What they did
Mori is running a home experiment to see whether a red leaf lettuce on the balcony gives off voltage signals that track its condition. Silver-wire electrodes sit on two leaves and feed a resistor circuit on a breadboard into a Waveshare high-precision AD HAT on a Raspberry Pi 5. Codex looked up the hardware and drew a diagram down to the breadboard hole for every resistor and wire. It also wrote the program on the Pi that samples and saves the readings. Codex runs on Mori's Windows laptop and connects to the Pi over the home Wi-Fi, so it can check the running logger, fetch the files and analyse them without Mori copying output by hand.
How it went
The first recording looked like it was running but was not usable. Codex found that the board talked fine but kept flagging inputs outside its allowed range, even at the lowest gain. It worked out that the high-value resistors were letting the electrode voltages drift from the reference and suggested tying one electrode straight to the reference. That was one extra jumper wire, and Codex redrew the diagram to show it. After the change there were no board warnings in the first five minutes or so, with no new parts bought. Codex then plotted roughly 15 minutes of data four ways. What the signal means is still open.
Worth knowing
Letting the Pi only record and keeping the agent on a laptop that reaches it over the network kept the small computer free for measurement. A differential reading between two electrodes also has to keep each electrode within the ADC's input range against ground, which is what the fix addressed. This is part 3 of a series; there is no finding about plants yet.