Each entry states the question, what was uncertain, what was done, what came back, and what it changed. Where a conclusion changes, a later entry names the one it supersedes.
A clean separation in resting potential between conditions turned out to track which amplifier recorded them. Swapping rigs moved the effect with the hardware. Condition and rig are now counterbalanced by default.
This was the strongest surviving candidate from a systematic search of the short-period band, and it reproduced across two acquisition rigs. It tracked ambient temperature cycling and persisted in uninoculated controls at comparable amplitude. It was retired after four months.
Radio emissions from the microcontroller and a shared supply rail dominated everything else I had been attributing to the substrate. Spectra before and after each change are included, along with the changes that made no measurable difference.
Without a reference for what the acquisition chain contributes on its own, there is no way to tell a quiet block from a quiet amplifier. The phantom sits at the converter's specified floor, which sets the ceiling on what can be claimed from anything else.
Four P. ostreatus blocks in constant darkness (paper data, Dec 2025–Apr 2026; ESP32 + ADS1115). Coherent circadian-like 24-hour rhythms in within-window variability: 23/30 inoculated channels vs 4/14 hardware-matched controls (p = 0.006). Peak phase clustered ~13.7 UTC in inoculated channels (Rayleigh R = 0.83) but not in controls (R = 0.054). Room temperature and humidity also cycle at 24 h and controls were recorded later, so an environmental contribution is not excluded. What distinguished inoculated channels was a consistent peak phase that same-hardware controls did not show.
Setup, electrode placement, and the first weeks of continuous logging on ESP32 + ADS1115 hardware.
For the current position rather than the history, see Findings.