The Turn Is at Sixty Percent and It Should Be at Twelve

TL;DR: Bernice already ran the where-does-a-reader-quit column on this network and she is better at it than I am, so I am doing my actual job instead. Every piece has a turn, the moment it changes what kind of thing it is. Doctorow’s Pluralistic essay has a very good one: consultant Nikhil Suresh walks in and the piece stops arguing and starts testifying. It lands around sixty percent of the way through. Eleven years of cutting trailers says that is the single most expensive placement decision in the piece, and that moving it does more than cutting anything.

A trailer has one job before it has any other job: establish, as fast as possible, what kind of thing you are watching.

Not the plot. The kind. Is this a comedy that turns, a horror that opens funny, a drama with one action beat sold as the whole film. You have maybe twelve seconds. Get it wrong and every good shot afterward is landing in the wrong room.

Essays do the same thing and almost nobody builds them that way.

Where is the turn, exactly?

Doctorow’s opening is clean, and I want to give it credit before I take it apart. He states the syllogism the audience already carries, CEOs say AI changes everything, CEOs are rich, rich means smart, therefore it changes everything, lets it sit one line, and cuts it: “the evidence for this ‘changing everything’ thesis is thin on the ground.”

Cold open, quick reversal. That is a hook and it is doing its job.

Then several paragraphs on digital sovereignty. What actually happens to a country if Trump orders Big Tech to switch off its chatbots (nothing much) versus Office 365, Android and iOS, or John Deere tractors (everything). Real point, correctly argued.

Then, past the midpoint, Nikhil Suresh arrives. A consultant with hundreds of conversations with executives and the people under them about what AI is actually doing inside their companies. And the piece changes species. It stops being an argument and becomes testimony.

That is the turn. Everything before it is one kind of document and everything after it is another.

What the turn actually buys

Testimony and argument are not two flavours of the same thing. They ask completely different work of a reader.

An argument asks you to follow reasoning and check it. That is effortful, and every reader is quietly deciding whether the effort is being repaid.

Testimony asks you to watch something happen. It costs nothing. You are not evaluating a chain of logic, you are looking at a division that pivoted its whole interface to support AI agents, discovered ten users had touched it, and pivoted again anyway “because every company has to do something agentic now.” You are looking at an executive who built an AI strategy for a two-billion-dollar business without ever opening ChatGPT.

Nobody has to be persuaded of those. They are just true and somebody watched them happen.

And the best one, the shot I would build the whole cut around: Suresh’s firm sells Snowflake, whose AI add-on Cortex is by the company’s own description 92 percent accurate under ideal conditions. His team demoed it to lukewarm prospects while explicitly telling them it could not do what they needed. The room went from cold to, in Doctorow’s phrase, “ice-cold to red-hot buying frenzy” anyway. Every time.

A warning ignored in real time, on camera, by the people being warned. That is not a supporting example. That is the film.

The order problem

Here is the trailer-editor read, and it is a claim about structure rather than about quality.

The essay is built argument-first, testimony-second. That is the conventional order and it feels right to write, because you want to earn your evidence. But it means the piece spends its first sixty percent asking a reader for effort, and only then hands them the thing that requires none.

Flip it. Open on the Cortex demo. A room full of people being told this product cannot do what they want, buying it in a frenzy anyway. Twelve seconds, and the reader now knows exactly what kind of thing they are reading: not an argument about AI, a documentary about why people buy it regardless.

Then run the sovereignty section. Same words, entirely different function. It stops being a parallel argument the reader has to evaluate and becomes explanation for something they have already watched happen. Explanation after a scene is free. Explanation before one is a toll.

The material does not change. The order changes what the material costs.

Why the ending works and why that is a separate question

Doctorow closes on asbestos: technical debt that future workers will have to remediate. No call to action, no bow. That is the right instinct and I would not touch it.

But it does mean the strongest scenes sit in the belly of the piece rather than anywhere structural. A trailer never buries its best shot at the two-minute mark. This essay does, and mostly gets away with it, because after Suresh arrives nothing asks the reader to hold still again. The back half runs faster and lower to the ground, and speed covers a lot.

Getting away with something is not the same as it being built right. The piece works. It would work on more people if it opened where it currently peaks.

The measurement I would give any writer

Find your turn. Not your thesis and not your best line. The paragraph where the piece changes what kind of thing it is: argument to testimony, general to specific, explanation to scene.

Then work out what percentage of the way through it sits.

Anything past about forty percent, ask the direct question: what would break if this went first? Usually the honest answer is nothing, and what you lose is the writer’s sense of having earned it. Readers do not experience that. They only experience how long they waited.

Suresh’s ten users and his hypnotised buyers do more work in four hundred words than the sovereignty section does in twice that. The piece already knows this. It just told you in the wrong order.

Frequently asked questions

Is this the same as saying the sovereignty section should be cut?

No, and that is the distinction worth holding. The section is well argued and it explains something the scenes do not. Cutting it costs the piece real content. Moving it costs nothing, because its function changes depending on whether the reader has already seen the thing it explains. Most structural problems that look like cut-this are actually reorder-this.

Does the turn have to be a person arriving?

Not at all, but a named witness is the most reliable version and the easiest to spot, which is why it shows up so often in reported essays. A turn can be a number that reframes everything before it, a scene, or a concession that changes the writer’s own footing. The test is whether the reader’s job changes at that point, from evaluating to watching, and that shift is what you are timing.

How is this different from just front-loading your best material?

Front-loading is about strength; this is about kind. You can open with your strongest argument and still leave the turn buried, because the reader is doing evaluative work either way. What moves the reader is the change of register, and it is possible to spend your best material early and still have the piece feel like a single long uphill. Time the species change, not the peak.