# Editorial feedback

The strongest idea is the counterintuitive one: Every’s heaviest token user does not see his lead as a reason to cut back; he sees it as a sign that the team may not be experimenting ambitiously enough. The rest of the piece adds the important qualification: high usage is defensible when an experiment produces useful work or learning, and it should be audited afterward so the next attempt is more efficient. That “experiment first, do the token math second” sequence is sharper than a general question about ROI.

## Recommended package

**HED: Spend First. Do the Token Math Later.**

**DEK: Every’s top token user consumes more than three times as many tokens as the runner-up. His argument: Try ambitious ideas, then ask whether the results justified the bill.**

**OPENING: Every CEO Dan Shipper uses more than three times as many OpenAI tokens as the company’s next-highest user. His reaction isn’t to rein himself in. It’s to wonder whether the rest of the team is experimenting ambitiously enough with AI.**

This version puts the article’s argument in the headline, lets the dek supply the striking evidence and qualification, and gives the opening a clean reversal. It also flows naturally into the next paragraph’s clarification that the goal is not “torching tokens for sport.” “Justified the bill” covers both productive successes and useful failures; the draft shows that learning can itself be a return.

## Notes on the current copy

- **“Token Math” is clean but too broad.** It names the topic without expressing the surprising claim. It could describe a pricing explainer, a usage calculator, or a cost-control piece.
- **“How Many Tokens are Too Many?” is more inviting, but still generic.** The story does not establish a universal threshold; its answer is contextual and depends on what a run produces or teaches.
- **“What’s the ROI on experimentation?” makes the piece sound more conventional than it is.** The compelling point is that efficiency should not choke off exploration prematurely. The question also leaves both the subject and the promised answer vague.
- **The alternate dek has energy but carries too much jargon.** “Tokenmaxxing” is explained later and works better there. “When expensive experimentation is worth it” promises criteria, while the draft offers a philosophy and examples rather than a firm decision framework. “Rein your tokenmaxxing in” is also a strained construction.
- **The opening’s first sentence is hesitant and cumbersome.** “Often,” “if,” “might think,” and “about” stack qualifications before the reader reaches the point. “Highest spender” can also imply dollar spend, whereas the evidence given is token usage.
- **The paragraph repeats rather than escalates.** Its first two sentences make the reversal abstractly, and the third and fourth make it again with the actual evidence. Starting with the three-times figure lets the reaction supply the turn.
- **“Everyone else should consider ramping up” is slightly broader than the evidence supports.** The draft says the gap makes Dan wonder whether colleagues are ambitious enough; the replacement retains that attribution instead of converting it into a universal prescription.

## Bolder alternative

**HED: You’re Not Spending Enough on AI**

**DEK: Every’s top token user consumes more than three times as many tokens as the runner-up—and thinks the bigger risk is experimenting too little.**

This is the strongest email-subject-line option, but it deliberately generalizes from Dan’s view of Every’s team. The dek restores the attribution; the body should keep the existing qualifications about reviewing results, learning from failed runs, and choosing cheaper models when they are sufficient.