Live office hours with YC's Paul Graham and Sam Altman (2013)

what are you working on? So we are building a multiplayer programming game for teaching people how to code. So like Codecademy, but actually a game.

A game? So how do you win the game? So you just beat more and more levels until you're an awesome developer. So you sort of get points somehow for it. It's competitive learning?

You learn a program and you get more points? Or competitive, so there's multiplayer. Is it like write code to kill a bad guy? Yeah, so the first level you got your guy, you write code to move him around and then you kill another guy, so it's like right in the middle that you're killing dudes. Alright, so you're motivated.

Yeah, instead of what is your name, what's the length of your name as a string, yeah, okay, that's cool, but like, no, killed that ogre. Yeah. It's like what you do right away. It's not just badges. Not badges.

Is it launched yet? It is. Actually, we launched it yesterday.

Oh. Well, beta. Things move fast around here. Check it out.

Did you launch because we told you you were gonna be in office hours? Actually, no. We have. But just a coincidence? It is. Somehow. Unfortunate because we didn't have time to prepare. The launch went crazy.

Yeah.

But they were I got home from the from the dinner last night, and I get on a hang out with these guys and they're just at the server terminal control c ing and restarting the server because it's under so much load. How's it doing now? Not much better, unfortunately. We can we can only serve a certain fraction of the traffic that we're getting and that's been going on for twenty four hours.

How did you start working on this? I wanted to learn to code about a year or two ago. I had been a semi technical co founder at my first startup and I tried Codecademy, tried a whole bunch of these and I just couldn't stay stick with it. It wasn't engaging enough for me. And so these guys, my two co founders were like, hey, why don't we actually make a game? So you were the original guinea pig?

I was. They thought you can't keep motivated using existing stuff. We'll make a game where you can kill people? Well, you know, I was it it wasn't just that. Our our first startup, the customers kept coming to us and saying, we keep using your your your company, your product because it's like a game. And we hadn't attended that at all. And What was it?

It's a a company to teach people Chinese characters. I see. And so we thought, well, if we can do that inadvertently, what would happen if we actually made a game?

So how far can you learn how to program by like, how much can you teach people? Right? Because I can remember the kind of crappy programming I did when I was in high school where I didn't really understand what I was doing. So if you look at the stuff that's on Topcoder or Hacker How can you force people to learn advanced concepts when all they really need is to have the right library calls? Right?

Can you make advanced concepts produce advanced weaponry?

So the software engineering part of learning to be a developer, that's something we can add later, focusing on core programming for now. Because you get motivated enough, it's like, Okay. Now I want to build this app. But then you have to get through different levels. Right? Presumably, you get more and more sophisticated.

Yeah. But you could get more and more sophisticated just by writing more and more code and getting access to the right library functions without actually learning any more about programming. Right? Couldn't you? How can you force them to learn more about programming in order to make more powerful weapons?

So you can have things like, Okay, your code needs to run this fast, in this one you need to learn how to use recursion. This is the only method available to you in this one. You need to figure out how to do an anonymous function passing method here.

And generally, if you make the levels hard enough, which you're able to do when they have a reason to complete it, they try really hard, then you can get them to do harder and harder stuff as natural progression of the game. Have you run beta users through this yet? Quite a few actually. What did you learn from it? Like, what went wrong?

Well, first thing that went wrong was that we we started at too high a bar because I had worked in a kind of a semi technical role in my first startup. We assumed.

a whole bunch of prior knowledge that was totally untrue for our beta users. So, you know, we started out like, you know, writing for loops, which we're like, oh, well, that's that's simple. And then we got people with no programming background and they didn't even know how to complete a line. They didn't know a concept of formal notation is the single biggest obstacle. Correct.

What's the most advanced concept you're teaching now?

So, so far, we have some dev levels where it was like, okay, you're gonna need to figure out the targeting strategy for your artillery, so you're gonna fire into the center of a group of dudes, and your soldiers, backed up with your artillery, have to avoid your shots, so you have to make sure that they don't chase into your line of shooting. What era of technology is this?

So it's a web game and you're doing everything in JavaScript. No, no, no. What era of combat was a fantasy. So you're a wizard and you're casting spells to control your soldiers and your heroes and that sort of I see.

How many users did you get? There's no Apache helicopters or anything like that. No, unfortunately. Don't have the art that Free industrial plus magic.

We keep saying do robots. We could do robots. Yeah. It's one fantasy you can make up anything. Yeah. Right? How many users did you guys get yesterday? So we we maxed out the server at 15,000 people.

We had 200 concurrent, but we really don't know because we were actually people were just getting four zero fours. Why not just spin up a bunch more servers? We weren't architected that way. We we didn't think we'd nearly this much traffic. We just posted it to Reddit. That's it. Yeah. We we posted it to Reddit and we got swamped.

Not even the main Reddit. We actually, people on the Reddit threads were just like they were repeated things. Were saying, you know, like, the hug of death, hug of death, hug of death, and, you know, four zero four not working. So that's they were scrambling all last night to do that.

Do you know if it's people that didn't know how to program before that are mostly doing this or if it's just people that wanna play a fun game? So the the people that know how to program already, they're like, okay. When's it on GitHub? When can we this is awesome. Let's get on here. Had, like, 20 people yesterday, oh, lurking me, being like, oh, man. I wanna wanna help out. And when can we pull?

When can we clone? Yeah. So we're thinking open sourcing in the next couple months to really capitalize on the interest. But most of the people, yeah, they're on the learn programming subreddit. They're like, don't know any programming. This is great. And they beat all their levels and we're like, Crap, we need more levels.

Just trying to focus on Do guys know anything about the gaming business? Do you know how to make games?

We're learning is.

the quick answer there. Okay. Because there's probably certain best practices in the gaming business and probably whatever they do would be the starting point. So if you're just wondering how much to open source, I don't know how much they open source things in their world, but whatever they do is probably the default thing to start with. Yeah.

We actually that was the first thing we did when we started the company was we realized, wow, none of us are professional game designers.

Let's find and talk to game designers. And so we've got this kind of core group of people that are advising us, mostly just telling us when what we've built sucks, but it's been very helpful thus far.

Is there anything they told you that changed what you were doing?

They said make robots instead people understand robots. Yeah. They said make robots. Because when you have like controlling your units via code, people think, Okay, that's natural if you're robots. When you say, Oh, it's a spell, you're a wizard, you're adding to the fantasy and it's like, Oh, it's a little bit hard to understand.

We'll see. Okay. So is it robots now? No. Oh, so that's It's just hard to change the art, right? Yeah, the art is. Yeah. Alright.

And are you guys now just gonna create as much content as you can while people are Yeah. We just finished the level editor, so now the hope is we can finally turn out three levels a week using our awesome live coding drag and drop thing as opposed to hard coding all the coordinates and be like, does it work now? Yeah.

So growth first. How many people do you have? Is it just you two? No. We have one guy that's manning the servers right now. He's keeping alive. Hopefully, keeping it Scott, keep it alive. This is the same team from the first startups.

We've been working together for six years now. How did you originally meet? So I was his roommate and I lived down the hall from my co founder. College. Yeah.

College roommate. Did you guys study? Were you guys programmers?

Yeah. So Scott and I did CS and George is the econ film guy. Yeah. I see. And then we graduated and we're like, oh, let's not get jobs. That's gonna suck. We're like, what are you talking about? And we did the startup, and then three months later, it You know what?

Was like, good A lot of the reason people start startups is because they don't want jobs. Yeah.

Honestly, if we were looking at someone's application and they worked for a long time for a large company, that's actually bad to us because the best startup founders probably could not stand that.

Three months at IBM didn't eat in. It was like sophomore year,.

no more. How are you guys gonna make money with this? So it's a recruitment model, basically.

The leads that we generate through the coding challenges provide us with the opportunity to qualify people before we even get in touch with a potential company.

You possibly train these people good enough to make them valuable employees? The recruiters we talked to said, yes, Absolutely. Companies are interested in the developers on your site, and we're interested, and so let us know when you have something. Recruiters famously say all sorts of crazy stuff. Yeah. They do. So Yes. That's yet to be validated.

Yes. Indeed. The other people in the space that we talked to also say the same things. People running coding challenges and doing placements and the boot dev boot camps and that sort of thing. Yeah. We we spoke to actually one Y Combinator company. We asked him how he had done his recruiting. He said, we sent group of qualified recruiters a spreadsheet.

I said, how how did that turn out? He said, oh, we had 50 placements in six months. So I said, okay.

It time? We're out of time. Alright. Nice meeting guys. Thank you.

Thanks.

Okay, guys. You guys guys. Wait. Wait. Come back. Back for a second. You didn't realize that, but that was your Y Combinator interview. You're in the next batch.

You are a plank. Right? Nice. Awesome job.

No pressure, I guess.

Nice job. We just sort of decided to do that on the fly. I mean, that was the that was the first time I talked to them.

Awesome. Hi. I'm Karen. Hey. I'm Finbar. We are making giveit100. com.

What what is it? Giveit100.

You look at me. Okay. We're making a video site where you sign up, you choose something that you want to get better at, and then you share a video of your progress every day. So, what would be a typical example? Like, what's the most is it launched now?

It's in private beta right now. Okay. So, what do you anticipate being the typical use case? Like, what sort of thing would people get better at? The most common one right now is dancing. Reason for that is.

because I made a video of myself learning to dance in a year and I put it online, ended up going You mean you spent a year learning to dance? I did spend a year learning to dance. And, what kind of dancing? Robot dancing.

Okay.

It's all robots today. Ron Code, you hear this? This is the new trend, robots.

So, I ended up getting several hundred emails from people who said, Hey, because I saw this video, it wasn't a video of an incredible dancer,.

but it was someone who started off not knowing how to do it and getting better. So was this a video you put on YouTube and a lot of people looked at it? Yeah. So was this what led to the startup? Yeah. Okay. So you made this video of yourself learning to dance, and then you thought if other people did something like this, it would encourage them to dance too? Yes.

What sorts of things are people mostly showing themselves besides dance learning?

What beta users are there, by the way? We have invite only beta 50 people. 50? Yeah. There's like 4,300.

or so on a waiting list. Yeah. There's.

a nine month old learning how to walk. There's a woman who's recovering from a multiple sclerosis exacerbation. She's relearning how to walk. There's people who are learning how to ride a unicycle, learning a new language,.

learning how to code, learning design. Why haven't you accepted the rest of the wait list?

Well, we're kind of just ironing out some kinks in the product and getting it to the stage where we think it's gonna be really engaging. Is it not engaging enough now? Well, is. We actually have some really awesome engagement stats. Well, if it's engaging enough now,.

you've earned.

out enough kinks. Sure, we have. I think really the major things that I'd like to see personally are the social sharing features. Because when we open the floodgate and have lots and lots of people come onto the platform, want to kind of maximize on that. And, you know, if a lot of them come on and share it and then leave, like, they could have got a lot more people to come in by that point.

So So they're not sharing it enough now? Well, there's no way for them to share it right now because it's totally a a private closed beta. Like, nobody else can see Right. We really just are we're kinda just experimenting on our first batch of people getting their feedback,.

and then we're gonna launch in the next couple weeks. Do they always make videos of their progress? Yes. It's video. That's how it works. We started off as a photo and video site, but then we cut out videos. We cut out photos because the videos were more interesting. So, how do people make videos of themselves learning to code?

Look how much faster I can type. Actually,.

there is someone who's learning how to touch type. But they sometimes talk into the camera, they talk about what's challenging, what they're struggling with.

They'll show actual code, they'll show what they actually built. How many views does an average video get out of, like, a potential 50?

So the view counts, we're seeing about a thousand views a day and we have roughly between twenty and thirty of our kind of small group of users are coming back to website every day. Twenty and thirty out of 50 come back to the every day. So unique of unique visitors. It's not like the same 20 to 30 every day. It's like people will kind of wait a few days and then upload a batch of videos at once.

The videos are hosted on your site, not YouTube? That's right. Yeah. But, we want to use YouTube. We want to piggyback off of YouTube as a marketing channel, the same way we did with my video. So, we'll take really compelling one hundred day challenges, and we will turn it into a viral video, put it up on YouTube and say, made with 100.

It seems like that would have been really important to test during the beta. Will people share these on YouTube and do they get watched? Well, I guess our test for it is my video, which has 3,000,000 views and was shared widely. But, you don't put it on YouTube.

Well, the video clips themselves are on our site and that's something that you can go on every day and see the same people every day, see their clips. No, Sam was saying you should have tested putting it on YouTube specifically. Oh, the clips, the ten second clips themselves.

I mean, think the kind of format that we have on the website where you kind of have this gallery of ten second clips and you can just kind of see them all and consume them all kind of in context and in sequence is really powerful. So you have a view with a page with a whole bunch of little videos on That's right. And you can see from the beginning to end person's progress. The story.

That's the most compelling part about that. If you envision Paul Graham, I'm Paul Graham, I am learning.

To pick which startups. I'm learning to pick.

I'm learning to pick which startups for Y Combinator for over one hundred days. Then you see day one, day two, day three, and then as you hover over each video, it just starts playing. So you can you can watch it for a second for ten seconds. We cap it at ten seconds because I have a short attention span and I'm building this for myself.

What do you think will be the most popular things? I don't mean the most popular things by number of people who do them. I mean, what will be the most popular things for third parties to come and watch? That's a really good question. Third parties. People who are not the people who are practicing.

You've said you built this for yourself. What do you want to watch? What are you excited about watching other people learn? I want to see a good story. I want to see someone who is struggling and and is against all odds, like, doesn't think they wanna do it. I wanna see, like, Phil Lipin at his 3AM hour saying, I'm out of money and I just got an email from this investor.

And I wanna see video of that rather than just hearing him talk about it.

I think like today we I don't think he would have used your system. No, seriously, not for like starting a startup. Maybe someone in this room will. More for learning how dance or something like that, right? But what do you think will be I mean, what specific type, what genre of stuff? Will it be people learning how to dance? Do you think that will be the most popular stuff?

Or will it be babies learning to walk? I think there's some We're seeing some Because the babies learning to walk part actually sounds pretty exciting. Like parents would love to be able to document their kid's progress. I'll tell you the thing, if you don't have kids, one of the big problems about being a parent is the memories of the current kid overwrite the memories of the more recent past.

I'm so sad. I can't really remember what my four year old son was like when he was three. I see three year olds and I think, Oh, yeah. I remember when he was like that, but only vaguely because my God, I got this four year old jumping up and down on the bed in my mind. They're very if you wait until you have kids.

I think there's going to be a number of real killer categories which will be very interesting. The children one is certainly very, very compelling when you see this kid crawling, learning to open a door, and then his parents hold his hands and he's taking baby steps. You'd.

have videos that implicitly have these, the structure of sequences, That's right. But they're not organized that way. They're just like on your iPhone, right, in chronological order. They're not like the series of the kid trying to say some phrase or something like that.

How good are the users at sticking with the whole hundred days of making a video every day? So a good question.

Out of the users that we have, there's an average of around 18 videos uploaded per user. So I guess we have some people who actually our kind of earliest implementation of the product was send us videos via Dropbox every day. So we have some people now who are actually up to kind of in the eighties. Yeah.

So there's a guy beatboxing who we've got him from day one through to day 85, I think, and he's pretty awesome. But he's like beatboxing. Beatboxing. Yeah. So Exactly. My day one. Okay. Exactly like that.

So he off and he's not very confident, not very good, but you really see over time the amazing improvement in him, and that's what we're trying to get at. Are people encouraging each other to stick with it? The point of this that the community will make you be more likely to Sure, absolutely. So that's definitely part of it as well.

So we have these kind of commenting and kind of propping features where people say, oh this was a really awesome day. Actually, I think our most commented on and most kind of liked video, we've got somebody learn to unicycle, and one day she uploaded a video where she had a bad fall and fell over and everybody was like, Oh, that looks really sore, but keep at it, keep going, you're getting there.

People make their stuff semi private?

Yeah. There's a feature to make all your videos private because a lot of people, they don't want to share when they're going through it, but maybe once they're good, they're Or they want a group of their friends to be able to see it, like their kid.

Yeah. That'd be an interesting thing for us to implement. You know, the kid thing, the kid walking is very different from someone teaching themselves how to unicycle. Don't over optimize too early. Let it grow into whatever it's going to grow into. Maybe it'll end up being kids are the big thing, or maybe not, who knows? But be empirical about it and don't wire in some outcome too early.

Is it time? Thank you, Esther. Alright, you guys. Sounds pretty good to you. Hey, Ryan. Hey, Paul. Hey, Ryan. Hi, Sam.

You don't know how odd this is for us. This is the brother of someone we funded in the past, and except for having a beard, he seems identical. It's very disconcerting.

Hopefully that's a good thing.

Yeah. He meant yeah.

Alright. What are you working on? It says, Oh, that's your username. What's the startup? It's called Flexport. We're the first licensed US customs brokerage built around a modern web application. A customs brokerage? Yeah.

Whenever you import a product from another country, you have to clear it through customs. What does a customs brokerage do? Collect tons of documents and organize them and file forms with US customs to clear your goods to show that this is a legal product and you pay the right taxes, etcetera. Is it one of these things where dealing with the government is so awful.

that you need a specialized group of people whose whole job Which is also a guild of licensed Oh, they have be licensed. Oh, yeah, heavily licensed. It's okay if can get the So the government trusts them? Yes. They're not going to lie? Correct. Right? You got get background So the government kind of rubber stamp the paperwork.

You go through an FBI background check as well to get the license. I see. What do you actually do? Do file forms for this employee? Yeah. Based on what the product is, there could be 120 different forms you have to file, so we have to take what the product is, determine which forms are needed, fill those out for the customer, file them electronically with the government.

So there are existing customs brokers. Yeah. Somehow.

you're going to take you're going to be an instance of software eating the world. Yeah. But you're going to eat customs brokers. Yeah. So what do you do? Is somehow scalable if you write software? What do you do differently than an existing customer?

Well, first of we don't use a fax machine, unless the customer really wants Wow.

Alright.

So, yeah, it's an online dashboard to allow you to organize all these documents and help you understand which documents are needed, and then we actually collect those documents for you instead of asking you to go get it,.

and then we'll file it for So I can just come to a website, type in what I'm importing, and have it come in The US, and you'll take care of everything else. Yeah.

Will experience for users be as simple as a customs broker, or are they gonna have to do a little bit more work? No, less work. Less work? Absolutely. We do the work for you. But doesn't the human's customs broker interview.

out of people what they're importing? Like the person says they're importing clothes? We're not removing the human elements. We're pairing you with a licensed customs worker. We have customs workers on staff, and it's more of a locations tool that enable it. It's sort of like Uber. Yeah, haven't used that analogy. It's kind of like teleborder for products instead of for people.

Okay. Yeah.

How much does someone pay a broker to import, like a million dollars of goods?

Usually it's not dependent on the value of the goods, but it's between $100 and $300 per shipment, and there's 30,000,000 shipments that enter The US every year and are filed with a customs entry. 30,000,000? Yeah. And then that's just tip of the iceberg. So really, mean the logistics How many of them use customs brokers? Use a customs broker.

Mean, you can have a if a big company will have a customs broker in staff Yeah, would think, like Apple probably used customs At a certain scale, you hire customs workers. We'd like to make it so you don't have to hire a customs broker because our software is easier to use than maintaining that division of your company. Do you have customers now? Yeah, we have three customers right now.

That are importing stuff. Yeah, trying actually have a waiting list of 300. The biggest oil company in the world signed up. We were a little scared of creating an energy crisis, so we told them to hold off for a bit.

We have a supertanker approaching Port Of San Francisco right How.

did you get those customers?

I was in the industry for twelve years, and I know a lot of importers.

Are you currently a customs broker?

Personally, I do not have a customs worker license. Have a customs worker that works for me, and they're kind of teaching me everything. So you guys, have you already been doing the manual version of this? Over the years I've probably imported about a thousand containers and cleared them through customs. Companies that I've worked with, worked for my brother's company is one of them.

I used work for my brother.

You do know how to do this yourself?

Yeah, sure.

Why has no one done this before? Well,.

why the existing companies haven't done it before is kind of obvious. If you go to a customs broker convention What software? What software? Forty years probably. Startups have done it? Well, first of I mentioned it's highly regulated. Hard to get the license. Actually, recently, it wasn't possible to clear a shipment except at your local port.

So if you built a software startup to do this, you could only help people importing into the Port Of Oakland, unless you had an office in every port. Really? In 02/2007 You're not doing the clearing though, aren't you? Just matching them up with a customs broker? No, no, we are a licensed customs brokerage. We actually do the clearance and file it electronically with customs.

Okay, so you guys are kind of the customs broker of record? Yes, we are the customs It's not quite like Uber. Yeah, I didn't quite get that analogy, to be honest.

You have the same mordant sense of humor as your brother, too. Office hours with him are always a little bit prickly. Alright.

So, how are you going to get all of the importers to switch to this? Presumably they have these long standing relationships with their customer brokers. Yeah. And, they kind of importers,.

to a large degree, have this figured out by definition. They've been doing it. They know how to import goods. But every time you import a product into The US, it's public record, that product, and my last company actually sells that data. We aggregate every time you import something. We've collected 300,000,000 of those shipping manifests and sell subscriptions to access it.

We know every single So you have customer list. What's that? So you have the customer list? Yeah, we have every importer in America database, so we can Wow. That's very convenient. It helps.

Wow. So how far along are you? Have you got sort of this beta.

version that kind Yeah, that's like the MVP product. It's a web app, you can sign up for it. We're not taking new users right now, but that's just a matter of me wanting to feel like everything's super tight and nice user experience. But it has the functionality. Initial version's gonna have to How many shipments have you done with your first three customers? How much what?

How many inbound shipments have you done? No. So the first clearance is happening in November. So we've got these guys lined up, ready to go. But the first shipment Actually, the government shut down three weeks ago, stopped us from They wouldn't take a new broker. The guy whose job it is to onboard us was furloughed, but he's back in his office and I filed the forms on Monday. Holds everything.

I was hoping to have that done and be able to come up on YC app and say, Hey,.

cleared our first shipment already, but See you idiots in the government, you are actually slowing down innovation.

How much do you make? Do you have a sense of, like, on average how much you'll make per customer? Yeah. So, well, for each clearance, the gross margin should be at 75%.

It doesn't take a lot of time to So you can be really hands on and hire a lot of people to help through the It depends. How much do people pay for one clearance? We're gonna charge $100 to do it. No matter the size of the shipment? Yeah, pretty much.

It's actually not about the size, but there's some things you might charge extra for, certain If you want to do something that has to clear with FDA, there's extra paperwork. We might charge extra for that. It seems that people would pay a lot more for a seven seventy seven full of iPhones than.

one little shipping container of Yeah, they're.

pay more in taxes to the government, right? But as far as the broker, it's still just one form or whatever that So the current brokers charge flat rates no matter what the shipment is? No. They charge more. And, you know, I'm kind of looking at it a little differently where the brokerage is just the way that we enter the much larger logistics space.

Because once I'm your customs broker, know everything about your supply chain. What would you do then? Sell you freight, warehousing,.

inspection, You hook them up with trucks or something like that.

Someday I would like to be doing all those things, but Are are other brokers doing that sort of these services after you get it into the country?

Yeah, but I don't know that they look at it as their primary way that they're going to make money. They would never go into it as a loss leader, for example. And I don't know if I'll do that either. I don't like burn rates, but So.

how much do you think you'll be able to make once the thing launches?

Yeah. Well, logistics globally is a $2,300,000,000,000.

industry. No, no. I just mean when you do the customs broker. I just like to say the word trillion. You know, really big markets are bad for start ups. Not good. Yeah. If you say a too big number, the investors don't believe You.

know what? It's really hard to say. I've kind of modeled that and say, Okay, we can make about $30,000,000 a year in profit just being a customs brokerage if you get, say 1%. I don't like to do that kind of compound analysis, but you know each customer is probably worth maybe 2,000 or $3,000 a year, and I think we can get as many customers. But the whole business is $3,000,000,000 a year?

30,000,000 is 1%? Yeah, exactly. The customs clearance business Are just It's about 5,000,000,000 a year, right, based on the number you said earlier? The whole US customs broker About.

three, yeah. Now, don't have exact figures for that, but based on the number of shipments that are cleared and what people charge for those shipments. Are you going to hire just an army of sales guys and go down that list? Possibly. My last company had an army of sales guys and it wasn't that fun to manage, but Things that aren't fun are still sometimes working.

Sometimes it's the way to make the most money, so I have to do it better than I did the last time, so it's more fun.

We've got end.

All right. Boy, that was interesting. Yeah. All right. Thank you. Thank you for coming. Thanks, guys.