Friday, April 18, 2008

My Link Blog needs some love

Link blogging is all the rage. Or at least that's what I tell myself because I'm working on a link blog aggregator idea. If you're not familiar with what a link blog is, its basically a collection of articles or article headlines from many sources. A link blog may be a manually selection of interesting articles or be automatically aggregated out of a collection of sources. Google Reader shares are a popular way of creating manually collated collections, and Google Reader public tags will create automatic feeds of all the blogs under a particular tag/topic.

So of course I have a few link blogs of my own. If you're subscribed to this blog in a reader you're missing these link blogs, which I have to admit are usually more interesting than what I write here. These link blogs cover various topics - I have link blogs on investing, software development, interesting articles by other bloggers, web infrastructure and so on. These are visible in the sidebar of this blog, or you can try them out by clicking on the links below and importing the feeds into your reader. You can also create your own if you use Google Reader.

My "interesting reads" link blog

Software Development - mainly Rails/Ruby related


Web infrastructure


Happy Reading!

Tuesday, April 8, 2008

Losses this Recession

Roubini, the IMF, Goldman Sachs, and a number of others are coming up with total credit losses due to the housing crisis similar my estimate in a previous post. Some of these estimates are rolling up losses on credit cards, auto loans, and commercial real estate along with residential housing losses. Others like Roubini are estimating a $1T loss just due to residential housing. Either way this level of losses will wipe out a good portion of Wall Street financial firm's market cap not to mention pension funds and insurance organizations.

A massive government bailout seems inevitable, probably at even larger scale than occurred with the RTC during the early 90's S&L crisis. This bailout is already under way, particularly in terms of inflating our way out of the problem. I do not anticipate a 15 year depression/recession such as occured in the US during the 1930's or Japan over the last 15 years, primarily because the Fed in the US has shown a strong bias towards an activist, inflationary strategy.

I do anticipate sharply higher commodity prices, a 20-40% decline in the dollar, multiple financial firm bankrupcies/rescues and house prices bottoming only after 25-50% declines.

With every decline comes a great shorting opportunity followed by a great buying opportunity. I just picked up "Anatomy of the Bear" a study of the last 4 bear market bottoms. This book looks at the actual historical record to refute popular mythologies on what characterizes bear market bottoms. Apparently both "buy and hold is your best strategy" and "Bottoms only occur when everyone thinks things will only get worse" are both false.

I'll post the Cliff Notes soon as I'm done.

Friday, April 4, 2008

Ingredients of Startup Success

Paul Buchheit of Gmail and FriendFeed fame recently posted the best synopsis of the ingredients of startup success I've seen.

There's one thing I'd add, using the same mountain climbing analogy. Often you put a team together, and it climbs partway up a mountain expecting gold at the top. About halfway up a number of things can go wrong.
  • The route may be impassable given the climbing techniques the team is capable of (the technology doesn't give the hoped-for benefit)
  • You may get a good view of the top and see that there is no gold, just a bunch of shrubs. (the market dynamics make it hard to make a profit)
  • The view halfway up the mountain lets you see the top of an adjacent mountain that does have gold at the top (Focusing on a different segment of the market is more profitable)
  • There's a weak link in the team that makes climbing slow and difficult (missing skill sets)
In my experience these types of problems are not exceptions, rather the norm. Successful teams recognize the issues quickly and make the changes necessary, whether that be changing market focus, hiring different skillsets, or changing the product to better meet the market's needs.

Saturday, March 29, 2008

I've officially moved Mark Andreesen's blog to the "humor" section of Google Reader.

Here's Comedy Central's wrap up of Cramer and Bush's comments on the recent troubles on Wall Street. At the end Bush pronounces "The economy is fine..." The economy must be totally fucked.

Friday, March 21, 2008

The sources of human misjudgement

Here's a transcript of a talk Charlie Munger gave a while back on the sources of human misjudgement. Munger is Warren Buffet's partner in Berkshire Hathaway. This is a good synopsis of a bunch of sources of material from Darwin to Skinner to Cialdini. Interesting for investing, marketing, and everyday life.

In the talk he refers to a set of mental tricks used by Darwin to reduce confirmation bias - I'm going to see if I can find the biography he refers to. I've been collating my own list of checkpoints before making investing decisions. The list constantly gets longer - so many errors to make, so little time.

Saturday, March 15, 2008

Bank mortgage losses - just beginning or almost over?

Interesting data in this link on recent trends in home ownership from the Census Bureau:

Back in 2006 I found it interesting that the US home ownership rate had risen to about 5% over the 100 year average (69% vs 64%). Knowing that after bubbles burst, things tend to revert to the mean, I thought I might be able to estimate what the banks will lose post-bubble. A back of the envelope calculation yields about $1.5 Trillion in lender losses if the ownership rate returns to the 64% historical average. This is based on 20% loss on foreclosure assumption and the median home price of $235K - see the Census Bureau report in the above link for the number of housing units involved. The loss rate could be pretty optimistic - most banks are assuming their subprime CDO losses are between 50-100%, but this is due to trying to mark them to market in an illiquid market, not any real assessment of what home prices will do. The contrary opinion is here:

Given how wrong S&P has been about just about everything, I'm taking their opinion as evidence that someone paid them to paper over a really big hole. If the $1.5T is correct, it means only about 20% of the $1.5T has been written down so far, which means the failure of Bear Sterns is just one of many to come.

There is evidence that one cause of the higher home ownership rates is a permanent demographic shift towards people marrying later in life and thus there being a higher number of single-occupant homes. This would increase demand for housing absent that trend. However we also know that there were several other now reversed causes such as low interest rates, a blind rush to profit on rising home prices, and loose lending standards. In previous bubbles, ownership rates and prices of the asset in question reset to historical averages only after undershooting their historical averages, which could mean the 1.5T is too low.

Once the 10% of US homeowners that have zero or negative equity believe that the price of their largest asset is headed south for the long haul many will just bail. No lowering of interest rates will matter because they just won't want the albatross. The only thing that might have an impact is if lenders forgive the principle amount impacted by the price decline. Which puts us back into at least a trillion in losses for the banks. A loss like that would wipe out the market cap of a good fraction of Wall Street. Think tech stocks 2001-2003. Or Japan over the last 15 years.

The Fed's recent actions suggest that they will do just about anything to try to prevent it. We'll have to leave the efficacy and side effects of a mass bailout to another post.

What do you think? What do you trade?

[disclaimer: I am wicked short C, FNM, NTAP, QQQQ, DSL, BSCI but have bailed out of my builder shorts as of three months ago. ]

Thursday, March 13, 2008

Positioning your product

This sign cracks me up every time I pass it on the way home. Reminds me of a few startups. First they try breakfast. But the cook's not that good. Few customers come back a second time. So they add expresso. But the expresso isn't that good the time they're trying to add balloons to the list the investors bail.

Do one thing. Do it well.

"One thing" usually means providing one solution to a problem for a group of people. Doing it well may mean having a critical mass of features. This is often a challenge in a first rev product. I try to find new product ideas and market segments where the team can solve one problem and do it well enough to make a real difference without taking on an un-doable number of features. The trick is finding a group of users or an application where we can make a difference (= beat the competition) with a subset of the ideal feature list.

In mature markets this is difficult. New markets that have shorter requirement lists are ideal. Sometimes old markets where a new "disruptive" technology can be applied will work too.

Someday I hope to meet Sweet Amy and ask her how it all went down.

Wednesday, March 12, 2008

A ranked feed for the Rails community

I've been working on a solution for two problems in my daily feed reading. First, there are just too many quality things to read. I need them organized by topic and ranked inside each topic . Second, Google Reader doesn't give me visibility to the comments in each blog without clicking through to the posting. What I want is one place where all my feeds, say on RubyonRails, are ranked, visible, and include all the comments. Ideally I can comment on the postings without exiting the feed reader interface. I've hacked together a partial solution you can try out at

This is a few dozen Ruby on Rails related feeds from my Google Reader collection fed into aideRSS. AideRSS ranks based on the number of delicious saves, Google Reader Shares, diggs, and comments for each post. Next that AideRSS output is piped into Friendfeed which provides one place where people can comment in a way that all comments are visible in one spot. The friendfeed feed can then be pulled into G.R., or just used as -is.

There are a number of problems with this hack. You'll still want to use your favorite feed reader. In the feed reader you're not able to contribute to the comments without exiting the reader, nor can you actually read the posting without clicking through to it. However if commenters use Friendfeed, Google Reader will embed comments in the feed. Today of course most of the comments will still be on each blog rather than centralized on this feed. I've seen some feed readers that pull in comments, and then there's Cocomment, but these various pieces need to work together.

If only Blogger/Wordpress/Typepad, Google Reader, AideRSS and Friendfeed would get busy and have a bastard child I'd have a solution.

I'm interested in feedback on this - does anyone else have this problem? What would be your ideal solution?

Thursday, February 21, 2008

Scaling a Facebook application - users and infrastructure

Here are some links and notes from recent presentations on how to scale a facebook application, both in terms of the technical infrastructure and the design of the application.

For the application design see this link to the great highscalability blog

In addition to the above, here are some notes I took from a talk Blake Commagere (developer of the popular Vampires Facebook application) gave two weeks ago at the Facebook developer meetup.

- Jealousy, Courtship, Selfishness, Sexuality motivate everything
- Help users meet someone they have chemistry with.
- Virality kicks in when you play to the users's selfishness. Tell the user what's in it for them
- Conversion of the invited is more important than the number of invitations. Try to convert the invited at the first invite, diminishing returns after that. So have to give the invited an appeal to selfishness as well.
- News feed is the most important viral mechanism - much more important than invites, as its viral beyond the invite list.

For the infrastructure discussion see this

The nginx + mongrel + memcached + rails stack seems to be the emerging best practice for rapid application deployment in the web community. I use almost the exact same stack for, except my whole configuration is on one Xen slice at Slicehost since I have less traffic. I've been pleasantly surprised at how well my 1GB Xen slice performs - its faster than the 2 core, 2GB dedicated machines I use in my test and dev environment. I'm assuming this is because the slice is on a 4 core machine with a larger RAID configuration behind it, and I'm grabbing more cores and spindles when heavy requests come in. Slicehost has been a great VPS provider, although I have only 4 months of history so far the price performance has been excellent. SearchArtGalleries is mostly disk bound when pulling complex searches out of Mysql. The nice thing about this stack is that each layer scales out easily, nginx being a good load balancer and static file server all in one. The exception is the database - wouldn't it be nice if scaling Mysql was as easy as adding mongrels?

Now back to trying to figure out how to get my app to 4 million users by the end of next week :-)

Thursday, January 31, 2008

Twitter Blocks on Amazon Web Services

This is really cool. I’m not sure why its cool, but it’s cool. Twitter is using Amazon’s web services to take a snapshot (I presume) of the twitter social graph and make it navigatable in 3D.

Sunday, January 20, 2008

Open source understands marketing

Maybe its just me, but lately I've noticed that many open source projects are out-marketing closed source companies. This is interesting because open source projects typically have no marketing teams, just a few engineers. Last week I was looking for a collaboration package to run on my startup's intranet. We want to share documents like PRDs and specifications, allow comments, and host blogs. I compared two options - microsoft sharepoint server and wordpress. Sharepoint server was attractive because we are running microsoft active directory so the authentiation would be integrated. Wordpress is well, free. There are lots of other alternatives, but these two illustrate the point. I wanted to get a quick feature comparison - typical thing you'd expect most potential customers might want to do. Like everyone else I start with Google and type in two queries - "wordpress features" and "sharepoint server features". I click on the first links that seem relevant on the first page here and here.

The wordpress page gives me one page of HTML that's easy to read in about two minutes and addresses the actual features of the product. The Microsoft link forces me to navigate a maze through several links before I'm presented with a link to a word document, which I then need to download and launch word. Then I need to weed through a 69 page long document to find the 1 page of features I want. The first 13 pages of the document try to convince me that I need to "Get started quickly" and "Collaborate Easily" - basically to convince me that I need a collaboration platform. As if I happen to be browsing the sharepoint server site for fun or something. And the requirement to have Word installed is a nice touch that says "We want to ensure anyone that is not already a customer never becomes one". In contrast the open source sites I read tend to provide fast and easy access to the information I want and broad platform support. When I'm evaluating a new piece of software or hardware I want to know four things:

1. What features does it have?
2. What's it going to take to get it running?
3. How much does it cost?
4. Who else is using it and what do they think of it?

I don't think IT departments evaluating products are much different. What proprietary hardware and software companies are getting wrong is they have forgotten who their customer is, they are not valuing the customer's time, and they don't understand where in the decision process the customer is when they arrive at the product website. Instead of giving me answers to the above questions, product literature tries to sell me the benefits of the general category of products. That's silly - especially for a product category like blogging software that everyone that hasn't been living under a rock already knows about.

What open source teams understand that many big proprietary software and hardware companies don't is the following:

1. Sell features, not benefits. Yes, this is the reverse of what you may have learned in business school, which probably explains why open source marketing is better - its written by engineers! Selling benefits is fine if you're selling a product or feature the buyer is unfamiliar with, or one they haven't decided they want. For most tech products the buyer is a technical person who is already convinced they want the category of product. They wouldn't be at your site if they didn't already want the type of product you're selling. Second, you're talking to another engineer, not a clueless consumer, so you don't need to explain the benefit of every feature - they already know. The only times to talk about benefits is when discussing the unique features of your product, which are going to be a very short list. Occasionally you may be selling a completely new category of products that no one is familiar with, and you have to explain why someone would want that category of products. Again, this is very very rare. One of the reasons why open source companies get this right is the engineer that wrote the software also writes the collateral. The moral is - the writer must be technical and understand what the buyer is interested in.

2. Keep it short. Each of the questions above should be addressed in 1 page of html, no more than two 15" screen lengths long. Expecting anyone to read a 69 page document to get a feature overview is just plain stupid. To keep things simple and short, ditch the document-by-committee method of creating documents that enterprises use - this leads to long meandering drivel. A document should have exactly one author. Editing is fine, but editors shouldn't try to word-smith the document.

3. Make trying out your product easy. People like to test things before they buy them. Most open source packages I've used from Linux to RubyonRails are dead simple to install, run on everything, and they're free to test out. Contrast that to most proprietary software which runs on 1 platform and is crazy hard to install due to lame downloader tools, poor compatibility, and broken DRM. The culprits are too numerous to name, but the guilty include Java and everything Microsoft.

4. Position yourself against other products. Potential buyers are interested in comparisons. Your literature must answer the question "how is this better than the competition?". This is also known as a positioning statement. There's no need to name the competition, but your feature list should highlight why your product is better. One of the best examples of this I've seen is the Right up front in bold letters it explains why a prospective user should try Rails. Lets compare that to Sun's java developer site. Lots of words, but no reason why I should use Java. So, I don't.

5. Integrate testimonials and community into everything. Prospective users should be able to quickly get to a list of other users, see real testimonials and see what users are saying on other sites.