-->

D#MZ

Market&ing Algorithms. Machine Learning. Fuck.

MBA from the Street.

One of the cruel realities that I’ve experienced as a consultant is that if you’re not asked for advice — advice will be frowned on.

Focus on the box. Looking outside of the box, by definition, is not focusing on the task at hand. This is not a balancing issue, it’s just a reality that companies can only really have one pundit. 

Capitalism (un)fortunately is the land of dictators. Travel with knowledge, and you’ll do OK.

A Growth Hackers way of choosing the right Wheels for his Car.

I roll with an Audi A4. It’s a neat little car that fits into parking spaces, and easy to drive… Especially in the city.

Usually people find wheels by google images. But I prefer finding a forum of people that have the same car as me, and similar passion of blowing money on depreciating assets. 

Since I’m @ Starbucks, and their internet connection might as well be dial-up. Surfing the forum, page by page, is a clear waste of time. Thus I programmed a quick little script that’ll download all the images of my car with aftermarket wheels on it <3

https://gist.github.com/D3MZ/5651537

With coverflow, it makes wheel shopping a breeze!

atravisscorry:

Lean Startup Machine #LSMTO @LeanToronto @INcubes (at [IN]Cubes)

atravisscorry:

Lean Startup Machine #LSMTO @LeanToronto @INcubes (at [IN]Cubes)

Ahh peak performance&#8230; A thing of beauty! Growth Hackers dream&#8230;. No problems.. Just smooth sailing&#8230;

Ahh peak performance… A thing of beauty! Growth Hackers dream…. No problems.. Just smooth sailing…

Laziness is expensive

The internet is attached together with shoelaces. It’s almost surprising that it works, but it does. Things fail, but always try to have them fail only once.

I can’t stress how important it is to track when things die. How would you know how to prevent problems if you don’t care about fixing them? This is how expensive a failure is for me (a 3-fold drop in performance):

If this was client facing, some jackass would be clicking the refresh button like 5 times now. 

Anonymous asked: Hi

hi hi

Bruteforce, Cloud stylez.

Way back when, I’ve wrote about scraping cities in N.America using Mechanize: http://demetriusmichael.com/post/45767080423/scraping-cities

60k cities is small time, and yes Bruteforce will always be satisfying. But it’s way easier to just do research. If you spend 10 seconds googling, you’ll see that some asshole already aggregated a list of coordinates by city across the planet (http://www.maxmind.com/en/worldcities), and we didn’t need to go through the process of destroying some other poor guys web server!

Like all lists on the internet, they need a little house cleaning. This one uses non industry standard encodings, so we have translate strings to UTF-8, or in my case, not give two shits and delete whatever doesn’t look remotely like English.

The other pain-point is that he uses country codes, rather than country names, so I have to normalize the data. This isn’t a big deal, but country names are more memorable and easier to query to meat sacks.

Here’s the code:

https://gist.github.com/D3MZ/5390631

That above script uses ~3GB of memory, and inserts maybe 10-50 cities / second on my laptop. Since the Earth has millions of cities and I’m impatient, I’m going to brute force it anyway —— On the cloud.

Now the fun.. 

  1. First, we spin up the most retarded server on the planet: image
  2. Run the “parallel” hook on the script.
  3. export/zip mongo.
  4. import wherever you want it.
  5. You’re done.

https://gist.github.com/D3MZ/5390801

*Zipping like a boss.

Symlink your zips to their parallel equivalents! Now you can kill the planet faster with your multicore energy consumption.

https://gist.github.com/D3MZ/5368270

Not happy&#8230; It&#8217;s dipping faster than I wanted&#8230; Stops at Break-even. This is Trading like a pro ladies.

Not happy… It’s dipping faster than I wanted… Stops at Break-even. This is Trading like a pro ladies.

It&#8217;s been a year since I&#8217;ve traded&#8230; bit rusty, but expecting the classic ABC, or  1-2-3. Depending if you&#8217;re French.

It’s been a year since I’ve traded… bit rusty, but expecting the classic ABC, or  1-2-3. Depending if you’re French.

Sane people will long here, but fuck it. we&#8217;re shorting&#8230; Red means i&#8217;m losing money :P

Sane people will long here, but fuck it. we’re shorting… Red means i’m losing money :P

Cloud the right way.

You remove a couple pieces from Jenga, and the building collapses. Kill a few cells, and nothing happens.

The problem is in the process. In school they want you to break down problems in a series of steps: Step A => B => C. So you program that way.

But what if you blow up Step B? What happens to your pathetic little process then? What if you do this:

A => Universe

B => Universe

Universe => C

Blow up B? No problem. C still lives. Want More C… Create more A, B, and/or C. 

Algorithms growing faster than I can scale&#8230; Red is the speed I&#8217;m paying for. Blue is the speed I need.

Algorithms growing faster than I can scale… Red is the speed I’m paying for. Blue is the speed I need.

This is probably the best photo I&#8217;ve taken all year.

This is probably the best photo I’ve taken all year.

*YAWN* Amateur Hour.

God is within definition.

Ask the right questions and get actionable answers. Add automation. A|B testing is for children, no one has time for that.

Dump Multivariate tests into machine learning algorithms, and keep laughing. 

Amateurs. 

Maybe start with a library based off of: http://www.csie.ntu.edu.tw/~cjlin/papers/liblinear.pdf