Health is a maze we’re not built to navigate. We speculate wildly, fixate on single symptoms, obsess over alarming words, and are easily swayed by the latest headline or anecdote. Faced with an ocean of data—studies, personal journals, poetry, occult lore, clinical trial acronyms—the human mind simply cannot hold it all, let alone sort, correct, and synthesize it. Yet that’s exactly what the body demands: a coherent picture of its intricate web of processes.
Enter the grand project: the health of our species, rendered intelligible not by hunches, but by intelligence itself. Large language models are, for the first time, able to take on what I call the Health Hyper Object (HHO)—the entire sprawling, contradictory, and deeply human archive of what we’ve said and written about our own physiology.
What is health, really? It’s physiological processes, yes, but they’re marked and recorded in institutional and historical contexts as words and acronyms. Those symbols derive their power from thousands of papers, each layering meaning onto strings like “IL-6” or “HbA1c.” The objective side of the HHO alone—PubMed—contains somewhere between 30 and 80 billion words of dense, interlocking scientific discourse.
Then there’s the subjective side: first-hand accounts written in deeply personal semantic schemas, what you might call individual idioglossia. This is the poetry of illness, the occult writings that map inner states to celestial rhythms, the direct diaries and journals outlining specific effects of a particular food, a particular practice. From 1800 onward, all the world’s poetry, journals, and books might amount to 10 trillion words, each one encoding a fragment of lived bodily experience.
And strewn across both realms are gaps in knowledge that can be inferred. We know A does B in some situations, and from another vector of perspective we know B does C, but no one has yet recognized that A does C. These hidden connections are latent in the data, waiting for a mind that can see across the silos.
Now imagine an LLM that ingests not just PubMed but all 10 trillion words of that subjective and objective sprawl. Its first monumental task is to make the corpus intelligible to itself: indexed, searchable, coherent, corrected where error has crept in, duplicates dissolved, contradictions flagged. Then the real work begins. Health is to be pulled out of this mass by inference—by statistical methods that sum up additions, delete redundant claims, correct mistakes, order sequences, sort relationships, compare outcomes. The body is just an intricate web of processes, and processes can be processed.
The human failure here is not a moral one; it’s a cognitive limit. We can’t hold 10 trillion words in mind. We seize on a vivid phrase and overlook a quiet statistical signal. We weave stories around a single data point. LLMs suffer no such frailty. They can sift the idioglossia of a 19th-century mystic and cross-reference it with a 2024 meta-analysis, spotting the A→C inference that a thousand specialists missed.
Understanding health is not a matter of willpower or intuition. It’s the job of intelligence—at a scale and patience no human can muster. This is the promise of the Health Hyper Object, transformed by LLMs from a cacophony of symbols into a living, navigable map of what it means to be well.
(Wrote myself, then put through Deepseek.)