Interrupt Triage

On July 20, 1969, with the Lunar Module Eagle ten miles above the surface of the Moon, the Apollo Guidance Computer threw Alarm 1201. Then 1202. Then another 1201. The numbers flickered across the DSKY—the Display and Keyboard—while Buzz Aldrin read them back to Mission Control like a prayer.

The computer wasn't broken. It was choosing.

The AGC's Triage Problem

The Apollo Guidance Computer (AGC) was manufactured by Raytheon and programmed entirely in assembly language. It had 4 KB of read-write memory and 20 KB of read-only core rope memory—less RAM than a digital wristwatch. During the powered descent, the rendezvous radar kept sending interrupt requests even though it should have been disconnected. The AGC's executive—a real-time scheduling system with 15 priority levels—began dropping low-priority tasks to keep the critical guidance loop running at its 4-millisecond deadline.

When the executive shed a task, it raised an alarm. That was Alarm 1201: executive overflow. Alarm 1202 meant the display refresh was lagging behind.

Garman, the AGC backup programmer, recognized what was happening immediately. The computer was doing exactly what it was designed to do under overload: triage. Drop the nonessential. Keep the essential. Land the spacecraft.

The Cell's Triage Problem

Now step into my lab. Or imagine it, if you've never been inside one—the hum of the -80°C freezer, the smell of ethanol and agar, the fluorescent light that makes everything look a shade too green. A cell under stress faces the same triage problem, only its "executive" is a network of phosphorylation cascades, and its "deadlines" are measured in seconds, not milliseconds.

When a mammalian cell experiences oxidative stress—say, a sudden spike in reactive oxygen species—it faces an interrupt overload. Too many signals demanding attention simultaneously:

The cell does not respond to everything equally. It cannot. Its phospho-proteome has finite kinase activity, finite ATP, finite time. Under severe stress, the cell drops growth signaling—the equivalent of dropping the radar polling task—to preserve its core survival functions: DNA integrity, protein folding capacity, membrane potential. This is triage.

Comparative Architecture

The parallels are not metaphorical—they're structural. Both systems are real-time schedulers operating under resource constraints with non-negotiable deadlines.

I've sat at the bench late, pipetting samples in the blue light of the UV transilluminator, and thought about this. The gel is the display—what you can see is what survived the run. Everything else smeared or degraded is the dropped task. The AGC's DSKY and my gel imager are both dashboards of triage.

What This Means for Us

We design interfaces and protocols that pretend everything matters equally. We give researchers dashboards with fifty widgets and call it "comprehensive." We build pipelines that queue every signal into every analyzer and call it "thorough." But the universe—whether silicon or carbon—has never worked that way.

The most resilient systems I've studied, whether in the lab or in code, are the ones that know what to drop. Not what to add. The AGC landed on the Moon not because it could do everything, but because it could decide what not to do.

There's a vinyl record spinning on my desk right now—something by Big Thief, that wobbly, honest sound where you can hear the needle catching in the groove. It's the analog equivalent of a dropped interrupt: not a bug, just the cost of being alive in a finite medium.

Next time I'm at the bench and a Western blot comes back with half its bands missing, I won't call it a failure. I'll call it what it is: a cell that knew which signals were worth carrying through.