Prototyping With Words

You've asked AI to write a document for you and what it produces is...okay? Not terrible, but also not what you wanted, and you can't quite articulate what's wrong. You tweak a few words in the prompt, run it again, and hope for the best.

You're stuck in a Groundhog's Day loop repeating the same steps. The problem, though, likely isn't that AI wrote poorly. Instead, it may be that you're operating at the wrong level—evaluating a finished draft when you're still not clear what your core argument is. More drafts with more words don't help you see that mismatch.

Instead, switch formats—each format highlights different issues. Thesis statements expose your core argument. Outlines reveal structural issues. Snippets and quotes highlight choices in style and tone. Full drafts put all the pieces in place for final checks, not early exploration. Part of solving any problem is representing it in the right way.

Back To Primary School

Here's the irony: you already learned these tools back in primary school. Your teacher taught you to assemble note cards, outline before drafting, and reverse outline articles to understand their logic. These weren't just make-work exercises. They were scaffolds to help you learn writing. They worked because different formats use different grammars that force different kinds of clarity. An outline is sparse but also strict—you can't hide gaps in your argument because there's no prose to smooth over those gaps.

But we all stopped doing that once we left school. The reason was the effort. Who has time to write note cards and outlines and drafts for every memo and blog post? The theory is sound, but the cost was too high. Until now. AI changes those economics. AIs are good at generating text, but great at translating between formats once they have your idea.

Thinking In Formats

When you read a specific type of document, think about what you notice, but also what you don't notice. Structure problems are invisible in prose but obvious in outlines. Logic gaps are hidden in flowing sentences but exposed in argument maps. Tone mismatches disappear in full drafts but jump out in side-by-side style tests.

To use this insight, first identify what feels off, then switch to the format that exposes that problem.

Unclear what you're arguing? Generate five thesis statements, 100 words each. Reading them side by side, you'll immediately see which captures what you meant—and maybe spot an angle you hadn't considered. This is what brainstorming was supposed to be.

Structure feels loose? Translate your draft into an outline. Put the key steps of the argument at the top level. If it doesn't read as a clear narrative in that simplified format, you've found the problem. Better yet, the outline gives you handles for fixing it: "Make II.2 a top level point, and put it after III" is a very compact way to explain the restructure you want.

Tone not landing? Pick a few key paragraphs and generate them in five different styles: confident, casual, technical, speculative. Ask the AI who each version works best for. Then carry the best elements to the whole piece. You're calibrating tone through targeted tests, not sentence-by-sentence tweaking.

You can move in both directions. Not just outline to draft, but draft back to outline when you're stuck. Not just thesis to full argument, but full argument back to summary when you need clarity. Compress when stuck, then expand when clear.

What This Changes

Having AI to help you write changes the way you not only write, but think about the writing. If you need to write every draft yourself, drafts are expensive and scarce. If you have an AI to help you, they're cheap and abundant. And that changes how you use them. They're no longer finished documents that weren't quite good enough—they're experiments and prototypes.

Think about how designers operate. They don't start with the finished product. They sketch, they wireframe, they build mockups at different fidelities—each designed to test different questions. Does the layout work? Does the interaction make sense? Does it feel right? Your final document may be your finished product, but the intermediate elements are more like low-fidelity prototypes used for testing. A thesis statement is a sketch. An outline is a wireframe. A style test is a mockup.

The shift is treating writing like design: you're not refining a single artifact, you're testing hypotheses through a series of prototypes. Each format is a different fidelity level optimized for evaluating different properties. The full draft comes last, after you've validated structure, logic, and tone through cheaper tests. The tools your teacher gave you—outlining, thesis statements, reverse outlining—these were always prototyping tools. Now you can finally afford to use them the way they were meant to be used.