In this instructional monologue, TJ Walker shares the precise prompting framework he uses to make AI-generated text sound conversational and human. He rejects vague directions and instead provides a concrete instruction template that emphasizes short sentences, breaking up longer thoughts, repeating key ideas, retaining repetition, and inserting natural spoken phrases. Walker stresses that this is an iterative process: the AI often re‑polishes the output, so the user must repeatedly push back, refine, and run the prompt multiple times. He also describes a validation step—reading the AI’s output aloud to detect any lingering stiffness—and adjusting until the text feels naturally spoken. Over time, practitioners develop pattern‑recognition skills that let them spot where to break sentences, add repetition, or insert conversational asides, leading to faster, higher‑quality results on the first try.
Chapters
Ultimize. At this point, you understand the concepts. Now I want to show you exactly what I do. Not theory, not general advice, not pie-in-the-sky stuff, but what I actually do. The actual instructions I give AI. Because this is where most people fall short. They understand what they want, but they don't tell the AI clearly enough. So here's the framework I use. I don't say, make this better. I don't say, make this more engaging. That's too vague. That's too general. That's too fuzzy and abstract. Instead, I say something like this. Rewrite this in a conversational speaking style. Use short sentences. Break longer thoughts into multiple sentences. Repeat key ideas for emphasis. Do not Remove repetition. Add natural phrases like, here's the thing, or let me explain. Keep the tone informal and spoken, not polished or written. That's it. That is specific. You can use that exact phrase. This is actionable. That gives the AI direction. Now here's what matters. You may need to run that instruction more than once because the AI will try to clean things up again. It will drift toward polished writing. You gotta really push, push hard, push back. You refine it. The AI works for you. Don't forget that. You say specifically, you say shorter sentences, more repetition, less polish. This is a process. It's not a single prompt. And here's something else I do. I compare. I take the AI version and I read it out loud. If it feels even slightly stiff, I adjust it. I don't accept good enough, almost good, because almost good still sounds artificial, and that's what we're trying to eliminate. Don't forget that. Now, over time, something interesting happens. You get faster at this. You start recognizing patterns immediately. You can look at a paragraph and instantly know where it needs to be broken where it needs repetition, where it needs a conversational aside, and eventually you start writing prompts that produce much better results on the first try. But it starts here: clear instructions, specific direction, and a willingness to refine. Oh yeah, you're going to need to refine because the AI is not going to do this perfectly on its own the first, second, or even 10th take sometimes. You have to guide it.