- Daniel Gross, a partner at YC, discusses various topics during office hours.
- He talks about the challenges of creating a startup community and suggests joining an existing community like Silicon Valley rather than trying to create one from scratch.
- Gross emphasizes the importance of talking to users when building a product, especially if the end customer is the one buying the product.
- He advises starting a marketplace app by spending time on one side of the marketplace to overcome the chicken and egg problem.
- Gross recommends building a foolproof product before seeking investment and views raising money as a distraction.
- He believes that the end output and user experience are more important than the tools used to create a product like AI and machine learning.
- Gross suggests leveraging machine learning when attacking the long tail becomes important.
- He advises considering raising money to add fuel to the fire and move faster, but not because of fear of being copied.
- Gross states that the level of concern about being copied depends on the product and market, but execution matters more than being first.

- The speaker advises someone to focus on a subfield they are passionate about rather than pursuing a commercially successful startup idea they are less excited about.
- To get a job at a startup without experience or a degree, the speaker suggests standing out by building projects and gaining acclaim in the engineering community.
- When considering whether to start a beta launch without a technical co-founder, it is important to assess whether technical expertise is fundamental to the company's success.
- Finding a technical co-founder can be difficult, but hiring early engineers and offering them equity can be a good strategy.
- To stay on the leading edge of technology and identify startup opportunities, the speaker recommends using bleeding edge hardware and software and observing how people solve problems in the future.
- Becoming a formidable founder is a learned skill that comes from building something and joining programs like Y Combinator.

- The speaker advises companies to only pursue machine learning if it is necessary for their business, not just because it is a trend.
- Google's coding practices for machine learning can provide useful guidelines on when and when not to use machine learning.
- Andrew Ng's machine learning course on Coursera is recommended for those wanting to get started.
- It is difficult for someone to raise money for a startup in Silicon Valley without an existing network, but Y Combinator can help with this.
- The biggest mistake first-time entrepreneurs make is not talking to their users enough.
- Y Combinator accepts founders aged 18 and younger.
- The speaker finds it thrilling to be on the other side and provide advice to help others avoid common pitfalls.

- Daniel Gross, a partner at YC, discusses various topics during office hours including the challenges of creating a startup community and the importance of talking to users when building a product.
- He advises starting a marketplace app by spending time on one side of the marketplace and recommends building a foolproof product before seeking investment.
- Gross believes that the end output and user experience are more important than the tools used to create a product.
- He suggests leveraging machine learning when attacking the long tail becomes important and advises considering raising money to add fuel to the fire and move faster.
- Gross states that execution matters more than being first and advises focusing on a subfield one is passionate about.
- He suggests standing out by building projects and gaining acclaim in the engineering community to get a job at a startup without experience or a degree.
- Gross recommends assessing whether technical expertise is fundamental to a company's success before starting a beta launch without a technical co-founder.
- Finding a technical co-founder can be difficult, but hiring early engineers and offering them equity can be a good strategy.
- He recommends using bleeding edge hardware and software to stay on the leading edge of technology and identify startup opportunities.
- Gross advises companies to only pursue machine learning if it is necessary for their business and suggests following Google's coding practices for machine learning.
- Y Combinator can help entrepreneurs raise money and avoid common pitfalls.

Daniel Gross and Craig Cannon go through questions submitted on Hacker News.

To submit a question for Office Hours, send an email to Macro@YCombinator.com with a few sentences of context and your question.

0:57 - Is it more valuable to build a startup community or join an existing one?

2:02 - How to get students to use your app?

3:06 - What are the most common methods to overcome the chicken/egg problem in a marketplace app?

3:57 - Should you raise money before building a product?

4:25 - How important is it to talk about the tools (e.g. AI/machine learning) that you use to build a product?

5:29 - Is there a point at which you should switch to machine learning?

6:47 - How important is it to raise money?

8:14 - How worried should startups be about people copying them?

10:00 - Should I focus on something I'm passionate about or a startup idea that I think could work but I'm not passionate about?

11:48 - How can a young person without a degree get a job at a startup?

13:44 - Should I launch without a technical cofounder?

15:55 - How much equity should you give to early hires?

16:15 - How can you be someone that lives on the bleeding edge of technology?

18:14 - What should I do to prepare to be a founder?

20:00 - We feel pressured to get into machine learning, should we?

21:19 - How feasible is it to fly to Silicon Valley to raise money?

22:41- What are the most common mistakes first time founders make?

23:43 - Does YC accept founders under 18?

He is a new partner at YC and we're doing office hours.

I started a Y Combinator company in winter 2010.

And what we built was kind of a search engine for all of your personal stuff online.

So instead of searching Gmail and Dropbox and Slack and ever and all in different places, you could have 1 app to kind of have a unified search experience.

And we built a really good product, but a pretty bad business model.

So we ended up getting acquired by Apple, where I worked for 3 years on a bunch of different machine learning efforts across iOS, OS X, and a few other platforms.

And As a couple of weeks ago, I rejoined Y Combinator, but this time as a partner.

So these are just questions submitted either through macro Y Combinator or HN.

So first 1, is it more valuable to grow a startup community where there hardly is 1 or help members of that community discover and become involved in a strong startup community elsewhere where there are more resources and connections?

I think it would be great if there were more startup communities around the world but they're really really hard to create.

As it turns out, for a good startup community to exist, you need a certain level of density that's just really hard to conjure yourself.

And so unfortunately, if you kind of want to apply the Occam's razor lens of do the simplest thing, you probably want to join an existing, thriving startup community, the best 1 being that I know of being in Silicon Valley, as opposed to try to bootstrap 1 on your own.

And did you move to the States for that exact reason?

I moved to the States because I got into Y Combinator, weirdly enough.

And I definitely reflect on the fact that even though Israel is 1 of the best startup communities outside of the United States, it still doesn't compare to Silicon Valley.

What do you feel is the best demographic to initially communicate to students, parents, counselors, etc.

If we are trying to ultimately get 16 to 19 year old high school students on our application and Why would you choose that particular demographic, whichever 1 you choose?

So in general, the best advice is to always talk to your users.

And so if you're building a product for teenagers, you should talk to teenagers.

If you're building a product where the end customer isn't going to be the 1 buying your product, an example would be enterprise software or if you're building something for babies, then you're going to want to talk to the parents, the people who have the purchasing ability.

But you still want to always talk to your users.

So this would be something you'd be doing in addition to.

And the reason why is if you built something really good, even though those that don't have purchasing power are the kind of the ones you spoke to, you've built something that they really want and they're then going to talk to their boss, a little bit harder to do if you're a baby, and try to get it.

If you're making a marketplace app, what are the most common methods for overcoming the chicken and egg problem in the very beginning?

For example, I think Airbnb did door-to-door in the early days.

The strategy that works best is to cheat effectively and spend a lot of time yourself doing something that doesn't scale by building 1 part of the marketplace yourself.

So if you use the Airbnb example, the founders were really diligent about making sure they had great supply in cities where they had a lot of usage.

If you talk to very early folks at Uber, they'll tell you that they spent a lot of time personally hand-holding, recruiting drivers, making sure that they had good supply on the network.

And so if you're trying to build a double-sided marketplace, 1 way to start is to spend a lot of time yourself doing the thing that won't scale and building 1 side, either the supply or demand side.

Do you think it is better to build a foolproof product and then look at investment to fuel expansion may take longer or should you get the first money and use it to build a foolproof product faster but potentially riskier.

For reasons that would be far too long to expand on in this office hours you should always build the product first and then raise money.

While important technologies are at a high risk of becoming a buzzword, If they already haven't, I would say they have.

To what extent do you agree with the fact that the end output or experience matters more than the tools, for example AI, machine learning, whatever, that the company uses to create that product?

Well at the end of the day we're trying to build things people want.

And so the end output is going to be how you'll be measured and whether you'll be successful or not.

If you build a product with a really cool algorithm that doesn't culminate in something people want, then you won't be successful.

The other thing to realize is there are many situations where a human or very simple heuristics can get you 99% of the value.

And so as much as I love machine learning myself, more often than not, you should just try to focus on making something people want and do it in the laziest way possible Which occasionally might mean a few simple heuristics that get you?

90% of the way there or a few simple heuristics plus a human which will get you 99% of the way there

And now Is there a point at which you think it's appropriate to switch over?

So for example, say you're a company and you're in the batch at YAC.

And I imagine at this point they're coming to you and saying, when should we apply this?

So usually It's when you've reached either, I guess there's 2 ways of thinking about this.

Either you want to are those few companies where machine learning is so fundamental to even your initial product, because it's so unscalable to try to do it with humans, that you got to do it from day 1.

An example would be a self-driving car company.

But if you're not in that camp, and you should just be honest with yourself if you're not, you want to start leveraging machine learning when attacking the kind of long tail starts to really matter to you.

Say you're working on Amazon.com, and it's your job to make the homepage have great products.

You could start off by just putting the top grossing products on the homepage and you'd have a pretty good experience.

At some point you start caring about the long tail and so you start wondering about things like personalization.

Well, it certainly wouldn't scale to hire a human to model amazon.com for every other human that uses it.

And so that would be a point to try machine learning.

But if you're not at the point where you've saturated the short head of the power law distribution of goodness, then it's not time for machine learning.

All right, so this question is a bit longer.

Subject line is, how important is it to raise money?

So my startup, so Kevin's startup, Onco Assist, has been going 4 years now with steady growth.

However, we have struggled to raise money.

We have had conversations with angel investors and VCs, but they have never come to fruition, meaning they've never raised.

Our active users, yeah, I think it's probably, yeah.

The active user base is growing, as is the revenue we generate from selling services to pharma.

We are concerned someone will copy what we are doing.

Are these legitimate concerns or will trying to raise money just be a giant distraction?

The question is whether you can afford the distraction or not.

Raising money because you've built something good and you want to move faster, that is to say, you want to add more fuel to the fire, is a very legitimate reason to do it.

And you probably want to try to do it as quickly as possible with the right amount.

Figure out what is kind of the next objective that you want to hit, figure out how much money you need in order to get to that objective, and then go out and raise that amount.

Raising money because you're afraid that someone will clone you is a little bit less of a compelling argument, just because that fear is always gonna be out there.

But certainly if you feel like you have something great and you just want to kind of multiply your greatness, then

raising money is a very legitimate thing to do.

In your experience, How worried should young companies be about someone copying them?

That's a really good question It very much Depends on the product and the market that you're in If there are certain situations where there's a first mover advantage.

And so if you don't get to that product first, you're not going to gain the network effect, say, that you would need in order to get really big.

There are other products, 1 can think of, say, Lyft or Uber, where the market is so large and also the network effects are somewhat local and not global, so that it doesn't really matter if someone copies you.

I think there are more things in the latter camp than the former, though I think we all tend to kind of intellectualize and believe that everything's in the former camp.

That is to say, if I'm not first, it won't matter.

But at the end of the day, I think ride sharing is a great example of how both companies got really big.

Admittedly, 1 is much larger than the other, but both got, you'd be pretty happy to be an early investor in Lyft.

Because the market is so large and the network effects are somewhat local.

I mean, even Web Search, which has a global network effect with the way search quality works, as more people use it, it gets better.

Google is not literally the only product in the market, even though it's the predominant 1.

So I guess at the end of the day, I wouldn't be too worried about this for the reasons I described.

And 1 other final reason is execution matters more, much more than being first.

Facebook was not the first social network.

So I wouldn't spend too much time worrying about it, and I'd just buckle down and try to build the best thing instead of the first thing.

Dear YC, I have a unique insight in a field in which I'm a domain expert.

I see a huge opportunity for a startup that follows through on this insight.

I have the right technical knowledge needed for doing that.

To follow up on the other question, nice and vague.

But my level of excitement for this field is a 6 out of 10.

I'm much more excited about a different, new subfield of my main field.

I would be doing a lot of R&D work in that subfield before I'll have some technology that can be commercialized, and that R&D work is very exciting and fun for me personally.

Should I follow through on the first idea, which I know I can lead to a very commercially successful startup, or should I use my savings to do R&D in the subfield that really interests me for a couple of years and see if a startup comes out of that?

You know, I think it was in 2008 that Elon Musk said that running a startup was like chewing glass and staring into the abyss.

because Tesla wasn't doing great and SpaceX had failed a few launches.

But that is to say, doing startups is really, really hard.

And I think, I kind of think the only thing that makes it worth it or that enables you to see it through is that you really love what you're doing.

And you know, a lot of people like to repeat Steve Jobs quote about making something insanely great To be evocative about how good it has to be but I also think he was Trying to imply that you have to be somewhat insane About the field that you're in So I would spend all of your time on something that you're passionate on because life is short.

Also startups are really hard So it's only worth doing if you love it.

Yeah, especially when you think about recruiting people

That's right That's a really good point, too.

Yeah, so all right I studied business for 2 years at a community college and switched my major to CS For a semester before dropping out to learn on my own I ultimately want to found a startup, but I want to work at a startup to gain experience first How does someone young without experience or degree go about getting a job at a startup?

For example, if I cold email founders at startups I love, how can I convince someone to give me a chance versus the Stanford grads they're looking at?

What about trying to work at companies like Uber or Airbnb that are later stage?

So you know put yourself in that founders shoes.

You just got an email from someone you don't know that doesn't have any credentials.

Really what you want to do is you want to figure out a way to stand out as much as possible.

So 1 simple answer is if you can figure out how to get a job at a larger company that may be more accepting of whatever credentials you have as a stepping stone to working at a high growth startup, great.

Another strategy that works well too is to just try building things.

You know, they don't even have to be revenue generating things.

You can build a library that you open source that a lot of people use.

And that alone, I think, will get you a lot of acclaim, especially in the engineering community, which I've seen be really, really effective at helping people get jobs.

You know, Resume gets passed around and people are more excited about the GitHub project that you have than whatever credentials you have.

And I think that's far more indicative, frankly, of how you're going to be as an employee because it shows that you are self-managing, you have the ability to get stuff done make trade-offs, and that's often not fully taught I think in

a college curriculum, and I don't think you really feel it when you're maybe whatever they're 20 years old But if you play the long game it also works out like it's something that you see lots of people do, like immediately asking for favors before they've even really gotten to know someone or provided value.

I'm trying to get into a beta launch of grocery delivery product, suncart.co in Qatar, and have not been able to find a technical co-founder over the past 6 months.

I got the product made from an outsourced company, so I assume that's a contractor or consultant.

Should I go out and start the beta launch without a tech co-founder and keep the hunt going simultaneously?

Finding a co-founder is in general really hard.

Finding a technical co-founder is even harder just because the supply is smaller.

And so I guess the first thing to ask yourself is whether a certain degree of technical acumen is going to be downright fundamental to your company success.

There's a lot of companies that are built on software platforms that are sufficiently abstracted such that you don't need the co-founders to be technical in order to build something great.

you're in the former bucket, your charter is clear.

You kind of have to continue your search.

If you can convince yourself that you're in the latter bucket, then 1 good strategy is to just try to find your first engineer, second engineer.

This could be a person that doesn't necessarily have to have the level of fervor and commitment to the product that you do, and you'd give them a hearty chunk of equity, but perhaps they wouldn't be necessarily your co-founder.

And this would just let you expand your search horizons to far more people.

I have, in general, not seen success when people have tried to not even have engineering in-house and have outsourced that as well just because there's that tends to Put you a little bit too far away from making iterative product changes when you constantly have to go back to a contractor who may not understand the context, the rationale behind the changes that you want to make.

So, you know, in a nutshell, it's worth thinking about whether you need a technical co-founder or not, depending on what you're building.

It's totally okay if you think you don't, in which case I would just try to hire an engineer or 2.

And because they're early, you should give them a decent chunk of equity, but perhaps they don't have to be a co-founder per se.

If you're considering someone in your first, say, 8 or so hires, If you don't have any significant traction or growth, that is to say it's not obviously working, then you should be giving them anywhere between 1% to 5% of the company.

So lots of founders in the valley will start things based on trends, as mentioned before.

And at that point, it tends to be too late to be a first mover and be innovative, which you could debate.

How can you be someone that lives in the future on

So I think, yeah, I think the answer to this question is actually pretty simple.

Just try to use all the bleeding edge hardware and software.

You know, today that would be something like maybe using the Oculus headset or the Vive Eve headset.

But I think it's actually a really good question because it points out the fact that living in the future is actually a really good way to simulate what good startup opportunities are right around the corner.

You know, an example of this is, you know, in 2007 or 8, having 2 computers and having this pain of having files that weren't instantly transferred between both computers.

And I think the twist is not only to live in the future, but to kind of observe the hacky ways people are working around problems in the future.

So if memory serves, I think if I remember correctly, the Dropbox Hacker News launch Thread had a lot of people saying, this isn't a real product or company.

I already use rsync, which is basically a open source terminal library you could use to sync files between my computer.

And they were somewhat, they were obviously wrong about the company size, but they were right.

They had solved that problem in a hacky way.

So I guess I would say I would live in the future and also be aware of kind of very crude things people are building because they're in this very small audience of people living in the future and they can solve their own problems and try to see if there's something that you can generalize.

And I think that's another strong point for being technical.

Like when you see your friends hacking things together like that, like most people aren't going to go into the terminal and get stuff done.

If I know I want to start a startup in X amount of time, what should I be doing in that time to prepare to be a founder?

How do you turn yourself into the formidable founder that everyone in the startup world aspires to be?

I think that a lot of these, when you look at a lot of these successful founders, the qualities that you perceive to be as formidable are not innate, they're learned.

And I have found for myself, I don't think I'm that formidable, but I certainly became more formidable as a result of doing a company.

And I don't think that there could have been a way where I could have learned that by watching a class or by learning something or by going to a course.

And so the best way to do this is to just try to build something.

And the best way to try to build something, I think, is to apply for Y Combinator.

So I guess my advice would be try to make something people want and join YC.

And to put it in perspective, no 1 is born as this formidable founder person.

Like if you go around YC, they're just normal people.

And similarly, like if you talk to these like big shot founders, even now, they're just normal people.

I think there's a halo effect that gets created for these people, but you forget.

And certainly, I think if you'd ask their friends, I'm sure they'd look with kind of rose-colored tinted glasses in the past and say they always seemed cool.

But when you talk to, I know a lot of people that today you'd say are very formidable, and in their early days, they put their pants on in the morning just like you and I do.

Yeah, venture investing would be a lot easier if you could tell that from the beginning.

So how can a company start working on machine learning models?

We have lots of data, but we are not sure what to do with all that data.

We feel pressured to start working with machine learning should we try to get into it.

So much like the previous related question on this topic,

I would not spend time doing machine learning because you feel like the industry is abuzz about machine learning.

I would do machine learning because you feel like your business absolutely depends on it.

1 thing to look into that I think recently came out is Google's coding practices for machine learning.

It's a really interesting document with a bunch of rules about when and when not to use machine learning at Google and how to do feature engineering.

I believe the first rule in this document, written by Google, probably the most formidable machine learning company on the planet, is rule number 1, you don't always need machine learning.

Feel free to launch a product without it.

And so, my advice would be to really ask yourself whether you need it or not and don't do anything because there's a trend.

I can't think of a single opportunity where that's worked out well for anyone.

And if you did want to get your feet wet, do you recommend any courses or anything like that online?

Yeah, if you wanted to get your feet wet with machine learning, Andrew Ng's course on Coursera, I believe, is quite good, educational, and easy to consume.

How feasible is it for someone to fly to Silicon Valley for a week or 2 and raise money for a startup?

Silicon Valley works because it's a network and just like a pathogen or a virus, if there's kind of an unknown entity coming into the network, it won't get accepted easily.

And so you have to become known in some way.

And so, 1 strategy could be trying to, from your local country, do a lot of content marketing, try to set up a lot of meetings through friends of friends.

If you're really lost and you have no network, I do think this is an area where Y Combinator adds a tremendous amount of value.

There's a lot of Y Combinator founders, myself included, that arrived here literally knowing no 1.

knew 0 people in California when I arrived, 7, 8 years ago.

And Y Combinator created a network of obviously investors that I could talk to through Demo Day, friends, And I think it's a very good vehicle for that.

So the easiest answer to that question is apply

And to add to that, it's a myth that everyone at YC went to Stanford and is from here.

And adding to that, like, if possibly you wanted to try something like this, I suppose you could have a network out here to some extent of friends and they can intro you.

But I think by and large, it makes sense to be out here.

What are the most common mistakes you have seen first time entrepreneurs and founders make?

It's that you don't talk to your users enough.

And I think the interesting point about this feedback is not necessarily to listen to it or watch it on this YouTube video and then immediately discard it.

It's to realize that generation after generation of people who try to build companies, no 1 seems to learn this lesson and people continually don't talk to their users enough.

So it's worth thinking about, you know, why we have to repeat this lesson over and over and over.

I think it's because it's far more comfortable to kind of try to paint a perfect painting than show it to the world.

And so, you know, when in doubt, assume that you're not talking to users enough and you need to talk to them more.

It really doesn't make sense to list out, I think, any more challenges people have or mistakes people make because this is such a paramount 1.

Most companies die because they don't make something people need.

See all Y Combinator transcripts on Youtube

Office Hours with Daniel Gross