The Painted Eye

Calibration Logbook — Astrometric & Occultation Station EYRIE-9 Keeper: S. Oort. Mission elapsed: 11 y 214 d. Heliocentric range: 49.7 AU. One-way light time to Network: 6 h 53 m.

---

Entry 8841 — 2089-03-02 — 04:11:09.226001 UTC (station clock)

The clock writes the time. I write everything else.

I should explain that, because someone will read this who isn't me, years from now, on a desk seven light-hours warmer than mine, and they'll need to know which parts of this log to trust. Every timestamp on every line was laid down by the maser. It is good to one part in ten to the sixteen. It does not get tired, it does not flinch, it does not care what happened in the hour it is stamping. The words after the timestamp are mine, and I am none of those things, so weigh them accordingly.

What happened in the hour it is stamping is this. At 03:52 a particle came in from somewhere that has been on its way since before there was a Network to point me here, and it went through the focal-plane detector and through the back of my left eye, more or less along the same line, which is a coincidence I'd have found funny twenty years ago. In the detector it killed a column of pixels. In my eye it made a flash — a hard blue-white streak, the kind we used to call cosmic-ray flashes and log as a curiosity. I've logged forty of them this month. The detector I can map around. The eye I cannot.

I'm telling the log because the log is the only one out here who'll know what the date meant before the date matters.

---

Entry 8842 — 2089-03-02 — 09:38:47.551approx

I've stopped trusting the last digits of my own clock-reading, so I'm writing approx when I read the display by eye instead of letting the system stamp the line. That's the first thing to say, and it's worse than the column of dead pixels.

The maser keeps perfect time. I am the part of EYRIE that does not.

The occultation is in ten days. I'll set it down plainly so there's no drama in it later: ten days from this stamp, the star catalogued as Network reference HD-eighty-eight-oh-four — old, quiet, exactly where it's said to be for a hundred years — will pass behind the dark companion we came all this way to weigh. The companion gives off nothing. We can't see it. We have never seen it. We know it's there the way you know a missing tooth is there, by the gap it makes in the things around it, and the only way to weigh it is to watch the rim of that star bend and dim as the dark thing crosses in front, and time the dimming against the maser, and read the shape of the curve.

The curve is buried in the seeing. To pull it out, every photon the telescope collects has to be measured against a fixed mark — a known position on the focal plane, accurate to the micron, that says this is where the star's image truly sits. That mark is the reticle. Twelve hairline crosslines etched and then filled by hand on a plate of fused silica, two microns wide, laid down so the centroiding software has a ruler that doesn't move when everything else does.

Mine moved.

---

Entry 8843 — 2089-03-02 — 14:02:11.application-noise-flagged

A dust-grain impact — micrometeoroid, fast, small, the station logged the strike on the radiator at 03:50, two minutes before the particle that took the pixels and the eye, and I'd been so busy with the eye I hadn't read the strike log until now. The impulse was nothing, a few microns per second, but it walked through the structure and reached the optical bench, and the bench rang, and the reticle plate, which has sat true for eleven years, settled into a new seat in its mount. I can see the offset in the flat-field. The reference mark is no longer where the reference says it is. About eleven microns out, and rotated a hair I haven't measured yet.

Eleven microns. I keep wanting that to be small. It's a tenth the width of a hair, a length you'd never see, and it's five times the error budget for the whole occultation. The companion's mass comes out of the dimming curve, and the dimming curve comes out of the centroids, and the centroids are measured against a ruler that is now eleven microns wrong, which would put the mass off by enough that it would be a number, filed, believed, and false. A false number is worse than no number. No number, you go and get again. A false one you build on.

So I'll fix the ruler. That's the job. That has always been the job. I've repainted reticles before — three times in eleven years, when the silica fogged. There are spare plates. There's the matched pigment. There's the micromanipulator and the comparator scope and forty years of a steady hand.

Two of those things have changed since the last time. I'll get to which two.

---

Entry 8844 — 2089-03-03 — 07:20:00.000 (system-paced)

Transmission drafted for the next Network window. Recording it here too, because by the time they hear it the thing it's about will be decided one way or the other, and the log should hold what I told them as well as what I did.

EYRIE-9 to Network Optical. Mission day four-two-three-three. Reporting a combined event from two days ago: micrometeoroid strike to the optical bench, station clock oh-three-fifty, displacing the fiducial reticle approximately eleven microns plus an unmeasured rotation; concurrent radiation damage to the focal-plane detector, one column, mappable. The reticle displacement exceeds the occultation error budget by a factor of five. I intend to repaint the reticle on the spare plate per procedure C-nine before the HD-eighty-eight-oh-four event in nine days. Advise if there is reason not to. Be aware your advice arrives, at the soonest, four days after I've had to start. Oort.

I didn't tell them about the eye. There's no procedure C-nine for the eye. And I know what they'd say — abort the manual repaint, let the automated fallback take over — and I want to set down, while my hand is still mine, why that isn't the mercy it sounds like.

EYRIE has an automated calibration arm. It's a fine machine. It does not paint. It can position the spare plate and it can lay a deposited line, but a vapour-deposited line on this silica spreads at the edges — four, five microns of uncertainty in the very thing the line exists to make certain — and the people who built this station knew that, and that is exactly why there's a human out here at all. I'm not the romance of the mission. I'm the spec. The hand was written into the design as the trustworthy part. A trained hand under the comparator can lay a filled line true to two microns where the machine can't get under five. That was the whole argument for sending a person across eleven years of dark. The machine keeps the time. The human keeps the ruler.

That was the argument. I'm about to test whether it still holds, on the one ruler that matters, with the one hand I've got.

---

Entry 8845 — 2089-03-03 — 19:44:53.approx

I measured my hand tonight, the way I'd measure any instrument I didn't fully trust, which is what it's become.

The comparator can track the tip of the brush against its own grid and read out the excursion. I clamped my forearm in the rest, the way I would for the real work, breathed the way I've breathed over fine work for forty years — out, hold the bottom of the breath, move on the stillness after the lungs are empty — and I drew a line I meant to be straight.

The excursion is four microns. On a good pass, with the arm clamped and the breath held and the lights low, the steadiest line I can lay wanders four microns off true.

The reticle spec is two.

I read that number three times. Four against two. I've spent my life being the two. I taught the two; there are people on three stations who centroid the sky against marks I laid down inside their tolerance. And tonight the comparator, which doesn't care and doesn't flatter, sat me down and showed me that the most trustworthy instrument on EYRIE-9, the one written into the design as the thing the machine couldn't match, wanders twice as far as the job allows.

It's the radiation. It's eleven years of it. It's the same dose that fogged the silica and killed the pixels and is quietly clouding the lens of my eye — it's been in my nerves and my muscles too, the whole time, the part of the exposure that doesn't make a flash you can log. The hand didn't fail at a moment I could timestamp. It failed across eleven years, seven light-hours from anyone, in a body the maser doesn't measure.

So. The two things that changed since the last repaint. My hand is one of them.

---

Entry 8846 — 2089-03-04 — 02:17:30.approx

The other is the eye, and tonight I'll write the eye, because I've put it off across three entries and the procedure I have to run uses it.

You paint a reticle by sight. The comparator magnifies, but the judgment of where the line wants to go, the reading of the etched groove you're filling, the catching of the moment the bead of pigment is about to bloom too wide — that's the eye against the eyepiece. My left has the cataract the dose has been growing for two years; it scatters a point of light into a smeared little sun now, useless for a groove two microns wide. My right is better. My right is what I've been working on for months without writing it down. But the flashes come through the right too, the streaks, and you can't time them, and a streak across the field at the wrong instant is a flinch, and a flinch at this scale is the whole error budget gone in the length of a particle's passage.

Here is the thing I keep arriving at and walking away from. The reticle is a painted eye. That's what it is, has always been — a human-made eye, lines laid down by hand to match the true geometry of the sky, so the telescope's blind machinery can be told where to look and trust the telling. I have spent my life painting the eye the instrument sees through. And the eye I paint with is going the way the eye I paint for went eleven microns out: slowly, from the same cause, on a curve I can read perfectly and can't bend.

I'm not writing this for pity. The log doesn't take pity and neither do I. I'm writing it because the next entries are going to be the work itself, and whoever reads them needs to know that the hand was at four and the eye was scattering and the flashes were coming, before they read what the hand and eye then did. Otherwise they'll read it as craft. It's not going to be craft. Craft is when the instrument's good. This is going to be something else.

---

Entry 8847 — 2089-03-05 — 11:09:48.approx

Inventory of the matched pigment, because scarcity is the other arithmetic and I've been avoiding it the way I avoided the eye.

The pigment isn't paint. It's a metal-loaded suspension formulated to a refractive index that disappears into the silica except where it must not, milled to a grain finer than the line is wide, and it does not exist between here and the inner system. They sent enough for the mission plus margin: enough for, the manifest said, a dozen repaints. I'm on my fourth. On paper that leaves eight. On paper.

The paper assumed two things that turned out false. First, that every repaint went right the first time — but a botched line can't be wiped and redone on the same plate, the wiping leaves a ghost the centroiding sees, so a bad line burns a whole plate and a whole charge of pigment, and across three repaints I've burned two plates getting one good one. Second, the manifest assumed the suspension would keep eleven years in a sealed cartridge, and it hasn't; the metal's settled and aggregated in most of the charges, milled too coarse now for a two-micron line, useless for anything but practice. When I sort the cartridges by what's still good, I have two spare plates and, if I'm honest about the loading, pigment for maybe three full reticles.

Three reticles. Four-micron hand. Two-micron spec. Seven days.

I could practise on glass that isn't a plate, and I will, but practice with the real pigment is the only practice that counts and the real pigment is the thing I can't spend. You see the trap. To get the hand back to two I'd need to paint and paint and read the comparator and paint again, and I don't have the pigment to paint and paint. The machine that doesn't care has set me a problem with no slack in it, which is the only kind of problem it knows how to set.

I sat with that a long time tonight. Long enough that the station's own small noises sorted themselves into a kind of order — the pump cycling, the reaction wheels trimming our slow tumble true, the maser ticking down toward the occultation it has known the time of since before I left home. Everything aboard keeps its tolerance but me.

---

Entry 8848 — 2089-03-06 — 05:55:12.approx

A reply came in the window — the first answer to the transmission I sent three days ago. Six hours and fifty-three minutes old, which is as fresh as anything gets out here; it's been crossing the dark since they sent it, and they sent it the moment my message reached them, before they'd had much time to think, which I don't blame them for, because out here thinking happens at the speed of the light between us and nothing they decide can reach me in less than seven hours.

Network Optical to EYRIE-9. Acknowledge bench strike and reticle displacement. Strong recommendation: do NOT attempt manual repaint. Deploy automated deposition per C-twelve and accept the wider fiducial uncertainty; we will model the residual. A wide ruler we can characterise. A repaint we cannot verify from here is an unknown we cannot. Repeat, do not hand-paint. Your call locally and we trust it, but on our side the numbers favour the machine. Stand by for —

The rest is the model they want me to run, and I've read it, and they're not wrong. That's what's hard. They're not wrong. A machine-laid line is five microns wide and bad, but it's known bad — they can sit seven light-hours away and subtract a five-micron blur they understand from the data and get a companion mass with honest, fat error bars on it. A line I lay by a four-micron hand and a scattering eye is, to them, a line of unknown badness, and unknown badness can't be subtracted, it can only be believed or not.

They think the machine is the conservative choice. From where they sit it is. They're trusting the design — and the design, eleven years ago, trusted my hand.

The whole mission turns on a number neither of us can see from where we are: how good is the hand, tonight, really. They've assumed it's gone, because that's the safe assumption from seven light-hours of ignorance. I'm the only one who can measure it. And the comparator says four.

Four is worse than the machine's five at honesty and better than it at the thing that matters, if I can prove the four, which I can't from inside my own arm. I can only prove it on a plate, with pigment I can't spend, against a clock that won't wait for the proof.

---

Entry 8849 — 2089-03-08 — 22:40:03.approx

Three days gone, four to the event, and I've done the only honest thing I could think of, which was to stop trying to be the keeper I was and start measuring the keeper I am.

I painted on practice glass — not the plates, the worthless cover slips, with the worthless practice ink that's the wrong index but the right viscosity. Forty lines. I read every one on the comparator and wrote the excursion. I'm not going to copy forty numbers here; the file's attached to this entry. The summary is this: my mean is four-point-one microns and it doesn't get better with warm-up the way it used to, it gets worse, because the small muscles tire now and the tiredness is faster than the warm-up. My best single line of the forty was two-point-three. My best was almost good enough, once, by luck, on glass that didn't cost anything.

There's a thing inside this I didn't expect, so I'll set it down. Around line thirty I caught myself painting the way I tell students not to — chasing the groove, correcting mid-stroke, fighting the wander. And every corrected line was worse than the wandering ones, because a correction is two errors where there'd been one. The hand that's failing fails least when you let it fail and don't argue. The straightest of the forty was one I laid in one breath without watching, trusting the forty years under the four-micron tremor, accepting that the tremor was there and not flinching from it.

I don't know yet if that's wisdom or just the only move left. The comparator doesn't grade wisdom. It graded two-point-three.

---

Entry 8850 — 2089-03-10 — 13:18:55.approx

Two days. I planned the real attempt tonight and then I did the arithmetic I'd been refusing, the one that isn't about the pigment or the hand but about what I'd be claiming if I won.

Say I paint the plate. Say I get a line and it reads, on the comparator, two-point-four — close, the best the hand has in it, better than the machine. I install it, I catch the occultation, I send the mass home with a two-micron error bar on the fiducial. They file it. They build on it. A textbook gets the companion's mass to four significant figures and my name in a footnote.

And the truth is the comparator that read my line two-point-four is read by the same eye and judged by the same nerves that laid the line. I can't measure my own ruler with a better ruler than my own hand made, because my hand made the best ruler on the station and the best ruler on the station is now me. The instrument that checks the work and the instrument that does the work have become the same failing instrument. I could lay a line that's truly five microns out and read it, honestly, sincerely, with a scattering eye and a trembling reading hand, as two — and never know. The error I'm most afraid of isn't the one in the line. It's the one in my reading of the line. The catch wouldn't be a measurement. It'd be a hope wearing a measurement's clothes, and I'd be the last person able to tell the difference, because I'd be the one wearing it.

That's the entry I didn't want to write. The machine's five-micron line is honest because everyone, including the machine, knows exactly how blind it is. My line's badness is the one badness in this whole station nobody can characterise — not them, not the comparator, not me.

---

Entry 8851 — 2089-03-12 — 03:30:00.000 (event minus 06:11:40, system-paced)

The star is where it's been for a hundred years. The dark thing is coming up on it now, the way it has every few years since before any of us, and it doesn't know we're weighing it and it wouldn't dim for us if it did. The maser is counting down to the dimming with that flat patience the universe keeps for everything that doesn't have to die.

I've made the decision, and I'm recording it as mine, in the order it came, because the order is the only honest thing I have left to give.

I painted the plate.

I laid one line in one breath, at the bottom of the lungs, in the stillness after, not watching, letting the four-micron hand do what forty years built into it and not arguing with the wander. Then I painted the other eleven the same way. I read them on the comparator with the better eye and got numbers between two-point-one and three-point-eight, and a mean I'm not going to dress up.

And then I did the thing the keeper I used to be would call desecration, the thing that's the opposite of everything a steady hand is for.

I didn't claim them.

I installed the plate, because a hand-laid ruler at three-and-change microns is still better than a vapour line at five, and a real measurement deserves the best ruler on the station even if the best ruler is a tired one. I'll catch the occultation. The data will come out of EYRIE clean and complete and I'll send every photon of it home.

But I'm not sending the fiducial as two microns. I'm sending it as what it honestly is: a hand-laid reticle, mean excursion three-point-one, laid by a keeper whose hand has been measured at four and whose reading eye scatters and whose check on her own work is the same failing instrument as the work — fiducial uncertainty, therefore, not two microns and not five but unknown above three, to be treated by the Network as a floor and not a value. I'm telling them the catch sits on a ruler I can't fully vouch for. I'm telling them the one thing the keeper they sent was sent precisely never to have to tell them, which is that the hand they trusted is no longer inside its tolerance, and that they should weigh the companion with my honesty folded into the error and not my pride.

They'll be able to use it. Fat error bars and known, you can use. That's the gift. Not the number — the size of my doubt about the number, measured and declared, so that whatever they build, they build knowing exactly how far down the trust goes.

---

Entry 8852 — 2089-03-12 — 09:41:53.approx

It dimmed.

Six hours ago the rim of an old star bent and faded by the amount a thing we'll never see told it to, and the maser caught every microsecond of the fade against a ruler I painted blind, in one breath, knowing it wasn't good enough to claim and laying it anyway because three honest microns serve the truth better than five and far better than a two I'd have had to invent.

I've sent it all. The curve, the photons, the plate's true error, my hand's true error, the eye. The transmission's seven hours out now, crossing the dark toward people who told me to trust the machine, carrying the news that the keeper is the part that failed and the measurement is good anyway, because we built it with a place for the keeper to fail honestly into.

The clock is already counting toward the next event. It'll come round in four years and some months, indifferent and exact, and the dark companion will cross the old star again, and EYRIE will be here, keeping its tolerance — the pumps, the wheels, the maser, the patient blind machine. Whether the keeper's hand is here to paint the ruler, or whether by then the only honest line is the wide machine-laid one with the badness everyone can see, I don't know. I've stopped being able to tell my best line from my hope.

So I'll write down the one thing I'm sure of and leave it where the next keeper, if there is one, will find it before the next event matters: the ruler was never the point. The honesty about the ruler was. The hand was always going to fail — every hand is, that's what hands are. The design was never really betting on a hand that wouldn't tremble. It was betting on a keeper who'd measure her own tremor and tell the truth about it across seven light-hours of dark, with nothing watching but a clock that doesn't care and a star that's about to dim.

The line is laid. It's not good enough. I've said so. The catch is real.

That's the whole report. The clock can have the timestamp. The rest is mine.