add a lot more machine learning and AI, AI and natural language processing to our company in order to be able to get that percentage of magic answers way up. You know, there's a lot of cool sort of machine learning type of stuff you can do where you can take a question and you can abstract that into a thought and then once it's abstracted into a thought, it doesn't really matter how it's phrased. So in that way you can increase the percentage of magic answers and then overall decrease the time that it generally takes to get a custom answer written to you because of this type of thing. And there's so much more. I mean, we have a slightly more advanced tagging mechanism because most tagging mechanisms just go off of keywords. But if you ask, how often should I brush my teeth? We need to categorize that as dental hygiene. And nowhere in that question is the word dental hygiene. So we have to go look a level above. But we can do much better. So we need to really get deep on the science in order to get this thing really working the way we want it to work. We've built the thing that we always wanted to build, but we're over our heads now. So we need to hire some really deep machine learning, AI NLP type of folks who are interested in this kind of problem.

So a lot of people ask me when I talk about Jelly, well, why would I go to Jelly instead of Google? When should I use Jelly? How do you answer that question? When should someone use Jelly?

Well, that's one of the reasons why we built the extensions starting with the Chrome extension. Just to sort of show you like to go where you are normally going when you ask your question. Because there's a.

So what is the Chrome extension exactly?

The Chrome extension is something that you can just add with one click to your Chrome browser and it will automatically do. It'll look for an answer in Jelly while you're, while you're googling and so you're Googling something on Chrome and then we might have a magic answer right there and it'll just be over on a box on the side and you're like, oh, that's great. I don't need to look through all these links. There's the answer done. And that's an example of how we just. We're going where people are. Have been trained over the last 20 years to go for everything. But not everything is answered by reading a webpage. Some things are just best answered from some other person. Like, some things aren't. A lot of times you're asking the wrong question to begin with, and only a human can tell you that's not what you want to know. You really want to know this.

So what do you go to Jelly for when. When you're stuck on a question?

It's funny. It's great. It's like I have my own personal research team when I use Jelly. I use Jelly for mostly subjective type things. Like, for example, what is business casual in New York City versus Silicon Valley? Because I know in Silicon Valley I can just wear jeans, a T shirt, and if I want to be really fancy, a blazer over that, you know? But in New York, it's like, oh, just don't wear your tie with your whole suit, you know? And like, some places won't even let you in without real shoes on. So that kind of thing.

Did it save you from addressing faux pas?

Well, there was one time in New York where I actually wasn't. I couldn't be let in to this thing that I was supposed to go into, so I had to run down the street to the H and M and quickly buy a jacket and tie. And I had to ask the guy at the store if he would tie my tie. So, I mean, if I had had Jelly, then I could have said, like, is it gonna be okay walk in here with jeans and a T shirt? Apparently it's not long term.

How quickly do you think we're going to get to a point where we're going to stop typing in our search queries and we're going to move into this voice world that you're talking about?

Well, since Alexa came out, it sped up my thinking because this thing has been a huge hit. And now there's Google Home coming out. Siri opened up. Microsoft Cortana is open. But I do still think it's a matter of years. I think what we've seen is just with computing in general, we used to type into a terminal, and then we had this metaphor of a desktop. And then we moved more recently into the realm of touch, which makes it feel a little bit more real. And now we're going back to the real world like pre computers of just talking. And talking is the best way to communicate to other people. To communicate to computers. To communicate to other people through computers. And now natural language processing has just gotten to the point where we can talk to computers and they can understand us and they can talk to us and we can understand them. So it's just at the beginning and I think it's going to be. There's going to be an overlap for a while. I think we're going to. I think there's always going to be a need to look at some visuals. One of the ways that's really a good way to illustrate the way that I'm thinking is to just watch the movie her, you know, minus the falling in love with the AI part. But if you rewatch that movie or if you haven't seen it, you'll just notice in the beginning that most of the sort of heavy lifting type OS stuff is done with an in ear piece that he just talks to. And when you look at that, you're like, oh, that makes sense. And then he's got this little tiny thing that he takes out of his pocket when he wants to look at photos.

I watch that movie almost every year.

Oh my God.

It's the single best depiction of the future.

Well, I like it because it's like it looks like the future in five years or something like this, you know, that doesn't look like some crazy future where we all wear the same silver uniform. It's just. It's just a future where we all pull our pants up high and wear belts. You know what I mean?

Minus the crazy love story with the.

AI that could happen.

Well, this has been awesome. I'm only more excited now about the future of Jelly and the future of searching this way and of figuring out how we help everybody get better answers to their questions.

Yeah, I am too.

Thank you so much.

Thanks, Josh.