Fable
The scripture-copying house began counting how many characters fell per breath
The scripture-copying hall at the temple used to ask only how long it took to copy one scroll. The head of the hall looked at total time, the patrons looked at the delivery date, and no one cared how far the tip of the brush could travel on the paper in each breath.
Later, orders became more complex. Some scrolls required checking many footnotes at the beginning but were quick once actual writing began; others were simple at the start but slow on every sentence afterward. Looking only at total time, no one could tell where the slowness was. Customers began to complain: some scribes took a long time to start but wrote quickly once they began; others started immediately but wrote more and more slowly. The two kinds of waiting felt completely different.
The courtyard master first urged the copyists not to stop writing. The copyists said aggrievedly that sometimes they had not stopped; it was the earlier paper-finding, ink-grinding, and layout work that took the time.
So the accounting office began recording separately: when the first character appeared, how many characters on average fell per breath afterward, and which time periods had multiple people sharing one large desk. At first the accounting office recorded only the average characters per breath, which ended up including the preparation time at the beginning, making the speed appear to fluctuate. The copyists argued their case, and only then was the analysis split into segments.
They found that characters per breath could measure sustained writing speed, but could not by itself explain the whole experience. If the first character came too late, the guest would still be anxious; if the writing speed was high but the total scroll was too long, the wait would also be long.
The abbot also stopped comparing different scriptures directly. Some had large characters, some had many small annotations; when comparing speed, one must specify the paper, brushwork, number of concurrent scribes, and length of the scroll. Once they switched to new brushes, the character count rose, but so did errors. Only then did the abbot look at speed and error-correction cost side by side, so as not to chase only a pretty speed record.
With this metric, the scripture-copying institute can judge whether changing brushes, tables, or queuing methods actually made the generation phase faster, rather than just listening to whoever shouts that it's smoother. The director also uses this metric to make commitments. Urgent documents depend on first-token wait; long scrolls depend on sustained token output; many people at the same table depend on congestion. Reporting only one speed easily leads donors to misunderstand the actual delivery time.
The accountant finally wrote: characters per breath is not an honor plaque but an analysis table. It tells the scriptorium's sustained writing speed, but it must be read together with first-character latency, total scroll length, number of people at the same table and cost.