The cost curve changes everything

As soon as prompts cost real money at scale, elegance stopped being aesthetic and became financial. A long prompt wasn’t “thorough”—it was a recurring line item. So optimization techniques emerged fast: compress labels, remove redundancy, batch tasks, and use smaller models wherever possible.

What’s easy to miss is that “cost work” also improves quality. Tight formats reduce ambiguity. Batch outputs force consistency. Smaller models punish vagueness, which is a weird way of saying they teach you to specify better.

A lot of modern prompt craft comes from that era’s pressure: if you can cut tokens and get better answers, you’ve found the sweet spot—clarity as a cost-saving strategy.