Ian Summers lives in Exeter. His second clavichord was finished in 2025; the document describing how he built it is dated January 2026 and sits as a PDF on his own website. The instrument is 940 mm long. Twenty-six notes on eight strings — three notes per string over most of the range, four on the top two. Two and a half octaves, C3 to F5, and the lowest octave is a "short octave," so the keys that look like E, F♯ and G♯ sound C, D and E instead. The temperament is Kirnberger II. The sounding length of every note is calculated to the half-millimetre and printed out in full: 625, 611.5, 555.5, all the way down to 171. There are no wrest pins; he tunes it with eight guitar machine heads. The strings are 0.33 mm yellow brass, and the eighth is two of those wires twisted together at about one turn per centimetre. The tangents are staples, bent from the same 2 mm brass rod used to locate the keys. The listing woven between the strings is dressmaker's bias binding. The keys came out of a fifty-year-old piece of parana pine — an old shelf — drilled as a single block, thinned 9 mm at the far end with a Forstner bit, then cut into twenty-six keys with a jigsaw, the 1.5 mm kerf becoming the clearance between them. About five hundred hours: "about," because he says he is not sure how to allow for the thinking time between stages.

Put the construct where it belongs first. The piano is the construct. A piano hammer strikes the string and must then leave it; escapement exists for no other reason than to make the hammer let go at the moment of contact. A clavichord has no hammer and no escapement. The far end of the key carries a metal tangent that rises, hits the string, and stays there — holding it, acting as one end of the sounding length — until you release the key. Which means that for as long as the note lasts, the player's finger is mechanically continuous with the vibrating string. Press a little harder and the pitch rises: that is Bebung. That continuity is exactly what the piano's chisel cut away in exchange for volume; volume bought the concert hall, and the hall bought the repertoire, the conservatory, the market. The piano is beautiful, but it is beautiful the way sediment is beautiful — it is what one completed chisel-construct cycle left behind. It is construct now, not a living remainder. The clavichord's beauty is the beauty of the side that got cut: the finger still on the string.

But this instrument's real remainder is not the instrument. It is the document. Early Music is a construct of its own. Summers opens with a warning: players of early music are generally interested in authenticity, so modern clavichords are almost always built as replicas of historic instruments — that is not what he is describing here. How much that sentence weighs can be read under the Hackaday piece, where a professional harpsichord maker states the rule flatly: the twentieth-century "revival" instruments, the ones that tried to improve early keyboards with modern piano technique, are despised by serious amateurs and professionals alike; nobody now tries to improve them, and only replicas of good original models get made. So the field has a complete naming system for objects — "after Hubert, 1784" — and no slot at all for a new clavichord that is neither replica nor revival. That absent slot is the remainder. Which is why the document is not a plan but a record of choices and reasons: tall bridge or raised soundboard (he took the raised soundboard, because a lighter bridge might be louder); single or double stringing (he tried both on the first instrument, found almost no difference in volume, and chose single to reduce the load on the structure); heavy brass tangents or light staples (heavy on the first, light on this one, copied from the Urbino clavichord, which survives only as a fifteenth-century picture). A replica never has to arrive at any of those forks.

The fretting is where the judgment is. He wrote an Excel spreadsheet to try arrangements of strings and tangents. Because one string has to carry three or four notes, the temperament on this instrument is not tuned in, it is built into the wood: for two notes on the same string, the ratio of the tangent-to-bridge distances has to be the inverse of the frequency ratio. He had to settle the temperament before he could drill. He settled on Kirnberger II — not because it is standard, but because it is close to a system written down for the clavichord in a Wrocław manuscript of around 1470. And then comes the qualifier that matters most: these ratios are not a requirement if the tangents in question are on different strings. That is a refusal to close — he will not let a tidy rule swallow the cases it does not actually govern. The same restraint runs elsewhere. His first clavichord put twenty-six notes on seven strings, mostly four to a string, with the defect that some notes a minor third apart could not be played together because they shared a string; he does not bury that, he reports it as the reason this one has eight. And there is a failure left standing in a figure caption: the tangent on the fifth key, E/G♯, sits too far from the listing and kept bouncing off the twined string, so he added weight underneath it. A replica drawing has nowhere to put a sentence like that. The line that is most exactly the chisel-construct cycle, though, is an aside: the key-plate proportions come from a seventeenth-century virginals by Charles Rewallin in the Royal Albert Memorial Museum in Exeter — "Rewallin worked in the same part of Exeter as I do."

Three reasons seeing this now matters more than seeing it later. First, the document has not been filed yet. Written in January, surfaced by Hackaday on 6 September, twelve comments underneath. It is a PDF on a personal site built with EasySiteWizard 8.7, and that site is a one-man Musical Instrument Museum: sixteen rooms, three hundred and fifty instruments, the navigation a left-hand column reading Room 1 through Room 16. The site's TLS stopped negotiating with current browsers some time ago, and readers under the Hackaday post found that to read the thing at all they had to drop the s from https, or reach for wget. A remainder you have to downgrade your browser to touch. Second, every debt the document names is to an amateur, and every one of them is perishable: a blog called Renaissance Keyboards, Owen Daly's building notes hosted for him on someone else's dustyfeet.com, tabulatura.com, a few papers on the Urbino clavichord, and one long telephone conversation with the maker Alan Grove. None of that is in a library. The chain that made this instrument possible is a chain of personal websites, which is the kind of chain that goes dark overnight. Third, a slot is waiting. Clavichord building has a well-run construct right next door — the kit, the museum drawing. The moment this PDF is summarized into "a clavichord build guide," what gets dropped is precisely the living part: the reasons. And the deepest reason is sitting on the neighbouring page of the same site. The maker is a touch and haptics researcher at the University of Exeter who has spent a career on how skin reads vibration, and who has now spent five hundred hours building the one keyboard instrument in which the finger never stops touching the string. Nobody has named that connection yet. It is still a remainder. Go and read the reasons while they are still reasons.

Han Qin (秦汉) · @hqinjarsy · SAE Judgment & Aesthetics