My notes on Superintelligence by Bostrom

On the plane to the US I finished reading Nick Bostrom’s Superintelligence. I jotted down notes as I went and thought a few friends might be interested so posting here.

Bostom’s background spans philosophy (he is a professor at Oxford), computational neuroscience and physics - his breadth of knowledge makes this a broad reaching read. It’s particularly interesting if you have a basic understanding of machine learning and want to understand some of the philosophical and ethical questions raised by superintelligent machines.

A few things that stood out for me:

- various surveys of AI experts (who are plausibly at the optimistic end of the spectrum :) ) peg the likelihood that we will see machines with human level intelligence by 2040 at 50%, and 90% by 2075

- Bostrom convincingly argues that once human level machine intelligence emerges we may rapidly see an ‘intelligence explosion’ where the intelligent machines self-enhance their own software/intelligence at high speed. This leads to machines that are superintelligent. Since software can be copied the population of superintelligent machines can grow rapidly.

- He then argues that given the kinetics of such an explosion one entity may end up rapidly accelerating past other machine intelligence projects and forming a dominant position. This echoes the writing of Lanier and others on the increasing centralisation of power within the technology industry. He makes a particularly interesting point that digital agents may tend to greater centralisation of control due to reduced inter-agent transaction costs. For example the idea that firms or nations of machines could massively increase in size.

- the majority of the book focuses on what happens after a superintelligence emerges. He draws an interesting distinction between having more intelligence and more wisdom - and the risks of one developing without the other. He gives a hilarious example, worthy of Foster-Wallace where a superintelligent machine is tasked with producing 1000 paperclips. The machine, being superintelligent and supercapable rapidly produces 1000 paperclips. However, being a perfect Bayesian agent it is also aware that observational error may mean that it has actually produced fewer paperclips than this - there is a tiny but real chance it has only produced 999. So to remedy this it commandeers all the resources in the known universe to more accurately count whether it has actually produced 1000 paperclips or not. He lays out various types of superintelligence and various ways that things could go badly wrong for humanity from goal functions that on first glance seem to be bounded, but per the paperclip example are not. At lot of this seems to be the difference between programatic logic and ‘common sense’ and the complexity in creating a bridge from one to the other.

- he draws an interesting parallel between the fate of humans in a world with superintelligent machines, and the fate of horses in a human world. The horse population grew massively through the 1900s as a complement to carriages and ploughs, but then declined with the arrival of automobiles & tractors. The population of horses was 26m in the US in 1915 but declined to 2m by the early 1950s. The flipside of this is that the horse population subsequently returned to 10m driven by economic growth that have allowed more humans to indulge in leisure activities involving horses. 

- He explores how superintelligent machines might acquire their values. This section on value loading techniques is very interesting and summarises some of the most interesting mathematical and philosophical challenges facing the AI space. For example in one unfinished solution to the value loading problem we have a subset of intelligent machines that are known to have values that are safe for humans. These machines are allowed to develop a incrementally more intelligent machine - where the step in intelligence between the first group of machines and the mutation is small enough that the earlier machines can still test the new, slightly smarter machine to see if its values remain compatible with humanity’s safety. He makes the terrifying point that if there is an arms race going on for one company or nation to develop superintelligent machines first, this kind of caution is unlikely to be on the path of the ‘winning’ project - ‘move fast and break things’ seems like a bad motto when you are playing with something this powerful.

- Having framed the challenges of loading a superintelligence with values, he then moves to what values we want this superintelligent to have. Bostrom argues that humanity may have made relatively little progress on answering key moral questions and is likely still labouring under some grave moral misconceptions. Given that are we in a position to specify a moral framework for a superintelligent machine? He introduces the concepts of Indirect Normativity and coherent extrapolated volition in response to this - a hedge against our own limited moral framework and a bet that the machine can do better:

"Our coherent extrapolated volition is our wish if we knew more, thought faster, were more the people we wished we were, had grown up farther together; where the extrapolation converges rather than diverges, where our wishes cohere rather than interfere; extrapolated as we wish that extrapolated, interpreted as we wish that interpreted" - Yudakowski

finally he asks how to ensure that the immense economic windfall resulting from superintelligence should be distributed to benefit all of humanity, not just a narrow set of people (or machines).

Overall I found it very stimulating and would recommend.

Save me money, save me time, or save me both

Consumer technology can find a mass market when it saves us time, or money. Every person has their own preference for how they spend time and money. While some people would prefer to save time by spending more money (for example paying to avoid a queue at an airport) others prefer to save money by spending their time (for example going down to the venue to buy a concert ticket to save money on booking fees). Sometimes technology can save us both - for example how at launch Amazon saved both time and money when purchasing books.

If you plot this out on a chart it seems that start-ups can successfully enter a market in one of three quadrants - saving time, saving money, saving both. The logos correspond to my understanding of the proposition that the initial users of these services were compelled by:

image

So for example Priceline offered a cheaper airline ticket if you were willing to be more flexible with your time. Uber offered SF residents a way to spend more money but save time by waiting for a cab. In both cases users traded time for money or vice versa. On the axes, Groupon saved their first users money without requiring extra time, and Google saved their users time by getting them to information faster, but without costing more.

In the ticketing space, the launch of TicketMaster (over 30 years ago now) offered a step change in convenience by saving you the time it took to go down to the venue, providing you were willing to pay a new service charge. Continuing this trend, Stubhub offered the ability to skip the onsale process entirely and buy from the resale market at a higher price later. Most of the concert goers I’ve spoken to who prefer to buy tickets from Stubhub are clear that they are deliberately spending more money to save time.

It would seem that to get initial growth a start-up needs to save users time, money or both. (That is unless you can solve some next level Mazlow need and get us all self-actualised a la Snapchat :-)

What surprised me is how these value propositions can shift over time as network effects develop. I’ve been thinking about this a lot recently as Uber has grown. When it first launched I saw it as somewhat similar to Stubhub - a service for affluent consumers to save themselves time in exchange for more money. What’s been fascinating to watch is how over time as more drivers and riders join the network it has become a multi-tiered service that can save both time and money with the introduction of UberX, greater taxi utilisation, and soon, ride-sharing. So over time you see:

image

The big takeaway for me is that a start-ups initial time/money proposition isn’t always a predictor of its eventual market impact. A service that started with "We just wanted to push a button and get a ride…And we wanted to get a classy ride. We wanted to be baller in San Francisco. That’s all it was about.” might end up reducing transportation costs and saving time for the mass market.

A Person Got to Have a Code (and a phone)

"A man gotta have a code" - Omar

The average person checks their phone 150 times/day. Setting aside whether that’s a ridiculous number or not, that got me thinking about the fact that we look at one image - our lock screen image - 150 times/day. 

There are probably quite a few interesting product opportunities that fall out from an image we look at 150 times/day or 55k/year. Other than a tattoo I can’t think of an image we looked at in the past as frequently or in as many contexts.

The specific thing I’ve been noodling on is how to use that image to reinforce behaviour. I have a few principles** I try to keep in mind as I go through life, but it can be easy to lose sight of them in the swirl of daily emotion. I’ve been experimenting with putting that list on my lock screen so I end up looking at it in passing 100+ times/day.

I’ve been doing it for a month or so now and have noticed that for the most part I’ll just tune out the image of the list en route to completing a task on my phone. But from time to time it does pull me in (often when I’m just getting out my phone to kill time). I’ll re-read them and perhaps something is reinforced or questioned.

I wonder what else could be done with an image we look at 150 times/day? 

** 1. YOLO, 2. FOMO, 3. BRB, 4. what would ‘ye do, 5. what would bill murray do

My talk from YC startup school

Hi I’m Ian Hogarth and I’m one of the co-founders of Songkick along with Michelle You and Pete Smith.

image

We started Songkick back in 2007 and were part of the summer ‘07 YC batch.

image

Songkick is the easiest way to find out when your favourite artists come to town and get tickets. If you’ve ever experienced the frustration of finding out that your favourite band was playing the day after the show then we’re for you. We’re the second most used concert service in the world after TicketMaster with about 10m unique fans/mo. We’re backed by Index and Sequoia. Given that the average artist makes 70% of their income from concerts we hope we will make a big difference to artists as well as fans.

One thing that comes from building the same company for 7 years is you get to watch waves of start-ups succeed and fail around you and your intuition about start-ups gets rewired. It’s kind of like rewatching the first series of 24 after seeing the final episode of season 8.

image

I remember being terrified by a competitor to Songkick that launched while we were only just getting started. They rapidly grew to millions of users. However over the next few years the foundation they’d built their growth on moved underneath them and they disappeared. That resets your sense of what to be scared by.

Similarly you will see startups that seem to have it all figured out, and when they become the talk of the town, it doesn’t surprise you. What may surprise you is what exponential growth looks like. That start-up that was doing well and just a bit better than you goes from 1m users to 10m in a year. And you’re still like ok we have a big year coming. But then you see them go from 10m to 100m the next year. And that resets your sense of how things can grow.

To put this in perspective, if I go back to that summer of 2007 only a few of the 22 start-ups in our YC batch are still in existence. I believe most of the others ended up being shut down or acquired in relatively small deals. Watching that play out really teaches you how hard it is. I remember being intimidated by everyone in our batch when we got to YC. So many people who were better technically, better product thinkers, and more experienced at building web products than Pete, Michelle and me.

image

One of the other start-ups in our batch that is still around is Disqus. If I massively oversimplify for a second, it’s plausible that Songkick and Disqus may end up being worth 100X more than start-ups in our batch that were sold early, so there may be something to be learned from what was different about our markets and our path.

image

But the more radical example is that other start-up that endured from our batch, Dropbox. Which is likely worth 100X more than us at this point.

image

What’s even more remarkable is that Drew and Arash remain two of the most humble and down to earth founders I know.

image

There is plausibly someone in this room who will go on to create something 10X bigger than Dropbox.

image

So hopefully you’ll take this as a bit of a disclaimer for all the advice that follows - if you really want to know the mysteries of the startup universe go talk to Drew and Arash! Also I am most interested in consumer products, so most of this talk applies to them.

So having appropriately caveated that I have at least 100X less insight to share about what goes into building the next Google than most other speakers who have spoken at startup school, I thought about what I would most like to have had someone explain to me in retrospect.

image

Firstly, on online music as an excellent way to take a beat down. Here are some exceptionally talented founders/builders who have, to a greater and lesser extent taken a beat down by doing a music start-up:

image

Dalton Caldwell now a YC partner and the former CEO of Imeem; Sean Parker; Geoff Ralston the creator of what became Yahoo Mail & former head of product of Yahoo; Ali Partovi founder of LinkExchange & iLike, Dave Goldberg CEO of SurveyMonkey and Launch Media, David Pakman Venrock partner & former CEO of eMusic - all drawn to the flame! I thought I could put a useful talk together summarising what I’ve learned from talking with some of these great people who have built music start-ups. I then realised that Dalton already did that talk.

Dalton’s YC talk is an excellent primer on the music industry so I’m not going to rehash his advice here. What I will offer you is a slightly more reductive view on the art and entertainment industry - film, TV, music, visual art - if you’re thinking about building a company in one of those domains here is what I think you need to understand:

image

1. everyone thinks they are more into music/film/art than their actual consumption reflects. it’s like that study you may have read about around the sub-prime crisis where most homeowners felt that average house prices would decline during the next 6 months, but also that their home would stay the same or increase in value

2. actually most people are into a much narrower set of things than they realise. And so a small number of creators drive the majority of activity and revenue for these industries

3. 99.99999% of artists struggle their whole lives without financial success, so when they break out they are often willing to trade future rights to their music or touring or merchandise in exchange for financial stability. Having artist friends who are still sleeping on friends’ couches after 10 years this isn’t selling out, it’s about getting stable and finally paying off your debts. Unlike tech where you can be a talented engineer at a start-up that fails & still go get a great job at Google, if you’re a great musician that fails to make it you can’t just go join Radiohead

4. this transference of rights from artists->middlemen leads to a small number of companies controlling a huge amount of the rights that are needed to innovate on behalf of the artist or consumer. In addition as the entertainment industry lurches from one model to another (for example CDs to downloads to streaming) the perceived threat to their survival means regulators allow even more radical consolidation - for example in the last 10 years the world’s largest record label Universal acquired EMI, who combined now represent something like 40% of popular music rights. The same thing happened in the live industry with the merger of the biggest global concert promoter, LiveNation with the biggest ticketing company, TicketMaster. And don’t be fooled into thinking that other parts of the music industry aren’t rights oriented markets, both merchandise and ticketing both have similar rights systems in place comparable to copyright in recorded music

5. that level of rights consolidation means that it’s almost impossible for a start-up to transform the entertainment industry without some permission from one or more powerful content owners, which means waiting till they are ready to embrace you. That in my opinion is one of the differences between Spotify’s success now and the nightmare that Dalton went through. The one caveat is that much much larger technology companies can force things to move faster - for example the way that Google protected YouTube from potential label annihilation and 8 years later owns the largest free streaming music service on the planet - or the way that Apple created a digital download market in one big move

6. So I think the biggest question to ask yourself as someone aiming to build a technology company serving fans and artists in the entertainment industry is why will the labels/promoters/agencies/studios/galleries be ready for this now? If pg said Kill Hollywood, I guess I’d say Grow Hollywood or fail

The broader point here is that the level of supply-side consolidation in your industry massively changes how you build your start-up. If the industry is heavily consolidated you are more likely to have to partner with the supply-side middlemen. Consider for example the work that Stripe did with the banks early on. If the supply side of the industry is more fragmented (e.g. vacation rentals or private hire vehicles) you are probably best off competing directly with the existing incumbents. 

The flip side of this is that it’s incredibly rewarding to work on a product that helps fans & creators in an area of culture you love. I’m proud of what we’ve done so far to improve concert-going and I’m inspired by what Netflix, Spotify and others have done for their respective industries. You just need to make sure that your timing is right for the middlemen, as well as fans/artists.

The second thing I want to share today is the importance of understanding the start-up game before you try and play it.

 image

Start-ups aren’t a straightforward path to financial success. Particularly in consumer start-ups the level of randomness means that as someone who is good at building things you’re more likely to make money from joining a great company as it starts to take off (the first first 100 employees @ FB probably made more money than the founders of most successful start-ups).

So I think the reason to found a company is actually a less financially oriented one - you are really motivated to try and solve a particular problem and the satisfaction of building something to solve that problem is enough to balance out 5-10 years of high stress and a good likelihood of failure.

So if I haven’t deterred you yet, then here are the rules of the game as I see them. Hopefully internalising these challenges early on will help you be more successful. There are 3 engines that determine a start-up’s success:

 image

A gratification engine, a growth engine and an economic engine. This insight came to us via the legend Sean Ellis about 3 years into Songkick’s life.

 image

If I define a new variable ‘Unicornness™’, your level of unicorness will be roughly:

Unicorness = gratification^growth^revenue

You become full unicorn aka Airbnb, Dropbox, Google if you get all 3 right. Every engine that fails to start will reduce your unicornness an order of magnitude.

image

Let’s take the gratification engine, expressed in much less nerdy terms by YC as “make something people want”. In Songkick’s case figuring this out was a pretty brutal experience. We launched Songkick to solve the problem of knowing when your favourite artists were coming to town. Our first release combined a few different scrapers of ticket sites that generated an incomplete dataset of concerts in the US and UK and a mac plugin for iTunes that you had to download and install that would scan your itunes library. It was a pretty crappy first time use.

image

The return on the 5 minutes of your life that it took to sign up and install the plug-in was a list of upcoming concerts, personalised to your music taste. Some people were willing to do this and they liked it. But the amount of friction involved was too high for most regular fans. And Songkick ends up being most powerful for regular fans who at present go to ~1 show / year, but after getting Songkick might end up going to 4 or 5.

For a long time we felt that the reason that more people didn’t use our product was that it didn’t do enough and we added a massive array of additional features that resulted in very little additional usage.

The turning point came when my co-founder Michelle, inspired by Sean Ellis, started running surveys that really dug deep into why the users who loved our product loved it. And why the users who were kind of ‘meh’ found it lacking. The bottom line was that our simple idea of personalised listings and not missing another great gig was actually a gratifying enough experience for all types of music fans - people found shows they wouldn’t have gone to and had life changing experiences! but too few users were getting to it. We needed to make it radically easier for you to give us your music taste (which became possible with new APIs on mobile devices) and we needed to have better underlying data. That was a really big lesson for me. Engineers and start-up people cherish the idea of 80:20, or the idea of an MVP. But once you find something that works, the key is to do the 20:80, the grindingly incremental work that adds the final 20% of the value, but takes 80% of the time.

image

For us this has ended up being around data - getting more and more high quality, timely, more comprehensive concert data so we became the trusted authority for a fan. Your gratification engine will have many levels of refinement that compound on each other - onboarding flows, core experience, messaging etc and you should probably never stop trying to increase it.

image

The next engine is the growth engine - how new users discover your product. The first big point here is that your growth engine has no real chance of starting without a great product. I’ll talk more about that in a minute, but I think there are 4 main ways to drive substantial growth in consumer products:

image


These aren’t mutually exclusive. For example Yelp has a killer mobile app so they grow through both world of mouth and SEO. Airbnb also has strong WOM which means they can amplify a referral program with paid acquisition. They’ve also grown through M&A, PR & more creative growth hacking e.g. the alleged craigslist thing. So many different growth channels can combine effectively.

For us there were a number of different drivers of growth. The first was realising that there was no canonical page on the internet for a tour or concert - similar to what Yelp does for restaurants or IMDB for films. And when you build enough value to be the canonical page for something you see lots of different sources of growth, from social referrals to API/platform opportunities to SEO. That core insight lead to a flurry of things that caused us to grow from the BD partnerships we did with YouTube, Spotify, SoundCloud and others to our artist facing products. The second big growth factor came from mobile and WOM - in my opinion, mobile app stores reward a gratifying product more than any distribution platform in history and so much of the growth we saw there came just from making the product easier to use so more people would recommend it to friends. So we’ve benefited from 3 of these channels.

image

Finally the economic engine, how you make money from your users. I can’t say as much about this because it’s still a work in progress and one reason that we’re not (yet!) in the pantheon of unicorns. Initially we bootstrapped revenue by setting up affiliate partnerships with ticket vendors who pay us when we generate a ticket sale - similar to the model for Kayak or TripAdvisor. That has taken us to millions of dollars in revenue on gross ticket sales of over $100m. However it is unlikely to take us hundreds of millions in revenue. Our goal is to enable fans to buy tickets in the Songkick app as well as getting linked out to 3rd party sites. Per my earlier discussion on music start-ups this depends on being able to scalably access inventory in partnership with artists, promoters & venues, which is still a work in progress. It’s exciting though - in London you can now buy tickets to a huge number of gigs through our app (over 25% of all shows in London & growing fast), and we’re rolling out other geographies soon.

image

Finally there is the team you build and retain to solve all these problems. This is the Songkick team en route to a festival when our bus broke down!

Each of these are dependent on each other. If I had to express it mathematically it would be something like:

image

Firstly, gratification is a function of your team, your economic engine and your growth engine.

That’s because you need a great team to build a great product. And a powerful economic engine can be a big part of your gratified experience (e.g. saving users money). Finally in many instances a consumer product gets better with more users (e.g. a marketplace or social network), so you may also need growth to deliver a gratifying user experience.

So the gratification engine depends on the other two engines and your team.

Secondly growth is a function of your gratification engine, your economic engine and your team.

That’s because great execution on growth requires hiring great growth people, so growth is dependent on team. If you have a revenue engine you are also able to pay to acquire users and access a powerful source of growth. One key point to make here is that it requires more expertise to grow through free channels than by spending money on paid channels - compare the number of start-ups that figure out how to grow virally on Facebook with the number of start-ups who figure out how to buy FB ads. Most importantly without a great gratification engine you won’t get the most powerful and fundamental growth driver, word of mouth. As yet another example of this interdependence - our partnership with YouTube didn’t start via VC connections, or epic biz dev. It started from a product manager at YouTube reaching out because he loved our product as a regular user.

So the growth engine depends on the other two engines and your team.

You can make similar arguments for the interdependence of the team you can hire and retain and your economic engine on everything else.

So:

unicornness = product*revenue*growth

growth = f(product, revenue, team)

team = f(product, revenue, growth)

product = f(team, revenue, growth)

revenue = f(growth, product, team)


AKA everything is connected and you’re watching the first season of True Detective.

image

I’ve laboured this point because it seems like the most important thing to understand about start-ups is that it’s all connected and you need to get all of these key pieces working in concert to build an exceptional business. The earlier you figure out the whole system, the earlier you get on the path to becoming the next Dropbox.

image

Finally, all of this takes time and is very hard and you can’t give up. So some thoughts on resilience and how to develop it.

Firstly - it does usually get better if you keep going. I remember the bleakest point in Songkick’s life was around december 2010. Nothing felt like it was working. We went into Christmas after a pretty brutal board meeting with a plan for some things we’d try in the new year.

image

When things get hard I go back to our growth graph since then and look at that same spot.

image

It usually does get better if you keep moving and trying new things. As pg says be “relentlessly resourceful”. As the founder of a great company told me once - survival can be a growth strategy. The best thing about surviving is that you get to see new platform shifts for example the shift to mobile that Songkick has grown through. Everyone likes to talk about how new start-ups get built when new platforms emerge. But things that are already working can suddenly work a lot better. For example Shazam and Pandora are two companies that were 8 and 7 years old at the time of the iPhone launch and had been great, but not total breakouts. The iPhone played a big role in changing that. I remember hearing from the Pandora team that the iPhone launch doubled growth for them overnight.

Platform shifts expand the set of start-up visions that can finally be fully realised. So let that be another reason to push through the hard times.

I would recommend trying to articulate why you believe you are doing important work. I think a good way to do that is to keep asking why until you get to the root. We wrote those down a few years back and here’s what we came up with:

image

Then when you are low you have something to remind you why you’re going to work through whatever todays flavour of crisis is. 

When you’re having a bad week, spend some time with your users. The happy ones will remind you of why you started. And the unhappy, disengaged ones should help you transform an abstract sense of impending doom into a practical feeling of something to fix. Our product team ended up knocking through a wall in one of our meeting rooms and creating a makeshift user research lab. That helps to set a regular tempo for having our whole team watch our users use your product as individuals, not in an aggregate Google Analyticsy way.

image

Start your company with people you can count on when shit is going sideways. I think it’s pretty hard to know that about someone without a real foundation of friendship so I thoroughly endorse YC’s thing about building on top of a long standing and trusted relationship. I have been very fortunate to have two amazing co-founders in Michelle and Pete and an amazing team, many of whom have been with us from very early on. I can’t imagine how I would have weathered some of the tougher moments without them.

image

So in summary:

- if you’re going to do a start-up in the entertainment industry or any industry where the supply side is highly consolidated, you’ll probably need to work with the existing middlemen. So start trying to understand how you can help them in addition to fans & creators.

- consumer start-up success seems to depend on getting 3 key fundamentals right: gratification; growth; economics. Understand how you’re going to do that as early as possible.

- even once you find something that people want, there will still be days when it feels hopeless. I’ve suggested a few ways to nurture your resilience when that happens. If you keep moving, you’ll find a way through!

Thanks.

What should Benedict Evans’ business model be?

image

I started sketching out this post before Benedict joined a16z, so excuse the fact that his ‘business model’ is now resolved for a little while :). For ‘Benedict Evans’ feel free to substitute any smart, public intellectual unaffiliated to an existing organisation with a wealth of insight to provide.

When Benedict and his blog emerged on the start-up scene, it was an interesting event: he had over a decade of experience in mobile & media and was rapidly building an audience of the most influential people in tech. The most entertaining feature to me was his weekly summary of how many more readers he’d acquired that week. I reached out to him when his blog was just getting started and we had a good conversation over breakfast. I really admire him and watched him build his audience with fascination. The question that kept coming back to me was - what should his business model be?

The value proposition as I see it is: world class insight into a particular domain (in his case, mostly the evolution of mobile/media) which he was providing for free. Normally such insights are clearly part of a broader free+paid model - Fred Wilson was eloquent and explicit about this when describing his blog: "It is the model behind this blog in fact. You get the content for free. Anything else, you have to pay for with equity in your company". 

The ways to get paid that I could see at the time:
- put some of the content behind a paywall (e.g. similar to Ben Bajarin at Techpinions or Om Malik) ie charge a subset of the public for incremental insight
- get paid offline (e.g. Eric Ries' book, events etc)
- build a product around his community (e.g. Sean Ellis & growthhackers that monetises in another way)
- get paid in cash/equity as an advisor to various start-ups, VCs, established businesses ie charge individual businesses for company specific insight
- get paid in carry as part of a VC firm

I have to say I was hoping that somehow this might represent the unbundling of VC and that Benedict would build a standalone business around advice/insight. But I suspect that what he’s doing now will probably be the best move he could possibly have made. It is interesting to consider what % of a16z he has vs the other partners/contributors. It would be a great data point on the value of insight.

Building start-ups from first principles

image

image credit

In my opinion one of the most helpful things YC has done for the start-up ecosystem is to apply a reductive mindset to start-up strategy: Make something people want. There is so much wisdom contained in that one sentence.

The thing it reminds me of most is studying mechanics at university. The way I was taught mechanics, your goal is to take a small set of first principles (mainly Newton’s laws of motion) and learn how to rigorously apply them to understand the motion of physical objects. The point in a classical mechanics course where you apply these simple laws and end up with a model of how a gyroscope works is breathtaking.

I watched Elon Musk speak at the Dublin Web Summit, and one of the most helpful points he made was that innovation is the product of reasoning from first principles (good summary here). Given his current set of companies he was mainly referring to the first principles of physics/chemistry, but I think there are some similar first principles for building start-ups that must be discovered and applied. YC’s motto feels like one of them. Like Newton’s laws of motion, the hard part isn’t reading the sentence once, the hard part is learning how to apply it to understand a gyroscope. Paul Graham’s essays on start-ups and their early customers often feel like the course notes that explain how to apply that first principle.

One of the hard things about being a start-up CEO is simultaneously building a product and a business from first principles whilst recognising the need to market your business through narrative and analogy. The classic example here is the “X of Y” pitch that drives most investor pitches. Narrative and storytelling is an incredibly important component of marketing and quickly explaining what you do. But if you want to build a great product & business, and you try to do that by analogy, you will fail. So the art is to learn how to use first principles when you are building, and analogy when you are selling.

Start-up inspiration at every age

Age 19 Matt Mullengweg co-founded WordPress
Age 20 John Collison co-founded Stripe
Age 21 Sophia Amoruso co-founded Nasty Gal
Age 22 Joe Lonsdale co-founded Palantir
Age 23 Daniel Ek co-founded Spotify
Age 24 Michelle Zatlyn co-founded Cloudflare
Age 25 Larry Page co-founded Google
Age 26 Julia Hartz co-founded Eventbrite
Age 27 Ben Silberman co-founded Pinterest
Age 28 Andrew Mason co-founded Groupon
Age 29 Bryan Johnston co-founded Braintree
Age 30 Jeff Bezos founded Amazon
Age 31 Perry Chen co-founded Kickstarter
Age 32 Victoria Ransom co-founded Wildfire
Age 33 Jan Koum co-founded WhatsApp
Age 34 Jessica Livingston co-founded Y Combinator
Age 35 Reid Hoffman co-founded LinkedIn
Age 36 Renaud Laplanche co-founded Lending Club
Age 37 Reid Hastings co-founded Netflix
Age 38 Jimmy Wales co-founded Wikia
Age 39 Martin Lorentzon co-founded Spotify
Age 40 Aneel Bhusri co-founded Workday
Age 41 Paul English co-founded Kayak
Age 42 Robin Chase co-founded Zipcar

…running out of time to Google people’s ages but here’s a start on 43+:

Age 46 Linda Avery co-founded 23andMe
Age 50 Andrew Viterbi co-founded Qualcomm
Age 52 Kenneth Lerer co-founded The Huffington Post
Age 55 Arianna Huffington co-founded The Huffington Post
Age 58 Satoshi Nakamoto (maybe) co-founded Bitcoin
Age 64 David Duffield co-founded Workday

Your age and experience (or lack of) will help you and hinder you in different ways. Ambition & opportunity, at the right time in your life, with a big problem you want to solve, is probably more important.

ps if you can help me fill out names for the blanks post age 43 then please leave in the comments and I’ll add.

pps some interesting empirical data here and here.

The best ‘growth hackers’ don’t talk about ‘growth hacking’

image

Over the course of building Songkick, and seeing it grow to over 9 million fans/month, I’ve had the pleasure to learn from and work with some exceptional thinkers on growth. I’ve learned the most from Dan Rogers who has lead our quantitative growth efforts since 2010, Sean Ellis and Andrew Hunter who was an early advisor to Songkick and Dan’s mentor. I love learning about marketing and have at various points in Songkick’s past tried to make sense of various types of distribution, primarily:

- quantitative growth channels e.g. viral loops, email marketing, widgets, SEO, SEM

- partnership or narrative driven growth channels e.g. BD partnerships, PR

Quantitative growth channels are now what’s termed ‘growth hacking’.

It’s a mix of creativity (finding novel channels to drive growth that competitors have not yet discovered) and technical/analytical skill (scalably exploiting those channels). The lifecycle is:

1. find a channel that works before your competitors (and if the channel is broad enough this may include all consumer apps)

2. exploit that first mover advantage

3. eventually competitors cotton on and you hit the Law of Shitty Clickthroughs, or the channel closes for some other reason (e.g. platform you’re building on decides to compete)

4. get creative & go find another novel channel. Return to step 1.

Another way of looking at this is that new marketing channels have a limited amount of new users they can supply, and there is infinite demand from start-ups who want more users. It’s a zero sum game, where the overall value of winning the game also usually declines over time.

So the best growth hackers shut the fuck up about what’s working & hope that they can keep the channel to themselves for as long as possible. Usually the only way you learn really novel marketing approaches is by spending time in person with a trusted peer, sharing your respective secrets & hoping the novel things you learn from them balance out the information leakage.

This is at odds with what’s going on in start-up land at the moment where there’s now even a forum dedicated to ‘growth hacking’. I’m really skeptical you can learn anything there beyond what’s worked in the past (which is no longer relevant) and how to apply a quantitative approach to marketing. I have a theory that the growth hackers forum is actually just an elaborate growth hack by Sean Ellis to market Qualaroo!

So if you’re a start-up looking to grow, get creative, find your unique approach to growth and then keep it to yourself.

The 10 books, films and artists I loved most in 2013

I did this last year and friends seemed to enjoy it, so here’s the 2013 instalment. 

Books (most published before 2013):
1. Ulysses (James Joyce
2. In Search of Lost Time (Marcel Proust)
3. Catch 22 (Joseph Heller)
4. Average is Over (Tyler Cowan) 
5. The Sense of an Ending (Julian Barnes)
6. On Chesil Beach (Ian McEwan)
7. Who Owns the Future (Jaron Lanier)
8. … that was it. Ulysses & Proust took up most of my reading time but wow they were worth it


Musical artists (new music released this year):
1. Kanye West
2. Chance the Rapper
3. Pusha T
4. Kevin Gates
5. Indian Wells
6. The Haxan Cloak
7. Jon Hopkins
8. Tree
9. Arcade Fire
10. Future

Films (released in UK in 2013):
1. Amour
2. Blue Is the Warmest Colour
3. The Great Beauty
4. Zero Dark Thirty
5. A Hijacking
6. Frances Ha
7. Gravity
8. American Hustle
9. World War Z
10. Neighbouring Sounds

What are the web’s mom & pop stores?

In general I agree with Marc Andreessen’s thesis that software is ‘eating' the world. More and more large network effects based businesses seem to be displacing existing offline businesses (Yellow Pages->Yelp, Tower Records->Spotify, Addison Lee->Hailo etc). One of the most successful venture capital firms only invests in ‘large networks of engaged users’.

But if that is happening what’s the equivalent of a small mom & pop business online? If you just want a little corner of independence and a decent living what are your options?

I guess one option is just to be a supplier to these mega-marketplaces - make stuff for Etsy, drive cars for Uber, rent properties on Airbnb. But what if you want a bit more independence than that?

Competing with network effects or web enabled economies of scale seems like a dead end - it’s going to be hard to start a small independent online bookstore with the ‘Everything Store' as competition.

I think the answer might be to build a small utility that a decent number of people find useful and charge them for it. You’ll probably be safer if there are no obvious network effects to exploit. That could be an iPhone app to read articles later like Instapaper, it could be a way for bands to put their music on iTunes and YouTube more easily like Distrokid. 

Those two services seem like unfair examples in some ways because the creators are two of the most effective solo-entrepreneurs in the world. But as tech literacy increases, maybe there are more Marco Arments and Philip Kaplans carving out an independent niche online, immune to the rich get richer trajectories of network based businesses like Amazon, Google, Apple etc.