A useless Twitter bot (probably) boosted my career

I still enjoy the results of the bot I designed that night that I could not sleep. Besides, it probably helped me land a new job.

enrique a.
Coinmonks
Published in
3 min readSep 9, 2018

--

How it all started

It was about a year ago, the last months of 2017. I was “between jobs”, had recently returned to my city after finishing my master’s degree in Europe. I had been studying Python for few weeks by that time, and I was enjoying it.

One night, after several hours of study and experimentation I finally decided to call it a night and go to sleep. Or so I tried, because I could not fall sleep, and idea was taking shape in my mind.

Since my first days studying computer science, I have always loved randomness. I felt inspired by projects like Corporate Bullshit Generator and Automatic CS paper among many others. I even created an android app to generate video game names (topic for another story). I also love genetic programming in part because of the random factor (also, another story).

The idea

Coming back to that night, I started visualizing an application to create “Random Art”. I didn’t know at first what that meant, but soon it became clear that the easiest way to go is generate images (think paintings). I don’t claim it is an original idea, I was re-thinking the infinite monkey theorem. A program creating an infinite number of images, some of them must be good or at least interesting right?

I tried to go to sleep and continue in the morning, but I couldn’t. I finally rose again from bed and started my design on paper. I wasn’t thinking about something super complex like style transfer using machine learning. I wanted to start more basic, simple forms, colors, repetitions, offsets. Something in the realm of Barnett Newman. Full disclosure: I didn’t know the artist, I only googled the style I was searching for.

The idea is pick a size of canvas, a color set (color, gray scale or black and white) and the quantity of shapes to draw. Each shape will be of a certain type (rectangle, polygon, etc), will have a distinct color and it may be “repeated” on X or Y axis for a number of time. All of this randomly chosen at run time.

The development

I finally managed to sleep and the next day I started coding what I already had on paper. I also came with the second part of the idea. Not only to develop the image generator, but also make it a Twitter bot. So I did my research: a little bit of Pillow for image manipulation, and tweepy to access the Twitter API.

A couple of days later the git repository and the twitter account were ready to spin. I did promote them in my facebook and Twitter accounts, but I don’t have a great audience for that kind of projects.

And that was pretty much the end of it, the bot has been running ever since

Is easier to create pleasant thing in black and white

The conclusion

To this day, the twitter account has practically no followers. However, I do enjoy watching the results every day, and I do “like” my favorite results, here are some examples:

The twist

I have reasons to believe this and another couple of github projects helped me to land my current job. I know for instance it got me to have the interview, after they were not considering my profile at the beginning. Now I program in Python every day, and I still use Pillow from time to time. Working on your own projects and ideas is a great way to invest time.

--

--

enrique a.
Coinmonks

Writing about Machine Learning, software development, python. Living in Japan working as a machine learning leader in a Japanese company.