# The Prompt Lab

> Prompts that actually work, and the ones that don't.

Canonical: https://promptway.com/prompt-lab

## What this is

The Prompt Lab is a working bench, not a lecture hall. We treat prompts the way a kitchen treats recipes: write it down, run it, taste it, fix it, run it again. Every piece here ends with something a reader can paste into a chat window tonight and judge for themselves by tomorrow morning.

We exist because most prompt writing online is either marketing for a tool or someone repeating a tip they read three weeks ago. Our readers are operators, writers, marketers, and small teams who use AI as a daily instrument. They want to know what changed in the output when you moved one sentence, not what "prompt engineering" means in the abstract.

What makes us different: we publish the diffs. The bad first draft sits next to the working version, and we explain why one outperformed the other. If we cannot reproduce a result, we say so and move on.

## What we cover

- Short, single-task prompts and why their shape matters
- Voice-matching: getting a model to sound like a specific writer or brand
- Structured output patterns (JSON, tables, briefs) that survive real workflows
- Prompt chains that hand off cleanly between steps
- Audits of public prompts, including our own past misses
- Side-by-side rewrites with the reasoning between them

## What we don't cover

- Theory papers and academic taxonomy of prompting
- Vibes pieces about how AI "feels" to use
- AI ethics essays, which belong somewhere with more patience
- "10 secret prompts the pros don't want you to know" listicles

## Articles

- [Structured Output That Survives a Model Swap: The JSON Scaffold I Actually Ship](https://promptway.com/blog/json-scaffold-survives-model-swap): Multi-model routing is normal now. Here is the schema-first prompt pattern that keeps objects valid when you change the brain underneath.
- [The 12-Prompt Eval I Run Before I Trust Any Model Upgrade](https://promptway.com/blog/twelve-prompt-eval-before-model-upgrade): A model wave is not a migration plan. Here is the fixed task set I score before I change a default, with a pass/fail rubric you can steal.
- [Upgrade Your Prompt Stack for Sonnet 5 and GPT-5.6 Without Rewriting Everything](https://promptway.com/blog/upgrade-prompts-for-sonnet-5-gpt-56): New models land and teams thrash their prompts from zero. Here is the lab method: keep the scaffold, re-test three failure modes, ship the diffs.
- [The De-Slop Prompt Stack: Six Prompts That Stop Claude and ChatGPT From Sounding Like a Robot](https://promptway.com/blog/the-de-slop-prompt-stack): Readers now punish writing that feels machine-made. These six prompts remove the tells, and they're built to be saved once and used forever.
- [The Constraint Goes First](https://promptway.com/blog/the-constraint-goes-first): Everyone writes prompts top-down: role, context, task, then a stack of rules at the end. That's the bug.