Sam Altman

All companies that grow really big do so in only one way: people recommend the product or service to other people.

Maybe bitcoin will be the world reserve currency, maybe it will totally fail, or maybe it will survive in some niche capacity. I don’t know how to weight the probabilities (although I think in the immediate term it's likely to go down), but I do have a thought about the metric to watch: growth in legitimate transactions. A currency without the major use case being legitimate transactions is going to fail.

I frequently get asked by non-technical solo founders if I know any potential hacker cofounders they should talk to.  These people give a passionate pitch for the idea and a long list of all the hustling they've done, customers they've spoken to, models they've built, provisional patents they've filed, etc.  Most of the time, they are thoughtful and hardworking.  But they've often been searching for their technical cofounder for many months, and things have stalled during that process. 

When people like this say "I'll do whatever it takes to make this business successful" (which they almost always say), I say something like "Why not learn to hack?  Although it takes many, many years to become a great hacker, you can learn to be good enough to build your site or app in a few months.  And even if you're not going to build the next version, if you're going to run a software company, it seems like a good idea to know a little bit about it."

Usually the response is something like "That wouldn't be the best use of my time", "I don't like it", or "I don't have that kind of brain".  (Earlier today it was "You don't understand, I'm the idea guy.  If I'm hacking, who will be talking to investors?", which is what prompted this post.)  But every once in awhile people think about it and decide to learn to hack, and it usually works out.

They’re often surprised how easy it is.  Many hackers love to help people who are just starting.  There are tutorials for pretty much everything and great libraries and frameworks.

As an important aside, if you try to learn on your own, it can be really hard.  You’ll hit some weird ruby error and give up.  It’s important to have someone—a friend, a teacher at a coding bootcamp, etc.—that get you through these frustrating blocks.

When hackers have to for their startups, they are willing to learn business stuff.  Business people should do the same.  If you're not willing to do this, you should remember that there are far greater challenges coming in the course of a startup than learning how to code.  You should also remember that you can probably learn to code in less time than it will take to find the right cofounder.

Speaking of cofounders, a word of warning: meeting a stranger for the express purpose of cofounders hardly ever works.  You want someone you've known for awhile and already worked with.  This is another good reason for learning to hack yourself instead of bringing on a cofounder.

You can build the first version of your product, and even if it's terrible (we had a non-technical founder in YC that learned to hack with Codecademy and was still able to learn enough to build a prototype), you'll actually be able to get real user feedback, iterate on something other than mockups, and perhaps impress a great hacker enough to join you.  Although you may never win a Turing Award, if you're smart and determined, you can certainly get good enough to build a meaningful version 1.

If you're a solo founder and you can't hack, learn.

One of the most interesting changes in venture capital going on right now is the separation of advice and money.  For a very long time, these have been a package deal.

Great advice is really important; some founders don’t appreciate this initially (I was guilty of it) but always learn to.  But great advice does not have to come from venture capitalists; it often comes from people like former founders.

There have been a few small indications of the advice/money separation over the past few years, but crowdfunding is now really making it happen.  Some companies can raise money on very good terms from investors that don’t know much about startups, and then give equity to the advisors they want to work with.

There are probably going to be big advantages and big disadvantages to this.  On the positive side, founders may end up with less total dilution and get to choose whatever advisors they want—not just the people that happen to manage institutional money.  Another big positive is that more competition (and more transparency) makes investors behave better.  On the negative side, advisors probably won’t work quite as hard for a company that they don’t have a lot of capital invested in.  Also in the negative column, this will probably further worsen founders’ disrespect for capital.  And perhaps worst of all, I expect a lot of people to lose a lot of money—startup investing is both hard and appeals to gambler’s instincts, and it’s easy to imagine it becoming the new daytrading.  At some point, of course, the pendulum will swing back.

Advisors will probably still put in some capital, but probably at a better effective price than people who just invest.  The hard part is that everyone thinks they are a great advisor and wants the special treatment.

The bigger force at work is the long-term trend towards founders having more leverage than investors.  This change in leverage has happened for a lot of reasons, but specifically, crowdfunding probably would not have been possible if companies needed as much capital to start as they did ten years ago.  Also, startups are cool now, so more people want to invest.

Quick and painless fundraising, without advice necessarily being part of the package, is what many founders want.  In a sense, VCs sell advice, but founders want to buy money.

Crowdfuding is an answer to this (also, the crowd is willing to fund things VCs are not, pay higher prices and on very clear terms, etc.)  Fundraising has not been an efficient market—VCs and angels have been able to corner it with laws, access, and it being the only source of advice.  But the Internet continues its never-ending march.

The best VCs are great, and they will probably continue to do well.  In fact, they’re so good that they could probably get away with only selling advice—they understand how to build big companies in a way that few other people in the world do.  They may have to adapt their strategy somewhat—for example, in response to being able to buy less ownership in earlier rounds, I suspect some firms will shift to writing much larger checks to the obvious winners in later rounds.

The mediocre and bad VCs will have to adapt or die. 

Maybe now we’re actually in the early stages of a startup bubble.  Valuations are trending up again.  And no one is talking about a bubble anymore, so it could be happening. 

Many companies still feel reasonably priced.  Yes, Facebook is up 70% in the last six months, but I think it’s likely still undervalued.  Lots of companies raising B rounds are doing it with large profits and at reasonable multiples.

But the very early stage feels somewhat out of whack.  Companies raising money at $15MM+ plus valuations with no traction and no real vision beyond starting a startup still strikes me as unsustainable (not to mention bad for the companies).

Lots of other signs point to a bubble—founders of Series A stage companies being angel investors, a significant uptick in the number of parties, soaring rents, soaring salaries, lots of new investors coming to valley, and MBAs starting companies as the fashionable thing to do again.

Even if this is the beginning of a bubble, it will likely go on for awhile longer—at least as long as the Fed keeps the stimulus going, and likely longer.  Maybe it goes on for another couple of years.

We can debate whether or not companies are overpriced, but it is an absolute certainty that at the first sign of real trouble, most investors will overreact and invest much less precisely at the time they should be investing more.

So the question is, if you believe that this may be the beginning of a bubble, what should you do now?

Fortunately, the answers are things you should be doing anyway.

First, you should not be too afraid.  The only thing that is cheap during a startup bubble is capital.  Everything else is relatively more expensive.  If you’re a real company, a downturn that you’re prepared for can be beneficial.  Good companies can make incredible strides during a bust.

Make sure you have enough money in the bank, and treat this money as the last money you’ll ever raise—at the very least, have a plan B to get to profitability without raising any additional capital.  If you need to raise more, this is a good time to do it.

Focus on a path to profitability.  Remember that, if your customers are mostly startups, revenue can dry up fast.

Resist the urge to ramp up to a crazy burn rate.  Be especially fearful of long-term commitments like expensive leases and people.  Don’t hire unnecessary people.  Stay as lean as you can.  Establish a culture of frugality.

Bubbles bursting also require startups to focus, which is a good thing to be doing in any case.  Cut products and features that are not working.  Focus relentlessly.

Of course, people are terrible at predicting bubbles and busts, so I’m likely wrong.  Luckily, I really do believe these are good things to do anyway.

After startups raise money, their next biggest problem becomes hiring.  It turns out it’s both really hard and really important to hire good people; in fact, it’s probably the most important thing a founder does.

If you don’t hire very well, you will not be successful—companies are a product of the team the founders build.  There is no way you can build an important company by yourself.  It’s easy to delude yourself into thinking that you can manage a mediocre hire into doing good work.

Here is some advice about hiring:

*Spend more time on it.

The vast majority of founders don’t spend nearly enough time hiring.  After you figure out your vision and get product-market fit, you should probably be spending between a third and a half of your time hiring.  It sounds crazy, and there will always be a ton of other work, but it’s the highest-leverage thing you can do, and great companies always, always have great people.

You can’t outsource this—you need to be spending time identifying people, getting potential candidates to want to work at your company, and meeting every person that comes to interview.  Keith Rabois believes the CEO/founders should interview every candidate until the company is at least 500 employees.

*In the beginning, get your hands dirty.

Speaking of spending time, you should spend the time to learn a role before you hire for it.  If you don’t understand it, it’s very hard to get the right person.  The classic example of this is a hacker-CEO deciding to hire a VP of Sales because he doesn’t want to get his hands dirty.  This does not work.  He needs to do it himself first and learn it in detail.  Then after that, he should lean on his board or investors to give him opinions on his final few candidates. 

*Look for smart, effective people.

There are always specifics of what you need in a particular role, but smart and effective have got to be table stakes.  It’s amazing how often people are willing to forgo these requirements; predictably, those hires don’t work out in the early days of a startup (they may never work).

Fortunately, these are easy to look for.

Talk to the candidates about what they’ve done.  Ask them about their most impressive projects and biggest wins.  Specifically, ask them about how they spend their time during an average day, and what they got done in the last month.  Go deep in a specific area and ask about what the candidate actually did—it’s easy to take credit for a successful project.  Ask them how they would solve a problem you are having related to the role they are interviewing for.

That, combined with the right questions when you check references, will usually give you a good feel about effectiveness.  And usually you can gauge intelligence by the end of an hour-long conversation.  If you don’t learn anything in the interview, that’s bad.  If you are bored in the interview, that’s really bad.  A good interview should feel like a conversation, not questions and responses. 

Remember that in a startup, anyone you hire is likely to be doing a new job in three to six months.  Smart and effective people are adaptable.

*Have people audition for roles instead of interviewing for them.

This is the most important tactical piece of advice I have.  It is difficult to know what it’s like working with someone after a few interviews; it is quite easy to know what it’s like after working with them 

Whenever possible (and it’s almost always possible), have someone do a day or two of work with you before you hire her; you can do this at night or on the weekends.  If you’re interviewing a developer, have her write code for a real but non-critical project.  For a PR person, have her write a press release and identify reporters to pitch it to.  Just have the person sign a contractor agreement and pay them for this work like a normal contractor.

You’ll get a much, much better sense of what it’s like working with this person and how good she is at the role than you can ever get in just an interview.  And she’ll get a feel for what working at your company is like.

*Focus on the right ways to source candidates.

Basically, this boils down to “use your personal networks more”.  By at least a 10x margin, the best candidate sources I’ve ever seen are friends and friends of friends.  Even if you don’t think you can get these people, go after the best ones relentlessly.  If it works out 5% of the time, it’s still well worth it. 

All the best startups I know manage to hire like this for much longer than one would think possible.  Most bad startups make excuses for not doing this.

When you hire someone, as soon as you’re sure she’s a star you should sit her down and wring out of her the names of everyone that you should try to hire.  You may have to work pretty hard at this.

Often, to get great people, you have to poach.  They’re never looking for jobs, so don’t limit your recruiting to people that are looking for jobs.  A difficult question is what you should do about poaching from acquaintances—I don’t have a great answer for this.  A friend says, “Poaching is the titty twister of Silicon Valley relationships”. 

Technical recruiters are pretty bad.  The job boards are generally worse.  Conferences can be good.  Hosting interesting tech talks can be good for technical hiring.  University recruiting works well once you’re reasonably established.

Don’t limit your search to candidates in your area.  This is especially true if you’re in the bay area; lots of people want to move here.

View candidate sourcing as a long-term investment—you may spend time now with someone that you don’t even talk to about a job for a year or more.

Use you investors and their networks to find candidates.  In your investor update emails, let them know what kind of people you need to hire.

As a side note, if I were going to jump into the mosh pit of people starting recruiting startups, I would try to make it look as much like personal network hiring as possible, since that’s what seems to work.  I’d love a service that would let me see how everyone in my company was connected to a candidate, and be able to search personal networks of people in the company (LinkedIn is probably good at this for hiring sales people but not very good at this for developers). 

*Have a mission, and don’t be surprised at how much selling you’ll have to do.

You need a mission in order to hire well.  In addition to wanting to work with a great team, candidates need to believe in your mission—i.e., why is this job more important than any of the others they could take?  Having a mission that gets people excited is probably the best thing you can do to get a great team on board before you have runaway traction. 

As a founder, you’ll assume that everyone will be as excited about your company as you are.  In reality, no one will.  You need to spend a lot of time getting candidates excited about your mission. 

If you have a good mission and you’re good at selling it, you’ll be able to get slightly overqualified people—although, in a fast-growing startup, they’ll end up in a role that they feel not quite ready for quickly anyway.

You should use your board and your investors to help you close candidates.

Once you decide you want someone, switch into closing mode.  The person that the new hire will report to (and ideally also the CEO) should be doing everything possible to close the candidate, and talking to her about once a day.

*Hire people you like.

At Stripe, I believe they call this the Sunday test—would you be likely to come into the office on a Sunday because you want to hang out with this person?  Liking the people you work with is pretty important to the right kind of company culture.  Only a few times have I ever seen a scenario where I didn’t like an otherwise very good candidate.  I only made the hire once, and it was a mistake.

That being said, remember you want at least some diversity of thought.  There are some attributes where you want uniformity—integrity, intelligence, etc.—and there are some where you want coverage of the entire range.

*Have a set of cultural values you hire for.

Spend a lot of time figuring out what you want your cultural values to be (there are some good examples on the Internet).  Make sure the whole company knows what they are and buys into them.  Anyone you hire should be a cultural fit.

Andrew Mason says “Values are a decision making framework that empower individuals to make the decision that you, the founder, would make, in situations where there are conflicting interests (e.g. growth vs. customer satisfaction)”.

Treat your values as articles of faith.  Screen candidates for these values and be willing to let an otherwise good candidate go if he is not a cultural fit.  Diversity of opinions and certain characteristics (e.g. you want nerds and athletes both on the team) is good; diversity of values in a startup is bad.

There are some people that are so set in their ways they will never get behind your values; you will probably end up firing them.

As a side note, avoid remote employees in the early days.  As a culture is still gelling, it’s important to have everyone in the same building.

*Don’t compromise.

In the grind of a startup, you’ll always need someone yesterday and it’s easy to hire someone that is not quite smart enough or a good enough culture fit because you really need a specific job done.  Especially in the early days, never compromise.  A single bad hire left unfixed for long can kill a company.  It’s better to lose a deal or be late on a product or whatever than to hire someone mediocre.

Great people attract other great people; as soon as you get a mediocre person in the building, this entire phenomenon can unwind.

*Be generous with compensation packages, but mostly with equity.

You should be very frugal with nearly everything in a startup.  Compensation for great people is an exception.

Where you want to be generous, though, is with equity.  Ideally, you end up paying people slightly below to roughly market salaries but with a very generous equity package.  “Experienced” people often have higher personal burn rates and sometimes you’ll need to pay them more, but remember that great companies are not usually created by experienced people (with the exception of a few roles where it really matters a lot.)

I am sure I will get flamed for saying this, but it’s the right strategy—if you want an above-market salary, go work at a big company with no equity upside.

Ideally, you want to pay people just enough they don’t stress about cash flow.  Equity is harder, but a good rule for the first 20 hires seems to be about double what your investors suggest.  For a company on a good but not absolute breakout trajectory, some rough numbers I’ve seen are about 1.5% for the first engineer and about 0.25% for the twentieth.  But the variance is huge.

Incidentally, a very successful YC company has a flat salary for effectively all of their engineers, and it seems to work well.  It's lower than what these people could get elsewhere, but clearly they enjoy the work and believe the stock is going to be worth a lot.  The sorts of people that will take this deal are the kind of people you want in a startup.  And unless something goes really wrong, at this point, these engineers are going to make way more money than that would have taking higher salaries elsewhere—not to mention how much better their work environment has been. 

You will likely have to negotiate a little bit.  Learn how to do this.  In general, materially breaking your compensation structure to get someone is a bad idea—word gets out and everyone will be upset.

*Watch out for red flags and trust your gut.

There are a few things in the interviewing/negotiating process that you should watch out for because they usually mean that person will not be successful in a startup.  A focus on title is an example; a focus on things like “how many reports will be in my organization” is an even worse example.  You’ll develop a feel for these sorts of issues very quickly; don’t brush them off.

If you have a difficult-to-articulate desire to pass, pass.

*Always be recruiting. 

Unfortunately, recruiting usually doesn’t work as a transactional activity.  You have to view it as something you always do, not something you start when you need to fill a role immediately.  There’s a fair amount of unpredictability in the process; if you find someone great for a role you won’t need for two months, you should still hire her now. 

*Fire fast.

I have never met a newbie founder that fires fast enough; I have also never met a founder who doesn’t learn this lesson after a few years. 

You will not get 100% of your hires right.  When it’s obviously not working, it’s unlikely to start working.  It’s better for everyone involved to part ways quickly, instead of hanging on to unrealistic dreams that it’s going to get better.  This is especially true for the person you have to let go—if they’re just at your company for a couple of months, it’s a non-issue in future interviews.  And everyone else at the company is probably aware that the person is not working out before you are.

Having to fire people is one of the worst things a founder has to do, but you have to just get it over with and trust that it will work out better than dragging things out.

*Put a little bit of rigor around the hiring process.

Make everyone on your team commit to a hire/no hire decision for everyone they meet, and write up their thoughts.  If you get it wrong, this is useful to look back at later.  It’s good to have a brief in-person discussion with the entire interviewing team after a candidate leaves.

Have someone take the candidate out to lunch or dinner.  Insist that everyone is on time and prepared for interviews/auditions.  Make sure every candidate leaves with a positive impression of your company.

Be organized—one person should coordinate the entire interview process, make sure every topic you want to cover gets covered, convene people for the discussion after all interviews are done, etc.  Also, have a consistent framework for how you decide whether or not to hire—do you need unanimous consent?

Remember that despite being great at what they do your team may not be great interviewers.  It’s important to teach people how to interview.

*Don’t hire. 

Many founders hire just because it seems like a cool thing to do, and people always ask how many employees you have.  Companies generally work better when they are smaller.  It’s always worth spending time to think about the least amount of projects/work you can feasibly do, and then having as small a team as possible to do it. 

Don’t hire for the sake of hiring.  Hire because there is no other way to do what you want to do.

Good luck.  Hiring is very hard but very important work.  And don’t forget that after you hire people, you need to keep them.  Remember to check in with people, be a good manager, have regular all hands meetings, make sure people are happy and challenged, etc.  Always keep a sense of momentum at your company—that’s important to retaining talent.  Give people new roles every six months or so.  And of course, continue to focus on bringing talented people into the company—that alone will make other good people want to stay.

Always be identifying and promoting new talent.  This is not as sexy as thinking about new problems to solve, but it will make you successful. 

Thanks to Patrick Collison, Andrew Mason, Keith Rabois, Geoff Ralston, and Nick Sivo for reading drafts of this.

A friend recently asked me for my list of current breakout companies.  I made it, and noticed that all of them had only one thing in common.  I then went back through the last three YC batches and looked at our top-ranked companies (although we get it wrong plenty of times, it’s reasonably predictive).  These companies had the same trait. 

This shared trait is a connection between the online and the physical worlds.  There are two main models for this—the Uber model, where you push a button on a website/app and something happens in the real world, and the Airbnb model, where you use a service to do something in the real world that would be possible but extremely inefficient.  In both cases, the key thing is enabling users to do things they do in real life much more easily—yes, you could have called a cab company, but it took a long time, the cab didn’t always come, you didn’t know when it was near, you had to have cash or get a nasty look from the cabbie, etc.

Even Facebook is a lightweight version of this—with Facebook, you bring your real world online so you can interface with it more often, more easily, and from anywhere.  Facebook has obviously been much more successful than services like SecondLife, which were entirely virtual, and services MySpace, which were less about real identity and real friends.  Amazon was perhaps the original example of bridging these two worlds.

As a corollary, new companies that exist only in the real or online world do not seem to do as well as companies that connect them.  This happens more often on the online side than the physical side, both because we generally don’t think of things like barbershops as startups and because most people don’t expect pure hardware companies without a strong Internet component to do very well.  And this bridge is often a matter of degree—many, if not most, online companies touch the real world in some way.  But the ones that do it the most seem to be doing better.

In many ways, this is a version of Marc Andreessen’s “software is eating the world”.  To eat the world, you have to operate in the world.  I think two big reasons it’s happening so much now are 1) smartphones are in everyone’s hands and 2) there is a level of personal comfort with putting real life on the internet that seems to have tipped about two years ago.

There are probably a lot of areas where this doesn’t apply—enterprise software and developer tools, for instance—and there are clear counterexamples like Google.  But as a general rule, it seems to be worth considering when thinking about new businesses.   

I’ve heard a lot of different theories about how things get done.  I’m interested in this topic, so I pay attention and see how the theories hold up. 

Here’s the best one: a combination of focus and personal connections.  Charlie Rose said this to Paul Graham, who told it to me.

It seems very accurate.  There are lots of good things that I keep meaning to do but never quite make it to the top of my list; I never make any real progress at all on those.  Conversely, I find that whatever I focus on most nearly always happens.  Small-ish startups seem to be able to do about three things at once, and usually they are whatever the CEO is focusing on.  Certainly anything that is not someone’s number one or two priority is unlikely to get done in the hectic world of a startup.

Most early-stage startup founders do a bad job of getting the company to focus on just two or three critical priorities—they chase whatever shiny new object appears that day.  This is somewhat expected—the sort of people that start companies generally like doing new things, not executing relentlessly on the same things.  But restraint is critical.  It’s very easy to justify taking on one more project by saying that it won’t be that time consuming.  Unfortunately, it will likely either be time consuming, or it won’t be worth anything.  Most founders know what to do; they just don’t know what not to do.

The Y Combinator version of focus is “write code and talk to users”.  For a startup that is just a few people, most other things are a waste of time (assuming the founders have already thought through the strategy of the company, and that “talking to users” also implies getting users).  For whatever reasons, many founders love to spend time on anything else—worrying about the details of corporate structures, interviewing lawyers, doing a really good job bookkeeping, etc.  All of this pretending-to-run-a-company gets in the way of actually running a company.  The best startups we fund come to office hours to talk about their product, how to evolve it, how to grow faster, and excited to show us new features their users want.  The worst come to talk—again and again—about everything else.

On the personal relationships part, most people eventually realize it’s hard to do really good things by yourself—most of them just require too much work.  Successful startups usually find that their biggest problem is hiring, and certainly hiring well is the highest-impact thing a founder can do for his startup (and the best thing an investor can do is fund great founders).  “Always be recruiting and promoting talented people” is very good advice.  Having good relationships with the people you work with is also very good advice.

I make it a point to meet and help as many smart people as I can; besides being fun and interesting, this is important to getting things done.  These are the people I tend to try hire or fund, and I think it’s the same for lots of others.  This is not really nepotism; I think working with your friends is a good strategy, and smart, effective people tend to like other smart, effective people.  The best hires I’ve made or seen other companies make are usually friends or friends of friends.  Partnerships and sales also rely heavily on personal relationships, for many of the same reasons.

It’s easy to not spend enough time on personal relationships—it seems in conflict with focus.  But it’s an important exception.  It’s also one of the most enjoyable parts of work.

When you combine extreme focus and great teams, magic happens.

“I’m an ambitious 19 year old, what should I do?” 

I get asked this question fairly often, and I now have a lot of data on what works, so I thought I’d share my response.

Usually, people are deciding between going to college (and usually working on side projects while they do so), joining a company, or starting their own startup. [1] [2]

The secret is that any of these can be right answer, and you should make your decision based on the specific circumstances of each option.  The critical point is that you want to do the thing that is most likely to get you on a path to do something great.

No matter what you choose, build stuff and be around smart people.  “Stuff” can be a lot of different things—open source projects outside of class, a startup, a new sales process at a company you work at—but, obviously, sitting around talking with your friends about how you guys really should build a website together does not count.

The best people always seem to be building stuff and hanging around smart people, so if you have to decide between several options, this may be a good filter.

Working on something good will pull you along a path where good things keep happening to you.  (In fact, this effect is so strong that there’s a danger of getting sucked into too many interesting things and getting distracted from what you really want to do.)

In making this decision, you want to take the right kind of risk.  Most people think about risk the wrong way—for example, staying in college seems like a non-risky path.  However, getting nothing done for four of your most productive years is actually pretty risky.  Starting a company that you’re in love with is the right kind of risk.  Becoming employee number 50 at a company that still has a good chance of failure is the wrong kind of risk.

If you stay in college, make sure you learn something worthwhile and work on interesting projects—college is probably the best place to meet people to work with.  If you’re really worried you’ll miss some critical social experience by dropping out of college, you should probably stay.

If you join a company, my general advice is to join a company on a breakout trajectory.  There are a usually a handful of these at a time, and they are usually identifiable to a smart young person.  They are a very good risk/reward tradeoff.  Such a company is almost certainly going to be successful, but the rest of the world isn’t quite as convinced of it as they should be.  Fortunately, these companies love ambitious young people.  In addition to the equity being a great deal (you might get 1/10^th of the equity you’d get if you join a tiny new startup, but at 1/100^th or 1/1000^th of the risk), you will work with very good people, learn what success looks like, and get a W on your record (which turns out to be quite valuable).  Spending a few years at a company that fails has path consequences, and working at an already-massively-successful company means you will learn much less, and probably work with less impressive people.

Incidentally, don’t let salary be a factor.  I just watched someone turn down one of these breakout companies because Microsoft offered him $30k per year more in salary—that was a terrible decision.  He will not build interesting things and may not work with smart people.  In a few years, when it’s time for something new, the options in front of him will be much worse than they could have been.

If you start a company, only do so if you have an idea you’re in love with.  If you’re hanging out with your friends trying to come up with an idea, I don’t think you should start that company (although there are many who disagree with me).  Starting a failed startup is less bad than joining a failed company as an employee (and you’ll certainly learn much more in the former case).  If you fail at an idea that you really loved and could have been great, you’re unlikely to regret it, and people will not hold it against you. Failing at a me-too copycat startup is worse.  Remember that there will be lots of other opportunities to start companies, and that startups are a 6-10 year commitment—wait for the right one

One big pro for starting a company is that it’s usually the way to learn the most in the shortest amount of time.  One big con is that it’s easy to start a company for the wrong reasons—usually so that you can say you’re starting a company—and this makes it easy to cloud your judgment. 

No matter what you choose, keep your personal burn rate low and minimize your commitments.  I have seen a lot of people miss great opportunities because they couldn’t afford a reduction in salary or because they couldn’t move or didn’t have the time.

Think about risk the right way.  Drew Houston gave a great commencement speech where he said you only have to be right once.  That’s true.  The risk is not getting on the path where you get to be right that one critical time. 

Thanks to Lachy Groom and Nick Sivo for reading drafts of this. 

[1] Sometimes, a 4^th option is being a VC.  This is usually a mistake—the best way to become a VC is not to grind your way up the ladder from junior associate intern.  Even if you want to be a VC, you’re much better off starting or joining a startup, and getting partner offers when you’re 28.  Plus, good founders want to work with an investor that has operational experience.

[2] Interestingly, no one is ever considering going into academia.

There’s a lot of debate about whether or not the pace of innovation has slowed, and if so, what that means.  After some particularly brutal pitches that felt like Saturday Night Live parodies of startups, I feel compelled to weigh in.  This question cannot be answered with a yes or a no; innovation has slowed dramatically in some areas and gone faster than most people would have thought possible in others.  It’s interesting to ponder why.

We were once great at innovating in the physical world.  Sometimes I stand in Manhattan and think about how amazing it is that everything around me was dug out of the ground, and made with things that were dug out of the ground.  Making an iPhone starts with digging silicon out of the ground, putting some impurities in it and making it into a chip. Then robots, also made with metal mined from the ground, assemble all the chips plus more stuff dug out from the ground to make a phone.

But recently, software (and mostly Internet software) has been the focus of innovation, and its importance is probably still underestimated—there are compounding effects to how it’s changing the world that we’re only now beginning to see.

It’s amazing how fast it’s happened.

In 1990, the Internet was 21 years old and only 2.8 million people had access to it.  (HTML/HTTP wouldn’t be released for another year.)   There are now over 2.5 billion Internet users—a nearly 100,000% increase in 23 years.  Also in 1990, there were 12.4 million mobile phone subscribers.  These were very basic phones, of course—even in 1999, there was no such thing as smartphone.  In 2012, the number of global smartphone users crossed 1 billion, and the number of mobile phone subscribers is much greater.  There are now approximately as many mobile subscriptions (not unique subscribers) as people in the world!

But innovation in the physical world (besides phones and computers) over the same time period has been less impressive.  We went from the first flight with a piston engine to the first flight with a jet engine in about 30 years; 30 years after that we landed on the moon.  Since 1939, jet engines have certainly improved incrementally, but I’m still waiting to travel by ramjet (which incidentally was patented in 1908).  When we made it to the moon in 1969—guided by computers with 64KB of memory and a clock speed of 0.043MHz, similar to a modern high-end toothbrush—people thought the rest of the solar system and the stars were not far off.  We are still waiting.

In the 1960s, oil was our primary energy source, followed by coal, gas, hydro, nuclear, and a tiny fraction of renewables.  Today, the order is…oil, coal, gas, hydro, nuclear, and a tiny fraction of renewables.

Instead of way better ways to generate energy, we talk about reducing usage.  I’m all for conservation, but there’s still something defeatist about it—weren’t we supposed to be generating huge amounts of cheap, clean energy by now? 

There are countless other examples of apparent lack of breakthrough progress in materials science, biotech, food, healthcare, etc.  I can’t order food from a Replicator.  In some ways we’ve even backtracked—I have not been able to travel faster than the speed of sound since I was a kid.   (To be fair, there may be great innovation happening now that is difficult to see.  Innovation often looks like the creation of idle toys while it’s happening, and only in retrospect can we see how important developments really are.)

There’s another important consideration in addition to software/physical—short-term/long-term (or incremental/radically new).  We’ve seen much more progress on things that can be incrementally improved in a short time frame than things that require open-ended, new development.  This is true for both software and physical things.  For example, there are a huge number of people working on making websites really nice-looking and easy-to-use—and they’re doing a fine job—but not very many people working on artificial intelligence.  There are a decent number of people working on more efficient jet engines, but not very many civilians working on replacing the jet engine with something new.

Incremental progress is good—it compounds, and it’s a way to develop great things eventually.  It works a much higher percentage of the time than step change innovation.  It’s certainly the model I’ve observed help many startups reach great success.  The iPhone, which I think is the most important innovation of 2005-2010, came about at least somewhat via compounding incremental progress.  But there were a few critical discontinuities, and Apple took as much time as they needed to figure those out.  They also had the luxury of one guy saying that they were going to do these crazy things like ship a phone with no keyboard running a real OS and require everyone to buy a data plan. 

Part of the reason Internet businesses have been so successful is that it’s easy to iterate quickly with incremental changes.  This makes them appealing businesses.  But there are some things incremental refinement will never get us.

On the software side, it’s relatively easy right now to raise money for a new website or mobile app that brings an existing, offline industry online.  In most cases, this is a real, but small, step forward.  It’s harder to raise money for software companies working on uncertain, long-term, high-reward projects (like AI).  When it comes to hardware, it’s medium hard to raise money for a consumer hardware company, even with a well-understood plan and a short timeframe to ship [0].  It’s very difficult to raise money for a new car company, a new rocket company, a new energy company, etc.

Right now in Silicon Valley, most investors are more interested in your growth graph than your long-term plan—i.e., more interested in the past than the future.  Some possible explanations for this are risk aversion and intellectual laziness.  (Actually, it’s a little better than that—compounding growth is an extremely powerful force, and if investors believe growth will keep going at the same rate, then they can be making a very wise decision.)  This focus makes it hard for companies to raise money if they’re doing things that won’t have a growth graph of any sort for years.

Looking at the past and projecting it forward is good for public market investors, who probably should judge a company on past performance.  But venture investors are supposed to be taking risks on unproven technology and ideas.  Perversely, in the current world, public market investors love to speculate about the future when pricing a stock (sometimes getting themselves badly burned, e.g. the Internet bubble) while most venture investors—the occasional $40+MM A round aside—seem to be leaning in Benjamin Graham’s direction [1].

A two-by-two matrix of progress in innovation (and ease of attracting investment) with software/physical on one axis and short-term/long-term on the other axis looks like this:

 Green is good, yellow is ok, and red is bad.

I don’t think it’s totally correct to say that innovation has slowed—what’s happened is that innovative energy has mostly been directed to short-timeframe opportunities in software.  This is at least mostly rational—honestly, over the last few decades there have just been way more successes in the computer industry than in biotech, cleantech, etc.   The virtual world is far, far less regulated than the physical world, which makes it a less risky place to start a new business.  (Although a very interesting side note is that many of the most successful Internet businesses have been ones that connect the virtual to the physical world in some way.)   And many founders, if you catch them after a few drinks, will often say that they want to get rich in a small handful of years, and timeframes on the Internet are much faster.  Many investors are even worse on this point.

And most importantly, computers are awesome.  Again, it’s not clear to me that we’re worse off focusing our collective energy here, although I do think we should do more on the long-term opportunities.  If the Internet is the most important development of the last 30 years, it makes sense to have most of our best people working on it [2]. 

And now we get to the fundamental issue with innovation today.  People—founders, employees, and investors—are looking for low-cost (and virtual is lower cost than physical), short-term opportunities.  But what causes the low-cost, short-timeframe preference?

For one, there is some deeply human appeal to getting rich quickly, which has probably been around forever.  Investing in software businesses has worked (sometimes), and will probably continue to do so for some time.  Software has been a great place—that did not exist during the physical innovation boom—to direct investment capital.

We haven’t had a really capital-intensive war in a long time, and while that’s obviously great, I am always astonished when I read about how much real innovation has involved the military or national security—it’s hard to get effectively unlimited budgets and focus any other way.  The Internet itself came from the military.

Regulatory uncertainty for non-Internet businesses is another issue, which can at least be improved—although it will be challenging to ever match the freedom of the Internet. 

Perhaps most importantly of all, we can use software to drive innovation in the non-software world by doing as much as possible in software.  We should focus on the intersection of software and everything else—for example, programs that let us do design in software can bring down cost and cycle time.  I believe that short sub-project times are almost always a win, and many companies in the physical realm seem to have long timelines without a lot of intermediate milestones. 

Another thing we can do is to better reward truly long-term investing.  Right now, there’s a focus on the extreme short-term—many investors care mostly about what earnings are going to be next quarter, and the rise in volume of weekly options has been impressive.  Unfortunately, what Wall Street thinks trickles down the pyramid to venture capitalists and even angel investors, so we see a shortening of time horizons everywhere.  This is probably difficult to stop entirely, but I’m sure that a very favorable tax policy on multi-year holdings (and an even less favorable one for short-term holdings) would help. 

Venture funds commonly have a ten year life; lengthening that to fifteen or twenty might help.  I’d also like to see companies move to five- or six-year vesting for founders and employees. 

Or maybe we need a new funding model for radical innovation.  Investment capital will look for the best risk-adjusted return; right now, lots of investors believe Internet companies are it.  As mentioned above, many of the things we consider to be breakthrough innovations with physical things happened before computers, when there was not such a readily available and attractive investment platform.  Investors are often making rational decisions by choosing a shorter-timeframe opportunity that requires less capital.  So maybe the government needs to spend more money on discovery (although right now, it’s decreasing funding for things like the National Labs and the NIH), and maybe there’s a philanthropic model that makes sense.

And there are probably a lot of other things we should be doing too. 

I am a proud child of the computer age.  I can’t imagine life without them, and I’m happy lots of people are working to make them better.  Although it makes sense that most innovation is happening incrementally on the Internet, it would be a shame for us as a species to lose the drive for the sort of discontinuous innovations that got us the Internet in the first place.  I can’t go live in a computer yet, so I’d also like the real world to continue to get better.

Thanks to Jack Altman, Patrick Collison, Kevin Lawler, Eric Migicovsky, Geoff Ralston, and Nick Sivo for reading drafts of this.

 [0] ]It used to be really hard, but preorders have made it easier.  For example, Pebble was unable to raise money from investors until they no longer needed it—they collected more than $10MM in effective preorders on Kickstarter.  But until then, investors dismissed them, saying “we don’t like hardware.”

[1] No disrespect to Color’s investors—they believed strongly in something radical, without a growth graph, and took a big capital risk.  I’d be happy to see more of that, but maybe for different sorts of companies.

[2] Talk about STEM—the study of science, technology, engineering, and mathematics—is mostly BS.  Learning to hack Rails in ten weeks is regrettably a good career move as of June 20^th, 2013.  Learning physics over ten years has a much worse economic payoff.  In today’s world, we have programming, and we have everything else.  At least in terms of a plan to become qualified for better jobs, STEM should be renamed CS.

As a side note to the side note, I don’t blame finance for stealing our good engineers.  The failure is that non-CS engineering career paths don’t look very attractive right now.