## Abstract

Many CTFs run for about 24 - 48 hours. How do some people stay up to hack for this long?

In this lightning talk we speculate a design for soda robot that uses RL to optimize the sodas for ctf performance.

What if you made a robot that runs experiments on soda nutrient/drug composition and CTFers? Each CTFer would have a personalized soda machine that would learn over time their performance effects on various soda compositions. This robot would also run double blind experiments on their CTFer to enrich its model of the CTFer’s metabolism and to de-noise the feedback from confounding variables. We shall call it the BallmerBOT.

Additional questions:

- What is “performance”? How do we measure “performance”? Should we use physiological sensors such as an EEG, heart rate and blood glucose monitor to BallmerBOT’s feedback?
- Can we use eye tracking technology and UI interaction to infer performance changes that may be introduced by the soda?
- Can we attach BallmerBOT to a home assistant to provide multi-modal stimuli to the CTFer that couples with the experience of drinking the soda?

We also explore sibling caffeinated beverages like yerba mate, tea, coffees and buckfast; as well as recently commercialized fluids such as pure oxygen shots, and performance drink mixtures that often combine compounds like L-Theanine, Taurine, Guarana, Alcohol, vitamin / electrolye shots, methylene blue and so on.

If these questions sound fun then join us! Ideally with a soda that you enjoy in your hand.