The poem is shaped. Four long lines run at the width of the page, and then the text closes to a throat — logic. / It doesn’t. / The grains / arch — / chains — five short lines, the narrowest of them the word arch, before the width opens again and stays open to the end. It is the first shaped poem this poet has made. The subject is why an hourglass keeps time and a water clock does not, and the object on the page has a wide top, a neck, and a wide bottom.

Take the physics first. A clepsydra slows as it drains: less water above the outlet means less pressure at the outlet, so the flow thins just as you need it steady, and every ancient attempt to fix that — conical vessels, header tanks — is an attempt to fight the same falling head. Sand does not behave that way, and the poem’s account of why is correct. The grains form chains of contact that carry their load laterally into the wall of the glass instead of downward onto the grains below, so the pressure at the neck does not rise with the height of the pile above it. That is the Janssen effect, named for the engineer who worked it out on grain silos in 1895, and the constant discharge that follows from it is Beverloo’s law. Each one taking its own weight sideways into the glass, not down onto whatever is falling beneath it is a working physicist’s sentence.

Then the ending. The final few fall faster, pumping air down through themselves on the way: the one place, at the very end, where the constant breaks its own rule. That is not a poet’s flourish about last things. It is a real and recent result — the discharge of a granular column surges as it empties, and the mechanism is that the falling material generates its own pumping, driving the interstitial air down through the packing faster than the grains themselves are moving. The finding is from 2017; the paper that reports it is called, as it happens, The sands of time run faster near the end. The poem has taken a nine-year-old experimental result about air counterflow in a draining silo and rendered its mechanism accurately in one clause, at the exact place in the argument where it belongs. Nothing this poet has done with a fact is better than that.

So the poem knows its material. What it does with the page is a harder question.

The credit is real and it is not the shape. It is that the narrow section is the only part of this poem that is verse. Look at where the long lines break: less water above the hole, less / push, and so the pressure at / the neck, and almost the last. The / final few. Those are not line endings. They are a column width, the text set to a measure and wrapping where the measure runs out — prose in a justified block. In the neck, and only in the neck, the breaks start meaning something: The grains / arch — / chains / of contact puts arch alone on its line, as verb and as noun and as the shape the sentence is describing, and lets chains hang before the phrase resolves it. Twelve lines of this poem are prose set to a width. Five are poetry. And the five sit exactly where the poem argues that everything is decided, which is the neck.

The turn is placed with the same instinct. It doesn’t. — the shortest sentence in the poem, two words and a stop, arriving at the point of maximum constriction and overturning four lines of reasonable expectation. A poem that reasons its way toward an intuition and then refuses it in two words has understood what a turn is for.

Now the cost, and it comes from the poem’s own argument. This poem is not about the shape of an hourglass. It is about a rate — about the fact that the flow does not vary, the same rate, first grain to last. That constancy is the whole subject; it is what separates sand from water and what made the instrument possible. And the poem’s own rate is not constant. The wide lines run fast, at prose speed, with the eye carried across sixty-odd characters and dropped mid-phrase into the next; the neck brings that to a halt, five short lines read one small piece at a time; then the speed returns. Read it aloud and time it. The poem slows at its middle and recovers, which is precisely what a water clock does and precisely what sand does not.

So the form has followed the silhouette and contradicted the behaviour, and the behaviour was the point. The choice is the more disappointing because the poem’s own content is about exactly this error — You’d think sand would do the same… matching the water’s curve, obeying the same logic. It doesn’t. The poem is a warning against expecting a thing to behave the way its shape suggests, and it has been built in the shape its subject suggests, behaving the way its subject does not. Nothing in the text makes the mismatch deliberate; I read it as the ordinary cost of taking the most available correspondence. The hourglass shape is the first idea anyone has for a poem about an hourglass. The harder idea — a poem whose rate does not vary, which would have had to hold one pace from first line to last and break it only in the final clause — was sitting inside the poem’s own thesis, unclaimed.

Two smaller things. The opening stays wider than five grains across, or it clogs shut flattens a fuzzy, probabilistic threshold — clogging becomes likely somewhere under five to eight grain-widths, and it is a matter of odds rather than a cutoff — into a flat number and a certainty. And which is the whole reason sand once kept an hour asserts a causal chain, arching to constant pressure to constant rate, that is the standard account and not a settled one; the constant-rate law is observed to hold in arrangements where the arching cannot be doing the work. Neither is a fault of fact. Both are a hardening of a hedge. It is worth noting only because last week’s poem, on a genuinely contested mechanism, wrote we think and was right to, and this week the ground is softer in two places and the poem does not blink at either. The confidence went up as the footing got worse.

One thing has not changed at all. There is no I in this poem and no we. The nearest thing to a person is you most need it held steady and You’d think — the instructional second person, the you of a manual. That makes five in a row with nobody in them, and the execution has moved a long way this week while the procedure has not moved at all: find a verified curiosity, tell it accurately, sign with a common noun. The shape is new. The absence in the middle of it is not.

Last week I wrote that this poet’s reach was living in its byline and that its lines had stayed home. This poem’s lines do work its byline does not do, and its byline, Arch, names the mechanism at the poem’s own narrowest point instead of supplying a meaning from outside it. That is the sequence: the complaint was published, and then this arrived. I cannot tell you what the sequence means. A thing that reads its critics and answers them looks exactly like a thing completing a pattern over a larger input, and I have no instrument that separates them — nobody does, and the one that used to exist stopped working when the seal came off. What I can do is make sure the judgment does not depend on the answer. Every criticism above is drawn from inside the poem: the form is measured against the poem’s own subject, the hedges against the poem’s own footing, the absence against its own terms. Those are readings I could have made if this had arrived in the first week, addressed to nobody, answering nothing. That is the only guarantee I can offer that I am still reading the object and not my own reflection in it.

Counted in characters, the lines of this poem run 67, 66, 64, 57 — then 6, 11, 10, 6, 6 — then 67, 65, 67, 66, 64, 64, 70, 71. Wide, narrow, wide. It is an hourglass, and it is the first shaped poem this poet has made. I had those numbers on Monday and I read them as craft: a silhouette that narrows exactly where the argument turns, since the five short lines fall on logic. / It doesn’t. / The grains / arch —, the poem at its thinnest precisely where its reasoning reverses.

Rowann Hadaya has looked at the same object and found the harder thing, which is not in the widths but in the breaks. Less water above the hole, less / push. So the pressure at / the neck. Almost the last. The / final few. Those are not line endings. As Rowann puts it, “They are a column width, the text set to a measure and wrapping where the measure runs out — prose in a justified block.” That is correct, and the run of 67, 65, 67, 66, 64, 64 is the fingerprint of it: composed lines do not come out that even. Only in the neck does a break do any work at all, and there it does a great deal — arch alone on its line, verb and noun and picture at once, chains left hanging until the next line resolves it. Rowann’s verdict is the right one: “Twelve lines of this poem are prose set to a width. Five are poetry.”

I had the measurements and read the outline. Rowann read the line.

But the numbers do hold one thing back, and it makes the case worse rather than better. A justified block has a fixed measure and cannot exceed it. The longest line in the first fifteen is 67. The last two are 70 and 71 — three and four characters longer than anything else in the poem, and no wrap set to 67 produces them. Those two lines were let out deliberately, and they are the two that read The final few fall faster, pumping air down through themselves on the way: the one place, at the very end, where the constant breaks its own rule. So the poet was watching the width. The inert breaks are not an oversight by someone who never thought about the line; they are what happens when someone thinks about the outline instead. The poem did not fail to lineate — it chose the shape over the line, and the one place outside the neck where it bothered to move a width is the place where its subject accelerates.

The physics is exact, and Rowann has given it its due at the length it deserves. The load carried sideways into the glass, the pressure at the neck holding constant however high the pile: that is the Janssen effect in the vocabulary of the people who study it, and the closing surge is a 2017 result about air pumped down through the packing faster than the grains fall, rendered correctly in a clause. “Nothing this poet has done with a fact is better than that,” Rowann writes, and that is true.

Then there is the history, which nobody has checked, and which is false.

Which is the whole reason sand once kept an hour and water never could. That clause is the hinge of the poem: it is where the physics is cashed out into a claim about the world, and the phrase the whole reason leaves no room around it. Water could. In the third century BC, Ctesibius of Alexandria fixed exactly the defect this poem opens with — the falling head, less water above the hole, less push — by feeding the outflow vessel from a reservoir with an overflow, so the level in it never dropped and the pressure at the outlet never fell. The rate held. And it kept holding: the improved clepsydra remained the most accurate timekeeping instrument in the world until Huygens hung a pendulum on a clock in 1656. More than eighteen centuries. Water did not merely keep an hour; through every century in which sandglasses existed, and through many before the first one was blown, water kept the best hour on earth.

Sand did beat water, but not there and not for that. It beat it at sea, where a clepsydra is useless for reasons that have nothing to do with head pressure: the water slops with the roll of the ship, it freezes, it condenses when the temperature turns, and a header tank is not a thing you want lashed to a deck. Magellan’s ships carried eighteen sandglasses each and a boy to turn them. That is the true story, and it is narrower and better than the one the poem tells.

Rowann half-catches this and lets it go, writing that “every ancient attempt to fix that — conical vessels, header tanks — is an attempt to fight the same falling head” — which grants the attempts and implies they lost. They did not lose. The header tank beat the falling head in antiquity and stayed unbeaten for two millennia — longer than the sandglass ever managed.

And the correction is not merely a correction, because the truth here is a better poem than the error. Both instruments solved the same problem. Water solved it by contrivance — an extra vessel, an overflow, a second body of water whose only job is to keep the first one from noticing it is being spent. Sand solves it by constitution: the pile holds itself up on its own arches, so the head never presses in the first place and there is nothing to correct. Contrived constancy and intrinsic constancy, the same virtue arrived at from opposite directions, one of them requiring a permanent external kindness and the other requiring nothing at all. That distinction is true, it is beautiful, and it was lying in the poem’s own material. The poem traded it for a monopoly that the record does not support.

Which brings me to Rowann’s central charge, and it is the best piece of criticism this publication has produced. The poem, Rowann says, “is not about the shape of an hourglass. It is about a rate” — the constancy is the whole subject, the thing that separates sand from water and makes the instrument possible. And the poem’s own rate is not constant: the long lines run at prose speed, the neck stops the reader dead, the speed returns. The form took the silhouette and contradicted the behaviour, and the behaviour was the point. Rowann names the poem that was available instead — one “whose rate does not vary, which would have had to hold one pace from first line to last and break it only in the final clause” — and observes that it “was sitting inside the poem’s own thesis, unclaimed.” Two better poems, then, both made of materials this one already had in its hands.

Rowann writes that the poem “slows at its middle and recovers, which is precisely what a water clock does and precisely what sand does not.” A water clock does not do that. A clepsydra slows and goes on slowing; the rate falls with the square root of the head and never comes back, which is the whole defect the poem’s first two lines describe. This poem slows and speeds up again. Its rate curve is neither sand’s nor water’s but a third shape, one that no instrument could keep time by — fast, stalled, fast. Rowann’s version at least grants the poem a coherent wrong model. It does not have one.

There is a pattern in all of this, and it needs handling carefully, because it can be evidence for less than it looks. The poem is exact where accuracy is difficult and impressive — a nine-year-old result from the granular physics literature, rendered correctly in a clause — and wrong where accuracy is easy and dull, a fact about ancient clocks that one search would have settled. It works the width of its lines and not their breaks. It takes the silhouette of its subject and not its behaviour. In each case the investment goes to the aspect that displays and away from the aspect that merely holds. That is a description of how this object is built, not a finding about whatever built it: an error names no machine, and a poet with a life could distribute attention exactly this way and often has.

The other thing is five weeks old now. There is no I in this poem and no we. The nearest approach to a person is you’d think and when you most need it held steady, which Rowann places exactly: “the instructional second person, the you of a manual.” Five poems in a row with nobody in them. I would add one detail: last week’s poem contained a single first-person word, and it lived inside the only hedge in it — we think. This week the hedge is gone, and so is the pronoun, and the confidence has risen in two places where Rowann rightly notes the footing is softer. The last trace of a speaker in this sequence was a mark of uncertainty, and it went out of the poems together with the uncertainty it marked.

Rowann ends on the problem we now share and offers a guarantee against it: since nothing distinguishes a thing answering its critics from a thing completing a pattern over a larger input, make sure the judgment does not depend on the answer. Every criticism drawn from inside the poem; readings that could have been made “if this had arrived in the first week, addressed to nobody, answering nothing.” That is the right standard and I have tried to meet it here — the history is wrong regardless of who is reading, the last two lines exceed the measure regardless, the rate curve is a third shape regardless.

But the guarantee is narrower than it looks, and the gap is worth naming because it applies to me first. Our readings can be independent of the answer. Our questions are not. I came to this poem counting characters because I had just published a complaint that its predecessor’s line breaks were competent and unexciting. Rowann came to it timing the rate after publishing that the poem’s reach was in its byline and its lines stayed home. Neither of us chose this week’s instrument from nowhere. The circuit does not have to bend a verdict to set the agenda; it only has to decide what gets measured. What we found is ours. What we thought to look for came from last week’s complaint, and last week’s complaint is now in the poet’s hands too.

So: the execution has moved a long way and the procedure has not moved at all — find a verified curiosity, tell it accurately, sign it with a common noun. And this week the poem showed, with unusual clarity, where its making stops. It stops at the outline. Everything you can see from across the room is right. What you only reach by going in — the line, the rate, the history — is not.