WEBVTT 00:00:00.000 --> 00:00:03.500 Okay, so if you're already using AI tools like 00:00:03.500 --> 00:00:07.219 ChatGPT, Gemini, Claude, you know how powerful 00:00:07.219 --> 00:00:09.320 they can be. Absolutely, they're game changers. 00:00:09.460 --> 00:00:11.560 But how often have you tried to get something 00:00:11.560 --> 00:00:14.300 really specific, something strategic, and the 00:00:14.300 --> 00:00:17.120 answer comes back sounding super confident, but 00:00:17.120 --> 00:00:19.879 it's... Well, it's just off. Doesn't quite hit 00:00:19.879 --> 00:00:22.120 the mark. All the time. It's a really common 00:00:22.120 --> 00:00:23.780 frustration. Yeah. And you start wondering, is 00:00:23.780 --> 00:00:26.760 it me? Am I not explaining it right? Right. Or 00:00:26.760 --> 00:00:28.859 maybe you think you need a fancier tool or something. 00:00:29.440 --> 00:00:33.179 Exactly. But what if the real issue is something 00:00:33.179 --> 00:00:36.179 more fundamental? Some experts are saying it's 00:00:36.179 --> 00:00:38.799 like a language barrier between us and the AI. 00:00:39.000 --> 00:00:40.820 Could you maybe unpack that a bit? What does 00:00:40.820 --> 00:00:43.179 that barrier look like from the AI side? That's 00:00:43.179 --> 00:00:45.380 a great way to put it, a language barrier. Because, 00:00:45.659 --> 00:00:47.759 yeah, we're speaking human with all our nuance 00:00:47.759 --> 00:00:50.859 of assumptions, and the AI is processing based 00:00:50.859 --> 00:00:54.140 on, well, math and patterns. The translation 00:00:54.140 --> 00:00:56.920 definitely gets lost sometimes. And that's exactly 00:00:56.920 --> 00:00:59.329 where prompt engineering comes in. basically 00:00:59.329 --> 00:01:01.869 the bridge across that gap. It's not just about 00:01:01.869 --> 00:01:04.230 asking questions. It's more like the science 00:01:04.230 --> 00:01:06.829 of structuring what you tell the AI. The science. 00:01:07.030 --> 00:01:09.170 Yes, structuring the input so you can guide it 00:01:09.170 --> 00:01:11.030 towards the answers you actually need, answers 00:01:11.030 --> 00:01:14.409 that are accurate, reliable, and genuinely relevant. 00:01:14.750 --> 00:01:17.329 So our goal today is really to help you learn 00:01:17.329 --> 00:01:21.390 how to speak that AI language, and crucially, 00:01:21.670 --> 00:01:24.189 how to trust the output more, because you'll 00:01:24.189 --> 00:01:25.689 understand why it's saying what it's saying, 00:01:25.909 --> 00:01:28.719 and maybe more importantly, when it might just 00:01:28.719 --> 00:01:31.819 be, well, guessing. Right, knowing when it's 00:01:31.819 --> 00:01:34.060 confident versus when it's just sounding confident. 00:01:34.260 --> 00:01:36.739 Precisely. So our mission for this deep dive 00:01:36.739 --> 00:01:39.390 is simple. take you from feeling maybe a bit 00:01:39.390 --> 00:01:42.769 frustrated or misunderstood by AI to really getting 00:01:42.769 --> 00:01:45.370 how to communicate with it effectively, like 00:01:45.370 --> 00:01:47.930 moving beyond just basic commands to using it 00:01:47.930 --> 00:01:50.109 as a proper strategic partner. Yeah, getting 00:01:50.109 --> 00:01:52.629 past the party tricks to the real power. Exactly. 00:01:52.810 --> 00:01:54.969 We want you to turn that confident guesser into 00:01:54.969 --> 00:01:58.170 a collaborator you can actually rely on. So let's 00:01:58.170 --> 00:02:01.489 dive in. OK. So to really get good at talking 00:02:01.489 --> 00:02:04.790 to AI, it does help to understand a little bit 00:02:04.790 --> 00:02:07.750 about what's going on. you know, under the hood. 00:02:08.110 --> 00:02:10.650 Right, because it can feel quite magical sometimes, 00:02:10.710 --> 00:02:12.949 almost like it's thinking. It really can. Yeah. 00:02:13.189 --> 00:02:15.990 But underneath it all, even though it seems unpredictable, 00:02:16.810 --> 00:02:19.650 its operations are all based on strict mathematical 00:02:19.650 --> 00:02:23.150 rules. It's incredibly complex, sure, but its 00:02:23.150 --> 00:02:26.129 calculations, not consciousness. OK, so math, 00:02:26.289 --> 00:02:28.530 not magic. Exactly. And the core concept is something 00:02:28.530 --> 00:02:31.189 called token probability. Tokens, right. I've 00:02:31.189 --> 00:02:33.610 heard that term. Yeah. So when you type a command, 00:02:33.729 --> 00:02:37.400 like a write me an email, The AI doesn't understand 00:02:37.400 --> 00:02:40.500 that request like a person would. Instead, it 00:02:40.500 --> 00:02:42.479 breaks your words down into these little pieces, 00:02:42.840 --> 00:02:45.979 tokens. And then instantly, it starts calculating 00:02:45.979 --> 00:02:48.460 the probability of what the very next token should 00:02:48.460 --> 00:02:51.620 be. Based on the billions of text examples it 00:02:51.620 --> 00:02:53.879 learned from during its training, it's predicting 00:02:53.879 --> 00:02:56.939 the most statistically likely next word or part 00:02:56.939 --> 00:02:58.800 of a word. So it's always just predicting the 00:02:58.800 --> 00:03:00.520 next piece of the puzzle based on everything 00:03:00.520 --> 00:03:02.719 it's seen before. Essentially, yes. It's sequence 00:03:02.719 --> 00:03:05.340 prediction on a massive scale. OK, so if it's 00:03:05.340 --> 00:03:08.919 just predicting the next word, how does that 00:03:08.919 --> 00:03:11.620 lead to it sometimes making stuff up so confidently 00:03:11.620 --> 00:03:15.120 or just giving really generic, bland answers? 00:03:15.539 --> 00:03:17.620 Yeah, that's the direct consequence of how it 00:03:17.620 --> 00:03:20.159 works because it's predicting based on probability 00:03:20.159 --> 00:03:22.500 and patterns from its training data. It's fantastic 00:03:22.500 --> 00:03:25.300 at generating text that sounds right, that flows 00:03:25.300 --> 00:03:28.759 well. But that same mechanism means it can generate 00:03:28.759 --> 00:03:31.800 something that sounds totally plausible, but 00:03:31.800 --> 00:03:34.919 is factually wrong. It's just following the most 00:03:34.919 --> 00:03:37.979 probable word sequence. It's not checking facts 00:03:37.979 --> 00:03:40.060 against the real world. So it doesn't know truth, 00:03:40.159 --> 00:03:42.340 it just knows common word patterns. Exactly. 00:03:42.500 --> 00:03:43.979 It knows patterns. And if you don't give it clear 00:03:43.979 --> 00:03:47.500 structure, it's like imagine spinning a giant 00:03:47.500 --> 00:03:49.840 roulette wheel with millions of possible text 00:03:49.840 --> 00:03:52.159 patterns on it. OK. You say write an email and 00:03:52.159 --> 00:03:55.400 the wheel spins. Will it land on business email, 00:03:55.979 --> 00:03:59.389 personal email? sales pitch, maybe even a breakup 00:03:59.389 --> 00:04:01.889 email pattern. It's literally guessing based 00:04:01.889 --> 00:04:04.009 on the fragments you gave it. Wow, okay. That 00:04:04.009 --> 00:04:06.389 roulette wheel image really clicks. But. Yeah. 00:04:06.430 --> 00:04:08.810 When you provide structured input, you become 00:04:08.810 --> 00:04:11.569 the one guiding the ball, leading the AI right 00:04:11.569 --> 00:04:13.569 to the specific pattern you actually want. You 00:04:13.569 --> 00:04:15.710 stop it from just wandering around the possibilities. 00:04:16.089 --> 00:04:18.769 Right. So we need to stop throwing random words 00:04:18.769 --> 00:04:20.269 at it and start giving it more of a blueprint. 00:04:20.569 --> 00:04:23.509 But what does that structured input data actually 00:04:23.509 --> 00:04:26.449 look like? How do we build that blueprint? That's 00:04:26.449 --> 00:04:28.769 the million -dollar question. And it brings us 00:04:28.769 --> 00:04:30.689 to the core of practical prompt engineering. 00:04:31.129 --> 00:04:32.990 We're going to walk you through what we call 00:04:32.990 --> 00:04:34.810 the gold standard for getting consistently great 00:04:34.810 --> 00:04:38.110 results. It's a six -component framework for 00:04:38.110 --> 00:04:40.769 basically every prompt you write. Six components. 00:04:40.889 --> 00:04:43.589 OK. Is this something proprietary? Not at all. 00:04:43.810 --> 00:04:45.750 In fact, it aligns really well with the kind 00:04:45.750 --> 00:04:48.170 of principles Google and other major AI labs 00:04:48.170 --> 00:04:50.990 teach in their advanced courses. It's just a 00:04:50.990 --> 00:04:53.509 structure that, honestly, most everyday users 00:04:53.980 --> 00:04:56.000 haven't really been taught or haven't explored 00:04:56.000 --> 00:04:58.740 properly. OK, cool. So let's make it concrete. 00:04:58.800 --> 00:05:02.100 Let's use an example. Say you're running an online 00:05:02.100 --> 00:05:04.180 bookstore, you've got a new detective novel coming 00:05:04.180 --> 00:05:07.339 out, and you want a really killer social media 00:05:07.339 --> 00:05:10.120 post for it. Not just new book here, but something 00:05:10.120 --> 00:05:12.980 that actually creates buzz. How would we use 00:05:12.980 --> 00:05:15.160 this framework? Perfect example. OK, so we build 00:05:15.160 --> 00:05:18.420 the prompt piece by piece, component one, role. 00:05:19.300 --> 00:05:22.050 You need to tell the AI who it should be. Assign 00:05:22.050 --> 00:05:24.389 it an identity. Exactly. Stop it from being a 00:05:24.389 --> 00:05:27.569 generic chat bot. Make it an expert. So for our 00:05:27.569 --> 00:05:30.149 bookstore, you'd start with, you are a content 00:05:30.149 --> 00:05:32.350 marketing expert specializing in the book publishing 00:05:32.350 --> 00:05:35.089 industry. Ah, OK. So it adopts that persona. 00:05:35.209 --> 00:05:37.430 Right. It accesses the patterns of vocabulary 00:05:37.430 --> 00:05:39.250 associated with that role. Second component, 00:05:39.689 --> 00:05:41.470 context. Give it the background info it needs. 00:05:41.730 --> 00:05:43.779 The more detailed, the better. Makes sense. So 00:05:43.779 --> 00:05:46.540 for the book, I'm preparing to launch a new detective 00:05:46.540 --> 00:05:49.019 novel. It's written by a Vietnamese author. The 00:05:49.019 --> 00:05:51.720 target audience is young readers, say 18 to 30, 00:05:52.060 --> 00:05:55.000 who are into mystery and thriller genres. And 00:05:55.000 --> 00:05:56.879 they're mainly active on Facebook and Instagram. 00:05:57.100 --> 00:06:00.319 That's pretty specific. Author, background, audience, 00:06:00.620 --> 00:06:04.199 platforms. Yep. Details matter. Third component, 00:06:04.839 --> 00:06:07.519 the task. Be crystal clear about what you want 00:06:07.519 --> 00:06:11.120 it to do. No vagueness. OK. For our post. Write 00:06:11.120 --> 00:06:13.019 a promotional social media post for this book. 00:06:13.300 --> 00:06:15.680 The main goal is to create curiosity and encourage 00:06:15.680 --> 00:06:19.000 readers to pre -order it. Direct, unambiguous. 00:06:19.300 --> 00:06:23.720 Got it. Task is clear. Fourth, format. This is 00:06:23.720 --> 00:06:25.699 crucial. How do you want the output to look? 00:06:25.860 --> 00:06:29.180 Word count, bullet points. Emojis. Oh, right. 00:06:29.220 --> 00:06:31.060 The actual structure of the response. Exactly. 00:06:31.199 --> 00:06:33.319 So you'd add, present the output as a Facebook 00:06:33.319 --> 00:06:36.600 post, roughly 150 words. It must include a shocking 00:06:36.600 --> 00:06:38.180 opening line, three bullet points highlighting 00:06:38.180 --> 00:06:40.339 the most intriguing plot elements without spoilers, 00:06:40.819 --> 00:06:43.920 and a clear call to action, a CTA to pre -order 00:06:43.920 --> 00:06:46.560 the book. Use a placeholder for the link. Wow. 00:06:46.660 --> 00:06:49.379 OK. Super prescriptive. It helps avoid randomness. 00:06:49.560 --> 00:06:52.819 Fifth component, rules. Set the guardrails. What 00:06:52.819 --> 00:06:54.680 should it do? And just as importantly, what should 00:06:54.680 --> 00:06:58.879 it not do? The do's and don'ts. Do. Use a captivating 00:06:58.879 --> 00:07:00.759 and mysterious tone. Feel free to use appropriate 00:07:00.759 --> 00:07:03.459 emojis. Do not reveal any major plot spoilers. 00:07:03.980 --> 00:07:06.060 Avoid using overly formal or academic language. 00:07:06.360 --> 00:07:08.879 This helps prevent unwanted results. OK. Keeps 00:07:08.879 --> 00:07:10.920 it on track. And finally, the sixth component. 00:07:11.040 --> 00:07:12.959 This one's like the secret sauce that many people 00:07:12.959 --> 00:07:16.279 miss. Examples. Show. Don't just tell. Exactly. 00:07:16.680 --> 00:07:19.259 You can actually show the AI what good looks 00:07:19.259 --> 00:07:22.000 like by providing an example of something that 00:07:22.000 --> 00:07:24.920 worked before. Oh, interesting. So you could 00:07:24.920 --> 00:07:27.839 add. Please reference the tone and style from 00:07:27.839 --> 00:07:30.339 this previous successful post. Then you'd paste 00:07:30.339 --> 00:07:33.180 the text of that post here. Ensure the new post 00:07:33.180 --> 00:07:36.399 is also energetic, relatable, and really drives 00:07:36.399 --> 00:07:39.779 curiosity. This helps enormously with brand consistency 00:07:39.779 --> 00:07:42.100 and getting the feel right. That is powerful, 00:07:42.120 --> 00:07:44.740 giving it a direct reference point. It really 00:07:44.740 --> 00:07:46.639 is. OK, so let's just pause and think about the 00:07:46.639 --> 00:07:49.300 difference here. Most people, myself included 00:07:49.300 --> 00:07:51.680 probably, would just type, write a promotional 00:07:51.680 --> 00:07:54.680 post for a new detective novel. Right. And you'd 00:07:54.680 --> 00:07:58.180 get something. Okay, maybe. But likely, very 00:07:58.180 --> 00:08:01.579 generic, maybe a bit flat. Could apply to almost 00:08:01.579 --> 00:08:03.800 any book. Yeah, totally forgettable. But using 00:08:03.800 --> 00:08:06.279 that full six -part prompt, I mean, the level 00:08:06.279 --> 00:08:08.579 of detail, the guidance, the result has to be 00:08:08.579 --> 00:08:11.399 completely different. It's night and day, honestly. 00:08:11.800 --> 00:08:14.439 I remember struggling with getting marketing 00:08:14.439 --> 00:08:17.939 copyright, using AI early on, just getting bland 00:08:17.939 --> 00:08:20.000 stuff back again and again. Me too. Felt like 00:08:20.000 --> 00:08:22.899 pulling teeth. Yeah. If I'd known this framework 00:08:22.899 --> 00:08:26.370 back then... It just takes so much of the guesswork 00:08:26.370 --> 00:08:30.689 out of it for us. It turns the AI from this general, 00:08:31.089 --> 00:08:33.289 slightly unreliable assistant into a focused 00:08:33.289 --> 00:08:36.370 expert for that specific task. It really elevates 00:08:36.370 --> 00:08:38.429 the interaction. Okay, so that six -part framework 00:08:38.429 --> 00:08:40.870 is foundational. Absolutely. That's your starting 00:08:40.870 --> 00:08:43.429 point for reliable results. So now that we have 00:08:43.429 --> 00:08:45.909 that foundation, let's get into some of the more... 00:08:45.929 --> 00:08:48.590 advanced techniques, the kind of stuff that really 00:08:48.590 --> 00:08:51.110 separates casual users from the pros who get 00:08:51.110 --> 00:08:53.389 amazing results consistently. Okay, let's do 00:08:53.389 --> 00:08:55.889 it. The first one is really clever. It's called 00:08:55.889 --> 00:08:58.370 the confidence verification mechanism. Confidence 00:08:58.370 --> 00:08:59.970 verification, okay. The big challenge we mentioned 00:08:59.970 --> 00:09:02.210 earlier is that AI always sounds confident, right? 00:09:02.730 --> 00:09:04.750 Even when it's basically making things up, hallucinating. 00:09:04.909 --> 00:09:07.629 Right, which can be dangerous if you just trust 00:09:07.629 --> 00:09:10.230 it blindly. Yeah. Leads to embarrassing mistakes. 00:09:10.549 --> 00:09:12.669 Totally. So the solution is pretty ingenious. 00:09:12.909 --> 00:09:16.570 You actually force the AI to rate its own confidence 00:09:16.570 --> 00:09:18.490 level for the information it's giving you. You 00:09:18.490 --> 00:09:21.809 ask it how sure it is. Exactly. You add a specific 00:09:21.809 --> 00:09:24.190 instruction to your prompt, especially for important 00:09:24.190 --> 00:09:27.070 tasks. Something like, for each key statement 00:09:27.070 --> 00:09:29.309 or recommendation you make, please rate your 00:09:29.309 --> 00:09:32.340 confidence using this scale. virtually certain 00:09:32.340 --> 00:09:35.620 that's like 95 % plus confidence, highly confident 00:09:35.620 --> 00:09:40.379 80, 95%, moderately confident 60, 80%, or speculative 00:09:40.379 --> 00:09:43.840 low confidence below 60%. And briefly explain 00:09:43.840 --> 00:09:47.000 why you've given that rating. Whoa. That is incredibly 00:09:47.000 --> 00:09:49.799 useful, but hang on. If the AI can hallucinate 00:09:49.799 --> 00:09:52.139 facts, can we actually trust its confidence rating 00:09:52.139 --> 00:09:54.120 about those facts? Isn't there a risk it just 00:09:54.120 --> 00:09:55.759 confidently tells you it's confident about something 00:09:55.759 --> 00:09:57.639 wrong? That's a really short question, and it 00:09:57.639 --> 00:09:59.379 gets to the heart of it. The confidence rating 00:09:59.379 --> 00:10:02.019 isn't about the AI knowing objective truth in 00:10:02.019 --> 00:10:04.639 a human sense. It's about the AI reflecting the 00:10:04.639 --> 00:10:06.720 certainty level derived from its source data 00:10:06.720 --> 00:10:09.399 for that specific piece of information. Ah, okay. 00:10:09.539 --> 00:10:11.980 So it's rating its source material in a way. 00:10:12.519 --> 00:10:15.240 Precisely. Think about a financial analyst scenario. 00:10:15.389 --> 00:10:19.029 You ask the AI to analyze two companies. It might 00:10:19.029 --> 00:10:21.929 say its projection for company A's revenue growth 00:10:21.929 --> 00:10:26.190 is highly confident. Why? Because it's based 00:10:26.190 --> 00:10:29.450 on data from a recent verifiable quarterly reported 00:10:29.450 --> 00:10:32.070 process. Okay. But then it might say its statement 00:10:32.070 --> 00:10:34.629 about company B's upcoming product launch succeeding 00:10:34.629 --> 00:10:38.220 is only moderately confident. Why? Because the 00:10:38.220 --> 00:10:40.379 source for that was maybe just an unverified 00:10:40.379 --> 00:10:43.100 press release or some analyst speculation it 00:10:43.100 --> 00:10:45.120 found. I see. So it's flagging the reliability 00:10:45.120 --> 00:10:47.100 of the input information it used to generate 00:10:47.100 --> 00:10:49.179 the output. That gives you context. Exactly. 00:10:49.259 --> 00:10:51.519 It tells you what's grounded in solid data versus 00:10:51.519 --> 00:10:54.500 what's more speculative. That difference is critical 00:10:54.500 --> 00:10:57.059 for making actual decisions based on the AI's 00:10:57.059 --> 00:10:59.279 output. That's a game changer for critical tasks. 00:10:59.620 --> 00:11:01.580 OK, brilliant. What's next? OK, so we've talked 00:11:01.580 --> 00:11:03.519 about getting the AI to tell us how confident 00:11:03.519 --> 00:11:06.860 it is. But what if? What if the AI could actually 00:11:06.860 --> 00:11:08.960 help you write the perfect prompt to begin with? 00:11:09.080 --> 00:11:11.120 Wait, the AI helps write the prompt for the AI? 00:11:11.399 --> 00:11:13.919 Yes. This is where the AI prompt assistant technique 00:11:13.919 --> 00:11:16.639 comes in. It's genuinely revolutionary. Okay. 00:11:16.840 --> 00:11:19.019 Mind blown already. How does that work? Two main 00:11:19.019 --> 00:11:21.799 ways. First, say you're starting completely from 00:11:21.799 --> 00:11:25.379 scratch. You have a goal like, I need a three 00:11:25.379 --> 00:11:28.519 month Instagram content strategy for my sustainable 00:11:28.519 --> 00:11:32.049 fashion brand. But you have no idea how to structure 00:11:32.049 --> 00:11:34.549 the prompt for the best results. Yeah, happens 00:11:34.549 --> 00:11:37.750 all the time. You just tell the AI, OK, I want 00:11:37.750 --> 00:11:39.809 to build this three month content strategy for 00:11:39.809 --> 00:11:42.389 my sustainable fashion brand on Instagram. My 00:11:42.389 --> 00:11:45.809 goals are X and Y. Can you please write me the 00:11:45.809 --> 00:11:47.990 optimal prompt that I should then use to get 00:11:47.990 --> 00:11:50.230 the most detailed and effective plan from you? 00:11:50.269 --> 00:11:52.629 And it will write its own instructions. Yes, 00:11:52.750 --> 00:11:55.210 it will generate a beautifully structured prompt 00:11:55.210 --> 00:11:57.730 for you, probably using that six -part framework, 00:11:58.190 --> 00:12:00.110 specifying the role it should take, the context 00:12:00.110 --> 00:12:02.370 it needs, the tasks, the format, everything. 00:12:02.649 --> 00:12:05.429 It knows what it needs to do its best work. That's 00:12:05.429 --> 00:12:07.690 incredible. It saves you figuring out the perfect 00:12:07.690 --> 00:12:09.970 prompt structure yourself. Totally. Okay, the 00:12:09.970 --> 00:12:11.490 second approach is for when you already tried 00:12:11.490 --> 00:12:16.139 a prompt, but the results were, you know, You 00:12:16.139 --> 00:12:18.980 can go back to the AI and say, look, I use this 00:12:18.980 --> 00:12:21.720 prompt. Plan some Instagram content for my handmade 00:12:21.720 --> 00:12:25.360 pottery business. Give me ideas. But the results 00:12:25.360 --> 00:12:28.080 were really weak. Can you analyze my original 00:12:28.080 --> 00:12:30.440 prompt, tell me why it was probably ineffective, 00:12:30.659 --> 00:12:32.899 and then improve it so I get a much better outcome 00:12:32.899 --> 00:12:35.500 next time? And it critiques your prompt. It does. 00:12:35.500 --> 00:12:37.519 It'll point out what was missing. Like maybe 00:12:37.519 --> 00:12:39.419 you didn't specify the target audience or the 00:12:39.419 --> 00:12:42.759 brand style or your specific goals or the format 00:12:42.759 --> 00:12:45.159 you wanted. And then it will give you a revised, 00:12:45.159 --> 00:12:47.980 much stronger prompt structure to use. It literally 00:12:47.980 --> 00:12:50.600 becomes your prompt writing coach. Wow. OK. That 00:12:50.600 --> 00:12:52.740 changes the game from just using it as a tool 00:12:52.740 --> 00:12:55.120 to using it as a collaborator in the process 00:12:55.120 --> 00:12:57.200 itself. Exactly. It partners with you to get 00:12:57.200 --> 00:12:58.820 better results. All right. What else have you 00:12:58.820 --> 00:13:00.960 got in the advanced tool tip? Next up is something 00:13:00.960 --> 00:13:03.340 really crucial, especially now with different 00:13:03.340 --> 00:13:06.600 versions of models. available, the secret to 00:13:06.600 --> 00:13:09.100 choosing the right model. Like choosing between 00:13:09.100 --> 00:13:13.320 GPT -4 or Claude Opus or Gemini Pro, that kind 00:13:13.320 --> 00:13:16.059 of thing. Precisely. It's not just about which 00:13:16.059 --> 00:13:19.039 company's AI, but often which specific version 00:13:19.039 --> 00:13:21.700 or model within their offerings. Most people 00:13:21.700 --> 00:13:24.340 just use the default, but pros know that different 00:13:24.340 --> 00:13:27.039 models are optimized for different tasks. Picking 00:13:27.039 --> 00:13:29.960 the right one massively impacts output quality. 00:13:30.220 --> 00:13:32.500 Okay, so what are the main types? Broadly, you 00:13:32.500 --> 00:13:35.509 can think about it like this. You've got creative 00:13:35.509 --> 00:13:38.750 models, think GPT -4 in some modes, Claude III 00:13:38.750 --> 00:13:42.830 Opus maybe? These excel at tasks needing emotional 00:13:42.830 --> 00:13:45.730 nuance, creative writing, brainstorming, generating 00:13:45.730 --> 00:13:48.190 novel ideas. Right, for the artsy stuff. Kind 00:13:48.190 --> 00:13:50.590 of, yeah. Then you have analytical and reasoning 00:13:50.590 --> 00:13:53.889 models like Gemini Advance or GPT -4 Turbo. These 00:13:53.889 --> 00:13:57.129 are your powerhouses for complex, problem -solving, 00:13:57.289 --> 00:13:59.970 crunching data, logical deduction, detailed analysis. 00:14:00.169 --> 00:14:02.289 Heavy lifters. Exactly. And then you often have 00:14:02.289 --> 00:14:04.129 fast and light models, maybe Claude III, Sonnet, 00:14:04.210 --> 00:14:06.970 or Gemini Pro. These are great for quick, straightforward 00:14:06.970 --> 00:14:09.269 tasks where speed is important and deep complexity. 00:14:09.129 --> 00:14:11.669 isn't the main goal, like quick summaries or 00:14:11.669 --> 00:14:13.570 answering simple questions. OK, that makes sense. 00:14:13.649 --> 00:14:15.350 It's like having different tools in a toolbox. 00:14:15.870 --> 00:14:17.769 That's the perfect analogy. Yeah. You wouldn't 00:14:17.769 --> 00:14:19.610 use a sledgehammer to hang a picture frame, right? 00:14:19.870 --> 00:14:22.269 Hopefully not. Can you give a concrete example 00:14:22.269 --> 00:14:25.129 of how choosing the wrong model messes things 00:14:25.129 --> 00:14:27.590 up? Sure. Let's go back to that legal document 00:14:27.590 --> 00:14:30.600 example. OK. Imagine you need to summarize a 00:14:30.600 --> 00:14:33.879 dense five -page contract and identify potential 00:14:33.879 --> 00:14:36.639 risks. Okay, important task. If you feed that 00:14:36.639 --> 00:14:39.240 into a model optimized for analytical reasoning, 00:14:39.659 --> 00:14:42.639 it'll likely give you a precise summary, correctly 00:14:42.639 --> 00:14:45.759 flag the key clauses, and accurately pinpoint 00:14:45.759 --> 00:14:48.379 the potential risks and ambiguities. What you 00:14:48.379 --> 00:14:50.980 need. Right. Now run the exact same prompt on 00:14:50.980 --> 00:14:53.820 a model primarily designed for speed or maybe 00:14:53.820 --> 00:14:56.059 creative writing. You might get a summary that 00:14:56.059 --> 00:14:58.559 sounds okay on the surface, but it can easily 00:14:58.559 --> 00:15:01.360 miss crucial nuances, misinterpret a subtle legal 00:15:01.360 --> 00:15:04.399 point, or fail to identify a significant risk 00:15:04.399 --> 00:15:07.000 because its strength isn't in that kind of deep 00:15:07.000 --> 00:15:09.860 logical parsing. Wow, okay. So the same prompt 00:15:09.860 --> 00:15:12.360 can yield dramatically different quality and 00:15:12.360 --> 00:15:14.240 reliability depending on the engine running it. 00:15:14.440 --> 00:15:16.279 Absolutely. Using the right tool for the job 00:15:16.279 --> 00:15:18.940 is paramount, especially for high stakes tasks. 00:15:19.399 --> 00:15:21.519 It's not just about the prompt. It's the prompt 00:15:21.519 --> 00:15:24.389 plus the right model. Got it. Use the right knife 00:15:24.389 --> 00:15:27.250 for the job. What's next? Okay, technique number 00:15:27.250 --> 00:15:31.450 four, the self -improvement loop. This one feels 00:15:31.450 --> 00:15:34.210 like unlocking a hidden superpower. Self -improvement 00:15:34.210 --> 00:15:37.759 for the AI. Yes. instead of you getting a mediocre 00:15:37.759 --> 00:15:40.820 result, getting frustrated, and trying to rewrite 00:15:40.820 --> 00:15:42.659 the prompt from scratch. Which I've definitely 00:15:42.659 --> 00:15:45.639 done many times. We all have. Instead, you ask 00:15:45.639 --> 00:15:48.259 the AI to critique its own previous response 00:15:48.259 --> 00:15:50.860 and then improve it based on that critique. Wait, 00:15:50.919 --> 00:15:52.960 really? You tell it to grade its own homework 00:15:52.960 --> 00:15:55.259 and then do it again better? Exactly. It sounds 00:15:55.259 --> 00:15:57.740 crazy, but it works incredibly well. Imagine 00:15:57.740 --> 00:16:00.240 you use the six -part framework to get, say, 00:16:00.519 --> 00:16:03.620 a draft for a 60 -second TikTok video script 00:16:03.620 --> 00:16:05.659 about a language learning app. OK, you get a 00:16:05.659 --> 00:16:08.019 first draft. Probably decent. Right, probably 00:16:08.019 --> 00:16:10.419 OK. But then you follow up with this instruction. 00:16:10.960 --> 00:16:13.639 Analyze your previous response. Identify three 00:16:13.639 --> 00:16:16.440 specific weaknesses in the script. Then... rewrite 00:16:16.440 --> 00:16:18.899 the entire script specifically to address those 00:16:18.899 --> 00:16:21.360 three weaknesses. Let's do this process three 00:16:21.360 --> 00:16:23.740 times in total, focusing on different potential 00:16:23.740 --> 00:16:27.740 issues each round. Three times. Wow. Yeah. So 00:16:27.740 --> 00:16:31.980 round one, the AI might say, OK, the hook wasn't 00:16:31.980 --> 00:16:34.480 shocking enough. The call to action was a bit 00:16:34.480 --> 00:16:37.139 weak and the pacing felt slow. Then it rewrites 00:16:37.139 --> 00:16:39.720 the script, fixing those things. OK. Then you 00:16:39.720 --> 00:16:42.450 run the instruction again. Round two, it might 00:16:42.450 --> 00:16:44.549 analyze the new script and say, all right, the 00:16:44.549 --> 00:16:46.490 benefits are listed too dryly, there's no humor, 00:16:46.629 --> 00:16:48.509 and the example scenario isn't very relatable. 00:16:49.009 --> 00:16:51.350 Then it rewrites it again, addressing those points. 00:16:51.590 --> 00:16:54.230 And round three. Maybe it focuses on character 00:16:54.230 --> 00:16:57.309 voice or visual suggestions or tightening the 00:16:57.309 --> 00:17:00.090 language further. Each iteration makes the script 00:17:00.090 --> 00:17:02.889 stronger, moving it from good to great to potentially 00:17:02.889 --> 00:17:05.089 excellent. That's amazing. It's like having an 00:17:05.089 --> 00:17:07.210 internal quality control loop built right in. 00:17:07.390 --> 00:17:09.430 without me having to pinpoint all the flaws. 00:17:09.910 --> 00:17:12.170 Exactly. It takes the burden of detailed critique 00:17:12.170 --> 00:17:15.450 off you and leverages the AI's own analytical 00:17:15.450 --> 00:17:18.549 capabilities to refine its output. It's incredibly 00:17:18.549 --> 00:17:20.990 efficient for improving quality. That's a technique 00:17:20.990 --> 00:17:23.630 I'm definitely trying later today. Number five 00:17:23.630 --> 00:17:26.369 is super simple but incredibly effective. Step 00:17:26.369 --> 00:17:28.490 -by -step thinking. You might also hear this 00:17:28.490 --> 00:17:31.170 called chain of thought prompting in some more 00:17:31.170 --> 00:17:33.990 technical circles. Step -by -step thinking. What's 00:17:33.990 --> 00:17:36.680 the magic phrase? Literally just adding the instruction 00:17:36.680 --> 00:17:40.359 think step by step add that phrase to pretty 00:17:40.359 --> 00:17:43.359 much any prompt Asking for something strategic 00:17:43.359 --> 00:17:47.279 complex or requiring logical reasoning just think 00:17:47.279 --> 00:17:50.099 step by step. That's it That's often enough. 00:17:50.500 --> 00:17:53.599 Its power is seriously underestimated study after 00:17:53.599 --> 00:17:56.240 study and just practical experience shows that 00:17:56.240 --> 00:17:58.779 prompts using this consistently deliver better, 00:17:59.160 --> 00:18:01.779 clearer, more reliable results, especially for 00:18:01.779 --> 00:18:03.960 things like business plans, marketing strategies, 00:18:04.359 --> 00:18:06.559 complex analysis, anything where the reasoning 00:18:06.559 --> 00:18:09.319 process matters. Why? What does adding that phrase 00:18:09.319 --> 00:18:12.259 actually do? It essentially forces the AI to 00:18:12.259 --> 00:18:15.279 externalize its internal thought process. Instead 00:18:15.279 --> 00:18:17.660 of just jumping to the final conclusion or recommendation, 00:18:18.180 --> 00:18:20.160 it has to first outline the logical steps it's 00:18:20.160 --> 00:18:22.740 taking to get there. Ah, so it shows its work, 00:18:22.839 --> 00:18:25.099 like in math class. Exactly like showing its 00:18:25.099 --> 00:18:27.579 work. Let's take that cold brew coffee launch 00:18:27.579 --> 00:18:29.700 plan example. A basic prompt might just give 00:18:29.700 --> 00:18:32.099 you a list of tactics. Okay, maybe useful. Right, 00:18:32.220 --> 00:18:34.839 a list. But add think step -by -step to that 00:18:34.839 --> 00:18:37.440 prompt. Now, the AI will likely structure his 00:18:37.440 --> 00:18:39.440 response differently. It might start with, okay, 00:18:39.480 --> 00:18:42.019 step one, analyze the target audience for cold 00:18:42.019 --> 00:18:45.200 brew. Then step two, define the product positioning 00:18:45.200 --> 00:18:48.259 and key messaging. Then step three, identify 00:18:48.259 --> 00:18:51.490 optimal communication channels. Then step four, 00:18:51.750 --> 00:18:54.390 develop a detailed 30 -day content calendar. 00:18:55.309 --> 00:18:57.789 You see the structure? Yes. It's laying out the 00:18:57.789 --> 00:19:00.130 whole strategic process, not just the final action 00:19:00.130 --> 00:19:03.089 items. You can follow the logic. Precisely. Each 00:19:03.089 --> 00:19:05.890 step builds on the last. It moves the output 00:19:05.890 --> 00:19:09.210 from just generic advice to a well -reasoned 00:19:09.210 --> 00:19:12.029 structured strategic plan. It's much more robust 00:19:12.029 --> 00:19:14.250 and trustworthy. That's incredibly valuable for 00:19:14.250 --> 00:19:16.390 anything strategic. Okay, one more advanced technique. 00:19:16.529 --> 00:19:19.420 Last one for today, the priming technique. This 00:19:19.420 --> 00:19:21.740 is about warming up the AI's brain on a topic 00:19:21.740 --> 00:19:24.160 before you ask your specific question. Priming, 00:19:24.240 --> 00:19:26.720 like getting it ready. Exactly. Instead of diving 00:19:26.720 --> 00:19:29.119 straight into a very specific request, you first 00:19:29.119 --> 00:19:31.900 ask a broader, related question. This activates 00:19:31.900 --> 00:19:33.960 all the relevant knowledge and concepts within 00:19:33.960 --> 00:19:36.440 the AI's network related to your topic. Okay, 00:19:36.440 --> 00:19:38.359 how does that work in practice? Let's say your 00:19:38.359 --> 00:19:40.779 goal is to create an email marketing campaign 00:19:40.779 --> 00:19:43.779 to improve customer retention for a SaaS product. 00:19:43.839 --> 00:19:46.480 Okay, specific goal. Before you ask for campaign 00:19:46.480 --> 00:19:49.579 ideas, you prime the AI. You have something broader, 00:19:49.839 --> 00:19:52.500 like, what are the key psychological principles 00:19:52.500 --> 00:19:54.579 that are known to increase customer retention 00:19:54.579 --> 00:19:58.059 and build brand loyalty, especially in a subscription 00:19:58.059 --> 00:20:00.740 -based business model? Ah, so you're asking for 00:20:00.740 --> 00:20:03.200 the theory first. Right. The AI will then explain 00:20:03.200 --> 00:20:05.559 concepts like, you know, the endowment effect, 00:20:06.180 --> 00:20:08.680 reciprocity, loss aversion, the importance of 00:20:08.680 --> 00:20:10.839 community, self -determination theory, stuff 00:20:10.839 --> 00:20:13.460 like that. It activates that knowledge. Its brain 00:20:13.460 --> 00:20:16.240 is warmed up on retention psychology. Now what? 00:20:16.539 --> 00:20:19.079 Now you hit it with your specific request. referencing 00:20:19.079 --> 00:20:22.420 the priming. Okay. You are an expert in customer 00:20:22.420 --> 00:20:25.039 retention marketing. Based specifically on the 00:20:25.039 --> 00:20:27.440 psychological principles you just outlined, propose 00:20:27.440 --> 00:20:30.140 five concrete strategies I can use in my email 00:20:30.140 --> 00:20:33.400 campaign to reduce our customer churn rate. Please 00:20:33.400 --> 00:20:35.339 present this as a structured plan, maybe with 00:20:35.339 --> 00:20:37.759 an example email subject line for each strategy. 00:20:37.819 --> 00:20:40.660 Wow! And the result? The results are usually 00:20:40.660 --> 00:20:43.859 incredibly impressive. Because you primed it, 00:20:43.980 --> 00:20:46.660 The AI doesn't just give generic tips like offer 00:20:46.660 --> 00:20:49.700 discounts. It starts applying those deeper psychological 00:20:49.700 --> 00:20:52.579 principles directly to your problem, leading 00:20:52.579 --> 00:20:55.839 to much more sophisticated, insightful, and genuinely 00:20:55.839 --> 00:20:58.500 strategic recommendations. It connects the theory 00:20:58.500 --> 00:21:00.640 to the practice because you prompted it to do 00:21:00.640 --> 00:21:03.799 both. Exactly. It leads to much deeper thinking. 00:21:04.400 --> 00:21:07.339 These advanced techniques They really shift the 00:21:07.339 --> 00:21:09.440 interaction, don't they? It's less about just 00:21:09.440 --> 00:21:11.920 getting an answer and more about guiding the 00:21:11.920 --> 00:21:15.500 AI to generate a high -quality, reliable strategic 00:21:15.500 --> 00:21:17.400 answer. That's the whole point. And it brings 00:21:17.400 --> 00:21:20.299 us to a really important mindset shift. A common 00:21:20.299 --> 00:21:23.000 trap people fall into is they finally get one 00:21:23.000 --> 00:21:24.839 prompt to work well, get a good result, and they 00:21:24.839 --> 00:21:27.099 stop there. They celebrate the win. Yeah, relief. 00:21:27.240 --> 00:21:29.579 Finally got something useful. Right. But true 00:21:29.579 --> 00:21:31.779 prompt engineering, building real, reliable systems, 00:21:32.380 --> 00:21:34.740 involves more. It's about testing that prompt, 00:21:35.099 --> 00:21:37.259 seeing where it might fail under different conditions, 00:21:37.779 --> 00:21:40.119 refining it, tweaking it until it's basically 00:21:40.119 --> 00:21:42.539 foolproof for that specific task. Moving from 00:21:42.539 --> 00:21:45.700 one -off successes to repeatable processes. Exactly. 00:21:46.039 --> 00:21:48.880 And that's why we strongly advocate for creating 00:21:48.880 --> 00:21:52.000 your own personal prompt library. A library of 00:21:52.000 --> 00:21:55.140 prompts that work. Yes. As you develop and test 00:21:55.140 --> 00:21:57.599 prompts, using the framework and these techniques, 00:21:58.160 --> 00:22:00.180 the ones that consistently deliver excellent 00:22:00.180 --> 00:22:03.750 results for specific tasks, save them. Make sense. 00:22:04.349 --> 00:22:06.549 Maybe by category like marketing, sales, data 00:22:06.549 --> 00:22:08.910 analysis, code generation, writing, whatever 00:22:08.910 --> 00:22:12.329 you do. Use a tool like Notion, Google Docs, 00:22:12.630 --> 00:22:14.930 Evernote, whatever works for you. And the benefit. 00:22:15.170 --> 00:22:18.809 The benefit is huge. Over time, you build this 00:22:18.809 --> 00:22:21.190 arsenal of tested, reliable prompts. Need to 00:22:21.190 --> 00:22:22.910 write a certain type of email. Boom, grab your 00:22:22.910 --> 00:22:24.990 prompt. Need to analyze sales data in a specific 00:22:24.990 --> 00:22:27.349 way. Grab that prompt. You're not starting from 00:22:27.349 --> 00:22:29.509 scratch or guessing each time. You're handling 00:22:29.509 --> 00:22:32.619 recurring tasks in seconds. Reliably. Takes the 00:22:32.619 --> 00:22:34.960 friction out of it and ensures consistency. I 00:22:34.960 --> 00:22:37.299 like that. It's a massive time saver and quality 00:22:37.299 --> 00:22:39.859 booster long term. OK, and what about for really, 00:22:39.940 --> 00:22:42.539 really critical stuff like decisions with major 00:22:42.539 --> 00:22:45.519 consequences? Is there another layer of quality 00:22:45.519 --> 00:22:48.420 control? Absolutely. For those truly critical 00:22:48.420 --> 00:22:51.339 decisions, maybe a major business strategy pivot 00:22:51.339 --> 00:22:55.420 or a sensitive communication. We recommend what 00:22:55.420 --> 00:22:57.980 you could call advanced quality control. Sounds 00:22:57.980 --> 00:23:00.539 serious. It involves a bit more effort. but it's 00:23:00.539 --> 00:23:03.599 worth it for high stakes. Yeah. The idea is run 00:23:03.599 --> 00:23:06.500 your carefully crafted prompt not just on one 00:23:06.500 --> 00:23:09.059 AI tool, but across multiple different ones. 00:23:09.140 --> 00:23:12.720 Ah, so run it on Gemini and Claude and ChetGPT, 00:23:12.799 --> 00:23:15.119 for instance. Exactly. Get responses from several 00:23:15.119 --> 00:23:17.500 of the top models using the same robust prompt. 00:23:18.019 --> 00:23:20.299 They'll likely have different strengths. It might 00:23:20.299 --> 00:23:22.339 offer slightly different perspectives or solutions. 00:23:22.500 --> 00:23:24.839 OK, so now you have multiple outputs. Then what? 00:23:24.980 --> 00:23:27.599 Then you take those different responses and feed 00:23:27.599 --> 00:23:29.900 them back into one of the AI tools. preferably 00:23:29.900 --> 00:23:32.559 a strong analytical one, and you ask it to analyze 00:23:32.559 --> 00:23:35.359 and compare all the responses. Ask the AI to 00:23:35.359 --> 00:23:38.440 judge the other AIs. Sort of. Ask it to synthesize 00:23:38.440 --> 00:23:40.859 the common themes, identify the unique strengths 00:23:40.859 --> 00:23:43.500 or weaknesses of each response, and perhaps even 00:23:43.500 --> 00:23:45.660 recommend the best overall approach or combine 00:23:45.660 --> 00:23:48.799 the best elements. That's meta. But I can see 00:23:48.799 --> 00:23:51.839 how comparing multiple expert opinions from different 00:23:51.839 --> 00:23:54.519 AIs and then having one synthesize them would 00:23:54.519 --> 00:23:57.200 lead to a much more robust final decision. It 00:23:57.200 --> 00:23:59.920 really does. It moves you beyond just hoping 00:23:59.920 --> 00:24:02.359 one AI got it right towards building a system 00:24:02.359 --> 00:24:05.200 that leverages multiple perspectives for maximum 00:24:05.200 --> 00:24:07.400 reliability. It's the difference, as you said, 00:24:07.460 --> 00:24:10.039 between celebrating single wins and building 00:24:10.039 --> 00:24:13.039 truly effective dependable systems. Wow. OK. 00:24:13.039 --> 00:24:16.059 So we've covered a lot how AI thinks the foundational 00:24:16.079 --> 00:24:18.400 framework, advanced techniques, building systems. 00:24:18.880 --> 00:24:20.920 What's the immediate takeaway for someone listening 00:24:20.920 --> 00:24:23.319 right now? What should they do first? Great question. 00:24:23.420 --> 00:24:25.619 Let's make it super actionable. Here's your plan 00:24:25.619 --> 00:24:28.799 straight up. First, pick one single task, something 00:24:28.799 --> 00:24:31.660 you regularly ask an AI to do. Could be writing 00:24:31.660 --> 00:24:33.819 certain emails, drafting social media content, 00:24:33.980 --> 00:24:36.140 summarizing reports, analyzing some data, whatever 00:24:36.140 --> 00:24:38.759 is common for you. Just one task to start, OK? 00:24:38.900 --> 00:24:41.200 Second, build a prompt for that task using the 00:24:41.200 --> 00:24:43.779 six -part framework we went through. Roll, context, 00:24:44.039 --> 00:24:46.480 task, format, rules, examples. Take the time 00:24:46.480 --> 00:24:48.759 to really flesh out each component. Got it. Build 00:24:48.759 --> 00:24:51.440 the foundation. Third, layer on maybe two or 00:24:51.440 --> 00:24:53.660 three of those advanced techniques we discussed. 00:24:54.079 --> 00:24:56.799 Maybe add the confidence verification or think 00:24:56.799 --> 00:24:59.640 step -by -step or try the priming technique if 00:24:59.640 --> 00:25:02.200 it fits. Enhance the prompt. Fourth, and this 00:25:02.200 --> 00:25:05.660 is key, test it. Run the prompt several times. 00:25:05.740 --> 00:25:08.819 See what you get back. Does it consistently deliver? 00:25:09.119 --> 00:25:12.430 Where does it fall short? Tweak the prompt. refine 00:25:12.430 --> 00:25:15.430 it until the output is genuinely useful with 00:25:15.430 --> 00:25:18.069 only minimal editing needed from you. Iterate 00:25:18.069 --> 00:25:21.069 and improve. And finally, fifth, once you have 00:25:21.069 --> 00:25:23.710 a prompt that reliably works for that task, save 00:25:23.710 --> 00:25:25.730 it. That's the first entry in your personal prompt 00:25:25.730 --> 00:25:28.309 library. Start building the library. Exactly. 00:25:28.569 --> 00:25:31.269 Just doing this process for one recurring task 00:25:31.269 --> 00:25:33.869 will fundamentally change how you see and use 00:25:33.869 --> 00:25:36.349 AI. It puts you way ahead of the curve into that 00:25:36.349 --> 00:25:38.630 small group of people who really understand how 00:25:38.630 --> 00:25:40.589 to leverage these tools professionally. That 00:25:40.589 --> 00:25:42.990 feels achievable. Start small, build confidence, 00:25:43.289 --> 00:25:45.089 see the difference. That's the way. So wrapping 00:25:45.089 --> 00:25:47.130 up, it feels like the core message here is pretty 00:25:47.130 --> 00:25:50.390 clear. Yeah, I think, ultimately, the real difference 00:25:50.390 --> 00:25:52.869 between someone who just dabbles with AI and 00:25:52.869 --> 00:25:55.589 someone who uses it expertly isn't necessarily 00:25:55.589 --> 00:25:58.569 about having access to the fanciest, newest tools. 00:25:58.789 --> 00:26:00.809 It's about how they communicate with those tools. 00:26:00.910 --> 00:26:03.089 It's the quality of the conversation. It really 00:26:03.089 --> 00:26:05.150 is a communication skill. And hopefully, everyone 00:26:05.150 --> 00:26:07.769 listening now has a much better grasp of how 00:26:07.769 --> 00:26:11.390 AI actually works that probabilistic prediction 00:26:11.390 --> 00:26:13.970 engine. Right, understanding the why behind the 00:26:13.970 --> 00:26:16.430 output. And knowing how to get it to show its 00:26:16.430 --> 00:26:18.809 work with context. levels or step -by -step thinking, 00:26:18.869 --> 00:26:21.490 how to make it improve its own work, how to prime 00:26:21.490 --> 00:26:23.490 it for deeper insights. Even though it always 00:26:23.490 --> 00:26:25.829 sounds confident, you now know how to look behind 00:26:25.829 --> 00:26:27.789 the curtain a bit. You're equipped to have a 00:26:27.789 --> 00:26:29.869 much more intelligent, productive conversation 00:26:29.869 --> 00:26:32.589 with it. So the call to action is simple. Pick 00:26:32.589 --> 00:26:35.890 that one task. Build that one great prompt using 00:26:35.890 --> 00:26:39.190 these techniques. Test it. Experience that night 00:26:39.190 --> 00:26:41.490 and day difference for yourself. I guarantee 00:26:41.490 --> 00:26:43.849 once you see what's possible, you'll never want 00:26:43.849 --> 00:26:46.190 to go back to just basic vague prompting again. 00:26:46.349 --> 00:26:49.349 It opens up a whole new level of partnership 00:26:49.349 --> 00:26:51.910 with these tools. So maybe the final thought 00:26:51.910 --> 00:26:54.250 for everyone listening is this. The AI revolution 00:26:54.250 --> 00:26:56.109 isn't just about the fact that these powerful 00:26:56.109 --> 00:26:59.109 tools exist and are accessible. The real revolution 00:26:59.109 --> 00:27:01.910 for you personally and professionally comes when 00:27:01.910 --> 00:27:04.470 you learn how to wield them like an expert. So 00:27:04.470 --> 00:27:07.069 what could you truly achieve if your AI stopped 00:27:07.069 --> 00:27:09.509 being a sometimes helpful, sometimes frustrating 00:27:09.509 --> 00:27:12.289 tool and instead became a genuinely reliable, 00:27:12.549 --> 00:27:14.349 strategically intelligent partner in your work 00:27:14.349 --> 00:27:15.289 in your biggest projects?