earl

You can go far in life by following a very successful boss. And there may well be more promotion opportunities / more responsibility available at yahoo under Marissa than staying at Google. Or perhaps they've already made a million or so and own a home; why not take a flyer and see where it goes? I think you massively underestimate personal loyalty to someone who it sounds like went way out of her way to look out for these folks' careers. Because if someone did that for me I'd be loyal.

He could pay a pound extra to see if his algorithm is correct...

I bet your prediction is wrong. I've expanded on it elsewhere, but fb is sitting on a couple enormous businesses.

1 - they can launch the first serious adsense ($10B/year pre tac) competitor because they see most people move about the internet on fb connect;

2 - the ad exchange (fbx) is brilliant: you can bring your 3rd party data in and run ads on fb properties. This allows everyone to reuse their retargeting / behavioral targeting data and target in a guaranteed brand safe environment. But no data leaves: fb is a data sink.

3 - adding to the above, they have real demographics for brand advertisers, so they can diversify from the dr people.

4 - adding to #2, they're even better positioned for 2nd screen. Tons of people want to connect desktop/laptop display with mobile, but fb doesn't have to do any guessing or prediction: fb knows all the data because of the shared login. Thus they'll monetize mobile better not just because they know dems, but because they can extend online 3rd party data to mobile via fbx. If nothing else, buckets of dirt cheap impressions for your targeting/bt lists.

edit: 5 - how could I forget gambling? Zynga moves into real money gambling / poker (some relevant court cases have apparently been recently decided) and fb continues to collect a 20-30% tax.

.02 from someone who does data analysis / ML and has an undergrad in math:

A math undergrad is a nearly total waste of time. If you want to do data analysis, learn to do data analysis. Real analysis and proofs and abstract algebra and diffeq are fascinating yet utter wastes of time.

If you want to learn useful stuff, take: applied calc, applied linear algebra, applied probability, applied stats, applied ML. And nothing else. All the skills you build proving stuff are, at best, tangential. And you don't need a bit of it to do things like derive gibbs samplers or EM samplers.

If you can take an activity such as someone interacting with ads, model that as a graphical model, derive a gibbs sampler and implement it, you'll be worlds ahead of even most grad students. Focus on getting there and it should be great for your career. But a math degree is just a shiny distraction.

Oh, and one more thing: if you want to work in a particular area, then a degree in that area may be useful. Eg bio, genetics, chemistry, fluids, optics, etc.

edit2: Please don't think I'm arguing against you getting a math degree. If you want one, get one. Just understand it's probably at least 80% intellectual masturbation instead of job skills. For example, as far as I can see in production machine learning, there is virtually no integration. Oh there are integrals all over the place but none of them are feasible and they'll all have to be numerically evaluated or worked around. So all those integration skills people spend endless hours learning are basically useless because the real world has virtually zero nice integrals. So understanding how to manipulate integrals and how to numerically evaluate them is very valuable, but not well covered in virtually all calc classes.

edit3: consider looking at the community colleges in your state. They often are stupid cheap and offer online courses.

Plus externally monetized by the hardware you have to buy.

It's worse than that; it's not just no recurring revenue. Sparrow is stupid cheap. Comparable software such as eudora cost $50 - $100 in the late 90s. Compare to $7 for the mac app or $2.10 for the ios app.

And they're pinched between a platform vendor who wants a good email client on the platform; Google who gives gmail away for free because it's strategically important / they monetize it with ads; and people who are nearly unwilling to spend developer-sustaining amounts of money barring a giant hit, particularly for software requiring ongoing development.

The app store for ios has exerted massive negative price pressure on apps, including creating a norm of free or $1 apps. I know it's basic psychology, but it still amazes me that tons of people using a $200 phone that costs on the order of $80/mo to use bitch endlessly in reviews and negatively rate apps for charging the ginormous price of $3. Or something equally out of line. This is exacerbated by shitty discovery in the app store. There's massive returns to being on the top app in category X list, and that's very difficult to do at higher prices. Plus you can't create paid upgrades without creating a new product in the app store, forcing you to start from zero with marketing and list rank. I predict the mac app store will likewise exert downward price pressure on app prices.

Sparrow is really really cheap. A generation ago I used eudora, which cost $50 to $100, for email. So between the mid to late 90s and 2012, the price fell something like 80 to 90 percent to $10 for the mac program. The above is worsened by the rise of decent free email clients since it diminishes the audience.

If you do the math, you have to sell a fuckton of $2.10 cent apps to afford salaries for good developers: say four people at $200k/year fully loaded (salary, health insurance, ss, medicare, office space, internet, diet coke, coffee, macbooks). That's 380k sales just to tread water. And we didn't deduct marketing expenses which are probably significant: you'll have someone fulltime doing seo / sem / banging on blogs and reviewers to get organic / etc. Plus, again, Apple has created a norm where apps get updated forever for free. So that $2.10 is close to being a lifetime value for a user. That means you probably can't use sem for user acquisition: a wildly optimistic 10 percent clickthrough and purchase rate (an order of magnitude high for most things) and 20 cents per click (also wildly cheap) puts you at $2.00.

I presume Apple does this because of classic economics: increase demand by decreasing the price of complements. But if developers can't get paid, users can't have nice things. And I'm not an ios or mac dev, but looking at the interactions in their apps, they seem super custom. That means you have to go to the bottom of frameworks, increasing development costs.

Worse yet, say someone looked at sparrow dying and wanted to clone it, because it looks like there's a business there. Now they'll be pinched by people being cheap as shit on the upside and google migrating some of the best features into their presumably free iphone client on the low side. So I think google has now poisoned the well for better gmail clients on mac/ios.

Dude, whatever you think of Ballmer, he started as the first business manager at msft in 1980 and worked there in roles of increasing responsibility until he was made ceo in 2000. There's not many people with that track record of experience.

That must be part of the reason apple built iwork. And there would probably be antitrust issues if ms used their near monopoly in office productivity software to aid their declining near monopoly is OS software.

on-demand doesn't matter.

If lots of your infrastructure is in ec2, you may need a good db server inside ec2 that is used constantly. eg if you have a read heavy cassandra workload, you need ssds.