“Perhaps clarity is less about more electrodes and more about less noise.”
– João Sato

We arrived at this conversation through a generous bridge: Rodrigo, a fellow resident at Cloudwalk and doctoral student at UFABC, pointed us in the direction of his advisor. The name was João Ricardo Sato, and from the very first email, we sensed something rare: a disposition that is both rigorous and kind.

On Friday, the three of us met — João, Eduardo and I — to talk about brains, signals, and the rough art of decoding poetry from voltages. Or more specifically, Creativity in vitro, my current obsession to find lyrical residues of “Aquarela” inside raw EEG signals using the OpenBCI Cyton + Daisy headset.

On quality and the illusion of more

João was gentle but precise: “quality is more important than quantity.” Fourteen channels are seductive, but the Ultracortex headset, with its dry electrodes and curly hairs in the way, tends to betray its promise. João suggested something that I didn’t want to hear — and yet I needed to: sometimes fewer, cleaner channels are better than a full array of noise.

He mentioned the Emotiv EPOC X as a potential option — sleek and compact, with 14 channels and relatively lower noise than OpenBCI’s dry pin-based approach.

And then he went further, pointing us to research-grade devices from the heart of Europe:

• The X.on, by Brain Products — a wireless EEG system with 7 active dry electrodes, designed for mobile research. Developed in Munich, it’s used in academic and clinical environments where precision matters more than count.

• The Unicorn Hybrid Black, by g.tec — an 8-channel EEG headset combining dry electrodes with open access to raw data. Based in Austria, the system balances ease-of-use with signal quality, making it a compelling alternative for artistic-scientific fieldwork like ours.

These systems don’t seduce with high channel counts. Instead, they offer a quiet promise: less noise, more truth.

On preparation as ritual

Signals need a body that welcomes them. He suggested using Nuprep gel to clean the skin and reduce impedance. I nodded. (Since then, I shaved part of my head. Ritualistically. I apply gel with reverence.)

On failure as a threshold

“Maybe you won’t be able to recognize words. But perhaps phrases. Or better: categories of meaning.” João wasn’t pessimistic — he was precise. There are physiological limits to what OpenBCI can give. So instead of chasing spectral ghosts of each word, maybe I should teach my model to differentiate six lyrical phrases, repeated fifty times. And if the classifier performs better than 1/6 by chance, that would already mean something.

But even here, the brain tires. Habituation sets in. The EEG doesn’t lie — it flattens. So one minute of looping audio for an hour? Probably not. We’ll need rhythm. Silence. Variability. Human calibration.

On the alpha sine wave

Here’s a trick from the neuro lab: want to know if your EEG is working? Close your eyes.
If you see a sine wave between 8 and 12 Hz in the occipital channels, that’s the alpha rhythm saying hello.
If it vanishes when you open your eyes, you’re alive — and your headset is listening.

I’ve implemented this now as my default calibration ritual.
Alpha as a gatekeeper.

On metrics that speak medicine

When it comes to classification, João reminded us of something elegant: ROC curves. In biomedical domains, they are intuitive. Sensitivity. Specificity. Not just numbers, but reflections of error, prediction, precision.

Final note

I left the meeting with a full heart and a disciplined mind. João is the kind of mentor you wish you had — and now I do consider him one, even if unofficially. I hope to keep him updated, to share progress and bumps in the road. For now, I send him waves — both cerebral and emotional — encoded with gratitude.