show and tell for hackers in DC.

Round 28: The Cloverfield Ultimatum

19 Jan 2016

Here are the Hackpad presentation notes from Round 28: The Cloverfield Ultimatum.

Aaron Schumacher

Sonic Histograms

Aaron is back in DC from a stint in NYC, and his first stop was at DC Hack and Tell!

Inspired by CSV Soundsystem, Aaron wanted to make songs and sounds out of data.

Using the tuneR R package, he was able to turn data points into individual tones and then play them in aggregate so you can hear the width of a flower petal, or perhaps the GDP.

(At Hack and Tell Round 22, Elaine Ayo made a similar project, where she used shark attack data to generate music and sounds.)

Neal Oden

Autumn Leaves

Neal Oden brought us a really cool art project that had the DC Hack and Tell crowd oohing and ahhing. Because industrial turntables that turn very slowly are super expensive, he built a cheap version using lego.

Combining his custom turntable rotates once every three hours with a camera that takes pictures every five seconds, he used Python to transform thousands of images of an autumn leaf turning.

Not only does this creative process create beautiful still images, but together it makes a stunning movie.

The video can be found here and the Python source code here.

Sarah-Jaine Szekeresh

Tune Into News

Sarah-Jaine Szekeresh made a twitter bot to help her multitask listening to the news and music at the same time.

Tweeting at the Twitter Bot @TuneIntoNews with the word “play” to check it out!

The bot checks NPR for top news headlines and pulls a key word from the title, then uses natural language processing and connects with Spotify to find related songs by the key word, then gives you a song and title that more or less fit together.

Written in Python, the bot connects to the NPR and Spotify APIs.

Rushi Shah


Markdown is great, but Rushi Shah loves LaTeX. What if you could write your blog in LaTeX and publish it to a PDF? PDF makes the most sense for this particular use case since the mathematical formulas and strict formatting of LaTeX are so important here.

Rushi wrote it in Haskell, of course, and you can try a live demo at

Check out the code at

Jim Webb

Recurrent Neural Networks

Jim Webb played with Recurrent Neural Networks to create new words from the Oxford English Dictionary.

By splitting text into individual characters and determining the relationships and patterns between the letters, it tries to fit patterns and find new text from the old.

So by training the Recurrent Neural Network on Shakespeare, you can generate new Shakespeare. By feeding it Obama speeches, you get new Obama speeches.

It gets the grammar and syntax mostly correct even if it doesn’t understand what the words themselves mean. The output and meaning though, is mostly incomprehensible and silly.

To work, it needs lots of input. Right now, Taylor Swift’s discography isn’t large enough to generate new songs, but eventually Jim will try again.

Steve Trickey

Hacking Kids’ Brains

Steve Trickey wants to HACK kids brains!!

… by teaching children to code at an early age! Language fluency is so important at an early age, and to get kids interested in coding, leaning heavily on games and stories is key.

Scratch has a really cool concept of “Remixing” code, which is like forking a repo but with a more visual way of giving credit and acknowledgement to the original creator. It’s a great way for kids to create and work collaboratively, and projects with a lot of remixes get promoted on the Scratch home page.

His slides are here.

Marie Whittaker

Economic Intelligence Dashboard

Marie Whittaker is working for the DC government putting city data to work for DC planning and economic development.

She can answer questions like “How many units of affordable housing were delivered last year” and “How important is the tourism industry to DC’s economy,” and visualizes the data on the Deputy Mayor for Planning and Economic Development (DMPED) website.

Her formula is straightforward and the results super impactful: get the data, put it in an easy to access place, and make it pretty. Then scrape, reshape, and design the visualizations, host it on Tableau public, embed the data on a Github pages site, and get feedback on Github and Twitter.

All of the data is made publicly available at

See also:

Some interesting insights that came out of her analysis were:

Brian Cohen

MXNet / Deep dream

Brian Cohen used a wrapper of a new library (MXnet) with bindings to many languages including Python; converted it to Julia. Google’s Deep dream uses neural networks to analyze images and from this, visualizes the relationships between images to show what the neural network “sees.”

Travis Hoppe

PDF Steganography

Did you know Travis Hoppe is a secret agent? It’s true!* He’s hiding messages in plain sight (steganography) and dropping them in specially designed PDFs.

By creating a new font that lies about the mapping of character codes (for example, A -> 65), the text you copy from the PDF becomes different than the text you get when you paste it!

Right now it’s a Caesar-style mapping (simple substitution) of the letters, but by mapping multiple fonts, it’s possible even to make an unbreakable one-time pad!

Thanks to everyone who presented, everyone who attended, @svthmc for the writeup, and @jessicagarson for the live tweets.

A shoutout to our sponsors WeWork for hosting!

Round 29 is being planned, stay tuned for the date and time. In the meantime, you can sign up to present. See you all next month!