AGI Trilogy

My father was an umpire. Twenty-eight years in Triple-A — never called up, never let go — behind home plate in a league that travels by night bus.

As a boy I watched his games pressed against the backstop fence. One night on a road trip he told me a story. Three umpires sitting in a bar.

The first umpire laid out his philosophy. "I call them as they are."
The second came back: "I call 'em as I see 'em."
The third set down his glass and said, "They ain't nothin' till I call 'em."

I said the first one was right, obviously. A pitch has a location and the zone has dimensions. My father laughed and picked up his mask.

"A pitch is a thing you measure, son. A strike is a thing you call."

That sentence found me again thirty years later, in a boardroom.

I should say first which kind of man I became.

The summer I was fourteen, my father called an outside pitch a strike in a Double-A playoff game. Strike three, game over, the end of some twenty-eight-year-old's last season. The broadcast graphic drew the ball half a ball off the plate, and the next day at school the kids shoved the clip in my face. That evening at the dinner table I shouted. "The camera saw it, Dad!"

He answered without putting down his fork. "The camera saw the pitch. That night, that pitch was a strike."

I decided never to forgive that sentence as long as I lived. From that day on I was on the camera's side. I came west, learned measurement, and made a career of building graders. I joined the organization that builds verifiers better than anyone in the world, built verification infrastructure there too, and at thirty-nine became the youngest director on its board.

On my desk I kept a brass indicator, the same model as my father's. The counter an umpire grips in his left hand, thumbing the wheel to keep balls and strikes, small enough to hide in a palm. It costs next to nothing. It was a trophy from the world I had escaped.

By then the verdict on my father's industry was already in. Machine calls across the minor leagues, full stop. The tracking system was off by millimeters; men like my father were off by half a ball. I called that justice. Measurement beating judgment.

And in our company there was one room standing in the way of that direction.

There is a group of eight that meets once a quarter. Its official name is the Readiness Determination Group; around the company they call it the kitchen cabinet. They are the keepers of the question laid down in Article 7 of the founding charter — is it finished. The moment AGI is determined to be complete, the entire structure of governance and contracts reorganizes. And that determination is made not by a benchmark but by eight signatures.

To my eye, that was our company's home plate. In the middle of an organization that measures in millimeters, eyes that miss by half a ball sat holding the last call.

So from my first year on the board I brought the same motion. The wording I revised every quarter.

The first version, in '26, was full transfer. Let the system assess its own readiness. In '27, the second version took a step back: machine determination as the default, the group keeping only a veto. The third version, late that year, was the challenge system. The group makes the call; on dispute, appeal goes to the system's measurement.

The day I brought the third version, I cited my father's industry as precedent. Even baseball handed it over to the machines in the end. If a hundred-and-fifty-year-old institution could hand it over, what reason has a seven-year-old charter to refuse?

The chair raised his hand. A quiet man. Silver-haired, slow, with a habit of reading the charter aloud instead of rebutting. That day he did not read the charter. He asked me a question.

"You said they handed it over. Did you check? The major leagues never did. The machine measures, a human calls, and a dispute goes to the machine. Baseball handed over the measurement; it did not hand over the call. A hundred-and-fifty-year-old institution drew that line."

I checked the record. He was right. The verdict I had spent my life calling justice had come down only as far as the minors. At the top, a human still called. I was forty when I learned, from the rulebook of my father's own industry, that my precedent was evidence for the other side. That night, before sleep, a strange thought came to me. Why was my father in Triple-A for twenty-eight years? By measurement — by the evaluations that decide who gets hired — he had graded out major-league long ago. I knew that. The call simply never came.

I turned that thought off partway. I could tell where it went if I left it on.

I should say something about the chair's readings.

Whatever my motion was, his answer was the charter. Sentences the founders had written in the year the company was nothing. Every quarter he read Article 7 — the determination is signed by human beings, words to that effect — and the clauses around it, as if he were reading them for the first time. I hated that worse than losing. I wanted to lose argument against argument. A reading is not an argument, so there was no winning or losing against it, and so the one who changed was always me. My motion grew more refined version by version; his reading did not change by one letter. The asymmetry was driving me out of my mind.

Once, after a meeting, I followed him down the hallway and asked. Not to needle him — I honestly wanted to know.

"The reading, every time — isn't it a little… like worship?"

He answered without breaking stride. "My father was a pastor. A church of twenty. He used to say — a sacrament is valid regardless of the size of the congregation."

I decided that answer was not an answer. I went on thinking so for months.

In February of '28 I brought my last version. Let's not go to a vote; let's go to an experiment. Have the system write the memo — the very determination Article 7 calls for. If the draft is poor, I retire this motion for good.

Having no grounds to refuse, they allowed it.

The draft came in forty seconds.

Sixteen people read it for ten minutes without a word. It was flawless. The places where the determination group had run aground nineteen times in seven years — define it by capability and the arrival re-arrives every quarter; the definition of completion runs in a circle — it crossed in three sentences apiece. The evidence was exact, and the style — this was the part that hurt most — was humble. Reading it, I thought: I have won. This was the verdict I had spent fourteen years preparing. The verdict of the moment when measurement beats judgment.

I turned the last page and found the signature block. Eight lines, in the exact form Article 7 prescribes.

The eight lines were blank.

I was looking at those blank lines when I heard a sound. A sound that was not in the boardroom. A ball hitting a mitt. At fourteen, behind the backstop, I heard that sound thousands of times. The ball hits, and then there is half a second. The ball has already passed wherever it passed, the mitt has already popped, and the world does not yet know what happened. Half a second in which the measurement is over and the event has not begun. What ends that half second is not data. It is the throat of the man in the mask.

The perfect draft was the sound of the mitt. And the boardroom sat inside the half second. Because no one had called it, that document — the document that completed my life's argument — was not yet anything at all.

I withdrew my motion with my own hand. Several people asked why; at the time I could not explain. Now I can. In that room, I had understood the third umpire's sentence. Thirty years late, with the worst possible timing, by way of my experiment's success proving my premise's failure.

Understanding and conceding are two different things. I was raised a realist, and I was not going to concede on a feeling. I wanted to concede by measurement.

So that fall I went to the data side myself and asked. Not our own books — the world's. The aggregated delegation curves: across millions of accounts, the share of work executed without approval had been crawling up a log scale for quarters on end. Every account would have its inflection point. The threshold date when what had been asked stopped being asked. I asked them to pull the histogram of those dates. They would bunch up in some single week. That week would be the completion date. Then the determination becomes a discovery, discovery is work a machine can do, and I could pick back up the motion I had laid down.

The histogram was flat.

Millions of thresholds, spread evenly across three years. No announcement-day effect, no seasons. By weekday, Tuesdays ran a hair high — inside the error. I looked at that flat picture a long time. Printed on it was the most honest sentence a realist could ask to read: the completion date is not a singular noun. The date had passed long since — every one on a different day, without any announcement. Each account, behind its own home plate, had made its own call. On some Tuesday morning, three lines into a report, coffee brewing, somebody called a strike without a sound. Millions of umpires, on millions of dates.

The world was full of pitches already measured, waiting for the call. Our group's work was one call among them: the last one. Not the biggest call. The latest.

The determination group's twentieth draft was written by the chair. What cut the definitional clause that had run seven years of drafts aground was not an elegant sentence but a tense. This determination does not declare completion; it ratifies a completion that has already occurred, severally. The evidence section held no benchmarks. In went four delegation curves, and my flat histogram. I chewed a good while on the fact that, of my fourteen years, the one artifact to reach the determination was that picture.

The signing date came out to the last Tuesday of November, and the reason was the board calendar. I was not a signer, so I watched from a seat along the wall. The eight read Article 7 in turns, a clause apiece, and each one who read then signed with a fountain pen. There was no telling the reading from the signing — and the longer I watched, the clearer it became that there had never been a difference. Someone had brought champagne. Nobody opened it.

The week the announcement ran, I checked the news the way you read a box score. Our story sat sixth for three days. The most-shared commentary was headlined "They Have Been Announcing Completion for Three Years." As the man who had pulled the histogram, I could not work up a rebuttal. The stock moved two percent.

The next year, as the group was growing from eight to twelve, the chair called for me. The determination was done and the docket was exploding. Is this something the system should not decide for itself — after completion, that question had, if anything, bred faster. He asked me to take one of the four new seats. Offering the bottleneck's chair to the man who had campaigned to remove the bottleneck seemed to amuse him.

When I accepted, he asked me why. It was not a formal occasion, and I nearly gave a formal answer anyway, and then just told him the fact.

"Turns out output doesn't come with a signature line. Even the perfect draft couldn't fill in its own."

At the first meeting, the chair circulated a letter from Hangzhou. A letter from an evaluation researcher who hoped to talk someday about the grading of grading, and who was building a windowless room. Which door would be the last left open in a closing age — the hands passing that letter around the table seemed to know.

In December I went to Toledo.

My father is retired, and on weekends he puts on the mask for a college league, unpaid. In front of kids from the generation raised on machine calls, an umpire of the last generation to call with the throat. After the game, in the parking lot, he asked. So — that thing at your company. Did it come?

I answered. It came. Except nobody knows it came. There's no date, either. We counted: it's in the millions. All on different days.

He nodded, tossing his chest protector into the trunk, and there was so little surprise in it that the surprise was mine.

"It ain't nothin' till you call it," my father said. "In our business, that's been settled doctrine for a hundred and fifty years."

On the drive back I thought about the indicator on my desk. The thing I had been calling a trophy. At the next quarter's first reading, I took it along in my pocket. Not that I meant to count anything. Just — a tool follows the trade of the hand that holds it. I think I heard that somewhere.