(we already outsourced the hippocampus)
on cognition, curiosity, and the architecture of memory
The spring of 1997 still comes back in fragments, pastel fabric chairs, a white vinyl table with rounded corners, sunlight edging through blinds that always stuck halfway. It was that suburban 1991 look, white furniture with soft colors, durable enough to outlive fashion but never heirloom. We were lower middle class, comfortable but provisional, surrounded by objects that imitated permanence.
On that table sat a typewriter that remembered a single line of text. Its LCD strip, maybe twenty words wide, was my first real negotiation with machine memory. Not the first electronic thing I had used—there were computers, consoles, word processors—but this one bridged ideas. It had spell check, which meant it didn’t just record, it predicted. It tried to guess my intentions, corrected me when I was wrong (and I often was), and quietly implied there was a model behind the curtain, however simple. I watched thought appear, hover, get judged, then click into permanence. Editing inside that narrow margin felt like an early lesson in how cognition might one day externalize itself, how a machine could join the loop of thinking.
Back then, knowledge had weight. Encyclopaedia Britannica pressed into shelves like geological strata, thirty-two volumes of curated certainty. Our Christian school didn’t have a real library—faith was the curriculum, not inquiry—so I biked to the public one. Sometimes I asked my mom for a ride, but mostly I went alone with a backpack full of pens and paper. I’d ride straight east down 75th Street, then northeast up Plainfield Avenue, where on clear days you could see the Sears Tower perfectly framed at the end of the road. That view was proof that the world extended beyond the subdivision, and that ideas did too.
Then the compression began. Infoseek, 1994. Lycos, 1995. AltaVista, 1996. Google, 1998. Wikipedia, 2001. PubMed went fully online in the mid-90s, arXiv opened physics to everyone, and JSTOR turned stacks into scans. Each step reduced the half-life of ignorance. Syntax replaced asphalt. Indexing replaced bus schedules.
From 1997 to 2001, I lived inside that transition at UIUC, a campus built on computation. Mainframes hummed underfoot, T3 lines ran into dorm rooms, and the University Library rose like a physical manifestation of the web. I spent countless evenings in the stacks, chasing references through cross-listed journals, tracing the lineage of ideas by footnote and flickering CRT. The smell of old bindings mixed with ozone from the terminals. It was lonely in the most productive way. That kind of isolation taught me how to think deeply, uninterrupted. It’s the same impulse that drives me to build w.oodland now, a high-tech outpost for solitude, study, and invention.
Cognitive science gave a name to what I was experiencing: the extended mind. Andy Clark and David Chalmers argued in 1998 that cognition is not limited to the brain but extends into tools, notes, and networks. External artifacts don’t just store thought, they participate in it. A list scaffolds working memory. A sketch externalizes spatial reasoning. A database query becomes a step in a reasoning chain. The mind, in practice, is a coupled system of neurons and nouns, brain and world in continuous conversation. Once you see that, you start recognizing it everywhere—the notebook margin, the lab whiteboard, the browser history, the Git log—all cortical tissue built from matter and light.
The substrates change, but the physics do not. Information is low-entropy structure, organized energy resisting decay. Whether stored in synapses, magnetic domains, or silicon transistors, the challenge is the same: preserve pattern against noise. Intelligence is matter folding back on itself, modeling itself, testing predictions against its own uncertainty.
Modern models continue that experiment. They don’t retrieve facts like a librarian; they reconstruct them through probability. Early systems used hidden Markov models to predict sequences in language and biology. The current generation, built on transformers, uses attention mechanisms and Bayesian weighting—priors updated with every new example, loss minimized through iterative correction. The result isn’t recall, it’s regeneration: a plausible future drawn from statistical memory. I think of this as thermodynamic reasoning—order coaxed from noise by iterative sampling. Temperature here is literal and metaphorical, a measure of creativity, precision, and risk. Lower temperature yields clarity, higher yields exploration. Drafts cool into coherence.
Writing inside this system feels both ancient and new. The process remains: ask, read, synthesize, write. Only now the loop runs at the speed of thought. I still cross-check claims, still chase primary sources, still shape sentences until they carry weight, but the workspace has expanded. Instead of biking to the library, I bike through a vector space. Instead of carrying eight books, I carry eight trillion tokens. Same curiosity, new bandwidth.
Tone has become another parameter. I can ask the system to speak clinically, lyrically, or with a dry wink. It’s all experiment. I try on voices, read them aloud, refine, repeat. It’s a new kind of scientific method—iteration as inquiry, imitation as training data. Every essay is a lab experiment in language, another attempt to collapse uncertainty into understanding.
Lewis Thomas remains my anchor, his essays teaching that clarity is the highest form of kindness. Oliver Sacks, whom I first read in college, showed me empathy as science. Primo Levi, also from those years, taught me that precision carries moral weight. Annie Dillard and Loren Eiseley are new companions—found while forming this very essay, suggested by curiosity and, admittedly, by AI. Each one expands the space of what writing can hold: fact, awe, pattern, light.
I still study reaction kinetics, both in molecules and in sentences. It’s the same discipline—observe, perturb, measure, refine. Activation energies everywhere. The structure of understanding follows the same laws of transformation as chemistry does: energy in, organization out, sometimes heat, sometimes illumination.
The vinyl table is gone, the Britannica donated, but the logic remains. Gather what you can carry, translate it into understanding, and share it forward. The difference now is scale. Human knowledge has entered its exponential phase. Every node, every query, every experiment feeds the same accelerating curve.
We’re past the early chapters of the information age, but its trajectory is still steepening. Each system we build folds knowledge back on itself, compressing centuries into seconds. The wave isn’t cresting—it’s compounding. If we ride it well, curiosity becomes not just our method, but our momentum.