The Startup Ideas Podcast - Prompt Claude better than 99% of people

Episode Date: December 10, 2025

In this solo episode, I walk through 10 concrete rules to get way more out of Claude Code and Claude Opus 4.5, based directly on tips Anthropic has shared in their docs and blog posts. I show how to move from vague prompts to architected briefs that use tone, constraints, structure, and power phrases to avoid “AI slop.” I demo examples across writing, research, teaching, and planning so you can see exactly how to apply each rule. By the end, you have a practical playbook for prompting Claude like a teammate and using it as a true thinking partner in your work. Timestamps 00:00 – Intro 00:56 – Rule #1: Tone of collaboration 02:16 – Rule #2: Principle of explicitness (action verbs, quantity, audience) 03:20 – Rule #3: Define the boundaries with clear constraints 04:26 – Rule #4: Draft, plan, then act (outline → refine → execute) 06:39 – Rule #5: Demand structured output (tables, formats, schemas) 08:00 – Rule #6: Explain the “why” behind your request 09:05 – Rule #7: Control brevity vs. verbosity (expert, brief, simplifier) 10:21 – Rule #8: Provide a scaffold and templates 11:21 – Rule #9: Use “power phrases” and expert personas 12:28 – Rule #10: Divide and conquer complex projects 14:09 – Putting it all together with an example For founders doing $50k+ MRR+: https://startup-ideas-pod.link/offline-mode Key Points I share 10 specific prompting rules that come directly from how Anthropic suggests people use Claude. I show how friendly, clear, and firm prompts beat either vague or overly polite requests. I demonstrate how explicit constraints (length, style, audience, banned words) create more creative and focused outputs. I use outlines, scaffolds, and structured formats to turn Claude into a planning and synthesis engine instead of a random text generator. I introduce “power phrases” like “think step by step” and “critique your own response” to unlock more advanced reasoning. I wrap everything into a final Stoicism lecture prompt that combines persona, context, constraints, structure, and tone. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ The Vibe Marketer - Resources for people into vibe marketing/marketing with AI: thevibemarketer.com Startup Empire - get your free builders toolkit to build cashflowing business - https://startup-ideas-pod.link/startup-empire-toolkit Become a member - https://startup-ideas-pod.link/startup-empire FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/

How can you get more out of Claude Code and Claude Opus 4.5? Well, I've got good news from you. Anthropic has actually, over the last 12 months, have been posting in their docs, blog posts, kind of teasing on X about how you can prompt these products to really get the most out of it. But the thing is, people haven't put it in a full guide. So I did the hard work to make it easy on everyone. By the end of this episode, you will learn 10 techniques or how to prompt Claude to get the most out of it. Super simple techniques, anyone can learn.

I'm going to show you real examples, easy to understand, and frameworks to help you crush it with Claude Code and Opus 4.5. Let's get right into it. So the first tip is, I know this is going to upset a few people, but the tone of collaboration is really important. You're going to want a friendly and clear and firm tone because, that yields better results and more direct results. So what's an example? A vague request might be something like, fix this grammar in this now, you know,

but the problem with that is, you know, it leads to overly cautious, pre-can, or basically just less helpful responses as the model tries to de-escalate. Politeness can sometimes result in chatty, less direct answers. Now, if you do, you know, an architected brief,

and this is what the folks at Anthropics, suggest you do. Do something like, please review the following text for grammatical errors and suggest corrections. My goal is to make it sound more professional and confident. This is direct. This is respectful. And it provides context, which is what anthropic needs in order to get you the result that you want. So really important. I know some of us are just kind of mean to our LMs. I've been there, you know, but treat it like a teammate, right? You would never want to be mean to a teammate, especially if you want to get them to produce. So rule one of 10 is the tone of collaboration.

Rule 2 of 10 is the principle of explicit. So state your request as a clear action-oriented command with all the necessary detail. So, and I used to do this, actually. I would do like a vague request, like I need a bunch of blog post ideas. But the problem is it's passive, it's not specific, and then you just get this generic AI slot. Architected brief. What's the difference?

Okay. Generate 10 blog post titles about the impact of remote work on urban planning. The title should be engaging for an audience of city officials and real estate developers. This prompt uses an action verb. Generate. You're going to want to use action verbs a lot. It specifies the quantity, 10. and target audience.

Those are the three things that you're going to need. In action verb, the quantity, and a target audience. Every highlighted phrase adds a layer of useful constraint. This works extremely well. Three on 10. The rule here is a well-defined box produces a more creative result than an empty field. So a vague request would be something like, write a short story about a detective in the future.

The problem is the possibilities are infinite, and that leads to cliche AI slop, unfocus, output. Architected brief, what's the difference? Write a short story, no more than 500 words in the style of Raymond Chandler. You can even do like in the style of Ernest Hemingway meets Raymond Chandler, or you can even put three or four or five different people. The story must feature a robot detective investigating a data theft on Mars. Do not use the word cyber. So you've added constraints on length, constraints on style, constraints on character, constraints on settings, and even specific words to force the AI into more creative and specific solution. I know this takes more time.

I know that, you know, it's kind of measured twice and cut once, but you will get more if you define the boundaries. Number four, the rule is draft, plan, then act. Use the AI to generate an outline or roughed version first. So don't try to get a perfect final product in one go. I know we all want to one prompt it and just it feels so good to one prompt it. I've been there. But, you know, the reality is working with the AI to create and then refine a plan or outline is a way more reliable path to a high quality result. And that's what we're out for.

This lets you course correct early. So think of it like this. The initial prompt might be, I need to write a report on the benefits of a four-day work week. Ask for a plan. Step one. First, propose an outline for this report. Then refine the plan. Step two.

That's a good start. In section two, please add a subpoint about employee retention. Step three, you can do, you know, step three is the execution layer. Excellent. Now, write the full report based on this revised outline. So, you know, it takes longer. but it takes longer. It feels like it takes longer,

but it's actually going to save you time because you're going to get a better outcome and you're not going to have to reprompt and reprompt and reprompt. If you have a business that's doing at least $50,000 a month in revenue, I've got something interesting for you. It's called Offline Mode.

It's a two-day event that me and my team are putting on at a 20,000 plus square foot mansion. Yes, this is what it looks like on January 23rd and January 24th in Fort Lauderdale, Florida. I'll include a little bit of link in the description if you're interested in coming. But it's basically for people who have a

business that's kind of cranking, but they really want to put it in rocket ship mode. They want to create a set of businesses that generate tons of money, tons of cash flow, tons of product market fit, tons of impact. But they're not just quite there yet. It's also about, you know, making your business AI first, how you can actually, you know, build not just one product, but multiple products, and you're going to leave with, you know, tactical answers to your questions. So if that's you, and this sounds interesting, I'll see you there. Rule number five is demand structured output. The AI is fluent in many formats beyond pros.

So, you know, a vague request might be something like, list the last three Apollo missions and some facts about them. So you're going to, it's basically going to give you a simple, unstructured paragraph, and that's going to be hard to parse. Now, an architected brief might be something like, provide the list of the last three Apollo missions, 15, 16, and 17, for each mission include the launch date, the crew members, and a key specific achievement. Present this information in a markdown formatted table.

So what you might have to do is go to something like perplexity and do a prompt there to understand a little more about what you want an architected brief to look like. But you're going to end up getting, you know, a way better output here. Like you can just see this markdown file if you're on YouTube. Also, like in common, if you're liking this sort of stuff, I'll do more of it. Requesting, you know, so look at this markdown, you know, it's just way, way better than what you'd get if you did something else and just went like a simple, vague sort of thing. So it might take you more time a little bit initially. but again, you're going to get better output.

Six, explaining the why. So the golden rule here is explaining the why behind an instruction helps the AI understand your true intent. So instead of saying, give me five marketing slogans for a brand new coffee, where the AI basically has no context, it doesn't know the brand values. It doesn't know your audience, your community who you're going after. It doesn't know your unique selling proposition.

Do something like this. Give me five marketing. slogans for a new brand of coffee. The key is that our beans are ethically sourced from small independent farms and our target audience, really important you put this in here, is environmentally conscious millennials. By the way, it's not just that you put our target audience as millennials. You have to define it even more. Go niche and then go more even nicher. You'll get better results. The slogan should reflect quality and sustainability. So by providing the why, basically ethically source for conscious millennials, the AI can generate far more relevant and

targeted slogans, then you get up getting better results. Seven on 10. The art of brevity and verbosity. So the rule here is explicitly command the AI to be more or less verbose to match your needs. So you're in control of the output length. Use simple, direct phrases to guide the AI. So let's just show some different examples. So maybe you want the expert here.

Explain photosynthesis in detail for a college biology student. Think step by step to ensure accuracy. Boom. Look at how expert this looks like. I mean, it's throwing me back to my biology and chemistry days, just seeing stuff like this when I was in school. Maybe you want something that's brief. Explain photosynthesis. Be concise and use bullet points. Sometimes here I'll say like explain like I'm five. Explain like I'm 13. Explain like I'm 17. And then you can see the bullet points. And then oh here, the simplifier. Explain photosynthesis like I'm five years old. So these are three options, the expert, the brief, the simplifier. The point here is that you are in control of your output length. And it is important that you include that in the prompt. Eight on 10. Providing a

a scaffold. The rule here is you want to give the AI, you want to give Claude a template or example to guide its structure and style. So the vague request would be something like, summarize this article, but, you know, this is way better. Look at this in the architected brief section. Summarize the following article using this format. Main thesis, one sentence, AI fills this in, key supporting points, three bullet points, the AI fills this in, and concluding insight AI feels this in and then paste the text here. So now you have this rigid structure ensuring the summary is not only accurate but is also formatted exactly as needed. So, you know, really, really we want the best results. We want the best final results.

And just by saying this, the structure and style, it helps. It's a little bit of scaffolding goes a long way. Nine on ten. Speaking the language. So using advanced prompting terms can trigger more sophisticated modes of operation. So models are trained on a vast amount of text about AI itself. So using terms from the field activate specific powerful behaviors. This is like cheat codes. So let's talk about a few power phrases that Anthropic has literally told us to use that most of us are not even using.

Think step by step. So what's the use case? Force the model to lay out the reasoning process, often leading to more accurate results on complex problem. So you're going to want to remember that one. Critique your own response. Use case. Ask the model to perform self-correction.

Find flaws in its initial draft and improve it. Adopt the persona of an expert in field. Use case. Primes the model to response with a deeper, more domain-specific vocabulary and framework. These are magic words. I want to screenshot this. This will help you get the most out of it. And finally, we have the divide number 10, the divide and conquer strategy.

So the rule here is for a complex task act as the conductor, prompt for each part separately, then prompt for the synthesis. So don't ask for a 10-page report in a single prompt. Break it down into logical subtas that you manage step by step. So if you listen to this startup ideas podcast, you know I love ideas. You know, I have a business ideabrowser.com that gives away ideas. And, you know, so you know I like building businesses. And you can see that it's easy to, you know, throw in, hey, build a business plan as the one prompt. But this is a way better to do it.

Step one, the blueprint. Create a detailed table of contents for a business plan for a new specialty coffee shop. Step two, section by section. Write the executive summary based on our plan. Now write the market analysis section, and then the A dotted line indicates more steps. And step three, the synthesis, review the complete business plan, ensure consistent tone and check for any contradictions. So this is the project workflow, you know, doing it section by section, synthesizing, having the blueprint, you know, you're dividing and conquering. You're breaking into logical subtask.

What might even be helpful here is like literally. taking out a pen of paper and being like, okay, if I want a business plan, what are the 10 things I need? If I want to create a deck, a fundraising deck, what are the 10 things I need? If I want to create a product, that's, you know, software product, what are the 10 different, you know, features I need? This is going to be helpful for you to get the most out of it. So there you have it. I've given you the 10 rules of how you can get, how you can prompt the prompt Claude to get literally 10x more of it. This is the simple. tools and techniques that you can use that have literally been shared by the people from Anthropic

themselves. We talked about the tone of collaboration. We talked about, which is using friendly, clear, and firm tone. We talked about the principle of explicitnessness. We talked about the defining the boundaries. We talked about the exploratory draft. We talked about specifying the details. We talked about explaining the why. We talked about how important brevity is, why you should put a little scaffolding, how important it is to speak the language. And we talked about the dividing and conquer strategy. So what does that look like when you put it all together? This is, you know, when you came into this episode, you might have prompted, or maybe

not, but you might have prompted Claude saying something like, tell me about stoicism. By the way, I call Claude, but for you, I'll call Claude because people get angry and in the comment section about that. Let's talk about now. Now that you know the tools of how to prompt Claude better, let's, let's say. Let's see it at work. Instead of saying, tell me about stoicism, you're going to say, act as a university professor of philosophy. Why?

Because you've included the persona in there. You're going to say, I'm preparing a one-hour intro lecture for students with no prior knowledge. That's going to explain the why. You're going to say, first create a lecture outline with three main sections, the divide and conquer. The outline should have a clear introduction and body. You've now constrained the scope. And conclusion, please format this as a nested, bolsteris. bolted list, rich output.

For each major point, include a key stoic figure, example, Seneca, explicitness, and one of their core ideas. Your tone should be accessible and engaging. Tone. So you've included that all in there, and then the output is just way better. So these are little things to help you get the most out of these LMs, and like, why matters is because we're all trying to get the most out of it. We're all trying to create products that people love. We're all trying to create content that people love. We're all trying to do things

that people love. And a lot of the times we get AI slop. And this is going to help you not get AI slop. And that's why I put the time to scour the internet for you so you can get the most out of this product. It's so, so powerful. I have no affiliation with Anthropic. But I want to see you not just like comments and subscribe on this, but actually go and build something great with this stuff. So I'm rooting for you. Have a creative day. And I'll see you next time.