F.R.I.S.K.I.E.S (ProtoHackNYC, 1st place)

N.B.: Anyi Sun, Jeremy Watssman and I took part in ProtoHack on April Fool’s 2017. The theme was “Evil Genius Ideas”. This was our winning idea.

Our initial research revealed that the vast majority of super-villains fail at global domination because of their hired help. 0% of comic book henchmen are successful at their central task. Human henchmen are expensive, unreliable, and inefficient — after a lengthy and expensive onboarding process, they are sent out into the field never to return again, or worse, returning in a small zip-loc bag.

Thanks to recent breakthroughs in robotics research, initially performed by our competitors at [REDACTED] and later stolen by us, we have a unique opportunity to improve upon the flawed henchmen model.

We made the decision early on to avoid designing human androids for several reasons. In general, human androids would share numerous flaws with human henchmen by virtue of being human shaped:

1. Human beings are apt at distinguishing real and “fake” human beings (see the “uncanny valley”). The range of motions and behaviors human androids would need to exhibit to successfully infiltrate human society prohibit expeditious development.

2. Human androids would be susceptible to the same societal checks placed on existing human henchmen (e.g., not being able to loiter around an area performing reconnaissance without drawing attention).

3. The idea of human androids would probably not have won us the hackathon.

Early in our design phase, we identified the most common use cases of henchmen, and the various tasks they would have to perform. These were:

1. Assassination
2. Surveillance
3. Security

Our list of evil use cases (green); early robot designs (magenta); early name for our product (cyan)

Our first robot assassin designs were of small rodents such as guinea pigs — after spending a week watching a small rodent nibble a man to death, our clients requested something more efficacious. We then unveiled a squad of elephants and hippopotamuses, which were very powerful and could easily eliminate any target, but lacked the subtlety required of a covert assassin.

Through further experimentation we reduced our shortlist of potential animal candidates to the following:

n = 80 participants on Amazon’s Mechanical Turk

We then conducted further studies to better understand how a potential human target was likely to respond to each of these animals. The ideal animal had to be likable enough to infiltrate the homes of targets and potentially assassinate them, but inconspicuous enough to conduct surveillance operations in varied operational environments.

We found that some otherwise ideal animals elicited feelings of disgust. Mice and pigeons, while stealthy and capable of carrying small explosive devices were often actively avoided by targets. The blast radius of an explosive charge carried by a robot mouse or pigeon was not sufficiently large to wound or maim a fleeing target.

Another class of animals had form factors that were were ill-suited to use necessary tools. For example, whereas squirrels were versatile enough to be both be seen as “cute” or “gross” depending on the context, their feeble arm-to-leg ratio meant that they had difficulty handling the recoil of the miniaturized AK-47s we had designed for them.

Of all the animals tested, dogs were the most likeable. We are currently looking into the feasibility of a dog model, but our initial testing revealed that the likability of dogs rendered them vulnerable to being spotted, approached and inappropriately fawned over during surveillance missions.

In the final consideration between cats and koalas, both of which exhibited ideal characteristics, koalas were discarded for being “too weird”.

We thus chose to develop a robot cat, which was convenient, given that we had already devoted significant resources to the development of this funny acronym:

System for
Eavesdropping and

or F.R.I.S.K.I.E.S (our original proposed name of “Cat Spy Network” was too conspicuous and led to the arrest of our first Head of Product).

Pilot testing of these robots took place in Istanbul, where a local culture of revering, respecting, and ignoring cats was deemed ideal for our first test run. Our initial efforts were very encouraging, and we were able to successfully stop a military coup, aiding an evil dictator who had styled himself as a caricature of a comic book villain.

While our robotics team worked on the physical product, our design team produced a clear, easy-to-use user interface for controlling these robots.

Using a mobile app, users are able to upload photos of targets, the location and time of operation, followed by which agents would use to identify them:

From left to right: choosing a service; uploading a target’s photo; selecting time of operation; selecting operative; selecting location of operation; confirming order.

Users receive a push notification upon completion of the task:

The app is currently only available on iOS, as our market research suggest that it is the preferred platform of evil geniuses.


Jeremy Watssman delivering our pitch:

Our slide deck:

Jeremy Watssman, me, and Anyi Sun