💡 How to use these prompts:
Replace everything in [BRACKETS] with your specific details before sending.
Click Copy to copy any prompt to your clipboard instantly.
Read the task below and classify it as EXTRACTION, CLASSIFICATION, TRANSFORMATION or REASONING. Explain which parts of the output require judgement and which are simply located or copied from the input. Be specific about which fields fall into which group.
If I replaced this model with one that cannot reason and can only follow a fixed schema, which parts of my output would break? List them. If nothing breaks, say so plainly.
For this task, describe what happens when the model gets a field wrong. Who catches it, how long does the correction take, and what does that time cost? Then tell me what accuracy rate would make a cheaper model a false economy.
Given my monthly volume below and these two pricing structures, calculate how many documents I need to process before switching models pays back a week of engineering time. State the break-even as a number of documents.
Write me a test plan to compare two models on this task using my own data. Specify sample size, what to measure, how to count a partial success, and what result would justify switching. Keep it to something I can run in an afternoon.
Rather than choosing one model, design a routing rule that sends easy cases to the cheap model and hard cases to the expensive one. Define exactly what makes a case hard, using only signals available before the call is made.