SXSW: Start-up Hatchery and Battleground

SXSW is a hot word at the Media Lab. So when asked to report on a story we “can’t report on in person,” I took the opportunity to learn more about an event I didn’t know much about, but still wanted to be at.

Like others, I used Storify in this assignment: Storify is a fantastic resource and was surprisingly seamless. SXSW is all over social media also, so finding sources was not a problem.

However, my initial goal of covering all of SXSW was quickly blown to pieces — there is just too much going on. I narrowed down the problem to covering start-ups at SXSW, which still turned out to be a problem in curating content from an ocean of Tweets. Furthermore, I found that it was difficult to find multiple perspectives of the SXSW start-up scene on social media. I’ve heard in person that it encourages a narrow, pitch-able view onto start-ups, and encourages this view through press and awards. But I wasn’t able to find much online sharing this same idea.

Posted in All

100K People 1 Pokemon

As a kid I would hide under the covers as a child playing Pokemon Red on my Gameboy. Gaming was occasionally social but mostly solitary. Almost two decades later, I’m returning to Pokemon with 84,000 other people.

Twitch Plays Pokemon is a collaborative gaming “social experiment” on, a streaming video platform. Viewers gives commands to control a single character in order to capture Pokemon and acquire badges. Created on Feburary 13th, 2014, the peaked at over 100,000 concurrent viewers and has ten million total views.


The Twitch Plays Pokemon Feed. Note that the right-hand chat moves at an incomprehensible velocity, though some players have developed a way to filter commands to see strategic comments. (

The anonymous Australian creator shared in an interview the technical details behind the experiment. Pokemon Red (151 ROMhack version) is mounted on the VisualBoyAdvance emulator. An IRC bot lists to button commands on the stream’s chat, which are then input into the emulator and shown on the stream’s overlay.

This mixture of javascript and python code allows us to address the question: how does collaboration scale? The flurry of commands on the right translates into game input depending on the voting mode, which itself is determined by votes. In anarchy mode, all commands are input at a rate of about one per second. In democracy mode, the most popular command is input every few seconds.

I’ve given it a shot, but what started as a minute of nostalgia gave way to an hour of frustratingly walking around the same plaza. In sixty minutes, we moved about ten steps and transitioned from anarchy to democracy to anarchy. Time-lapse clips document other stretches of time making no progress.

So why are tens of thousands of people playing? The friend introducing me to Twitch Plays Pokemon stated his motivations: “I thought this was an simple but powerful concept. Definitely a good way to spend a couple of minutes.” When asked about his expectations for the outcome of the experiment, he replied “It will be painful, but I wouldn’t be surprised if they beat the game.”

Miraculously, the community has progressed through the game. A Google Document recording current progress shows that the community has achieved four of eight badges and trained a decent set of Pokemon. An article on gaming blog Kotaku illustrates some strategies that players have developed to progress and counteract trolls, players that are intentionally counter-productive.

These efforts are documented and cultivated largely by an large community on Reddit that acts both Greek chorus and Roman senate. An active Twitter feed also comments on the current state of the game. Commentary on amusing frustrations and achievements have turned even into memes.

Screen Shot 2014-02-18 at 9.38.56 PM




What started as a small dedicated gathering turned into a large community sharing strategies, frustrations, and experiences. The experiment may not answer questions outright, but forces users to consider the merit of consensus-based rule and the role of trolls. And whether the experiment will triumph or fizzle out, we’ll test mathematician Émile Borel’s remark: “With an infinite number of monkeys and an infinite number of typewriters, one will type Shakespeare’s plays.”

Kevin Hu’s Media Journal

This exercise in self-awareness of our digital lives was informative and alarming. I’d like to comment on why, how much, and how I consumed media this past week, and what can be learned from this reflection.

Following Erhardt’s lead, I used RescueTime to collection data on my computer usage patterns. This was based on the assumption that, since my days are marked by staring at a monitor, my computer can tell me about my media footprint. That’s not the case, as we’ll see. A summary histogram of the six days since Thursday 2/6 is below, with applications/tabs on the X axis and hours spent on the Y axis:


Localhost:3000 is a development URL, so time in that category + sublime text 2 + iTerm are together the time I spend writing code. The second and eighth categories sum to the time I spend playing the video game League of Legends, which is scary (bothersome question: where do the rest of the 168 – x hours in a week go???). A list of the twenty categories in which I spent over thirty minutes is below:


For completeness, here is a stacked bar chart with days on the X axis and colors signifying different categories. We can see that, while my computer time varies, the distribution of time is somewhat stable.


From this tool, I supposedly spent little time on e-mail. That’s totally not true. I e-mail mostly on my phone and iPad, which is lost in this analysis. Furthermore, I would like to know not how much time I spent on an activity, but how many times I checked it. For example, I know that the little time I spend on the computer checking e-mail is likely due to habitual inbox checking rather than composing messages.

Speaking of my phone, RescueTIme does not capture the hours of pre-sleep phone Redditing costing half an hour a day and probably a substantial chunk of well-being.

But do I really spend no time reading articles or books? Here’s the moral of the rant: most important pieces of media I’ve consumed — those that enriched my understanding or changed a perspective — were not on the computer.


Three were by print: the papers “Visualization and Cognition” by Bruno Latour and “Modeling games in the Newtonian World” by David Hestenes (both via Bret Victor’s 60 over 60 list), and chapters 3-12 of Hobbe’s Leviathan (pages 20-90 in my copy). My print reading was six hours spread over Saturday and Sunday.

In contrast, I listened to the audiobook version of The Better Angels of our Nature by Steven Pinker (paired with Hobbes and recommended by my advisor Cesar Hidalgo) during my walks to and from the Media Lab. That’s 30 minutes every day, and summed to around three hours. Similarly, I read On Intelligence by Jeff Hawkins on my kindle for about 15 minutes (chapters 5-9) before diving into compulsive pre-sleep Reddit-ing.

But what about news? Sadly I realized that most of my knowledge of current events is through word-of-mouth or Facebook statuses. Very little is through reading extended pieces offering thoughtful analyses, which is alarming.

If I were to perform this experiment again, I would do more comprehensive and rigorous accounting of media consumption on all platforms. I would like to understand what types of news I do read and should read, and also how much time I spend on one device versus another.