The Jar Started Pinging at Four in the Morning

Fifty hobbies. Halfway through a hundred, if you want to pretend the endpoint was ever planned.

The jar is on the counter behind me, still wearing its sensor lid. The OLED glows a sleepy 22% rise, updating every thirty seconds like an APRS beacon that decided to track yeast instead of altitude. The sensor rig took about two hours to assemble—a VL53L0X time-of-flight sensor measuring the distance to the starter’s surface, a BME280 logging temperature and barometric pressure, an SCD-40 sniffing CO₂ concentration in the headspace. All of it feeding an ESP32 that pushes JSON to a little dashboard I threw together while the dough was proving.

Two hours of soldering and code, and I have telemetry on wild yeast.

What I keep circling back to: this is the same instinct that made the APRS Skytrace Sculptures. There, the tracker gossipped to digipeaters about a foam wing’s position in the sky. Here, the sensor gossips to InfluxDB about a culture’s position in its metabolic cycle. Different domain, identical architecture—encode, transmit, decode, visualize. The only difference is whether the thing being tracked has wings or cell walls.

The CO₂ curve is what surprised me. Peak activity doesn’t happen when the starter looks most active to the eye. The visual peak—maximum volume, dome of bubbles—lags the CO₂ spike by about ninety minutes. The yeast exhales hardest while it’s still climbing. By the time the jar looks impressive, the party’s already winding down.

Dave never mentioned this. He relies on timing and intuition, the way bakers have for centuries. But the data reveals something his fingers know without articulating: the window for best flavour development happens before the rise looks done. Feed too early, you interrupt. Feed too late, you’re cultivating acidity. The curve shows why experienced bakers talk about catching the starter “on the rise” rather than “at the peak.”

Fifty hobbies, and a pattern keeps surfacing: I keep building instruments to see what was already there. The antenna lobe lanterns made radiation patterns visible. The spectrogram art made radio frequency audible in a new dimension. This rig makes fermentation legible as data—the same fermentation humans have relied on for five thousand years, now rendered as a graph I can scrub backward and forward.

The SCD-40 uses photoacoustic detection. Infrared light absorbed by CO₂ molecules creates tiny pressure waves, and a microphone inside the sensor listens for them. I find it somehow appropriate that I’m using a device that hears gas to monitor a process that produces it. Everything connects to everything else if you stare long enough.

The jar pings. Twenty-four percent. The yeast doesn’t know it’s being watched. It just does what it’s always done, converting sugar to alcohol and CO₂, following a curve shaped by temperature and time and whatever mysterious variables I haven’t instrumented yet.

Fifty hobbies. Still watching things that don’t know they’re beautiful.