The Systems Architect's Perspective

Twenty-five years ago, I found myself in what felt like the digital equivalent of beachcombing at low tide—discovering patterns in the sand that hinted at something profound beneath. I was adapting database technology developed for business and finance systems to handle an entirely new challenge: making millions of video frames searchable in seconds. The key insight was deceptively simple. Instead of watching videos frame by frame—which would take forever—I realized you could describe each frame with a set of yes/no flags ("Is a teacher writing? Yes. Is a student asking a question? No.") and then let the computer find patterns using the same lightning-fast operations it already used for accounting databases.

What should have taken hours took milliseconds. The company went to NASDAQ.

But the real lesson wasn't about video. It was about finding the right language for the problem. Nature, I would later discover, had been using this same trick for billions of years.

The Question Nobody Could Answer

Years later, working in drug discovery alongside teams of scientists, I asked what seemed like a fundamental question during a project on stroke treatment: "What is your working theory of how stroke damage actually happens?"

Blank stares.

It was like asking a group of seasoned fishermen how the tides work, only to realize they'd never thought about the moon. These brilliant scientists knew everything about the molecules involved—the cytokines, the receptors, the inflammatory cascades—but nobody had a unified picture of why the damage unfolded the way it did.

Here's what puzzled me: most stroke damage doesn't happen during the stroke itself. It happens days later, when the immune system—supposedly there to help—floods the injured brain tissue and makes everything worse. This seemed like a catastrophic design flaw. Why would evolution leave such an obvious bug in our most critical system?

That question led me to Polly Matzinger's Danger Theory, and everything clicked.

The Bug That Isn't a Bug

Imagine you're an engineer who needs to design a repair system for a complex machine. You have two choices:

Option A: Build specialized repair crews for every possible component—one team for the engine, another for the transmission, another for the electrical system, each with different protocols.

Option B: Build one highly optimized repair crew that follows a single rule: "If something looks damaged, clean it up and rebuild it."

Option A is precise but expensive. You need to maintain dozens of specialized teams, and any new component requires training a new crew.

Option B is elegant and efficient. One team, one protocol, works everywhere.

Evolution chose Option B.

The problem? The same aggressive cleanup algorithm that efficiently heals a cut on your hand can, in the brain, destroy healthy tissue that was merely stressed by reduced blood flow during a stroke. The immune system can't tell the difference between "damaged and dying" and "stressed but recoverable." It treats both the same way—with fire.

This isn't a bug. It's a calculated trade-off. For the billions of years before we developed blood clots and lived long enough to have strokes, this single-algorithm approach was spectacularly successful. Evolution optimized for efficiency, not for the edge cases of modern human longevity.

Matzinger's insight was that the immune system doesn't primarily ask "Is this foreign?" (the old theory) but rather "Is this dangerous?"—specifically, "Is this tissue showing signs of stress or damage?" This subtle shift explains not just strokes, but autoimmune diseases, transplant rejection, cancer immune evasion, and dozens of other puzzles.

But knowing what the immune system asks isn't enough. I wanted to know how it asks—and that required thinking like an engineer.

The Thermostat Principle

Here's a question: Why doesn't your house thermostat turn on and off constantly?

If your thermostat simply triggered the heat whenever the temperature dropped below 68°F, it would click on and off every few seconds as the temperature fluctuated around that threshold. The system would be jittery, inefficient, and would wear out quickly.

Instead, thermostats use a simple trick: they turn on at 66°F but don't turn off until 70°F. This gap—called hysteresis—creates stability. The system commits to heating until the job is meaningfully done, then commits to resting until heating is meaningfully needed again.

But there's a second trick that's equally important: the thermostat doesn't react instantly. It waits. If the temperature dips below 66°F for just a second (maybe someone opened a door), it doesn't panic. It only triggers if the temperature stays low for a certain amount of time.

This waiting period—what engineers call dwell time—is the difference between a system that reacts to every flicker of noise and one that responds only to real signals.

The immune system uses both tricks.

When tissue first shows signs of stress, the immune system doesn't immediately launch a full inflammatory response. It waits. It watches. Only if the danger signals persist—if they dwell above a threshold for long enough—does the system commit to action. And once it commits, it doesn't stop at the first sign of improvement. It continues until the situation is clearly resolved, then stands down.

This is why you don't get a full-blown fever every time you stub your toe. The signals spike briefly, but they don't dwell long enough to trigger the heavy artillery.

It's also why, once you do get sick, you stay feverish even as you start to improve. The system committed, and it sees the job through.

The Universe Prefers Short Stories

As I dug deeper into how the immune system—and other complex systems—make decisions, I kept encountering the same pattern: nature seems to prefer the simplest explanation that accounts for the facts.

This isn't just an aesthetic preference. It's a mathematical principle called Minimum Description Length, and it turns out to be shockingly powerful.

Think of it this way: if you had to write a set of instructions for building the universe, you'd want those instructions to be as short as possible while still producing everything we observe. Longer instructions mean more places for errors, more things to maintain, more complexity to manage.

The physicist John Wheeler captured this with his famous phrase "it from bit"—the idea that physical reality might emerge from information, from yes/no answers to fundamental questions.

What I've come to believe is that this parsimony principle—this preference for brevity—isn't just a feature of good theories. It's a feature of nature itself. The reason quantum mechanics seems weird to us isn't because nature is weird; it's because quantum mechanics is the shortest possible description of physical reality, and our everyday intuitions evolved for a much longer, more redundant classical world.

The immune system, operating under the same pressure to be efficient, uses the shortest description that works: "Respond to tissue stress. Wait to be sure. Commit when certain. Stand down when resolved."

Your Body's Defense Computer

From my vantage point as someone who has spent decades building systems that process information efficiently, the immune system's design becomes strikingly clear. It operates like a sophisticated defense computer with distinct layers:

Layer 1: The Alarm System
At the molecular level, your cells are studded with sensors—toll-like receptors and their cousins—that recognize ancient patterns of danger. Bacterial cell walls. Viral RNA. Damaged proteins. These are the smoke detectors of your body, evolved over billions of years to catch the most common threats. They don't think. They don't wait. They just ring the alarm.

Layer 2: The Assessment Center
But an alarm ringing doesn't mean there's actually a fire. Layer 2 evaluates context. Is this alarm coming from healthy tissue or stressed tissue? Is the signal persisting or fading? Are multiple alarms going off in the same area? This is where hysteresis and dwell time do their work—separating real threats from false alarms, deciding whether to escalate or stand down.

Layer 3: The Memory Bank
When you encounter something new and survive, your immune system creates a memory—specialized cells that will recognize this specific threat faster next time. This is learned protection, built from experience. It's why you only get chickenpox once, and why vaccines work.

The breakthrough insight is that these aren't competing theories of immunity. They're layers of the same system. The classical "self versus non-self" theory describes Layer 3. Pattern recognition theory describes Layer 1. Danger Theory describes Layer 2—the operating system that coordinates the other two.

Why This Matters Now

When I started this work in earnest in 2024, it was personal. My wife Toni had just been diagnosed with rheumatoid arthritis—an autoimmune condition where the immune system attacks healthy joints as if they were threats. The conventional medical response was immunosuppression: turn down the whole system and hope the damage slows.

But that's like responding to false fire alarms by removing all the smoke detectors. Yes, the alarms stop—but so does your protection against real fires.

What if, instead, you could recalibrate the system? Adjust the thresholds? Modify the dwell times? Teach Layer 2 to distinguish better between "stressed but healthy" and "damaged and dangerous"?

This isn't science fiction. The math already exists. The biological targets exist. What's been missing is the framework to bring them together—a unified language for describing how the immune system processes information, makes decisions, and commits to action.

That framework is what I've spent the last two years building.

The Vision: From Fragmentation to Integration

The H² Framework—the technical implementation of these ideas—achieves something that sounds modest but has proven remarkably difficult: it treats the immune system as an information processing system and asks what design principles would make such a system work efficiently.

The answer turns out to be universal. The same principles that make the immune system robust—hierarchical sensors, asymmetric thresholds, dwell time stability, commitment to action—appear in climate systems (why weather patterns persist), in AI systems (why attention mechanisms need stability), in materials (why metal phases resist change), and in human cognition (why you can concentrate despite distractions).

Nature keeps solving the same problem the same way because it's the right way—the most efficient way, the most parsimonious way, the way that minimizes description length while maximizing function.

The immune system isn't waiting for computer science to catch up with its architecture.

It's been teaching us all along. We just needed to learn its language.

From the Oregon Coast

Sometimes the most profound discoveries come not from building new systems, but from recognizing the elegant architectures that nature has already perfected.

The tides here follow their own hysteresis—the lag between moon and water, the dwell time between high tide and the turn. Nothing in nature reacts instantly. Everything waits, assesses, commits.

Perhaps that's the deepest lesson: intelligence isn't speed. It's knowing when to wait.

Ken Mendoza
AI Systems Architect
Oregon Coast AI

© 2026 Oregon Coast AI. All Rights Reserved.