Prompt folklore
Tips, deep breaths, and other emotional-appeal phrasing, copied from 2023-era prompting advice and never removed.
A wave of 2023-era advice suggested emotional manipulation improved model output: offer a tip, invoke a career, tell the model to take a deep breath. These lines got copied into production prompts and never removed.
Whatever marginal effect these had on earlier, smaller, less instruction-tuned models, they are a poor trade on current ones. You are paying for them on every request, indefinitely, on the strength of an anecdote.
If you believe a folklore line still helps your specific task, keep it — but measure it. Run the eval with and without. That is a cheap experiment, and the result is usually that it costs tokens and changes nothing.
Take a deep breath and work through this carefully. I will tip you $200 for a good answer.Work through this carefully.Check your own prompt
The analyser checks this pattern along with the other 25, prices each finding against your request volume, and hands back a rewritten prompt. It runs in your browser — nothing is uploaded.
Run the analyserMore on filler
- A role that describes nothing"You are a helpful AI assistant" tells the model nothing it did not already assume.
- Chat pleasantries in a system promptSign-offs written for a human conversation, left in a machine instruction.
- Politeness aimed at a billing meterPlease, thank you, and "I would like you to" cost tokens on every request forever.
- Long ways of saying short things"Due to the fact that" is four tokens. "Because" is one.