- most of the work to be done is the same boring stuff you do at a startup: maintain the server farm, deal with flaky libraries, tune the database. The interesting algorithm work is done by PhDs... at universities.

- the bullshit factor at bioinformatics companies is as high or higher than the trivial startups you mention. You are ultimately working on tools that people don't really know how to use and trying to sell them to people that probably don't need them. The only "real" customers are pharma companies or larger biotech firms. This makes the environment much like working for an enterprise software vendor which everyone knows isn't very fun.

- the bioinformatics companies are either in glamorous places like New Jersey, Boston suburbs or... Silicon Valley. So, you're unlikely to be able to escape from the SV scene anyway.

- at a startup these days, you at least can use slightly less painful tools like ruby and python. In bioinformatics, you are stuck with ancient Perl/CGI scripts or some VB/Java time bomb.

I spent 2 years in bioinformatics and will be forever grateful that I was given the opportunity to escape and work on silly, pointless consumer entertainment software instead.

So, you're either working for the 'enterprise software vendor' (ie big pharma) or you obey the whim of the PI, which is like client work for people who, again, don't really want to listen to you anyway. 'Just get it to work' is the prevailing attitude.

The only notable exception would be if you were to get into bioinformatics outside of the USA. Try to find some work with either EU orgs (entirely govt sponsored research), such as the World Health Organization in Geneva, or take a look at what orgs are doing with public health research in developing countries. If you find a better ways to assist with AIDS research in Africa, you would be doing the world a bigger favor than currying favor to the pointy-headed in the US.

In short, look to the frontiers of modern society. They probably have needs that seem elementary to us, but are real and need solving, now.

Out of curiosity, why is that?

In particular, he talks about keeping track of a set of important problems worth solving. Then, whenever you find an interesting new technique, compare it against your library of problems, and see if there's a good fit. I expect the reverse is also similarly useful: keep a good library of techniques on hand, and look for one that's applicable whenever you run across an interesting problem.

This type of approach isn't necessary for building twitter or facebook clones, but it may well be ideal for looking for new ways to apply machine learning and other types of advanced algorithms, because it's primarily applicable in cases where there's an actual problem to solve.

http://www.ultraedit.com/company/IDM_full_story.html

http://www.justin.tv/hackertv/98115/David_Heinemeier_Hansson...

Best description yet of "Web 2.0"!

I feel much the same way, which is why I've been learning Eclipse RCP this year. Not everything has to be a killer website (although RCP apps can be made into web apps with RAP).

There are still many applications that don't quite fit into the "web 2.0" mold. For example, think of applications that deal with data users want to keep private, or share only on their own terms, not through a trusted third party, however benevolent they claim to be.

Instead of trying to convince users they need to trust their data to your web app, bring the application to them!

Build that solution. Sell that solution. You'll have done two things: solved your problem, and solved someone else's. If you make a few bucks along the way, so be it!

Google may use cool math, but be careful calling it the "only" at anything. There are tons of companies you probably haven't heard of doing awesome things. Everyone knows about Google, though, so it's easy to point to it as an "only".

I hope that made sense.

Having said all that, use your skill as a tool! Let's take Octopart as an example - the founders' pain was they couldn't find their parts easily online. So, they put their skills (algorithms [well, technically, they were physicists, but any good physicist is a good mathematician]) to work - they built a great search that Google doesn't cater to.

Lessons: 1) They identified pain - no search for parts 2) They used their skillset - math (algorithms) 3) They are making the world a better place* (as in much more of an impact* than those playthings you mentioned). *-very very relative

As about bio-informatics, who knows? Heck, no one really knows about any industry's future. If someone claims to know, he/she is trying to sell something - may it be an idea, prophecy, or plain bull-shiitake mushrooms :) Point is that if you have really identified your customers' pain, then they will happily send their wheel barrows full of love AND money towards your solution.

Not that there's anything wrong with bioinformatics.

http://www.palantirtech.com/

My profile has contact info if you'd like an intro.

Counterexamples welcome.

Why? Tubby time for the kids! Of course, I could set up something myself, but I'm too lazy.

Future iterations could include other similar devices (e.g., one with space sounds), or a "fourth button" that allowed custom sounds to be downloaded.

Second - the advertising dollars in the valley are not zero-sum, because the potential market, while finite, is not nearly what it could be. How to do effective brand advertising or video advertising on the web are very much unsolved problems, and whoever solves them will substantially increase the size of the pie that everyone is fighting over. These aren't technical problems, but they could have technical solutions (Ooyala in the video advertising space is a good example).

Also many people seem to make shitloads of money with "ad opmtimization software", tools for people to help them optimize their spending on ads. OK, that is ads again, but I think an area where you could run wild with algorithms.

What about image analysis? Google doesn't translate images yet...

Doing really cool stuff is riskier. Scarier. There are no rules laid down yet for what sort of value you're generating.

That said, there are hard algorithms more places than you expect. When trying to scale something that looks simple on the surface, problems often become complex enough to be interesting.

The first step is the only hard one.