Category Archives: Uncategorized

if history repeats itself… (post Facebook IPO predictions)

If history repeats itself, the Facebook IPO will mean:

– a bunch of second-tier social media companies go public to satisfy
investor demand for “social media allocations”
(facebook’s reportedly small float of 5b will make this more likely)

– high private valuations for social media companies will last at
least another year

– the press will write thousands of breathless stories that make it
seem like the future of western culture depends on new facebook
revenue streams (see coverage of google’s business model post-IPO)

– facebook will continue to do small talent acquisitions until they
have their “Google Video” moment and then like all mature tech
companies will start acquiring real businesses for 1b+ valuations. see
last paragraph of

– facebook will – in the eyes of the press – go from darling to “evil”
over the next 5 years. (again, see google coverage).

– the next 1-2 years will mark the end of this “patternson cycle” (see…
after that, really interesting innovation will start gestating for
the next cycle.


programming tools tested lately

python (love – still getting used to indentation and other syntax but
love the library support)
twitter search api – works great. wish they’d specify what the rate
limit is and let you put query strings in quotes for exact match.
historical searches would be great but guessing they don’t have the
technical ability / business desire to do so.
mongodb (love – first time i’ve done nosql and it’s liberating)
pymongo – made mongo+python super simple
osx brew installer (great, worked where downloading dmg’s didn’t)
matplotlib (looks great but overkill for my graphing needs)
mailgun (free version) – really nice smtp webservice/wrapper
textmate – great mac osx code editor, well worth the price
stackoverflow – again and again found the best code samples there
(disc: i’m an investor)
various cronjob UI’s for mac – didn’t work. ended up just having
scripts sleep for a while and not bothering with cron jobs.
mongohub – found this and other mongo front ends lacking but perhaps I
need to try more.

point of clarification: vested vs unvested options

There have been lots of articles about the Zynga "taking away employee's options" brouhaha (another one today in the NYTimes). One thing that bugs me is that many of these articles don't seem to distinguish between vested and unvested options. The difference is VERY important.

When you work at a startup, think of it as you get paid bi-weekly or monthly in two ways: cash (your salary) and options (the options in your option plan that vest that month).  In theory you could actually get option certificates issued to you bi-weekly or monthly the way you get cash sent to you for your salary, but that would create a lot of extra paperwork. So instead someone invented the idea that the employers tells the employee "you get N options over 4 years and they vest every month" or something like that.  Maybe to make things clearer they should have said "you get M options every month you work here" (where M would presumably be 1/48th of N).  This, however, would create a planning complications for the CEO, who needs to carve out options from the "option pool" which usually involves lots of negotiations with the VCs.

So if an employer comes to you and says they want to take away *vested* options, that's like saying they want you to return salary already paid to you. I don't think they could do it even if they wanted to and if they wanted to I think we'd all agree that was highly sketchy behavior.

If, however, an employer comes to you and says they want to take away *unvested* options that's like them saying they want to reduce your salary going forward.  That might be lame and unfair, and if anyone did it to me I'd probably quit, but it's very different than if someone tried to take away vested options.

The defensibility of network structures

Metcalf's law states that the value of a network is proportional to the square of the number of nodes in that network. This is true of networks where every node is connected to every other node. A more generalized formulation of Metcalf's law would be: the value of a network is equal to the number of connections (also known as edges) in that network. (An even better definition would try to incorporate a measurement of the value of each connection – e.g. two people who communicate a lot probably value that connection more than two people who don't).

It is widely understood that the resilience of chemical molecules or architectural structures is a function of not just their materials but also their structures. The same is true of information networks like social networks, marketplaces, and communication networks. Two networks with the same number of nodes (e.g. users) and same number of edges (e.g. relationships of Friending or Following) might have very different levels of resilience or – as is it's normally called in business contexts – defensibility.

Suppose we define the defensibility of a networked web service as:  The minimum number of users a competitor needs to capture in order to capture 80% of the value of the service.

I picked 80% somewhat arbitrarily. To measure defensibility more precisely you'd want to plot the distribution where one axis is number of users and the other axis is the number of edges each user has (I attempt this superficially here). Also note I am simplifying what Twitter and Facebook have evolved into as services. On Twitter, there are explicit edges (Following) but also lots of "soft edges", e.g. when someone @ replies a user she doesn't follow. Facebook has evolved from being a purely "undirected graph" (Friending) to being a hybrid network with the introduction of Liking and Following.

Networks like Facebook tend to have a low variation in the number of connections (Friends) per user compared to networks like Twitter where some people have many millions of followers but most people have less than 100. Academics would say Twitter is a far more "centralized" network than Facebook.

In that sense Twitter is far less defensible than Facebook. If a rival can capture, say, ten thousand of the top Twitter users, they might be able to capture 80% of the value that followers of those users get from the service.

Recently someone in charge of Google+ tweeted: "We’re about to pilot a ‘suggested user’-like mechanism on Google+. If you’ve got more than 100k followers on Twitter, DM me – lets talk!". This is a smart strategy that recognizes the primary vulnerability in Twitter's network structure.

Services competing with Facebook are better off trying to exploit clusters (e.g. geographic, demographic, interests) versus going after more "central" (popular) users. 

I am far from being an expert on the academic literature on social network analysis but from my research I haven't found anything that looks at the structures of networks from a business point of view. Interesting topics might be: the strength and weakness of various structures, strategies for attacking and defending those structures, historical case studies on how networks grew or decayed, and so on.  Perhaps someone can point me to relevant research if it exists.

“silicon alley”

People who have done tech in NYC for a long time think of the phrase “Silicon Alley” as referring to NYC during the dot-com bubble, from roughly 1995-2000. If you hear someone use the phrase to refer to today’s NYC tech scene, you can be pretty sure they know very little about the topic.

Client-side mobile speech recognition

Imagine if on your iPhone you had to type a whole paragraph, and then wait a few seconds for it to get sent to Apple’s server, and then get the text back to see if any words were mistyped or miscorrected. 

That is how speech recognition today works on mobile devices. It is all done server side (to try out a state-of-the-art example, download Dragon Dictation app on your iPhone, or try the built-in speech rec on an Android device). Perhaps Apple’s Siri will improve this (I hope!). But until speech recognition gets to be very close to 100% accuracy, the best way to improve the user experience will be to show each word and sentence as you speak and let you correct as it goes without waiting for the back-and-forth to web servers.

Open source projects like CMU’s PocketSphynx seem to provide sophisticated client side mobile speech rec. My understanding is that modern mobile devices don’t have the resources (processing/memory) to allow client-side speech rec to get nearly the accuracy levels as you can on the server side. At least not yet.


To understand Amazon as a business/stock, it is important to understand it is at least two almost completely different businesses.  

-Media, which accounts for ~45% of revenue and is growing ~10% y/y, and is going through a massive transition to digital goods where the main competition is going to be Apple, Google, Netflix, cable companies, etc.
– Everything else (known as “EGM” – “Electronics and General Merchandise”) which accounts for ~52% of revenue, is growing ~24% y/y, and where Amazon is mainly competing with specialty e-commerce sites and (primarily) offline retailers like Walmart.
The stat I find surprising is how much room there is to grow.  Amazon’s share of (global) e-commerce is ~8.3%.  E-commerce is only ~6% of global retail.  Hence Amazon’s share of total global retail is ~0.5%.


Notes from a Skillshare class on “Defensive Finance”

Tonight I taught a Skillshare class called “Defensive Finance”.  My colleague Eric Stromberg was kind enough to take notes.  These notes are a bit dry because I didn’t want to publish the real-world anecdotes I used to illustrate them. But perhaps they will be somewhat useful.

Defensive Finance

Vesting vs. Exercising 

  • Most companies have 4-year vesting
  • Vesting – occurs while working for a company
  • Exercising – If you leave and have vested a portion of equity, you have the right to exercise.
    • There is generally a 90 day exercise period before losing options. 
    • Standard advice law firms give is to not inform employees. So have to pay the strike price to own the common stock.
  • Founders get actual common stock. Use lawyers from one of big 4 startup firms (Gunderson, Fenwick, Cooley, Wilson Sonsini)
    • Founders buy shares at a nominal price, like $100
      • So, they never have to spend money on strike.
      • Get long term capital gains treatment
  • Single biggest mistakes is to omit founder vesting.
    • Many cases in venture capital where investors would look at cap table, and a big owner of the company would be someone who did not work there anymore, because he/she was a founder that left early. But founders vested immediately, so the long-gone founder owns a large chunk of the company
    • The argument for immediate vesting is that some worry is that if investors push you out its good to have stock vested, but this is rare.

Acceleration on Change of Control

  • Single Trigger- Vesting acceleration on acquisition
  • Double Trigger- Refers to getting acquired and fired. Have to have full acceleration in case of double trigger.
  • These are both things that founders should try work into the term sheet. Also good to get them for employees (especially double trigger).

Founders Contributing Capital

  • If some of the founders are contributing capital, separate labor component from their monetary component. If under ~$200,000, make it a convertible note. 
  • Have the note convert at next round of financing, or some discount to it. 
  • Cash equity should vest immediately at next financing, whereas sweat equity should not should not.
    • The document usually says something to the effect of “I give $50,000 at 5% interest rate to be converted to equity at the next financing.”

Liquidity Preferences

  • Need to know what % of company you own – total number of shares is a meaningless number.
  • Founder, Employee and Investor Shares
    • Founders have common shares
    • Employees have options on common shares
    • Investors have preferred shares
  • Let’s say a company raised $500 million on a $3 billion post valuation and investors were given 1X straight preferred. This means in the event of a sale, the investor gets the max of $500 million or the percentage ownership they bought 
    • So if  the company then goes down in value and sells for $1 billion, if you are the investor, which are you going to take? The $500 million.
    • If you are a founder and you thought you had 10%, instead of getting 100 million get 50 million because you get 10% of the remaining $500 million.
  • In last down market, companies took higher valuation in return for 2-4x preferences. This can be troublesome, especially for employees who think they have a certain percentage, but may not get anything after the liquidation preferences play out.
  • 1x is standard in this market, particularly in seed deals

Misaligned incentives

  • If have investors with different class of stock, can have misaligned incentives
    • One thing popular in entrepreneur friendly environment is uncapped convertible note.
    • Problem is that if I am an investor, my incentive is to get you a low valuation in the next round
    • You have investors who are (in theory, economically) trying to make the company worth less
  • A company had two acquisition offers. One was for $20 million in cash. The other $20 at hot private company at hot private company (for example, Twitter or FB)
    • Investors want the hot private company stock, because higher variance bet
    • Said to the founder: “if you sold the company for cash, would you take every dollar you made and invest it in the hot company?”
    • In some cases, may see founder get cash and investors stock

Board Issues

  • There is part of the term sheet is clause called “Board of Directors”
  • Control and ownership are generally negotiated separately. Control has to do with board seats.
  • Seed Structure
    • Most common structure is either just founders or something like 2 founders and 1 investors
    • Means that the founders control the board
  • Post-Series A Structure
    • Investors want more control
    • 1 founder, 1 investor , 1 mutually agreed upon independent
    • Or 2 founders, 2 investors, and 1 mutually agreed upon independent

Protective Provisions

  • These are additional rights that investors have that can override a board vote
  • Sometimes, in addition to board control rights, the investor has additional blocking rights, for example a the right to block a sale or financing
    • Some are reasonable, like founders can’t distribute more than $100,000
    • Others may include blocking rights on sale below, say, 3x of the valuation at which the investment was made
  • Bad investors will use things like blocking rights to not just block the sale, but to get more of the company
    • For example, saying the need 30% of a $100 million sale instead of 20%, because that’s the number they have to hit to have the deal make sense. 
  • Series B investor preferences are sometimes senior to Series A preferences
  • So, If a Series B investor puts in $5 million, and A puts in $2 million and the company sells for 5 million, the series B investor gets 5 million while the series A investor gets nothing.
    • It’s very common for series A investors to have blocking rights to a senior security. So that, for example, a series B investor can’t put a 1000X preference on top of the series A investor.
    • Try to avoid giving investors blocking rights on non-senior (parri passu) investments.

Additional Notes

  • Shares to advisors generally vest over, say, 2 or so years, and it is nice to give vesting at change of control
  • Every venture backed company is a C corp. LLC’s are very difficult for venture investors because of tax write-off
  • If there is a 50/50 split between you and your cofounder, you probably haven’t thought about equity split enough
  • 83B election is something you want to get immediately when you for a company
    • If don’t get it don’t get long term capital gains, and the clock starts ticking 12 months from when you file 83b
  • If someone else was helping you out during the early stages of a company, make sure they are in or they are out. And if out, make sure they sign a release.
  • Investors want to see founders that have enough remaining vesting that it can get you to the next milestone. At least a couple of years. 
    • Generally, start vesting founder equity at a reasonable milestone like when you left your job, not when you first came up with the idea.
  • Typically A round post option pool for employees is 10-20% (more often 10% lately)
    • Size of option pool has significant impact on founder ownership, and is something founders often under weight when they raise money, instead simply focusing on the dilution they will see from the investor.

bitcasa and convergent encryption

People have speculated how the new startup bitcasa can both encrypt files client side and dedupe files server side.  The CEO says they use a method called “convergent encryption.”  This encryption method uses the file being encrypted to generate the encryption key. That way only people with the unencrypted file can generate the key but the same files unencrypted will be identical encrypted (and therefore de-dupable).  It is believed by the security community to work as advertised, with two possible vulnerabilities:

1) “confirmation-of-a-file attack” – someone who gains access to your files can confirm whether you have a certain file.  For example, someone could verify you have a certain movie/music file or leaked document.

2) “learn-partial-information attack” – in certain cases (from what I’ve read those cases haven’t been strictly defined) an attacker could learn some information from a file if the attacker already knew other information in the file. Examples might be a government form where a lot of the text is known but some sensitive text (e.g. your social security number) isn’t.

I’m a fan of client-side encryption, and even with these “vulnerabilities” it seems to me what bitcasa is doing is a good idea and should be adopted at least as an option by other storage companies.

One big limitation that comes with encryption is the inability to do operations like searching text on the server side. This can potentially be addressed through a method called “homomorphic encryption.”