The City of Ten Thousand Scrolls
In the city of Kashi, the great vihara held ten thousand scrolls. A pilgrim arrived asking, "Where can I read about the rivers of the south?" The keeper Gajendra, an enormous and earnest man, walked through every shelf, opening scroll after scroll. By dusk he had read four hundred and found two relevant ones. The pilgrim, by then, had lost interest and gone for tea.
“A library without a map is a graveyard for questions.”
Disha Draws the Land of Meanings
A young cartographer named Disha came to help. She did not draw a map of the city. She drew a map of the meanings inside the scrolls. Every scroll about rivers she placed close together; every scroll about mountains she placed close together; rivers and mountains were near, since both spoke of water. Stars were far away, in another quarter. The map had no roads, only distances — closeness meant similarity. "Stand on the question," she said, "and the nearest scrolls are the answer."
“Place ideas near each other and finding becomes walking, not searching.”
The Cost of Walking Every Inch
But the map was vast. To find the nearest scrolls, Gajendra would walk the entire continent of meaning, comparing his question to every point. With ten thousand scrolls he could just manage; with ten lakh, he would die of old age. The pilgrims left again. Disha sat at the river and wept.
“Exact answers do not scale. The world is too large to compare against everything.”
The Weaver of Hubs
Then came Jali-acharya, the old weaver. "Do not walk every inch," he said. "Choose a few important hubs across the land — the great cities of meaning. Connect each scroll to its three nearest neighbours and connect the hubs to each other by long roads. When a question arrives, jump to the nearest hub, then walk only the short paths from there." It was a network of small worlds, with bridges between them. Now any question reached its answer in a handful of hops, even with a million scrolls.
“Approximate paths through a small world beat exact walks through a vast one.”
The Pilgrim With Two Conditions
A pilgrim came who wanted scrolls about southern rivers — but only those written after the last famine, and only in Sanskrit. The map alone could not enforce these. Disha added small tags to each scroll: a date, a language, a province. The walker, on arriving at a candidate, would read the tags first; if they did not match, he stepped past. So the search of meaning combined with the filters of fact. The pilgrim found his answer before the noon bell.
“Meaning narrows the search; metadata sharpens it.”
When the Map Was Redrawn
One day the kingdom adopted a new script, and Disha had to redraw the map of meanings entirely — the old coordinates no longer matched the new way of writing. The older Disha would have replaced the map overnight and broken every search until it was complete. The wiser Disha kept the old map alongside the new one, slowly migrated the scrolls in the background, and only when every scroll lived on both maps did she retire the old one. Not a single pilgrim noticed.
“Re-embedding is a migration, not a switch. Carry both maps until the journey is done.”
🪔 Deepak — the lamp of meaning · what this fable means in code
Gajendra the brute-force keeper is exact nearest-neighbour search — perfect, and untenable beyond toy scales. Disha's map of meanings is the vector embedding space, where similarity is geometry. Jali-acharya's network of hubs and short paths is HNSW (and its cousins, IVF and DiskANN) — small-world graphs that turn search into a few hops. The tags Disha added are filtered search and metadata indexing. The careful redrawing of the map under a new script is the discipline of re-embedding when you change embedding models — old and new must coexist until the migration is complete. Maps are how meaning becomes findable.