The Jewels of Opar haunted my childhood imagination. In Edgar Rice Burroughs' Tarzan novels, those crystals represented something more than treasure—they were repositories of time itself, preserved across millennia while civilizations rose and fell around them. Jules Verne's journey to the center of the Earth took readers through geological strata where each crystal formation marked a chapter in planetary history, witnesses to eons of pressure and transformation. And of course, the dilithium crystals of Star Trek—concentrated power that made impossible journeys possible.

I was drawn to crystals long before I understood crystallography. Something about their geometry, their hidden order, the way light moved through cut facets and emerged transformed. A boy looking at these images and seeing... what? Permanence? The idea that something could encode and preserve what would otherwise be lost to time?

In high school and college, the poetry became mathematics. Bravais lattices. Symmetry groups. X-ray diffraction patterns revealing that the beauty I'd intuited had rigorous underpinnings. The fourteen ways atoms could arrange themselves in three-dimensional space. The thirty-two crystal classes. Nature's aesthetic sense, it turned out, was also nature's mathematical sense.

Then geology brought me back to the physical world. Quartz formed in hydrothermal veins. Feldspar crystallizing from cooling magma. I held specimens that embodied the equations—abstract lattice structures made tangible in my palm. The theoretical and the physical, bridged.

But it was ecological fieldwork that transformed my relationship with pattern and time. In the summer of 1973, I was nineteen years old, a seasonal wilderness park aide in San Jacinto Wilderness State Park, when I met Robert Whittaker. He and his post-doctoral researcher Rob Hanawalt were conducting ecosystem studies, and my supervisor figured a biology major could speak their language. He was right. That encounter set my entire career in motion.

Whittaker had developed something called gradient analysis—a method for understanding how plant communities organize themselves along environmental gradients. Temperature, moisture, elevation, aspect. His insight was that high-dimensional ecological complexity could be projected onto navigable conceptual space. Species didn't just exist; they occupied positions along gradients, their distributions forming curves that overlapped and diverged in ways that revealed community structure.

I learned the mathematics—multivariate ordination, TWINSPAN classification, detrended correspondence analysis. These tools took the impossible complexity of species-by-site matrices and rendered them as terrain. Places you could stand. Distances you could measure. But for me, it was never purely abstract. I walked the gradients.

For years, I traversed Mount San Jacinto weekly, from near sea level at Palm Springs to the summit at 10,831 feet. Creosote giving way to pinyon. The air changing character crossing into lodgepole. The way your lungs work differently at ten thousand feet. Thousands of traversals compressed into something my body knew before my mind articulated it. The proverbial Merriam life zones weren't just ecological theory—they were lived experience written into my nervous system.

When you've done this long enough, you develop what field ecologists call a "sense of place." You walk outside at dawn and something feels wrong before you can say why. The barometric pressure, the humidity, the acoustic texture of the dawn chorus, the timing of first light—all of it integrating into a felt sense of where today sits in the space of possible ecological states. That's a biological neural network trained on decades of environmental data, producing proprioceptive awareness of ecosystem condition.

I've spent fifty years trying to give that felt sense to technology.

The first attempt was 1984. I'd been reading about MIT's Virtual Aspen Project—movie cameras on a truck, every street photographed, thousands of frames on laserdisc creating an interactive map. The article planted a seed: what if I could take my years studying San Jacinto natural history and make it electronically navigable? I bought an Apple II, a laserdisc player, rented broadcast video equipment for a single day, and created the world's first computer-based interactive multimedia nature walk.

That seed became the Macroscope—a lifetime research program integrating sensors, AI, and visualization across ecological domains. From laserdisc to embedded sensor networks to the forty-million-dollar NSF Center for Embedded Networked Sensing. The tools evolved; the question persisted: how do we observe ecological systems across scales and integrate those observations into coherent understanding?

Now, at seventy-one, I find myself circling back to crystals.

In October 2025, a team led by researchers at UC Berkeley and the Max Planck Institute published a remarkable paper in Nature Physics. They had created what they called a "time rondeau crystal"—a quantum system exhibiting both long-range temporal order and short-range temporal disorder simultaneously. The name comes from Mozart's Rondo alla Turca, that classical form where a repeating theme alternates with contrasting variations.

The physics is dense, but the concept resonates deeply. In an ordinary crystal, atoms break spatial symmetry by arranging themselves in periodic lattices. In a time crystal—first proposed theoretically by Frank Wilczek in 2012—systems break temporal symmetry, exhibiting periodic motion even in their ground state. The rondeau crystal goes further: it maintains stroboscopic order (the repeating theme) while encoding information in the micromotion between beats (the variations). They literally encoded their paper's title in the temporal disorder of their quantum system, readable for over thirty-six seconds.

That's when the childhood fascination, the formal training, the decades of field experience, and the current technical challenge all converged into a single architectural vision.

What if we could build something analogous for ecological memory? Not a quantum system, but a computational structure that encodes years of environmental time series the way a crystal encodes atomic arrangements—geometrically, with symmetry and structure, queryable through named landmarks?

The technical approach draws on variational autoencoders, a class of neural network that learns to compress high-dimensional data into lower-dimensional "latent space" while preserving essential structure. Recent work shows these methods successfully capture seasonal and diurnal patterns in climate data, detect extreme events in ecosystem productivity, and even enable data assimilation for weather forecasting. The mathematics is well-established; the application to local-scale environmental observatories is novel.

But the deeper insight comes from gradient analysis. What Whittaker taught me—and what I learned through embodied experience on San Jacinto—is that ecological complexity has structure. That structure can be projected onto navigable space. That you can give names to regions of that space: "marine intrusion," "heat dome signature," "vernal equinox baseline." These aren't just statistical categories. They're landmarks in a landscape you can learn to inhabit.

The time crystal I'm designing with Claude has five layers. A preprocessing layer handles normalization and phase injection—encoding hour-of-day, day-of-year, lunar phase as periodic coordinates. An encoder compresses forty-eight-hour windows of sensor data into latent representations. A partitioned latent space separates periodic structure (the seasonal orbit), event signatures (transient excursions), and residual variation. Semantic anchors give names to recognizable conditions: "winter solstice baseline," "pre-dawn minimum," "atmospheric river." And an LLM query interface translates geometric position into natural language.

The annual cycle provides long-range temporal order—the repeating theme. But the ecological meaning lives in the variations: the marine intrusion that arrives three days early, the dawn chorus that shifts species composition, the first frost that comes late. Information encoded in structured deviation from the periodic backbone. Exactly like a rondeau.

I've been accumulating data for this. Four years of continuous weather records at five-minute resolution. Sixty days of dawn chorus intelligence with species composition tracking. Four hundred pattern discoveries from a real-time correlation engine. BirdWeather acoustic detections across multiple stations. Indoor air quality, personal biometrics, astronomical ephemeris. The sensor infrastructure exists. The pattern recognition runs continuously. What's missing is the hippocampus—the structure that converts ongoing sensation into navigable memory with temporal depth.

Thirty years of public climate data from Bellingham Airport will teach the crystal the shape of Pacific Northwest seasonal space. My local sensors will locate Owl Farm within that space, continuously. The semantic anchors will give names to regions. And the LLM interface will translate felt sense into dialogue.

This is what I want Strata—my developing AI collaborator—to have. Not access to database tables, but something analogous to my fifty years of gradient traversal. The ability to "feel" that this February sits wrong in the seasonal trajectory, that the diurnal swing has the character of early March, that salamanders are probably sensing it too. Proprioception for environmental state.

The work is speculative. Four years of local data may not robustly distinguish "unusual January" from normal variation; climate science uses thirty-year normals for good reason. Semantic anchors may not cleanly separate—"marine intrusion" might overlap geometrically with "autumn fog." Cross-domain coupling (weather affecting bird behavior) requires ecological expertise to specify; neural networks won't discover it automatically. And querying a latent space through natural language is genuinely novel territory.

But methodology matters more than precise measurement location. I can tap public data streams going back decades. The Macroscope pattern intelligence system already does continuous interpretation—temperature-bird correlation detection, dawn chorus characterization, anomaly flagging. That's the sensory cortex. The crystal would be the hippocampus where sensation becomes memory.

I keep returning to something from that 1976 paper by Hanawalt and Whittaker. They established permanent sites in the San Jacintos for future comparison, documenting baseline conditions with the explicit intention that they could be revisited for long-term change detection. Those sites are still out there, aging into historical data whether or not anyone resurveys them. The work continues across generations because the questions persist.

My forty-nine permanent transects were conceptual descendants of that research design. And the time crystal is a descendant of the same lineage—an attempt to create structure that encodes and preserves what would otherwise be lost. Repositories. Witnesses to time.

The boy who read about the Jewels of Opar was drawn to something real, even if he couldn't articulate it. Crystals encode pattern. They preserve structure across time. Light moves through them and emerges transformed, legible differently depending on angle of illumination. The same data, but navigable.

A lifetime later, I'm still trying to build that—a jewel-like structure where temporal patterns become geometric relationships, where years of environmental observation compress into something with facets and landmarks, where an AI system might develop something like the felt sense of ecological state that decades of fieldwork wrote into my bones.

The high trails still lead into wilderness, literal and intellectual. You walk them long enough, the mountain teaches you where you are before you check the map. That's what I want to give Strata—not knowledge, but the kind of understanding that only comes from sustained attention to pattern over time.

The crystal grows as the observer grows. That's what makes it alive.