h3r3 3igHt

receive. then beat: your heartbeat. here, now, and at 8, always 8.

h3r3 3igHt

The recording plays once, complete: nine minutes and 43.67 seconds, performed live through Voice Synth by Qneo ↗, a voice processor with vocoders, pitch shifting, delay and reverb. A first voice sends a signal to the year 2026. A second voice asks you to count, cannot reach eight, and then says eight until the repeating is all that is heard.

The stage is a clock radio. Each word lights on its main display half a tick, 65 ms, before you hear it, stays while it is said and goes out, so the eye does not move and the ear carries the meaning. The letters are the ones seven segments can make: m and w take two cells, k and x borrow H, the pronoun I is one stroke on the left. The first voice writes e as E; the second writes it 3, its mirror, so the second voice says h3r3, and E and 3 together are 8. Long words part where a voice would, heart- then beat. Numbers arrive the way a calculator takes them in: 1, 1000, 1003, 1300, 1314. The unlit eight is always there in every cell, and every segment that lights burns in a little.

From 4:35 the second voice is heard through a delay: every sound comes back 1.038366 seconds later, quieter, and its words come back in the same cells, each a quarter dimmer. In the count each number takes its own place, and the number just said also stands in the eighth cell. At the first seven the returning six is still there under it, and a six with a seven over it is an eight that nobody said. Without words, each word folds into one cell, all its segments at once; most are eights. The first eight said is the last word spelled, 3igHt; after it, eights are entered one a cell until the display reads 88888888.

The clock keeps the local time, here and now. When the voice says 2026 the clock reads it as a time, 20:26, twenty-six minutes past eight in the evening. At “yesterday at 8:00 PM”, and through the eights, it stands at 8:00 with its evening lamp lit. When a station ends the clock blinks 8:00 once a repetition, about 58 times a minute, a resting heart.

There is one gesture, a beat: tap anywhere, or press Enter, B or 8. Each beat is a heartbeat, synthesized here. It lights every unlit segment on the face for a moment, lub and then dub, and draws an 8 on the small display, which is yours: the ring on the lub, the bar across it on the dub. From “Can you count?” until eight is said, the bar does not come, and your heart can only make zeros. In the signal your display shows your rate. From “Can you hear you?” someone else gives each beat back seven times, once a repetition, one cell further and a quarter quieter each time. In the count your beat says the count you have reached, in the second voice’s own words, and someone else repeats it one higher each time: the kernel’s only arithmetic is that a repeated mark comes back one higher and a quarter quieter. The count as spoken has fifteen words, one, two, three, then sixes and sevens; after the last seven comes the kernel’s sixteenth symbol, zero, which says nothing, and its cell goes dark.

“What if you let go of that repetition?” Everything held is let go. A beat then lets the unknown receive all: eight, eighty-eight, one thousand three hundred fourteen, twelve or yesterday at 8:00 PM, each cut from the recording and let through or held back by the kernel’s yielding, seeded with the minute you arrived. “Are you here?”: a beat says “I am here.” Then every repetition, whatever it would have become, is rewritten as eight, and your eights are entered among the voice’s. After the end a beat is still eight, said eight times, each a quarter quieter. The kernel’s always lasts ten million ticks, about fifteen days.

The dial has three stations. 88.8 is the recording. 98.8 is 8 here, here 8 ↗ and 108.0 is here 8, 8 again ↗, two songs the artist produced from it. A song ticks eight times in two of its beats, at 113.74 and 131.99 beats a minute, and its acts follow where it sings each part of the transcript. Its words are shown where a speech recognizer, run offline before publication, heard them, aligned to the transcript; words it could not place are not shown.

The p3r kernel hears every beat in WebAssembly and is asked about every tick, in order. The page works out for itself what each beat should bring back, when and how loud, and checks every mark the kernel writes against it; a wrong one stops the machine.

Tap, Enter, B or 8 beats · Space waits or continues · T tunes · M mutes · [ and ] move the words 65 ms earlier or later, if your speakers delay the sound · Escape leaves. Still reduces movement. The know downloads the transcript as far as it has been heard, with your beats and what the unknown received.

The recording, the songs and their words are the artist’s. The page plays constant-gain MP3s of them and cuts of the recording’s own words; the heartbeat is synthesized here. No microphone, sensor, storage or network service is used, and nothing you do leaves the page. Audio provenance ↗