Main idea

Turning spoken thought into readable text is not a finishing touch. Continuous speech repeats itself, doubles back, mixes two ideas, gives an example before stating the problem. Grouping what belongs together, cutting the repetitions, rephrasing, finding the thread again: that is a heavy load, and a slog to redo by hand.

But that load is not the same thing as arbitration. Deciding that a nuance is missing, that a sentence sounds too certain, that a sequence rings false — that remains a decision, and it comes after the shaping, on its result.

Two kinds of work, then, and not one job done more or less well.

Why it matters

As long as the two are conflated, delegating the shaping looks like delegating the text. Once separated, the question becomes precise: you can hand over the first and keep the second.

It also restores the value of a task treated as secondary when it absorbs a large share of the effort — and which, when absent, leaves material nobody will reread.

Nuances and limits

The separation isn't watertight: cutting one sentence or moving another can change what the text asserts. That is why delegated shaping demands a reread, not a simple sign-off.

And excellent shaping can mask weak thinking — readability is no proof of solidity.

Open questions

• At what point does a rephrasing stop being shaping and become arbitration in disguise?