DMCF · TX-002THE CALENDAR

The sheet draws the HOLDING (AMEND-042): the whole feed, no gate, no day-blocks, one continuous pass of ~1,074 h — so a sheet is a 24 h window cut out of that pass, and the first record is usually already playing at midnight. It resolves the same record at the same second as holding.html and systems.html; LAYER switches to the canon air pool and its eight day-blocks instead. One sheet per day per layer, ever.

Drawn as a recorder drum: twelve rows of two hours, and every record airing that day traced as its OWN loudness envelope — fingerprint-v1, the station's admitted instrument — at its actual airing position. A drone airs as a long level band; a percussive piece as jagged terrain. Height is loudness; the ink is brightness — the record's spectral centroid on the same instrument's mel scale, dark records in ink, mid in sepia, bright in blue. Two records can trace the same envelope and sound nothing alike. The weight is tonal focus — a pitched, sustained record draws fine and exact; a spectrally flat one (a printing press, a water wheel) draws heavy. The wobble is harmonic drift — the hand-tremor every stroke has always had, now meaning something: a record that modulates draws unsteadily, a harmonically still drone draws straight. Four channels, each chosen to be orthogonal to the last. On the first thirty sheets the ground itself is sepia, so there the mid band folds into ink and brightness reads in two bands rather than three — no trace ever shares the ruling grid's pen, and each sheet's token declares which encoding it used. The clock elects the day; a mint merely claims it. Yesterday has already happened; tomorrow is an exact, checkable prediction.

Deterministic while the pool stands: the sheet is a pure function of (date, feed). Any ingest re-draws every future day — and that re-drawing is itself the record. Scheduler constants are the station's own — 97 · 7919 · 2654435761. When the shown day is today, the red dot is the station: the record airing at this second, at its measured height. Click the sheet: the COLUMN CHORD — the twelve tracks the carriage line penetrates, ±5 s, played together. Double-click: the HALLUCINATION — the dot slips to one of those lines for ten seconds.

Press and hold: THE MATCHED PAIR. The other two gestures play the station against itself. This one plays it against its control group — the held, owned records that carry no assigned system and have never aired. The record under the finger is paired with the closest record in that silent set on the four channels this sheet draws: the two would trace the same line. They sound blind, one then the other, at the same phase of their own duration, then together; the order is drawn from the second the finger went down. Name the one that airs. Only the aired record has a place on the sheet — its partner has never had a position to mark, and that absence is the experiment. Admissible means within 2× the duration and within ±1 on ink, weight and wobble (widened to ±2, then ±3, only if nothing qualifies, and the tolerance used is printed); the winner is the least mean difference between the two loudness envelopes, each resampled to 32 points and normalised to its own maximum. Timbre is reported and never used to select — the four drawn channels say the sheet cannot tell the pair apart; the cepstrum says independently whether the ear should be able to. If the aired record cannot be named above chance, the gate is not acoustic — which is what three failed routes already imply, and which AMEND-047 left to route 1, unexecuted, under a checkpoint that closes the timing programme on 2027-08-13. PAIRS downloads the session's verdicts with the protocol and the selection terms beside each one. It is evidence, never a mint: no token, no ledger line, and nothing is added to the sheet.

Point at the sheet and it explains itself. Hovering reads all four channels at that exact second off the same files the trace is drawn from, so the reading and the curve cannot disagree. Readings live in their own column, never on the paper: the live box is pinned at its top and each hover stacks underneath it, so nothing you are reading moves while you read it. Height is the only channel read per point, so it alone shows the instant; ink, weight and wobble are the record's, which is what actually draws the line. A 0 is not a value — it means the bucket is silent and that channel has no reading there, and the column says so rather than inventing one. Wheel to zoom at the cursor, drag to pan, 0 to fit, + / − by steps. Zoom does not enlarge a summary, it re-draws: ×2 re-samples the envelope, ×4 inks the trace point by point instead of once per record — the brightness data was always per point, the paper was too small to show it — PLOT VIEW is the only export that carries the zoom: it plots the current crop, magnified to fill the sheet, because detail reaches paper by cropping and not by adding points — at ×8 on a full sheet the median segment is 0.06 mm against a 0.3 mm nib, which the pen cannot draw and the plotter's own linemerge would erase. Cropped, the same detail measures 0.5 mm at ×8 and 1.3 mm at ×24, and costs ~170 paths instead of 2,300. A plate is captioned NOT A MINT and takes no ledger line. ×6 names every record and ties the seams — the joins where two adjacent plays sound alike anyway (repair() enforces credit and series; it knows nothing of sound) — and ×8 turns the height channel into the spectrum itself, low / mid / high as three curves. Every export drops back to ×1 first: what you take away is the sheet, never the moment you leaned in.

Rarity is measured, never declared: the grey figures are percentiles over the 181 days around the shown one, computed in this session. A rare day is rare because the schedule's condition producing it is rare.

Both channels are projections of the frozen instrument, never new measurements. The envelope is sampled once per ~8 s of audio (a row resolves about that), so the longest records are no longer drawn from one reading a minute. Brightness is absolute — ~105 Hz per step, saturating near 1 kHz — because unlike loudness it only means anything compared across records. A record with no fingerprint draws a low dotted line in ink: no shape is shown rather than one invented.