↩ ExploRational · A Catalogue of Divergent Rationality
explorational · a divergent move
Memory arranged for recombination proposes juxtapositions no intention would have made. What happens when your fragments are filed to collide — and the box starts talking back?
For three centuries the educated European reader kept a second, exterior memory. Erasmus's De Copia (1512) told students to rule notebooks into heads — loci, the "places" of classical rhetoric — and to copy in every striking phrase their reading yielded; law students at the Inns of Court commonplaced cases the same way, under heads they would argue from for the rest of their careers. The point was never storage. A head is a promise that two fragments copied years apart will one day sit on the same page.
A second lineage wanted the filing itself to think. Ramon Llull built paper wheels — concentric, lettered, rotatable — to force combinations of concepts no meditation would assemble. Leibniz, nineteen, took the hint in the Dissertatio de Arte Combinatoria (1666): an alphabet of human thoughts whose recombinations could be enumerated outright. And Niklas Luhmann, who filed ninety thousand slips over forty years, made the strongest claim in the plainest prose: his slip-box was a communication partner. He asked it things. It answered with juxtapositions he had no memory of preparing — and he credited the box, not himself, with the surprise.
the signature instrument
File a fragment, spin the wheel, judge what the box proposes. Twelve public-domain slips are already filed, so the bench works before you have written anything. The displayed values are counts or ratios of events in this box — and nothing grades quality.
The slip-box
01 · capture
02 · the collision wheel
one mulberry32 draw per spin · uniform over the eligible pairs — each exactly as likely as each · no self-pairs, no weighting, no model of which pairs are "good" · same seed + same slips + same filter replays the same sequence · repeats flagged from a real log of past draws
what do these have to do with each other?
03 · the bridge log — the box's yield
spending removes the entry so it can go be a paragraph somewhere; the keep still counts in the hit rate — a judgment made is a judgment made.
04 · structural readouts
distinct pairs drawn ÷ pairs possible, over the slips currently in the box. Filing grows the denominator quadratically; spinning grows the numerator one pair at a time.
kept ÷ drawn — a ratio of your own keep-or-pass judgments, disclosed as exactly that. It is not a grade of you, the box, or the thoughts.
how many slips carry each tag. Peaks are where collisions get likely; the "different tags" filter forces draws across them.
plain text — copy it out; this page never sends anything anywhere.
filed to collide
A tag is a bet, placed cheaply now, that two fragments will matter to each other later. Deliberate thought cannot place that bet, because deliberation retrieves by relevance: ask your memory for material on persistence and it returns what you already associate with persistence — the essay you were going to write anyway. A box retrieves by adjacency. Whatever sits under the head sits there whether or not it fits your current intention, so consulting the box means being interrupted by your own past selections, half of which you no longer remember making.
That is the divergence engine of this move. It opens options by recombining: the stock of fragments stays fixed while their pairings multiply, and the pairing — not the fragments — is where the new thought lives. Erasmus's students did not reread their notebooks page by page; they raided them across heads, and the raid put a church father next to a proverb next to a point of law. A commonplace thinks the way a city street does: by forcing proximity between things that would never have sought each other out.
what to try
File five slips of your own — sentences you would have copied into a notebook anyway. Watch the sub-line: the starter box's 12 slips make 66 possible pairs; yours make it 17 slips and 136. Five slips — 42% more — and the space of possible collisions more than doubles.
Spin ten collisions. Write a bridge only when the pair genuinely provokes one; let the rest pass. Then read the hit rate for what it is — a record of your own keep-or-pass judgments, nothing more. Keep the bridges that surprised you.
Tick different tags only and spin again. The filter is real — it removes every pair sharing a tag, 11 of the starter box's 66 — so each collision now crosses your own filing scheme. The eligible-pairs count prints exactly what the constraint costs.
Watch coverage. Ten distinct pairs is 15% of a 12-slip box. File thirty more slips and the same ten draws touch just over 1% — C(42,2) = 861. The space of collisions outruns any deliberate program of comparison, which is why the wheel, not your diligence, has to do the sampling.
the box as partner
"Communicating with Slip Boxes" makes a claim usually read as charm and better read as engineering. A partner, Luhmann says, must be able to surprise you; the box surprises because its internal order has drifted away from yours. Slips are filed behind fixed numbers, not topics, so every slip keeps its neighbours forever; links laid down in 1965 still fire in 1985; and the owner's memory of what is in there decays while the arrangement does not. Query the box and it answers from its structure — which is no longer your structure. The surprise is a property of the filing system, not of the mood you consult it in.
Take the arithmetic literally. Ninety thousand slips hold C(90000, 2) = 4,049,955,000 possible pairs — four billion, in a box built by one careful reader. No deliberate mind visits that space; forty years of daily use samples a hair of it. So the division of labour is honest: the box proposes, vastly and mechanically, and the owner disposes, judging which proposals are live thoughts. Partner is the correct word for an arrangement in which each side does what the other cannot.
the honest bench
The wheel is one pseudo-random draw per spin — mulberry32, seed shown — uniform over the C(n,2) pairs of your current slips: every pair exactly as likely as every other, no self-pairs, no weighting, no model of which pairs are "good." Re-enter a seed and the same box replays the same sequence. The different-tags toggle is a plain filter, its cost printed beside it; repeats are flagged from a real log of past draws.
The numbers are counts of events in this box: slips filed, collisions drawn, bridges kept. Hit rate is kept ÷ drawn — a ratio of your own judgments and nothing else. Coverage is distinct pairs seen ÷ pairs possible. The bench proposes, never composes: no text on this page will suggest what two fragments have to do with each other. That sentence is the entire job left to you.
the move ↔ the box
| The move | The bench |
|---|---|
| a slip | one fragment worth keeping — small enough to file, sharp enough to survive being reread out of all context. |
| a tag | a filing that enables meeting: a cheap bet, placed at capture, that this fragment will matter to another one later. |
| the collision | the juxtaposition intention would never make — two slips forced onto one desk by arrangement plus chance, not by plan. |
| the bridge | the thought the pair provoked. The box convenes; the sentence is always yours. |
| the hit rate | how often the box out-thought you — the fraction of its proposals your own judgment chose to keep. |
| C(n,2) | why the box grows smarter faster than it grows: slips accumulate linearly, possible collisions quadratically. |
how this opening fails
risk · endless deferral
The catalogue's named risk for this move. Capture is work, collision is work, and both can substitute forever for the essay, proof, or decision they were supposed to feed. The failure is visible in the readouts: drawn climbing, kept climbing, nothing ever leaving the log. The bridge log exists to be spent, not curated — each entry is a draft owed to some piece of actual work, and the spend button is there to be used. A bridge still sitting in the log a month on was a souvenir. A supply chain that only accumulates inventory is a warehouse.
risk · hoarding without re-reading
A commonplace that is only ever written into never talks back; capture without collision is a diary with an index. The wheel is re-reading, mechanized — zero spins means zero surprise, whatever the slip count says. Erasmus's students copied in order to redeploy; Luhmann queried his box on most working days. If slips filed climbs while collisions drawn stays flat, you are not keeping a commonplace. You are keeping a record of intentions to think.
File your fragments to collide, and the box will propose what you would never have planned.
can you use it?
RECOGNITION — Which of these is actually keeping a commonplace? A: copying striking passages into a notes app you never reopen. B: filing fragments under heads and regularly forcing pairs of them onto one desk. C: listing twenty ideas before judging any of them.
B. Capture plus collision is the move. A is the near-miss — a diary with an index; a box that is only written into never talks back.
THE NEAREST NEIGHBOR — Random entry also uses chance to break intention. Where does the commonplace's chance draw from?
Your own filed past. The wheel samples pairs of fragments you once judged worth keeping — arrangement plus chance over your prior selections — where random entry pulls a stimulus from outside the stock.
PRODUCTION — Two hundred highlights sit in your read-later app; an essay is due and nothing is moving. Run the move before looking.
Tag the highlights under a handful of heads, draw random pairs across heads, and write one bridge sentence for each pair that provokes one — then spend a kept bridge as the essay's opening. Yours works if a pairing you never planned produced a sentence of your own, and that sentence left the log for the draft.