Jer Thorp at Eyeo

Jer gave a great talk at Eyeo today.

Jer says: I am notoriously bad at looking into the future.  I could have been employee #5 at Flickr (Ludicorp) but missed it.  I thought the web was a bad idea at first too.  So I’m going to talk about history instead of talking about the future.

This time last year I had just moved out of Vancouver with no fixed address.  I ended up taking a 6-week residency at Columbia and now I’ve been in NYC for a year.  I stayed because I was asked to work at the New York Times and got to invent my own title, Data Artist in Residence.  The choice to work there had everything to do with people and Mark Hansen in particular.  What matters is not the company but who the people are.  Working with Mark has been a blessing and we’ve had so much fun on this project.  I’m going to show you Cascade.  How stories published by the Times are shared through social spaces, in this case Twitter.

A view of Cascade

But any system in which things are being shared, this tool could apply.  Not just online, not just Twitter.  Maybe epidemiologists could study which diseases are being shared.  We look at separate events from the past and link them together in a cascade.  This is an early prototype.

What were the events that brought you to Eyeo today?  Like your dad brought home a Mac in the eighties.  We’re doing the same thing but over a shorter history.  Someone takes a url and shorten it to a bitl.y url.  Then it gets shared again, then again.  There’s lots of Times content being shared.  We wanted to build an exploratory tool.

Let’s start with Cascade view.  The side view.  The original sharing event on the left, all of the subsequent sharing events on the right.  Degrees of separation from the original sharing event.  The longer you wanted to Retweet you’ll be farther to the right.  This allows us to see the sharing activity of time very well, but there may be threads that we can’t see very well.  If two people share something at the same time, it’s smushed together.

So we have Radar view.  See the points around a central story to see how many threads of conversations there are about it.

In story view we take the cascades and cluster them around a story.  Around it like spokes are all of the major cascades about that story.  One of the big things about this project is to examine how a story is moving through space, if it’s something that’s likely to get shared a lot.

This is real-time data, it’s coming in live from the Times right now.  So here we have a very popular story, My Life as an Undocumented Immigrant.  I can go through all of the big stories of the last little while, go through each cascade and see how many events there were, how many tweets.  Let me pick one.  Google introduces a Facebook Competitor Emphasizing Privacy.  It’s too big to load so I’ll pick a smaller one actually.  Alyssa Milano.

When we have a caching system in place this will go a lot faster.  We get to see this conversation unfold.  Socially interesting stories have long lifespans, tech stories have shorter lifespans.

For each item in the cascade we can see who tweeted this, who their followers are, etc.  So we can see for each person, what story they tend to be influenced by or tweet.

One of the hugest cascades ever is the flight attendant who had a couple of beers and slid down the escape slide.

This is an exploratory tool.  We do these conceptual things and explode out of the conceptual space.  Don’t start in “reasonable”.  Start in “unreasonable” and then come back to reasonable.  There are some things we’re going to have to change as we move forward.

This runs on a five-screen video wall at the Times, we see a lot of time along the X-axis.  If you have a chance to come by the R&D department at the Times I can give you a look.  I gave away all the secrets to those animated transitions at my workshop yesterday… you can find it at

Mark and I talked about serendipity a lot in the early stages of our work.  I worked two years ago with Alex Beim on an accessible playground in Vancouver.  We ended up building an evolutionary tool that allowed us to evolve different landscape designs for the playground.  We wanted interesting path structures that satisfied the safety requirements. I shared this project with Dan Schiffman – who is expecting a baby this weekend and couldn’t come to speak at Eyeo.  He is here with us! (shows picture)

Never ignore an email because it has a boring subject line: Potential Freelance Job.  But he sent this to me and when I read it and I met with Jake Barton it was clear I wanted to be involved with him.  A tremendously difficult and rewarding project about 9/11.  It involves the embodiment of history.  Everyone who lived in New York in 2001 has a visceral reaction to a clear blue sky.  I noticed when I moved there last year that whenever these clear blue days came out everyone who was in NYC at the time recalls where they were ten years ago.

The way the memorial works is the footprints of the towers have become these reflecting pools.  Outside the edge are the names of the people who were killed in that tower.  One of the architects set a parameter that there would be no visible order to the names, that they would just flow, but that the relationships would be reflected in the adjacencies of the names.  Family members, friends placed together.  To embody the lives of these people not only through their names but through their relationships.  I think this will be one of the most meaningful aspects of the memorial.  But this is an extraordinarily complex problem.

We wanted to find the optimal solution to how this would work.  In this image we see the two pools, left and right, and the linkages that exist.  There are some very difficult adjacencies, and an impossible one where the two people died in different towers.  But the architects rotated the pools so the two names would be as close as physically possible.

They worked on the one with the first responders, the firefighters and police, and doing that taught them they would need an algorithm to do the rest.

The solution we ended up doing was a divide and conquer.  We built clusters of names linked by the adjacencies and figured the smallest size of cluster that would satisfy all of them.  One was huge and took up a side and a half worth of names.

So we made “lego bricks” of clustered names that could go along the pool until it found a place it could fit.  The big ones start first and the smaller bricks come later and fill that space out.  In the beginning all I wanted to do was get the area filled and get as many adjacencies as possible.  In the end we got about 98% of the adjacencies.  There was a group of mathematicians that solved for this same problem simultaneously that only got 80%.

We treated it as pieces of typography, not a grid of numbers, which is why it worked better in our case. In the walls of the memorial there are half inch expansion joints.  You might think we could line up the names so they never crossed a joint.  But the architects wanted it to flow.  We needed to look at which names had spaces that could cross the gaps.  The typeface dictated that some names would and some wouldn’t, so those names had to be grouped together loosely.

This is a math problem, it’s a design problem, but it’s also a human problem.  Building a memorial is one of the most human things you can do.  We wanted to make sure that people were allowed to be part of the process as well.

(He shows a picture of colored clusters, and the colors show each group).

We created a drag and drop interface in Processing so that humans could swap names around, and names were highlighted to show if they could reach across that half inch expansion joint. So people could go in and tweak everything the way they wanted it to be.  My piece was only a small par of the human hours that went into solving this problem.

The last thing I want to bring up is  We’re leaving these data streams.  It could eventually end up being a representation of our lives. gives people an opportunity to explore and interact with their own data from iPhones.  Location data for example.  Right now you have access to all this data, but when they fix it you won’t.  They will have taken away your access to this.  We created openpaths as a way for you to securely store this data for yourself.

This data represents stories.  It’s your own personal narrative.  Oh, that’s the place where I met my girlfriend.  I see myself traveling from my house to her house.  I see when I landed in New York City when I moved there.  Upload your data for that… to experience a year of your life in this way.

Researchers want this data, but they don’t own it.  Through this you can share it with the groups you want to share your date with.  Broker a relationship with these researchers.

This talk began with histories embodied in Times data, and then in the memorial.  But now it can be your histories.  This app runs in the background and is secure.  If your’e interested in openpaths talk with Brian House, he is here at Eyeo.

I’d like to remind you that everything I showed you today was built in Processing.  Ben and Casey are very reserved and they don’t tell you how much it f*ing rules.

It’s amazing to be at Eyeo, we had a dream to build this event to be the best conference in creative coding ever.  I hope everyone uses this as their stepping point.  I want people to be interviewing all of us 10 years ago, and for us to say, “Well, I was at the Eyeo Festival ten years ago, and it changed my life”.