# Opening feedback: Token Math

**Strongest idea:** Every’s CEO uses more than three times as many OpenAI tokens as anyone else at the company, yet his response is to ask whether colleagues are experimenting ambitiously enough. Randy’s 4.5 billion token attempt to make a 3D face gives that tension a vivid test: the result was “very janky,” but the audit exposed inefficiencies and led to a leaner setup and a benchmark. The piece’s useful rule is to make room for experiments, then examine their cost, results, and lessons. It does not establish a dollar return on Randy’s run.

**Framing and wording:** “Token Math” and “How Many Tokens are Too Many?” are broad; neither gives readers the surprising claim. “What’s the ROI on experimentation?” promises a calculation the draft does not provide. The alternate dek is closer to the theme but is long and leads with “tokenmaxxing,” which needs explanation. The opening’s “Often if … you might” delays the hook, while “highest spender” treats token volume as dollar spend. Keep the comparison explicitly about **OpenAI tokens**, and make clear that Dan is inviting ambitious work, not indiscriminate spending.

**Suggested replacement**

**HED:** Why Our CEO Wants Us to Burn More Tokens

**DEK:** A 4.5 billion token misfire shows what Every asks of an AI experiment: results, lessons, and a plan for a leaner next attempt.

**Opening paragraph:** Every CEO Dan Shipper uses more than three times as many OpenAI tokens as anyone else at the company. His reaction to the gap: Maybe the rest of us aren’t trying ambitious enough things with AI.

The short opening lets the next paragraph supply the essential qualification that this is not tokenmaxxing for sport. The dek’s 4.5 billion token example pays off in the later Randy section; moving that example up would make for a faster narrative opening if the article’s structure changes.