WEBVTT 00:00:00.000 --> 00:00:02.759 Most people treat AI like a glorified vending 00:00:02.759 --> 00:00:05.160 machine. You put in a question, you get an answer. 00:00:05.519 --> 00:00:08.119 Right. Very transactional. Exactly. But what 00:00:08.119 --> 00:00:10.800 happens when you realize you're actually sitting 00:00:10.800 --> 00:00:15.839 across from a $500 an hour consultant? Beat. 00:00:16.800 --> 00:00:20.000 Welcome to the deep dive. We are really glad 00:00:20.000 --> 00:00:21.969 you're here. So glad to be jumping into this. 00:00:22.070 --> 00:00:25.030 It's a huge topic. It is. Today, we're mastering 00:00:25.030 --> 00:00:28.109 the final highest leverage foundational skill 00:00:28.109 --> 00:00:31.570 of the Claude ecosystem. And that is prompt engineering. 00:00:32.030 --> 00:00:34.909 Yeah, which is just the ultimate force multiplier. 00:00:35.329 --> 00:00:37.590 Once you really get this, your daily workflow 00:00:37.590 --> 00:00:41.679 just totally transforms. So here is our roadmap 00:00:41.679 --> 00:00:44.020 for you today. First, we are going to unpack 00:00:44.020 --> 00:00:46.719 that consultant mindset shift. The foundational 00:00:46.719 --> 00:00:49.759 piece. Right. Then we will break down six specific 00:00:49.759 --> 00:00:51.859 power user prompting techniques. And these are 00:00:51.859 --> 00:00:53.960 like actual practical tools you can use immediately. 00:00:54.280 --> 00:00:55.979 Absolutely. And finally, we'll show you how to 00:00:55.979 --> 00:00:58.119 build a plug -and -play personal pumped library. 00:00:58.340 --> 00:01:00.380 So you essentially never start from scratch again. 00:01:00.600 --> 00:01:02.659 Which saves, you know, just an unbelievable amount 00:01:02.659 --> 00:01:05.180 of time. It really does. So let's start with 00:01:05.180 --> 00:01:08.829 that first mindset shift. The core idea from 00:01:08.829 --> 00:01:11.689 our sources is moving away from vague commands. 00:01:11.989 --> 00:01:15.329 You have to provide rich context. You really 00:01:15.329 --> 00:01:17.590 need to treat Claude like a brilliant consultant 00:01:17.590 --> 00:01:19.950 meeting you for the first time. That's the perfect 00:01:19.950 --> 00:01:21.870 way to look at it. Because, you know, they have 00:01:21.870 --> 00:01:24.329 all this world -class knowledge, but zero context 00:01:24.329 --> 00:01:26.870 about your actual life. Exactly. You wouldn't 00:01:26.870 --> 00:01:28.909 walk into a high -level strategy meeting and 00:01:28.909 --> 00:01:31.230 just demand a marketing plan. No, of course not. 00:01:31.430 --> 00:01:34.450 You'd be laughed out of the room. Right. You'd 00:01:34.450 --> 00:01:36.329 give them background. You'd tell them who you 00:01:36.359 --> 00:01:38.680 what the challenge is and why you're doing it. 00:01:38.819 --> 00:01:42.040 The who, the what and the why. That's the holy 00:01:42.040 --> 00:01:45.099 trinity here. It is, but I have to make a vulnerable 00:01:45.099 --> 00:01:48.420 admission right here. I still wrestle with treating 00:01:48.420 --> 00:01:51.500 it like a search engine, myself. My hands just 00:01:51.500 --> 00:01:54.500 instinctively want to type three keywords and 00:01:54.500 --> 00:01:56.780 hit enter. Oh, completely. It's so natural. I 00:01:56.780 --> 00:01:59.140 mean, we've had, what, 20 years of Google training 00:01:59.140 --> 00:02:02.140 us to do exactly that? We type keywords and expect 00:02:02.140 --> 00:02:06.480 magic. But an advanced AI really needs that deeper 00:02:06.480 --> 00:02:09.099 framing to shine. So let's break down those three 00:02:09.099 --> 00:02:12.379 elements. The who is establishing your role. 00:02:12.580 --> 00:02:14.159 It's your experience level, your constraints. 00:02:14.699 --> 00:02:17.159 Yeah, it grounds the AI in a specific perspective. 00:02:17.340 --> 00:02:21.120 Then the what? That outlines the exact project 00:02:21.120 --> 00:02:23.500 or deliverable. The tangible thing you need right 00:02:23.500 --> 00:02:25.840 now. And the why is your underlying goal. What 00:02:25.840 --> 00:02:28.479 does success actually look like? Right. Like 00:02:28.479 --> 00:02:30.439 the old way of prompting is just saying, help 00:02:30.439 --> 00:02:32.419 me write a blog post about prompt engineering. 00:02:32.639 --> 00:02:35.789 Which is Terribly vague. So vague. Yeah. It assumes 00:02:35.789 --> 00:02:38.270 the AI magically knows your audience and your 00:02:38.270 --> 00:02:42.469 tone. The output is always going to be kind of 00:02:42.469 --> 00:02:44.729 robotic and generic. Because it's guessing. Exactly. 00:02:45.349 --> 00:02:48.229 But the new way, you say, I am a freelance content 00:02:48.229 --> 00:02:50.830 creator building a premium course. That's your 00:02:50.830 --> 00:02:53.449 who. You instantly narrow the possibility space. 00:02:53.689 --> 00:02:55.849 Right. And then you say, I need an engaging blog 00:02:55.849 --> 00:02:58.110 post to attract students. That's the what? Giving 00:02:58.110 --> 00:03:00.629 it a measurable business objective. And finally, 00:03:00.729 --> 00:03:03.090 the why. You say, I want readers to feel excited 00:03:03.090 --> 00:03:05.590 and capable, so make it deeply practical. And 00:03:05.590 --> 00:03:08.569 you end with, let's think about this together. 00:03:08.990 --> 00:03:12.870 Yes. That collaborative invitation is huge. Level 00:03:12.870 --> 00:03:15.110 2 prompting isn't about just typing a longer 00:03:15.110 --> 00:03:17.710 command. It's about co -creating the solution 00:03:17.710 --> 00:03:20.110 with the AI. I find the mechanics of that really 00:03:20.110 --> 00:03:22.770 fascinating. Let me ask you this, though. Does 00:03:22.770 --> 00:03:25.349 giving all this upfront context actually change 00:03:25.349 --> 00:03:28.050 the AI's internal reasoning process or just the 00:03:28.050 --> 00:03:30.750 formatting of the output? Oh, it literally changes 00:03:30.750 --> 00:03:33.449 the probabilistic weights guiding the token generation. 00:03:34.030 --> 00:03:36.490 Wow. It activates entirely different neural pathways 00:03:36.490 --> 00:03:39.090 inside the model. You steer it away from average 00:03:39.090 --> 00:03:41.669 internet text toward highly -specualized knowledge. 00:03:42.430 --> 00:03:44.409 It literally thinks differently about the problem. 00:03:44.669 --> 00:03:47.169 Context literally rewires how the AI solves the 00:03:47.169 --> 00:03:49.860 problem for you. Yeah. Beat. Okay, so we have 00:03:49.860 --> 00:03:52.099 the mindset. We understand the who, what, and 00:03:52.099 --> 00:03:55.219 why. How do we actually steer this consultant? 00:03:55.780 --> 00:03:58.039 Our sources highlight six power user techniques. 00:03:58.400 --> 00:04:00.500 Let's walk through them. Let's do it. So technique 00:04:00.500 --> 00:04:03.840 one is role plus full context. Right. Assigning 00:04:03.840 --> 00:04:06.479 an expert persona. Exactly. You say act as a 00:04:06.479 --> 00:04:09.060 world -class conversion copywriter. That anchors 00:04:09.060 --> 00:04:11.759 the vocabulary the AI pulls from. It sets a much 00:04:11.759 --> 00:04:13.599 higher baseline for intelligence right off the 00:04:13.599 --> 00:04:16.839 bat. Technique two is chain of thought. Yeah, 00:04:16.980 --> 00:04:19.420 this one is massive for complex problems. Which 00:04:19.420 --> 00:04:21.920 means making the AI explain its logic step by 00:04:21.920 --> 00:04:24.399 step before answering. Why is that intermediate 00:04:24.399 --> 00:04:27.300 step so crucial? Because LLMs just calculate 00:04:27.300 --> 00:04:30.139 probabilities for the next word. When you force 00:04:30.139 --> 00:04:32.939 step by step thinking, it generates intermediate 00:04:32.939 --> 00:04:36.709 tokens. And those tokens actually help ground 00:04:36.709 --> 00:04:39.589 its final conclusion. It drastically reduces 00:04:39.589 --> 00:04:41.889 hallucinations. So you just tell it, think step 00:04:41.889 --> 00:04:44.370 by step. Literally just type that. Think step 00:04:44.370 --> 00:04:47.449 by step before giving your final answer. It works 00:04:47.449 --> 00:04:50.209 wonders. Technique 3 is few -shot prompting, 00:04:50.449 --> 00:04:52.850 meaning providing a few examples to show exactly 00:04:52.850 --> 00:04:55.160 what you want. Right. You give it a clear pattern, 00:04:55.339 --> 00:04:58.319 like paste in two of your past emails, then ask 00:04:58.319 --> 00:05:00.819 it to match that exact tone for a new one. It's 00:05:00.819 --> 00:05:02.819 perfect for brain consistency. Oh, absolutely 00:05:02.819 --> 00:05:05.319 perfect. Then we have technique four, surfacing 00:05:05.319 --> 00:05:07.540 hidden needs. This is all about finding your 00:05:07.540 --> 00:05:10.560 blind spots. You explicitly ask, what might I 00:05:10.560 --> 00:05:13.060 be overlooking? You ask it to find flaws in your 00:05:13.060 --> 00:05:16.779 logic. Exactly. It turns the AI into a strict 00:05:16.779 --> 00:05:20.399 senior partner. It reliably finds risks you totally 00:05:20.399 --> 00:05:23.379 missed. Technique 5 is collaborative dialogue. 00:05:23.800 --> 00:05:25.439 You make it a back -and -forth conversation. 00:05:25.819 --> 00:05:27.759 Yeah, instead of demanding an instant answer, 00:05:28.079 --> 00:05:31.079 you say, ask me clarifying questions first. It 00:05:31.079 --> 00:05:33.899 interrogates your premises. Right. You co -create 00:05:33.899 --> 00:05:36.699 the architecture together. It's so useful when 00:05:36.699 --> 00:05:38.639 you aren't even sure what you need yet. And the 00:05:38.639 --> 00:05:41.399 final technique, number six, structured output 00:05:41.399 --> 00:05:43.899 requests. This is just telling Claude exactly 00:05:43.899 --> 00:05:47.439 how to format the response. So asking for markdown 00:05:47.439 --> 00:05:49.839 or tables or specific bullet points saves you 00:05:49.839 --> 00:05:52.300 so much manual editing time later. So you can 00:05:52.300 --> 00:05:54.379 essentially stack all these techniques together. 00:05:54.639 --> 00:05:57.300 You use the roll, the context, chain of thought, 00:05:57.600 --> 00:06:00.720 hidden needs. It is like stacking Lego blocks 00:06:00.720 --> 00:06:02.939 of data. I love that analogy. That's exactly 00:06:02.939 --> 00:06:05.240 what it feels like. The roll is the green base 00:06:05.240 --> 00:06:07.829 plate. The context forms the load bearing walls. 00:06:08.009 --> 00:06:09.870 If you don't snap them together right, the structure 00:06:09.870 --> 00:06:11.949 just falls apart. Yeah, but when you do snap 00:06:11.949 --> 00:06:14.670 them correctly, you get deep reasoning and blind 00:06:14.670 --> 00:06:17.829 spot detection all at once. Whoa. Imagine the 00:06:17.829 --> 00:06:20.329 scale of insights when you stack step by step 00:06:20.329 --> 00:06:24.089 logic with hidden needs. Too sex silence. But 00:06:24.089 --> 00:06:26.410 let me ask a probing question here. When stacking 00:06:26.410 --> 00:06:28.509 all these techniques, is there a risk of over 00:06:28.509 --> 00:06:31.389 constraining the AI so it loses its creativity? 00:06:32.009 --> 00:06:34.930 Yes. That is a very real danger. Yeah. If you 00:06:34.930 --> 00:06:37.529 give it too many rigid formatting rules, it gets 00:06:37.529 --> 00:06:39.910 kind of mechanical. Like a robot filling in blanks. 00:06:39.970 --> 00:06:42.829 Exactly. That's why keeping the why completely 00:06:42.829 --> 00:06:45.839 open -ended is crucial. Give it guardrails, but 00:06:45.839 --> 00:06:47.779 let it choose the creative path. Too many rules 00:06:47.779 --> 00:06:50.920 stifle creativity. Keep your core goals open 00:06:50.920 --> 00:06:53.500 -ended. Precisely. We will be right back after 00:06:53.500 --> 00:06:56.160 this short break. Sounds good. That's the answer. 00:06:56.899 --> 00:06:59.040 Today's deep dive is proudly brought to you by 00:06:59.040 --> 00:07:01.399 our partners who share our mission of making 00:07:01.399 --> 00:07:04.540 highly complex topics accessible. They understand 00:07:04.540 --> 00:07:06.660 that mastering new analytical skills requires 00:07:06.660 --> 00:07:09.370 continuous, dedicated learning. We are grateful 00:07:09.370 --> 00:07:11.750 for their ongoing support of this show. We are 00:07:11.750 --> 00:07:14.410 back moving into our final segment today. Let's 00:07:14.410 --> 00:07:16.089 bring it all together. You know the mindset. 00:07:16.509 --> 00:07:18.790 You know the six techniques. But you shouldn't 00:07:18.790 --> 00:07:20.490 be typing these out from scratch every single 00:07:20.490 --> 00:07:23.720 time. No, please don't do that. It's a massive 00:07:23.720 --> 00:07:25.980 waste of cognitive energy. We need to build a 00:07:25.980 --> 00:07:28.560 reusable library. Right. A plug -and -play, personal 00:07:28.560 --> 00:07:30.959 prompts library. Yeah. You do this by saving 00:07:30.959 --> 00:07:33.060 your templates as artifacts. And an artifact 00:07:33.060 --> 00:07:36.620 is saved, standalone documents, or code snippets 00:07:36.620 --> 00:07:39.279 within the AI interface. Exactly. You just pull 00:07:39.279 --> 00:07:41.079 them up when you need them. Our sources give 00:07:41.079 --> 00:07:43.620 us five battle -tested templates. Template one 00:07:43.620 --> 00:07:46.079 is content creation. Right. For blogs, emails, 00:07:46.339 --> 00:07:48.860 sales pages, you define the expert role, your 00:07:48.860 --> 00:07:51.139 brand tone, your audience. But you also invite 00:07:51.139 --> 00:07:54.379 it to brainstorm angles first, right? Yes. Force 00:07:54.379 --> 00:07:57.279 it to explore creative directions before drafting 00:07:57.279 --> 00:08:00.019 the actual text. Template 2 is problem solving 00:08:00.019 --> 00:08:02.139 and strategy. This seems incredibly powerful. 00:08:02.379 --> 00:08:05.120 Oh, it's arguably the strongest tool here. You 00:08:05.120 --> 00:08:08.740 set the AI up as a seasoned consultant. You outline 00:08:08.740 --> 00:08:11.399 your messy situation, but the genius part is 00:08:11.399 --> 00:08:14.220 you ask it to ask you probing questions first. 00:08:14.600 --> 00:08:17.560 It analyzes your constraints step by step before 00:08:17.560 --> 00:08:20.379 recommending a strategy. Exactly. It forces you 00:08:20.379 --> 00:08:23.180 to face reality. Then Template 3. A learning 00:08:23.180 --> 00:08:26.040 and mastery plan. Perfect for a 30 -day skill 00:08:26.040 --> 00:08:29.519 challenge. You tell the AI your proficiency level 00:08:29.519 --> 00:08:31.639 and learning style. And it acts as a productivity 00:08:31.639 --> 00:08:34.340 coach. It builds a progressive curriculum just 00:08:34.340 --> 00:08:37.259 for you. Tailored exactly to your schedule. Template 00:08:37.259 --> 00:08:40.480 4 is review and optimization. This is where you 00:08:40.480 --> 00:08:43.759 ask, how can I make this 10 times stronger? Yeah, 00:08:44.019 --> 00:08:46.559 this one takes some bravery. You paste your work 00:08:46.559 --> 00:08:49.159 in and tell the AI to be a notoriously strict 00:08:49.159 --> 00:08:52.299 editor. You ask it to ruthlessly hunt for structural 00:08:52.299 --> 00:08:54.500 weaknesses. They have to ask for concrete examples, 00:08:54.500 --> 00:08:57.379 too. But be repaired, the feedback is going to 00:08:57.379 --> 00:08:59.820 be blunt. The fifth template is brainstorming 00:08:59.820 --> 00:09:02.320 and ideation. But when you're just totally stuck, 00:09:02.740 --> 00:09:05.230 you need divergent ideas. So you ask for eight 00:09:05.230 --> 00:09:08.070 to ten unconventional ideas. But not just a list. 00:09:08.409 --> 00:09:10.669 You ask for execution challenges and launch plans 00:09:10.669 --> 00:09:13.210 for each one. That weeds out the weak ideas quickly. 00:09:13.549 --> 00:09:15.509 Right. And then you have it ranked its top three 00:09:15.509 --> 00:09:17.809 recommendations. OK. So these templates sound 00:09:17.809 --> 00:09:19.909 incredibly efficient. But I have a reservation. 00:09:20.730 --> 00:09:22.909 Templates sound incredibly efficient. But don't 00:09:22.909 --> 00:09:25.470 they eventually lead to generic cookie cutter 00:09:25.470 --> 00:09:28.429 outputs if everyone uses these same five structures? 00:09:28.870 --> 00:09:31.250 That is a very valid concern. A lot of people 00:09:31.250 --> 00:09:34.009 worry about that. But the magic is in the customization. 00:09:34.850 --> 00:09:36.929 The who, what, and why variables your input are 00:09:36.929 --> 00:09:39.610 entirely unique to your life. The structural 00:09:39.610 --> 00:09:42.350 framework is shared, but the context is deeply 00:09:42.350 --> 00:09:45.409 personal. So the AI reacts to your unique data, 00:09:45.769 --> 00:09:48.590 not just the scaffolding. Exactly. Plus, the 00:09:48.590 --> 00:09:51.389 sources recommend creating a My Profting Style 00:09:51.389 --> 00:09:53.809 Guide. What goes into that master document? Your 00:09:53.809 --> 00:09:56.309 default professional mindset, your preferred 00:09:56.309 --> 00:09:58.929 tone, formatting quirks, and rules you always 00:09:58.929 --> 00:10:01.539 follow. How often should you update this personal 00:10:01.539 --> 00:10:05.100 style guide as these AI models get smarter? Treat 00:10:05.100 --> 00:10:07.960 it as an evolving system. Test and refine those 00:10:07.960 --> 00:10:10.120 templates every few weeks based on the outputs 00:10:10.120 --> 00:10:11.899 you're getting. Treat your style guide like a 00:10:11.899 --> 00:10:14.399 living document, tweaking it monthly. Perfectly 00:10:14.399 --> 00:10:17.000 said. It rapidly becomes your greatest digital 00:10:17.000 --> 00:10:19.559 asset. Let us step back and look at the big picture 00:10:19.559 --> 00:10:22.919 here. Prompt engineering is not a coding language. 00:10:23.320 --> 00:10:25.840 It is high -level communication. It really is. 00:10:25.980 --> 00:10:28.419 It's shifting from demanding instant answers 00:10:28.419 --> 00:10:31.779 to actively co -creating solutions with a brilliant 00:10:31.779 --> 00:10:34.919 consultant. That is the ultimate unlock for AI 00:10:34.919 --> 00:10:37.840 productivity. It forces you to clarify your goals 00:10:37.840 --> 00:10:40.419 before asking for help. It absolutely elevates 00:10:40.419 --> 00:10:42.299 your baseline thinking. So we have some homework 00:10:42.299 --> 00:10:46.019 for you today. Go find three old, lazy prompts 00:10:46.019 --> 00:10:49.240 you used recently. Rewrite them using the complete 00:10:49.240 --> 00:10:51.899 who, what, and why framework. Yes, set up your 00:10:51.899 --> 00:10:55.000 personal AI mastery project today. Start building 00:10:55.000 --> 00:10:57.419 those custom templates into your library. And 00:10:57.419 --> 00:10:59.080 I want to leave you with a final thought to mull 00:10:59.080 --> 00:11:02.840 over. If you have a brilliant $500 an hour consultant 00:11:02.840 --> 00:11:05.080 sitting across from you right now, what is the 00:11:05.080 --> 00:11:06.740 million dollar problem you've been too afraid 00:11:06.740 --> 00:11:08.940 to ask them to solve out to your music?