WEBVTT 00:00:00.000 --> 00:00:01.459 You know, it often feels like most of us, when 00:00:01.459 --> 00:00:03.899 you use something like chat GPT, it's like we've 00:00:03.899 --> 00:00:07.799 been handed the keys to this incredible futuristic 00:00:07.799 --> 00:00:10.619 spaceship. Yeah. But we only ever seem to use 00:00:10.619 --> 00:00:13.259 it to order pizza. That's a great way to put 00:00:13.259 --> 00:00:15.240 it. We sense the power, right? This thing could 00:00:15.240 --> 00:00:17.760 go interstellar, but somehow we feel like we're 00:00:17.760 --> 00:00:20.519 stuck in first gear. Exactly. there's that feeling 00:00:20.519 --> 00:00:22.519 you're missing something deeper, just scratching 00:00:22.519 --> 00:00:25.420 the surface. And it's moving beyond just asking 00:00:25.420 --> 00:00:27.539 for answers, isn't it? It's more about, like, 00:00:27.539 --> 00:00:30.440 engineering thought itself. That's the core idea. 00:00:31.519 --> 00:00:35.179 Welcome, everyone, to the Deep Dive. Today we're 00:00:35.179 --> 00:00:37.340 really digging into how you can shift your use 00:00:37.340 --> 00:00:41.100 of ChatGPT from a simple answer machine to, well, 00:00:41.200 --> 00:00:43.420 a true extension of your own thinking. We're 00:00:43.420 --> 00:00:45.100 pulling from Max Anne's Beyond the Answer Machine, 00:00:45.280 --> 00:00:47.960 and the goal today is pretty ambitious. It is. 00:00:48.320 --> 00:00:50.200 We want to give you a comprehensive playbook, 00:00:50.299 --> 00:00:53.320 really. We'll cover the basics, the best practices. 00:00:53.520 --> 00:00:55.600 Then we'll get into project -based workflows, 00:00:56.039 --> 00:00:58.859 specific problem -solving prompts, critical thinking 00:00:58.859 --> 00:01:01.880 exercises. And even how to teach the AI how to 00:01:01.880 --> 00:01:04.260 think. Right, which all comes down to prompt 00:01:04.260 --> 00:01:06.560 engineering. That's just the art of crafting 00:01:06.560 --> 00:01:08.840 really effective instructions for the AI. It's 00:01:08.840 --> 00:01:10.859 key. Okay, so let's start with that foundation. 00:01:11.260 --> 00:01:14.739 The guide emphasizes four core practices that 00:01:14.739 --> 00:01:17.829 many people... uh tend to overlook yeah these 00:01:17.829 --> 00:01:20.030 are fundamental if you want to get serious first 00:01:20.030 --> 00:01:23.650 up specify your output format don't just ask 00:01:23.650 --> 00:01:26.010 for stuff command the structure you want meaning 00:01:26.010 --> 00:01:28.549 meaning tell it you want a checklist or a table 00:01:28.549 --> 00:01:31.230 with specific columns like feature benefit use 00:01:31.230 --> 00:01:35.189 case maybe you need valid json data or a youtube 00:01:35.189 --> 00:01:38.230 script outline be explicit it needs that direction 00:01:38.230 --> 00:01:41.269 ah okay so instead of just ideas it's give me 00:01:41.269 --> 00:01:43.510 five ideas formatted as a list with summaries 00:01:43.510 --> 00:01:46.849 got it exactly second is The job title trick. 00:01:47.049 --> 00:01:49.170 Even though these models are smart, giving the 00:01:49.170 --> 00:01:52.150 AI a specific persona or job title is still incredibly 00:01:52.150 --> 00:01:54.189 useful. Like telling it to be a specific kind 00:01:54.189 --> 00:01:56.650 of expert. Precisely. Act as a financial advisor 00:01:56.650 --> 00:01:59.790 with 20 years experience. Or you are an expert 00:01:59.790 --> 00:02:02.329 health coach for busy parents. Or even something 00:02:02.329 --> 00:02:04.310 fun like you're a drunken pirate if you need 00:02:04.310 --> 00:02:06.629 entertaining coffee. It frames the knowledge 00:02:06.629 --> 00:02:09.050 and the tone. Okay, that makes sense. It immediately 00:02:09.050 --> 00:02:12.430 focuses the AI. And the third one, I think this 00:02:12.430 --> 00:02:15.250 might be the most critical, be ruthlessly specific. 00:02:15.689 --> 00:02:17.930 Oh, absolutely. It's the classic garbage in, 00:02:17.969 --> 00:02:21.159 garbage out principle, right? Vague prompts get 00:02:21.159 --> 00:02:24.000 vague answers. An amateur asks, give me video 00:02:24.000 --> 00:02:26.819 ideas. Right. A power user says, give me five 00:02:26.819 --> 00:02:30.060 video title ideas under 50 characters focused 00:02:30.060 --> 00:02:33.319 on creating FOMO for productivity content, targeting 00:02:33.319 --> 00:02:35.680 short form video platforms. See the difference? 00:02:36.139 --> 00:02:38.979 Context, constraints, detail. That's gold. Okay. 00:02:39.039 --> 00:02:41.860 And the last one. Master iteration loops in a 00:02:41.860 --> 00:02:45.520 single prompt. This is powerful. You make the 00:02:45.520 --> 00:02:48.340 AI improve its own work within one command. How 00:02:48.340 --> 00:02:50.280 does that work? Like ask it to draft a marketing 00:02:50.280 --> 00:02:53.560 email, then in the same prompt, tell it to critique 00:02:53.560 --> 00:02:56.240 the weaknesses of that draft, and then rewrite 00:02:56.240 --> 00:02:58.340 the email based on its own critique. Wow. So 00:02:58.340 --> 00:03:00.539 it becomes its own editor. Basically, yeah. A 00:03:00.539 --> 00:03:02.960 whole creative team in one prompt. So what's 00:03:02.960 --> 00:03:05.400 the immediate shift someone can feel just by 00:03:05.400 --> 00:03:08.060 applying these basics? Being explicit format, 00:03:08.419 --> 00:03:11.719 persona, detail makes the AI far more useful 00:03:11.719 --> 00:03:14.060 right away. Okay, that's the foundation. Now 00:03:14.060 --> 00:03:17.580 let's level up. Moving from single prompts to 00:03:17.580 --> 00:03:23.129 more systematic workflows using chat GPTs. Projects 00:03:23.129 --> 00:03:24.770 feature. Yeah. You mentioned this lets you create 00:03:24.770 --> 00:03:27.729 specialized AI assistance. Exactly. Projects 00:03:27.729 --> 00:03:29.969 are like dedicated workspaces. And the magic 00:03:29.969 --> 00:03:33.689 ingredient is custom instructions. Think of these 00:03:33.689 --> 00:03:36.550 as the AI's permanent prime directive for that 00:03:36.550 --> 00:03:39.289 specific project. So it remembers context within 00:03:39.289 --> 00:03:41.990 that project. It provides persistent context. 00:03:42.310 --> 00:03:45.409 The AI learns your specific needs, your style, 00:03:45.590 --> 00:03:48.870 your goals for that task over time. It gets smarter 00:03:48.870 --> 00:03:51.449 about helping you specifically instead of starting 00:03:51.449 --> 00:03:53.969 fresh. every time. That sounds like a huge efficiency 00:03:53.969 --> 00:03:55.889 boost. Can you give some examples of these projects? 00:03:56.150 --> 00:03:58.229 Sure. A really popular one is the personal journal 00:03:58.229 --> 00:04:00.629 assistant. Here, your custom instructions tell 00:04:00.629 --> 00:04:03.360 the AI its role. Listen, give feedback, maybe 00:04:03.360 --> 00:04:05.979 point out blind spots, offer support, or even 00:04:05.979 --> 00:04:07.919 tough love if you asked for it. And you feed 00:04:07.919 --> 00:04:10.979 it personal context. Critically, yes. Your name, 00:04:11.099 --> 00:04:13.400 maybe your job, your goals, recurring challenges. 00:04:13.560 --> 00:04:15.520 Then you can just brain dump your thoughts. And 00:04:15.520 --> 00:04:18.420 because the AI has all that background. It gives 00:04:18.420 --> 00:04:21.079 you a more objective, informed perspective. Like 00:04:21.079 --> 00:04:22.839 a super patient sounding board that actually 00:04:22.839 --> 00:04:25.519 knows you. Precisely. Another great one is the 00:04:25.519 --> 00:04:28.160 explain this for me research analyst. Essential 00:04:28.160 --> 00:04:31.220 if you need to grasp complex stuff quickly. Like 00:04:31.220 --> 00:04:33.139 for research papers or dense articles? Yeah. 00:04:33.279 --> 00:04:35.720 The custom instructions make it an expert at 00:04:35.720 --> 00:04:38.240 breaking things down with analogies. And you 00:04:38.240 --> 00:04:41.060 specify a very structured output format. What 00:04:41.060 --> 00:04:44.339 kind of format? Things like a simple explanation 00:04:44.339 --> 00:04:46.759 for a five -year -old, a high -level summary, 00:04:46.939 --> 00:04:49.220 key takeaways, a good analogy, what's new about 00:04:49.220 --> 00:04:52.220 this, future implications, a quick soundbite, 00:04:52.259 --> 00:04:55.339 and crucially... Why should someone care? So 00:04:55.339 --> 00:04:58.199 you could feed it a super technical paper, say, 00:04:58.319 --> 00:05:01.259 on a new AI game engine. And bam, you get an 00:05:01.259 --> 00:05:04.100 instantly digestible structured breakdown. No 00:05:04.100 --> 00:05:06.620 waiting through jargon for hours. How do these 00:05:06.620 --> 00:05:08.800 custom assistants really make everyday tasks 00:05:08.800 --> 00:05:11.300 easier? They just remove so much mental load 00:05:11.300 --> 00:05:13.120 and decision fatigue because they already know 00:05:13.120 --> 00:05:15.420 your context. Right. Okay, we've got our AI assistant 00:05:15.420 --> 00:05:18.680 set up. Now let's talk specific copy -pastable 00:05:18.680 --> 00:05:21.680 prompts for real -world problems, the life -changing 00:05:21.680 --> 00:05:23.730 stuff. Absolutely. Let's start with something 00:05:23.730 --> 00:05:26.449 practical. The wellness optimizer. The prompt 00:05:26.449 --> 00:05:29.709 is super direct. List three simple but powerful 00:05:29.709 --> 00:05:32.350 wellness habits for sleep, digestion, and mental 00:05:32.350 --> 00:05:35.550 clarity. Break them down. Morning, midday, evening. 00:05:35.709 --> 00:05:38.629 Each habit must take less than five minutes for 00:05:38.629 --> 00:05:41.250 a very busy schedule. Ah, so it cuts through 00:05:41.250 --> 00:05:43.250 the noise and gives you actionable, quick wins 00:05:43.250 --> 00:05:46.009 that fit a real life. The more vague advice. 00:05:46.449 --> 00:05:49.269 Exactly. Now, what about learning? So many people 00:05:49.269 --> 00:05:51.930 get stuck with, where do I even start? Or they 00:05:51.930 --> 00:05:53.769 read something and it just goes in one ear, out 00:05:53.769 --> 00:05:55.930 the other. Yeah, the retention problem. AI is 00:05:55.930 --> 00:05:58.410 a game changer here. For the where do I start 00:05:58.410 --> 00:06:02.209 paralysis, use the 30 -day master plan. You basically 00:06:02.209 --> 00:06:05.110 ask the AI to be a world -class curriculum designer 00:06:05.110 --> 00:06:07.389 for whatever hobby you want to learn. How specific 00:06:07.389 --> 00:06:10.470 do you get? Very. Create a detailed 30 -day learning 00:06:10.470 --> 00:06:13.629 plan for hobby, say, drone photography, taking 00:06:13.629 --> 00:06:16.529 a beginner to intermediate. Assume 45 minutes 00:06:16.529 --> 00:06:19.350 practice daily. Include theory, practical exercises, 00:06:19.730 --> 00:06:22.370 weekly goals, gear recommendations, and a final 00:06:22.370 --> 00:06:24.209 capstone project. Wow, that's like getting a 00:06:24.209 --> 00:06:26.410 personalized course outline. It really is. Someone 00:06:26.410 --> 00:06:29.189 used this for drone photography, and the AI laid 00:06:29.189 --> 00:06:33.389 out week one, hovering landing, week two, edit 00:06:33.389 --> 00:06:37.910 a 30 -set clip, week three. Shoot a 60 -sec sequence 00:06:37.910 --> 00:06:41.600 straight from camera. Week four, plan and edit 00:06:41.600 --> 00:06:44.060 a two, three minute mini film. It just removes 00:06:44.060 --> 00:06:46.199 all the guesswork. OK, brilliant. And for the 00:06:46.199 --> 00:06:48.680 retention issue, the in one ear, out the other 00:06:48.680 --> 00:06:50.800 problem. That's where active recall comes in 00:06:50.800 --> 00:06:53.500 using the learning reinforcement quiz. Simple 00:06:53.500 --> 00:06:56.060 prompt. Act as my study partner. I just studied 00:06:56.060 --> 00:06:58.860 topic. Ask me 10 challenging questions. Don't 00:06:58.860 --> 00:07:01.279 give answers until I try all 10. Then provide 00:07:01.279 --> 00:07:03.439 a detailed answer piece. So it forces you to 00:07:03.439 --> 00:07:05.439 retrieve the information, not just passively 00:07:05.439 --> 00:07:07.459 reread it. Exactly. It makes learning stick. 00:07:07.620 --> 00:07:09.060 And if you want to go deeper, there's the secret. 00:07:09.540 --> 00:07:22.819 How does that work? So it leads you to the understanding 00:07:22.819 --> 00:07:25.560 instead of just giving it to you. Mm -hmm. It's 00:07:25.560 --> 00:07:28.259 a fantastic way to build real intuition. You 00:07:28.259 --> 00:07:29.819 know, I still wrestle with prompt drift myself 00:07:29.819 --> 00:07:32.519 sometimes, getting the wording just right. But 00:07:32.519 --> 00:07:35.420 this Socratic approach, it really pushes past 00:07:35.420 --> 00:07:37.980 surface -level understanding for me. So how do 00:07:37.980 --> 00:07:40.160 these prompts truly accelerate personal growth? 00:07:40.480 --> 00:07:42.939 They remove friction and turn passive learning 00:07:42.939 --> 00:07:46.220 into active, tailored practice. Much more effective. 00:07:46.639 --> 00:07:49.439 Okay, this next part. This feels like where things 00:07:49.439 --> 00:07:53.079 get really profound. Using AI not just for answers, 00:07:53.199 --> 00:07:56.560 but... to ask better questions, becoming a sparring 00:07:56.560 --> 00:07:59.100 partner for your own thoughts, a personal red 00:07:59.100 --> 00:08:01.339 team, as the guide calls it. Yeah, this is shifting 00:08:01.339 --> 00:08:03.579 gears. You stop seeing it as just an infosaurus 00:08:03.579 --> 00:08:05.560 and start using it as an intellectual challenger. 00:08:05.699 --> 00:08:08.180 First up, the assumption destroyer. What does 00:08:08.180 --> 00:08:10.660 that do? Every belief, every plan we have rests 00:08:10.660 --> 00:08:12.480 on hidden assumptions. This prompt digs them 00:08:12.480 --> 00:08:15.139 out. You say, I believe your belief. Act as a 00:08:15.139 --> 00:08:16.980 masqueradogician. What are the hidden assumptions 00:08:16.980 --> 00:08:19.819 I'm making? Then give me the strongest steel 00:08:19.819 --> 00:08:22.540 man argument against my belief. So if I said, 00:08:22.579 --> 00:08:24.740 I believe remote work is always more productive, 00:08:25.019 --> 00:08:27.540 it would first point out assumptions like everyone 00:08:27.540 --> 00:08:30.819 has a good home office setup or all tasks suit 00:08:30.819 --> 00:08:33.639 remote work. Exactly. And then it would present 00:08:33.639 --> 00:08:35.740 the strongest case for why office work might 00:08:35.740 --> 00:08:38.580 be better in some scenarios. Not to prove you 00:08:38.580 --> 00:08:41.179 wrong, but to make your own thinking more robust. 00:08:41.480 --> 00:08:44.059 Makes sense. What's next? The devil's advocate 00:08:44.059 --> 00:08:48.639 or a premortem. This is vital for de -risking 00:08:48.639 --> 00:08:51.879 anything new. You lay out your plan. I'm planning 00:08:51.879 --> 00:08:54.259 to detailed plan. You're my trusted advisor, 00:08:54.500 --> 00:08:56.820 but play devil's advocate. Give me the most compelling 00:08:56.820 --> 00:08:59.600 data -driven arguments against doing this. Oof. 00:09:00.340 --> 00:09:02.500 Sounds like it could be a dose of harsh reality. 00:09:02.799 --> 00:09:04.840 It often is. Like, someone planning an indie 00:09:04.840 --> 00:09:07.759 game launch on Steam used this. The AI came back 00:09:07.759 --> 00:09:10.799 with brutal stats. 19 ,000 new games last year. 00:09:10.940 --> 00:09:14.159 Median lifetime earnings under $500. Massive 00:09:14.159 --> 00:09:16.500 marketing challenges. It forces you to confront 00:09:16.500 --> 00:09:18.840 the downsides up front. Necessary, though. Okay, 00:09:18.879 --> 00:09:21.240 what else? The consequence mapper. This explores 00:09:21.240 --> 00:09:23.259 the ripple effects of decisions' second and third 00:09:23.259 --> 00:09:25.840 order consequences. How does that work? You say, 00:09:25.940 --> 00:09:28.659 I'm thinking about making a decision, like quitting 00:09:28.659 --> 00:09:31.639 my job to start a business. Map out the likely 00:09:31.639 --> 00:09:34.539 second order consequences of consequences and 00:09:34.539 --> 00:09:36.899 third order consequences, both positive and negative. 00:09:37.159 --> 00:09:39.919 So for quitting the job, second order might be 00:09:39.919 --> 00:09:43.120 income stress. Third order could be kids seeing 00:09:43.120 --> 00:09:46.000 entrepreneurial resilience. Or maybe relationship 00:09:46.000 --> 00:09:48.500 strain. Precisely. It helps you think through 00:09:48.500 --> 00:09:52.259 the longer term, less obvious impacts. Then there's 00:09:52.259 --> 00:09:54.799 the personal blind spot finder. This one's intense. 00:09:55.159 --> 00:09:57.899 How so? It uses the AI's memory within a specific 00:09:57.899 --> 00:10:00.559 project, you ask. Based on all our previous conversations 00:10:00.559 --> 00:10:03.740 in this project, act as an executive coach. Identify 00:10:03.740 --> 00:10:06.259 my top five recurring blind spots or cognitive 00:10:06.259 --> 00:10:09.759 biases. Ooh, that requires some bravery. It really 00:10:09.759 --> 00:10:12.669 does. This is not for the faint of heart. You 00:10:12.669 --> 00:10:14.610 need to be ready for potentially blunt feedback. 00:10:14.730 --> 00:10:17.250 It's powerful for self -improvement, but yeah, 00:10:17.389 --> 00:10:19.730 needs emotional readiness. And there's one more 00:10:19.730 --> 00:10:22.370 advanced one here. The outside view forecaster. 00:10:22.549 --> 00:10:25.269 Yeah, this tackles our natural optimistic bias, 00:10:25.470 --> 00:10:28.090 drawing on Daniel Kahneman's work. You state 00:10:28.090 --> 00:10:30.730 your plan. I'm planning to launch a new podcast, 00:10:30.929 --> 00:10:33.860 maybe. Contrast the inside view, my hopes and 00:10:33.860 --> 00:10:37.039 specific plans, with the outside view. Find real 00:10:37.039 --> 00:10:39.899 -world data, statistics, base rates for similar 00:10:39.899 --> 00:10:42.740 projects. What is the statistically most likely 00:10:42.740 --> 00:10:45.200 outcome? So it grounds your personal optimism 00:10:45.200 --> 00:10:48.039 in historical data for similar ventures. Often 00:10:48.039 --> 00:10:50.830 less rosy, I imagine. Often, yes. but incredibly 00:10:50.830 --> 00:10:53.570 valuable for realistic planning. So how do these 00:10:53.570 --> 00:10:55.990 critical thinking prompts fundamentally change 00:10:55.990 --> 00:10:58.250 our approach? They force us to think critically 00:10:58.250 --> 00:11:00.649 and consider perspectives way beyond our own 00:11:00.649 --> 00:11:03.429 assumptions. Okay, now shifting slightly. Let's 00:11:03.429 --> 00:11:05.809 talk about using ChatGPT as a kind of cognitive 00:11:05.809 --> 00:11:08.149 toolkit for managing everyday negative thoughts. 00:11:08.789 --> 00:11:11.210 Structuring self -reflection. Right. And this 00:11:11.210 --> 00:11:13.909 needs a huge flashing disclaimer right up front. 00:11:14.009 --> 00:11:16.490 Absolutely. This is not therapy. It is not medical 00:11:16.490 --> 00:11:18.830 advice. It is never a substitute for professional 00:11:18.830 --> 00:11:21.470 mental health support. This is purely a framework 00:11:21.470 --> 00:11:23.889 for structured self -reflection on non -clinical 00:11:23.889 --> 00:11:26.470 everyday stress using principles from cognitive 00:11:26.470 --> 00:11:29.940 behavioral therapy or CBT. Understood. For personal 00:11:29.940 --> 00:11:33.159 insight only. So how does the CBT workflow prompt 00:11:33.159 --> 00:11:36.539 work? It guides you through challenging, unhelpful 00:11:36.539 --> 00:11:40.379 thoughts. Exactly. It turns the AI into a compassionate, 00:11:40.519 --> 00:11:43.299 objective journaling guide. It walks you through 00:11:43.299 --> 00:11:45.919 the process step by step, asking questions one 00:11:45.919 --> 00:11:48.279 at a time. What kind of questions? First, what 00:11:48.279 --> 00:11:50.960 was the situation? What emotions did you feel 00:11:50.960 --> 00:11:53.360 and rate their intensity? What were your automatic 00:11:53.360 --> 00:11:56.799 thoughts? Then it helps you identify common cognitive 00:11:56.799 --> 00:11:59.159 distortions in those thoughts, things like all 00:11:59.159 --> 00:12:02.019 or nothing thinking, catastrophizing, mind reading. 00:12:02.179 --> 00:12:04.860 After identifying distortions, it guides you 00:12:04.860 --> 00:12:07.519 to create a more balanced, realistic reframe 00:12:07.519 --> 00:12:10.139 of the thought. Then, crucially, it asks you 00:12:10.139 --> 00:12:12.700 to propose one tangible action step you can take. 00:12:12.860 --> 00:12:15.860 And it finimes with a simple, positive one -sentence 00:12:15.860 --> 00:12:18.240 affirmation. You can give an example. Sure. Say 00:12:18.240 --> 00:12:21.120 a YouTuber sees their views dip. Situation. Low 00:12:21.120 --> 00:12:24.360 views on new videos. Emotions. Anxiety 710. Self 00:12:24.360 --> 00:12:27.240 -doubt 810. Automatic thoughts. My new videos 00:12:27.240 --> 00:12:29.159 are failing. I'm losing my touch. Everyone's 00:12:29.159 --> 00:12:32.419 bored. AI helps identify distortions. All or 00:12:32.419 --> 00:12:35.059 nothing. Failing. Mind reading. Everyone's bored. 00:12:35.360 --> 00:12:38.399 User that creates a reframe. Growth isn't always 00:12:38.399 --> 00:12:41.080 linear. Some videos underperform. I'm still creating 00:12:41.080 --> 00:12:45.039 valuable content. AI suggests an action. Analyze 00:12:45.039 --> 00:12:47.580 my top five performing videos for common elements 00:12:47.580 --> 00:12:50.659 and offers an affirmation. I grow through every 00:12:50.659 --> 00:12:53.620 experiment, one video at a time. So beyond the 00:12:53.620 --> 00:12:55.480 practical steps, how does this method impact 00:12:55.480 --> 00:12:57.500 well -being? It helps break negative thought 00:12:57.500 --> 00:12:59.919 loops and leads to actionable clarity, which 00:12:59.919 --> 00:13:02.860 can be really empowering. Makes sense. Okay, 00:13:02.940 --> 00:13:05.220 before we move to the advanced stuff, let's take 00:13:05.220 --> 00:13:08.279 a quick moment for our sponsor, Midroll Sponsoread. 00:13:09.039 --> 00:13:11.480 And we're back on the deep dive. We've covered 00:13:11.480 --> 00:13:13.870 a lot of ground. from foundations to specialized 00:13:13.870 --> 00:13:16.269 assistance and critical thinking. Now let's talk 00:13:16.269 --> 00:13:18.330 cheat codes. Quick shortcuts. Yeah, these are 00:13:18.330 --> 00:13:20.750 super handy for quick results. First one, ELI 00:13:20.750 --> 00:13:23.590 5. Explain like a 5. Simple enough. Just take 00:13:23.590 --> 00:13:26.490 your complex topic followed by ELI 5, like quantum 00:13:26.490 --> 00:13:29.450 computing ELI 5. It dumbs it right down to the 00:13:29.450 --> 00:13:31.850 core concepts. Great starting point. Nice. And 00:13:31.850 --> 00:13:35.600 the opposite. TLDR. Too long. Didn't read. Paste 00:13:35.600 --> 00:13:38.159 in a long article, email, whatever, add TLDR, 00:13:38.379 --> 00:13:41.299 and you get a quick, concise summary. Huge time 00:13:41.299 --> 00:13:44.820 saver. Okay. What about jargonize? Humanize. 00:13:44.919 --> 00:13:47.360 These are cool. Humanize takes technical text 00:13:47.360 --> 00:13:49.980 and makes it plain English. Jargonize does the 00:13:49.980 --> 00:13:51.820 reverse, takes simple notes and makes them sound 00:13:51.820 --> 00:13:53.740 more professional or technical. You can even 00:13:53.740 --> 00:13:56.840 chain them. Humanize a dense report to understand 00:13:56.840 --> 00:13:59.620 it. Then jargonize your simple summary into a 00:13:59.620 --> 00:14:03.009 formal brief. Clever. And the last one. The Feynman 00:14:03.009 --> 00:14:05.690 technique. Ah, yeah. Named after the physicist 00:14:05.690 --> 00:14:08.769 Richard Feynman. You ask the AI, explain complex 00:14:08.769 --> 00:14:11.549 topic to me using the Feynman technique. It forces 00:14:11.549 --> 00:14:14.350 the AI to explain it simply using analogies as 00:14:14.350 --> 00:14:16.350 if it were teaching it. It really helps you identify 00:14:16.350 --> 00:14:18.889 gaps in your own understanding. So how do these 00:14:18.889 --> 00:14:21.590 simple tags make interacting with AI more efficient? 00:14:21.889 --> 00:14:24.590 They instantly reformat responses, saving you 00:14:24.590 --> 00:14:27.309 time and significant mental effort. Quick wins. 00:14:27.549 --> 00:14:30.190 Got it. Okay, now for the really advanced stuff. 00:14:30.490 --> 00:14:32.230 We're talking strategies that don't just ask 00:14:32.230 --> 00:14:34.769 for output, but actually teach the AI how to 00:14:34.769 --> 00:14:37.409 think about a problem, manipulating its cognitive 00:14:37.409 --> 00:14:39.350 process. This is where it gets really exciting. 00:14:39.490 --> 00:14:41.090 These aren't just clever prompts. They're like 00:14:41.090 --> 00:14:44.629 cognitive algorithms based on AI research. First, 00:14:44.809 --> 00:14:48.500 Tree of Thought 2T Exploration. Think of it as 00:14:48.500 --> 00:14:51.399 the multiverse approach. You force the AI to 00:14:51.399 --> 00:14:54.379 explore multiple distinct solution paths before 00:14:54.379 --> 00:14:57.559 picking one. The prompt structure is key. Solve 00:14:57.559 --> 00:15:00.600 complex problem. Explore three distinct and creative 00:15:00.600 --> 00:15:03.720 solution branches. For each branch, outline your 00:15:03.720 --> 00:15:06.340 chain of thought. Critically evaluate each branch 00:15:06.340 --> 00:15:08.740 on a 0 -10 scale for correctness, efficiency, 00:15:09.080 --> 00:15:12.200 and creativity. Finally, select the highest scoring 00:15:12.200 --> 00:15:15.460 branch and justify your choice. So for a big 00:15:15.460 --> 00:15:18.620 problem like job loss due to AI, it wouldn't 00:15:18.620 --> 00:15:20.720 just give one answer. Right. It might explore, 00:15:20.799 --> 00:15:23.700 say, a rapid reskilling path, an AI dividend 00:15:23.700 --> 00:15:26.240 path, and maybe an augmentation first work design 00:15:26.240 --> 00:15:28.259 path, scoring each one based on the criteria 00:15:28.259 --> 00:15:30.879 you set. It's amazing for complex strategy where 00:15:30.879 --> 00:15:32.980 there's no single obvious answer. That's powerful. 00:15:33.159 --> 00:15:35.919 What's next? Self -consistency voting. This boosts 00:15:35.919 --> 00:15:38.779 accuracy, especially for factual or problem -solving 00:15:38.779 --> 00:15:41.080 tasks. You ask the AI to solve the same problem 00:15:41.080 --> 00:15:43.360 multiple times using slightly different reasoning 00:15:43.360 --> 00:15:46.360 paths each time. How does that work? The prompt 00:15:46.360 --> 00:15:49.820 would be like, solve task, generate five independent 00:15:49.820 --> 00:15:52.620 reasoning paths to get the answer, return the 00:15:52.620 --> 00:15:55.600 final answer from each path, then hold a vote 00:15:55.600 --> 00:15:58.299 and choose the majority answer. Provide a definitive 00:15:58.299 --> 00:16:01.700 explanation based on the consensus. It leverages 00:16:01.700 --> 00:16:04.440 the wisdom of the crowd within the AI itself. 00:16:04.779 --> 00:16:06.820 Interesting. Then there's the reflection self 00:16:06.820 --> 00:16:09.000 -critique loop. You called it the Gordon Ramsay 00:16:09.000 --> 00:16:12.500 prompt earlier. Huh. Yes. Because it's tough 00:16:12.500 --> 00:16:15.120 on its own work. You turn the AI's critical eye 00:16:15.120 --> 00:16:18.100 on its own output. Prompt. First, answer question 00:16:18.100 --> 00:16:21.399 task. Then, enter critique mode. Identify all 00:16:21.399 --> 00:16:24.700 flaws, errors, or gaps in your own draft. Finally, 00:16:24.860 --> 00:16:27.159 produce an improved final answer that fixes all 00:16:27.159 --> 00:16:29.509 the issues you identified. And it actually finds 00:16:29.509 --> 00:16:31.870 flaws in its own first attempt. Oh, yeah. Often 00:16:31.870 --> 00:16:34.669 quite harshly. Then it produces a significantly 00:16:34.669 --> 00:16:37.250 better V2. It's like built -in quality control 00:16:37.250 --> 00:16:39.710 for creative or analytical tasks. Incredible. 00:16:39.809 --> 00:16:42.190 What about the personal consultant prompt, the 00:16:42.190 --> 00:16:44.129 automation audit? This one's super practical. 00:16:44.250 --> 00:16:46.269 The AI acts like an external consultant looking 00:16:46.269 --> 00:16:48.289 for automation opportunities in your life or 00:16:48.289 --> 00:16:50.679 work. How do you set that up? Act as a senior 00:16:50.679 --> 00:16:53.580 automation consultant. Ask me up to 10 diagnostic 00:16:53.580 --> 00:16:56.039 questions to understand my role daily routines 00:16:56.039 --> 00:16:58.600 and uncover repetitive tasks in my job life. 00:16:58.799 --> 00:17:01.279 Based on my answers, rank the top five automation 00:17:01.279 --> 00:17:04.359 opportunities. Suggest specific tools or a tech 00:17:04.359 --> 00:17:07.180 stack for each. Provide a high -level, step -by 00:17:07.180 --> 00:17:09.519 -step implementation plan. So it interviews you, 00:17:09.559 --> 00:17:11.200 then gives you an action plan. Exactly. Someone 00:17:11.200 --> 00:17:13.920 use this who is a content creator. The AI identified 00:17:13.920 --> 00:17:16.519 bottlenecks like chaotic video ideation, manual 00:17:16.519 --> 00:17:19.039 sponsor tracking, etc., and suggested tools. 00:17:19.180 --> 00:17:22.579 like monday .com, RSS feeds, and automation platforms 00:17:22.579 --> 00:17:26.319 like N8N. Very cool. Okay, last one. The world 00:17:26.319 --> 00:17:30.539 engine. This sounds different. it is this showcases 00:17:30.539 --> 00:17:32.859 the ai's ability to maintain a persistent state 00:17:32.859 --> 00:17:35.460 for interactive storytelling or simulations the 00:17:35.460 --> 00:17:38.019 prompt structure is key you are a world engine 00:17:38.019 --> 00:17:41.519 genre g solar punk mystery maintain a dynamic 00:17:41.519 --> 00:17:44.319 world state based on my choices narrate the current 00:17:44.319 --> 00:17:47.079 scene vividly always offer me three clear distinct 00:17:47.079 --> 00:17:49.579 branching choices for my next action weave the 00:17:49.579 --> 00:17:51.819 consequences of my choices logically into the 00:17:51.819 --> 00:17:55.039 evolving narrative whoa Imagine the possibilities 00:17:55.039 --> 00:17:57.920 with this world engine. It's like building your 00:17:57.920 --> 00:18:00.039 own interactive narrative playground where your 00:18:00.039 --> 00:18:03.180 choices genuinely shape the story, potentially 00:18:03.180 --> 00:18:05.660 endlessly. It feels like we're just scratching 00:18:05.660 --> 00:18:07.980 the absolute surface of what these models can 00:18:07.980 --> 00:18:10.569 really do creatively. It really does feel like 00:18:10.569 --> 00:18:12.529 that. It's a glimpse into truly interactive, 00:18:12.730 --> 00:18:16.190 persistent AI experiences. So how do these advanced 00:18:16.190 --> 00:18:20.750 methods truly elevate the AI beyond just answering 00:18:20.750 --> 00:18:22.950 questions? They transform it into a strategic 00:18:22.950 --> 00:18:25.190 thinking partner, not just a content generator 00:18:25.190 --> 00:18:27.410 or answer bot. Okay, let's bring this all together. 00:18:27.549 --> 00:18:30.529 As we wrap up this deep dive, the central message 00:18:30.529 --> 00:18:33.970 seems crystal clear. ChatGPT isn't just an answer 00:18:33.970 --> 00:18:36.130 machine anymore. It's really a thinking machine. 00:18:36.329 --> 00:18:38.230 That's the absolute bottom line and the real 00:18:38.230 --> 00:18:41.220 power. It's all in the setup. Generic chats give 00:18:41.220 --> 00:18:44.319 generic results. But using projects with tailored 00:18:44.319 --> 00:18:46.859 custom instructions, that turns this general 00:18:46.859 --> 00:18:49.480 tool into a whole team of specialized assistants 00:18:49.480 --> 00:18:52.160 built just for you. And the goal shifts, right? 00:18:52.259 --> 00:18:54.099 It's not just about getting better answers from 00:18:54.099 --> 00:18:57.140 the AI. Exactly. It's about you developing better 00:18:57.140 --> 00:18:59.839 thinking habits. The act of structuring a problem 00:18:59.839 --> 00:19:02.519 clearly enough for the AI forces you to clarify 00:19:02.519 --> 00:19:05.000 your own thoughts. Prompt engineering is critical 00:19:05.000 --> 00:19:07.910 thinking practice. And iteration is key. Those 00:19:07.910 --> 00:19:10.430 advanced techniques like self -critique or tree 00:19:10.430 --> 00:19:13.549 of thought, they force the AI to refine, improve, 00:19:13.910 --> 00:19:16.890 and explore options. Quality comes from that 00:19:16.890 --> 00:19:19.710 process. And don't forget memory. Within a project, 00:19:19.849 --> 00:19:22.589 the AI remembers your context, your goals, your 00:19:22.589 --> 00:19:25.410 style. It gets smarter about you over time. That 00:19:25.410 --> 00:19:27.609 persistent context is maybe its most valuable 00:19:27.609 --> 00:19:30.640 asset for personalized help. Right. This deep 00:19:30.640 --> 00:19:33.119 dive isn't just theory, though. We really want 00:19:33.119 --> 00:19:35.420 you, the listener, to put this into practice. 00:19:35.799 --> 00:19:38.240 So we've got an immediate action plan, a challenge 00:19:38.240 --> 00:19:39.839 for you this week. Yeah, let's make this real. 00:19:39.940 --> 00:19:42.599 Four things to try. One, set up your first project. 00:19:42.779 --> 00:19:45.220 Pick one recurring task, journaling, health tracking, 00:19:45.500 --> 00:19:48.259 summarizing news, whatever, and build a dedicated 00:19:48.259 --> 00:19:51.359 project with detailed custom instructions. Really 00:19:51.359 --> 00:19:53.980 define that AI assistance role. Okay, number 00:19:53.980 --> 00:19:58.160 one, build a project. Two, be brave. Try the 00:19:58.160 --> 00:20:00.460 personal blind spot finder prompt within that 00:20:00.460 --> 00:20:03.599 project or another one. Ask the AI to identify 00:20:03.599 --> 00:20:05.980 your recurring biases based on your chats. It 00:20:05.980 --> 00:20:08.039 might be uncomfortable, but potentially transformative. 00:20:08.460 --> 00:20:12.920 Number two, find blind spots. Gulp. Three, experiment 00:20:12.920 --> 00:20:15.980 with one advanced technique. Just pick one people, 00:20:16.099 --> 00:20:18.619 maybe the self -critique loop on something you 00:20:18.619 --> 00:20:20.700 write or the devil's advocate for a plan you 00:20:20.700 --> 00:20:22.799 have. See how it changes your process. Number 00:20:22.799 --> 00:20:26.390 three. Try one advanced prompt. And four, run 00:20:26.390 --> 00:20:28.890 the automation workflow finder prompt. Seriously, 00:20:29.109 --> 00:20:31.410 audit your daily routines. Ask the AI consultant 00:20:31.410 --> 00:20:33.509 to help you find things you could automate. You 00:20:33.509 --> 00:20:36.210 might be surprised. Okay, four steps. Project, 00:20:36.349 --> 00:20:39.170 blind spots, advanced prompt, automation audit. 00:20:39.519 --> 00:20:41.440 That's the challenge. And maybe a final thought 00:20:41.440 --> 00:20:44.339 to leave people with. Stop treating ChatGPT like 00:20:44.339 --> 00:20:46.859 a fancy search engine. Start treating it like 00:20:46.859 --> 00:20:49.839 this incredibly patient, knowledgeable, and yes, 00:20:49.960 --> 00:20:52.200 powerful thinking partner. It's like a second 00:20:52.200 --> 00:20:54.839 brain that can challenge you, simulate possibilities, 00:20:55.119 --> 00:20:57.279 help you structure your thoughts. Because honestly, 00:20:57.420 --> 00:20:59.480 the people who master thinking with AI, not just 00:20:59.480 --> 00:21:00.819 asking AI for things, they aren't just going 00:21:00.819 --> 00:21:02.039 to be more productive. They're actually going 00:21:02.039 --> 00:21:03.700 to be smarter. OTRO Music.