WEBVTT 00:00:00.000 --> 00:00:02.580 Have you ever had that feeling, that little flicker 00:00:02.580 --> 00:00:06.559 of annoyance? You ask a powerful AI, you know, 00:00:06.599 --> 00:00:09.539 maybe chat GPT, for something simple. Yeah. Like, 00:00:09.699 --> 00:00:12.160 just a short email. Maybe a summary. And what 00:00:12.160 --> 00:00:16.460 comes back is confusing. Or super complicated. 00:00:16.539 --> 00:00:18.980 Right. Overly complicated, or sometimes just 00:00:18.980 --> 00:00:21.679 plain wrong. Uh -huh. And you spend, what, 10 00:00:21.679 --> 00:00:23.519 minutes trying to fix it, and then you just go, 00:00:23.579 --> 00:00:25.660 you know what? Forget it. I'll just do this myself. 00:00:25.739 --> 00:00:28.320 It's faster. It feels faster. Mm -hmm. And that 00:00:28.320 --> 00:00:32.770 happens because... Honestly, most of us, we're 00:00:32.770 --> 00:00:35.549 using these incredibly powerful AI tools kind 00:00:35.549 --> 00:00:37.810 of the wrong way. How so? We treat them like 00:00:37.810 --> 00:00:40.409 a slightly smarter Google, a talking search engine. 00:00:40.549 --> 00:00:42.590 And we're just missing the real potential there. 00:00:42.950 --> 00:00:46.039 OK. Welcome to the deep dive today. Our mission 00:00:46.039 --> 00:00:48.619 is really to get past that basic prompting that 00:00:48.619 --> 00:00:51.560 surface level stuff and unlock the actual efficient 00:00:51.560 --> 00:00:54.460 power hidden in generative AI Yeah, absolutely. 00:00:54.840 --> 00:00:57.460 We boiled it down to six simple skills and these 00:00:57.460 --> 00:00:59.899 aren't like coding skills or anything super technical. 00:01:00.240 --> 00:01:03.380 They're more like Shifts in how you approach 00:01:03.380 --> 00:01:06.340 it and these shifts can turn AI from being this 00:01:06.340 --> 00:01:10.000 frustrating roadblock into a genuinely efficient 00:01:10.000 --> 00:01:12.049 teammate One that actually gives you what you 00:01:12.049 --> 00:01:14.049 need. Okay, so let's map this out. We're going 00:01:14.049 --> 00:01:16.150 to cover the whole workflow, basically. Starting 00:01:16.150 --> 00:01:18.790 with building your toolkit, your AI team, if 00:01:18.790 --> 00:01:21.069 you will. Right. Then getting really clear on 00:01:21.069 --> 00:01:23.730 what you want before you pipe anything. Crucial. 00:01:23.810 --> 00:01:26.670 How to give clear instructions, the absolute 00:01:26.670 --> 00:01:29.870 necessity of checking the AI's work. Verification, 00:01:29.950 --> 00:01:32.209 yeah. Then using these tools together, stacking 00:01:32.209 --> 00:01:35.310 them, and finally adding that indispensable human 00:01:35.310 --> 00:01:37.750 touch, the polish. Okay, let's dive in. Skill 00:01:37.750 --> 00:01:41.129 number one. Which is... Stop trying to make one 00:01:41.129 --> 00:01:44.450 tool do everything. This is, I think, the biggest 00:01:44.450 --> 00:01:47.709 foundational mistake. People grab one AI, usually 00:01:47.709 --> 00:01:50.590 Chad GPT, and expect it to handle 100 % of their 00:01:50.590 --> 00:01:53.549 tasks. Right. But that's just, it's illogical, 00:01:53.609 --> 00:01:55.250 isn't it? Like, you wouldn't hire a fantastic 00:01:55.250 --> 00:01:57.349 baker to fix your motorbike. No. Or ask your 00:01:57.349 --> 00:01:59.530 accountant to design the company logo. Doesn't 00:01:59.530 --> 00:02:02.340 make sense. Right, and yet we somehow expect 00:02:02.340 --> 00:02:06.459 one single AI model to write great emails, analyze 00:02:06.459 --> 00:02:09.819 complex spreadsheets, draft code, and create 00:02:09.819 --> 00:02:13.259 stunning images. It's setting it up to fail. 00:02:13.500 --> 00:02:16.759 You need specialists, a team. Exactly, a team. 00:02:17.060 --> 00:02:20.939 I like to group them into three main categories 00:02:20.939 --> 00:02:23.699 that you really need. First up, the brains. Okay. 00:02:23.840 --> 00:02:26.280 These are your large language models, the LLMs. 00:02:26.639 --> 00:02:29.080 They're the generalists, the thinkers. Good for 00:02:29.080 --> 00:02:31.020 talking, writing, brainstorming, that kind of 00:02:31.020 --> 00:02:32.879 stuff. And knowing the specific ones helps, because 00:02:32.879 --> 00:02:34.780 they do have different strengths. Like, chat 00:02:34.780 --> 00:02:37.139 GPT is kind of the Swiss army knife, right? Good 00:02:37.139 --> 00:02:40.120 for first drafts, quick ideas. Collectible, but 00:02:40.120 --> 00:02:42.460 Gemini, because it's Google, tends to be better 00:02:42.460 --> 00:02:44.740 if you need something about current events or 00:02:44.740 --> 00:02:46.819 recent news. It's more plugged in. For sure. 00:02:46.919 --> 00:02:48.919 And if you need writing that feels really human, 00:02:49.400 --> 00:02:51.879 my personal favorite is Claude. I find it's consistently 00:02:51.879 --> 00:02:54.379 better for more, say, emotional writing, like 00:02:54.379 --> 00:02:57.400 a blog post. Or, and this is key, summarizing 00:02:57.400 --> 00:02:59.759 really long documents. Like, if you throw a 100 00:02:59.759 --> 00:03:02.400 -page PDF at it, Claude generally handles that 00:03:02.400 --> 00:03:04.360 long context better than others. Interesting. 00:03:04.520 --> 00:03:07.680 OK, so that's the brains. Group two. The researchers. 00:03:08.139 --> 00:03:11.699 And this is critical. You really shouldn't trust 00:03:11.699 --> 00:03:15.659 a brain AI for hard facts, ever. Right. You need 00:03:15.659 --> 00:03:18.819 a dedicated fact checker. Yeah. For me, Perplexity 00:03:18.819 --> 00:03:21.120 is a daily driver. It's essentially a much better 00:03:21.120 --> 00:03:23.340 search engine because it shows its work. How 00:03:23.340 --> 00:03:25.500 so? It gives you the information, but also shows 00:03:25.500 --> 00:03:27.699 exactly where it came from with direct links 00:03:27.699 --> 00:03:29.919 you can check, verifiable sources. It's a must 00:03:29.919 --> 00:03:32.539 have. OK. And then there's Notebook LM. Which 00:03:32.539 --> 00:03:34.340 is pretty fascinating. It doesn't search the 00:03:34.340 --> 00:03:36.620 web. It searches your stuff. My own documents. 00:03:36.819 --> 00:03:39.439 Your uploaded notes, PDFs, research papers. You 00:03:39.439 --> 00:03:41.340 can basically have a conversation with your personal 00:03:41.340 --> 00:03:45.539 library. Wow, okay. And the third group. The 00:03:45.539 --> 00:03:47.939 creators. These are for visuals. Yeah. The Journey 00:03:47.939 --> 00:03:50.699 is famous for those incredibly detailed, often 00:03:50.699 --> 00:03:52.659 beautiful images. But you have to use Discord, 00:03:52.740 --> 00:03:55.039 which is a bit of a hurdle for some. It is, yeah. 00:03:55.300 --> 00:03:58.879 But then there's ideogram. An ideogram is fantastic 00:03:58.879 --> 00:04:01.639 for one... really practical reason. Which is? 00:04:01.780 --> 00:04:04.280 It's actually pretty good at putting text, like 00:04:04.280 --> 00:04:07.439 words, onto images without messing it up completely. 00:04:07.699 --> 00:04:10.919 Ah, the classic AI text problem. Exactly. It's 00:04:10.919 --> 00:04:13.300 still a major weakness for most image AIs, but 00:04:13.300 --> 00:04:15.599 ideogram handles it much better. So for someone 00:04:15.599 --> 00:04:17.899 just starting out, what's the minimum toolkit? 00:04:18.139 --> 00:04:20.600 Keep it simple. Start with one brain, maybe chat 00:04:20.600 --> 00:04:24.639 GPT, one researcher, definitely perplexity, and 00:04:24.639 --> 00:04:27.889 maybe one creator. like ideogram, if you need 00:04:27.889 --> 00:04:31.189 images with text. So the big takeaway here seems 00:04:31.189 --> 00:04:34.629 to be that separating the thinking tools, the 00:04:34.629 --> 00:04:37.189 brains, from the fact -checking tools, the researchers, 00:04:37.310 --> 00:04:40.149 is absolutely fundamental. Absolutely. It prevents 00:04:40.149 --> 00:04:42.310 those confident -sounding factual errors right 00:04:42.310 --> 00:04:45.930 from the start. It avoids misuse. So why is separating 00:04:45.930 --> 00:04:49.029 tools by function brains versus researchers the 00:04:49.029 --> 00:04:51.350 foundational step for mastery? Because it stops 00:04:51.350 --> 00:04:54.290 powerful LLMs from confidently making stuff up 00:04:54.290 --> 00:04:57.250 when you need accuracy. prevents fabricated facts. 00:04:57.589 --> 00:04:59.310 OK, that makes sense. Which brings us neatly 00:04:59.310 --> 00:05:01.990 to skill number two. Clarify the problem. Know 00:05:01.990 --> 00:05:04.149 what you actually want. Yes. This is probably 00:05:04.149 --> 00:05:06.170 the single biggest time saver. What do beginners 00:05:06.170 --> 00:05:08.730 usually do? They open chat GPT and just type 00:05:08.730 --> 00:05:13.329 a vague thought, right? Experts. They pause just 00:05:13.329 --> 00:05:15.949 for like 30 seconds. They think first. Because 00:05:15.949 --> 00:05:18.050 if your question isn't clear, the answer you 00:05:18.050 --> 00:05:20.129 get back is never going to be clear. It sounds 00:05:20.129 --> 00:05:23.069 simple, but. Think about it like this. The I'm 00:05:23.069 --> 00:05:26.019 hungry analogy. If you call a friend and just 00:05:26.019 --> 00:05:29.139 say, hey, I'm hungry. What can they do? Not much, 00:05:29.319 --> 00:05:30.939 right? Right. Where do you start? But if you 00:05:30.939 --> 00:05:34.360 say, OK, I'm hungry. I'm feeling like Vietnamese 00:05:34.360 --> 00:05:38.139 food, but nothing spicy. And I've got about $100 00:05:38.139 --> 00:05:40.500 budget. Now what happens? Yeah, they can actually 00:05:40.500 --> 00:05:42.519 help. They have specifics. Exactly. They can 00:05:42.519 --> 00:05:45.519 act immediately. AI is like that helpful but 00:05:45.519 --> 00:05:48.199 naive friend. It needs specifics. So I always 00:05:48.199 --> 00:05:50.060 recommend this quick 30 second pause checklist 00:05:50.060 --> 00:05:52.779 before typing. OK, what's on it? First. What 00:05:52.779 --> 00:05:55.660 is the real, specific, final result I need? Not 00:05:55.660 --> 00:05:57.779 just write an email. More like... More like, 00:05:58.100 --> 00:06:00.459 write a polite email to my boss asking for next 00:06:00.459 --> 00:06:02.740 Friday off, making sure to mention that Project 00:06:02.740 --> 00:06:05.279 X and Report Y are already completed. See the 00:06:05.279 --> 00:06:07.379 difference. Yeah, much clearer. Okay, what's 00:06:07.379 --> 00:06:12.560 next? Second, who is this actually for? The audience's 00:06:12.560 --> 00:06:16.339 everything for tone is for your boss, an angry 00:06:16.339 --> 00:06:19.290 customer. You're friendly social media followers. 00:06:19.629 --> 00:06:23.209 That changes everything. Vocabulary. Formality. 00:06:23.629 --> 00:06:25.990 Right. Third. Third. What does a good answer 00:06:25.990 --> 00:06:28.810 look like, structurally? Do I want a bullet point 00:06:28.810 --> 00:06:31.290 list? A short, concise paragraph? Maybe a table 00:06:31.290 --> 00:06:34.209 comparing options? Define the format. So those 00:06:34.209 --> 00:06:36.769 details, the goal, the audience, the format, 00:06:36.810 --> 00:06:39.050 those are the constraints the AI needs to succeed. 00:06:39.069 --> 00:06:41.949 They are. And honestly, I gotta admit something 00:06:41.949 --> 00:06:44.589 here. I won. Even now, I still sometimes wrestle 00:06:44.589 --> 00:06:47.290 with prompt drift myself. and the conversation 00:06:47.290 --> 00:06:49.529 goes off track. It usually happens when I rush 00:06:49.529 --> 00:06:51.790 and skip this initial planning step. So it happens 00:06:51.790 --> 00:06:54.410 to everyone. Oh, yeah. We all fall into that 00:06:54.410 --> 00:06:56.850 trap of vagueness if we don't consciously pause 00:06:56.850 --> 00:06:59.310 and clarify first. So how does pre -defining 00:06:59.310 --> 00:07:01.689 the desired final result format save the most 00:07:01.689 --> 00:07:04.750 time? A clear problem is 90 % of the solution 00:07:04.750 --> 00:07:07.949 beat. It just cuts out all that back and forth 00:07:07.949 --> 00:07:10.069 editing. Right. It eliminates the need for endless 00:07:10.069 --> 00:07:12.069 revisions because you got closer the first time. 00:07:12.230 --> 00:07:15.430 Precisely. And if avoiding edits is the goal, 00:07:15.689 --> 00:07:18.709 that leads us right into skill three. Communicate 00:07:18.709 --> 00:07:20.329 well, giving clear instructions. You mentioned 00:07:20.329 --> 00:07:22.589 prompting. Yeah, and I kind of hate that word 00:07:22.589 --> 00:07:24.949 prompting. It sounds so technical, so jargon 00:07:24.949 --> 00:07:28.129 -y. Just think of the AI like a brand new assistant. 00:07:28.689 --> 00:07:31.189 Super smart, incredibly tireless, never gets 00:07:31.189 --> 00:07:35.790 bored, but also completely naive. They need really 00:07:35.790 --> 00:07:38.170 clear step -by -step instructions to do good 00:07:38.170 --> 00:07:40.589 work. You can't just hint at things. OK, but 00:07:40.589 --> 00:07:43.660 wait. A lot of the advice you see online involves 00:07:43.660 --> 00:07:46.939 using these really complex multi -part frameworks 00:07:46.939 --> 00:07:50.160 for prompts, like a six -part structure. Doesn't 00:07:50.160 --> 00:07:52.139 that feel like writing a whole project plan just 00:07:52.139 --> 00:07:54.259 to ask for an email? Is that really faster than 00:07:54.259 --> 00:07:57.040 just typing a simple request? Seems like overkill. 00:07:57.399 --> 00:07:59.680 It can feel like overkill up front. I get that. 00:07:59.860 --> 00:08:02.100 But it's about structure. And that structure 00:08:02.100 --> 00:08:04.300 actually saves you time down the line, especially 00:08:04.300 --> 00:08:06.519 for important tasks. It gives you a repeatable 00:08:06.519 --> 00:08:09.139 formula. So the six -part framework, what is 00:08:09.139 --> 00:08:11.829 it? It's basically a checklist to make sure you 00:08:11.829 --> 00:08:14.529 cover all the bases. Role, context, task, format, 00:08:14.810 --> 00:08:16.730 rules, and sometimes an example. Okay, let's 00:08:16.730 --> 00:08:18.689 see it in action. Like a bad prompt versus a 00:08:18.689 --> 00:08:21.509 good one. Sure. Bad prompt. Write a social media 00:08:21.509 --> 00:08:25.110 post about drinking water. Vague. Generic. What 00:08:25.110 --> 00:08:28.310 do you get back? Probably something boring. Totally 00:08:28.310 --> 00:08:31.600 boring. Now the good prompt. Using the framework, 00:08:31.920 --> 00:08:35.000 role. You are a friendly, knowledgeable health 00:08:35.000 --> 00:08:37.700 expert. Okay, persona set. Context. This is for 00:08:37.700 --> 00:08:39.980 busy office workers scrolling Facebook during 00:08:39.980 --> 00:08:41.820 their lunch break. Audience and platform. Got 00:08:41.820 --> 00:08:45.059 it. Task. Write a 100 -word post focusing on 00:08:45.059 --> 00:08:47.379 three surprising benefits of staying hydrated. 00:08:48.019 --> 00:08:50.779 Specific action. Format. Start with an engaging 00:08:50.779 --> 00:08:53.500 question. Use bullet points for the benefits. 00:08:54.379 --> 00:08:57.259 End with a simple call to action. Structure defined. 00:08:57.440 --> 00:09:00.379 Rules. Do not use complex medical jargon. Keep 00:09:00.379 --> 00:09:02.600 the language simple and encouraging. Boundary 00:09:02.600 --> 00:09:05.559 set? Example. Maybe use a style like this. Feeling 00:09:05.559 --> 00:09:08.159 that 3 p .m. slump. Maybe you just need water. 00:09:08.399 --> 00:09:10.840 Ah, so you give it a style guide too. That detailed 00:09:10.840 --> 00:09:13.279 prompt completely changes the game. Right. It 00:09:13.279 --> 00:09:16.460 turns that vague idea into a concrete, executable 00:09:16.460 --> 00:09:19.120 plan that the AI can actually follow properly. 00:09:19.460 --> 00:09:22.600 So beyond just complexity, what's the most important 00:09:22.600 --> 00:09:25.399 role that six -part framework fills for a beginner? 00:09:26.120 --> 00:09:29.500 It structures scattered thoughts, beat, turns 00:09:29.500 --> 00:09:31.919 ideas into an actionable project plan for the 00:09:31.919 --> 00:09:34.740 AI. OK, that makes sense. What other techniques 00:09:34.740 --> 00:09:37.759 help with communication? Technique two is one 00:09:37.759 --> 00:09:40.440 I really love, especially with newer AIs. Show 00:09:40.440 --> 00:09:43.019 it. Don't just tell it. I mean. Use the vision 00:09:43.019 --> 00:09:45.600 capabilities. Instead of spending five minutes 00:09:45.600 --> 00:09:48.740 trying to describe, say, the simple and clean 00:09:48.740 --> 00:09:50.899 aesthetic of a website you like. Which is hard 00:09:50.899 --> 00:09:53.220 to describe accurately. Right. Just take a screenshot 00:09:53.220 --> 00:09:55.559 and upload it. Say, match this style. The result 00:09:55.559 --> 00:09:57.820 is usually way faster and 10 times closer to 00:09:57.820 --> 00:09:59.740 what you wanted. Oh, that's smart. OK, technique 00:09:59.740 --> 00:10:02.840 three. Ask the AI for help. Seriously. If you 00:10:02.840 --> 00:10:05.720 get a boring answer, tell the AI, say, that last 00:10:05.720 --> 00:10:07.580 response was a bit dull. How could I rewrite 00:10:07.580 --> 00:10:09.799 my prompt to get something more creative or fun? 00:10:10.039 --> 00:10:12.779 Use its own knowledge against it, kind of? Exactly. 00:10:13.299 --> 00:10:15.299 Leverage its understanding of how it works best. 00:10:15.659 --> 00:10:17.659 It often gives great suggestions for improving 00:10:17.659 --> 00:10:21.159 your own prompts. and technique four. Do it again, 00:10:21.580 --> 00:10:25.600 but better. This is simple but powerful. After 00:10:25.600 --> 00:10:28.220 the AI gives you an answer, just ask it. Okay, 00:10:28.279 --> 00:10:30.539 read your answer again. Find three specific things 00:10:30.539 --> 00:10:33.440 you could improve. Then give me the new, better 00:10:33.440 --> 00:10:36.039 version. Huh, you make it critique itself. Yep. 00:10:36.500 --> 00:10:38.879 You basically build in a quality control loop 00:10:38.879 --> 00:10:40.740 without doing the editing yourself initially. 00:10:41.000 --> 00:10:43.539 So by using these frameworks and techniques, 00:10:43.919 --> 00:10:46.980 we shift from just asking for stuff to actually 00:10:46.980 --> 00:10:49.799 directing a process, directing the AI. That's 00:10:49.799 --> 00:10:51.259 the perfect way to put it. You're the director. 00:10:51.759 --> 00:10:54.080 And now that we have the content created, we 00:10:54.080 --> 00:10:57.080 absolutely need to talk about safety verification. 00:10:57.259 --> 00:10:59.600 Right. The trust issue. OK, let's take a quick 00:10:59.600 --> 00:11:01.759 pause here. And when we come back, we'll tackle 00:11:01.759 --> 00:11:04.340 verification and how to combine these tools effectively. 00:11:05.039 --> 00:11:07.950 Mid -roll sponsor break. Welcome back to the 00:11:07.950 --> 00:11:10.029 deep dive. We've talked about choosing tools, 00:11:10.389 --> 00:11:12.490 clarifying intent, and communicating effectively. 00:11:12.990 --> 00:11:15.250 Now let's hit skill number four, verification, 00:11:15.809 --> 00:11:17.690 addressing that big fear. Yeah, this is probably 00:11:17.690 --> 00:11:20.049 the number one concern people have, right? Using 00:11:20.049 --> 00:11:22.570 wrong information from AI. And the hard truth 00:11:22.570 --> 00:11:26.409 is AI can be a very, very confident liar. They 00:11:26.409 --> 00:11:29.279 call it hallucination. Right. Which is just a 00:11:29.279 --> 00:11:31.360 fancy word for making stuff up. Right. But it 00:11:31.360 --> 00:11:33.539 does it with such confidence. Yeah. So your mantra 00:11:33.539 --> 00:11:36.960 has to be, trust, but always check. OK. So how 00:11:36.960 --> 00:11:39.399 do we check? Method one. Use the researcher tool 00:11:39.399 --> 00:11:42.919 we talked about earlier. Never, ever take a specific 00:11:42.919 --> 00:11:46.480 fact or number from a brain AI like ChatGPT at 00:11:46.480 --> 00:11:49.039 face value. And never. Not if accuracy matters. 00:11:49.679 --> 00:11:53.480 So if ChatGPT says, studies show 80 % of office 00:11:53.480 --> 00:11:56.289 workers experience back pain, What do you do? 00:11:56.409 --> 00:11:58.389 Check it somewhere else. Exactly. Copy that exact 00:11:58.389 --> 00:12:00.850 sentence. 80 % of office workers have back pain 00:12:00.850 --> 00:12:03.549 and paste it straight into perplexity. And perplexity 00:12:03.549 --> 00:12:06.230 will? Perplexity will go find actual studies 00:12:06.230 --> 00:12:08.809 and sources. It might come back showing, well, 00:12:08.850 --> 00:12:12.549 study A found 60%. Study B over here found 75%. 00:12:12.549 --> 00:12:15.429 And crucially, it will give you the links. Ah, 00:12:15.570 --> 00:12:18.529 so you get the real data plus the proof. Precisely. 00:12:18.750 --> 00:12:20.789 Now you have reliable information, not just an 00:12:20.789 --> 00:12:23.149 AI's assertion. OK, that's solid. Method two. 00:12:23.210 --> 00:12:25.740 Force the source. Ask the AI directly within 00:12:25.740 --> 00:12:27.899 the chat. Where did you get that specific piece 00:12:27.899 --> 00:12:29.820 of information? Can you give me the name of the 00:12:29.820 --> 00:12:32.179 original article or study or maybe a link? And 00:12:32.179 --> 00:12:34.539 what if it avoids the question? That's your red 00:12:34.539 --> 00:12:37.379 flag. If it gives you a vague answer like, as 00:12:37.379 --> 00:12:40.080 an AI language model, I access information from 00:12:40.080 --> 00:12:43.200 a vast database. The classic dodge. That's a 00:12:43.200 --> 00:12:44.919 huge warning sign. I probably fabricated it. 00:12:45.080 --> 00:12:48.159 Oh, and a little trick here. Start the conversation 00:12:48.159 --> 00:12:51.440 by explicitly giving the AI permission to not 00:12:51.440 --> 00:12:53.220 know. What do you mean? Tell it right at the 00:12:53.220 --> 00:12:55.019 beginning. Hey, if you're not sure about something 00:12:55.019 --> 00:12:56.600 or you don't know the answer, please just say, 00:12:56.639 --> 00:12:59.220 I don't know. Ah, that lowers the pressure for 00:12:59.220 --> 00:13:01.960 it to invent an answer. Exactly. It preempts 00:13:01.960 --> 00:13:03.779 some of the fabrication tendency. It's smart. 00:13:04.039 --> 00:13:07.059 OK, method three for verification. Get a second 00:13:07.059 --> 00:13:10.059 opinion. This is super easy. Take the answer 00:13:10.059 --> 00:13:14.220 you got from one AI, say chat GPT, copy it, and 00:13:14.220 --> 00:13:17.279 paste it into a competitor. Go ask Claude or 00:13:17.279 --> 00:13:19.600 Gemini, can you review this answer? Are there 00:13:19.600 --> 00:13:22.000 any mistakes or important things missing? Use 00:13:22.000 --> 00:13:25.940 AIs to check other AIs. Yep. They're often surprisingly 00:13:25.940 --> 00:13:28.259 good at catching each other's errors or omissions. 00:13:28.460 --> 00:13:31.480 It takes maybe 10 seconds and significantly boosts 00:13:31.480 --> 00:13:34.059 your confidence in the result. Why is deliberately 00:13:34.059 --> 00:13:37.700 asking the AI to show its source so essential 00:13:37.700 --> 00:13:40.980 now? It directly challenges the AI's tendency 00:13:40.980 --> 00:13:45.259 to make things up with total confidence. BEAT 00:13:45.259 --> 00:13:47.740 confronts that confident fabrication head on. 00:13:47.899 --> 00:13:50.639 Okay, so verification is non -negotiable for 00:13:50.639 --> 00:13:53.379 building trust. That leads us to skill five, 00:13:53.840 --> 00:13:56.539 combine your work, stacking the tools. Right, 00:13:56.700 --> 00:13:59.500 because the real magic, the real efficiency boost, 00:13:59.899 --> 00:14:01.740 happens when you get these specialized tools 00:14:01.740 --> 00:14:04.840 working together. Think of it like a band again. 00:14:05.000 --> 00:14:07.360 Okay. You don't just want a drummer. You need 00:14:07.360 --> 00:14:10.379 the drums, the guitar, the bass player, the singer, 00:14:10.899 --> 00:14:13.879 each specializing. but playing together in harmony. 00:14:14.139 --> 00:14:16.039 And we're talking about manual stacking here, 00:14:16.139 --> 00:14:18.000 right? Where we are the ones coordinating the 00:14:18.000 --> 00:14:20.220 different tools. Exactly. You are the director, 00:14:20.600 --> 00:14:22.659 passing the work from one specialist to the next. 00:14:22.980 --> 00:14:25.820 Automation with things like Zapier or N8N is 00:14:25.820 --> 00:14:28.240 possible later, but master this manual stacking 00:14:28.240 --> 00:14:30.360 first. Let's walk through an example. Say the 00:14:30.360 --> 00:14:32.820 goal is a well -researched blog post about hobbies 00:14:32.820 --> 00:14:34.759 that reduce stress. How does the stacking work? 00:14:34.899 --> 00:14:37.840 Okay, step one, the researcher. You go to perplexity, 00:14:38.080 --> 00:14:41.580 ask it. Find five science -backed hobbies proven 00:14:41.580 --> 00:14:44.519 to reduce stress. Please include links to the 00:14:44.519 --> 00:14:47.259 studies or sources. Proplexity does that, gives 00:14:47.259 --> 00:14:49.100 you the list with verification. Okay, research 00:14:49.100 --> 00:14:53.019 done. Step two. Step two, the brain. You take 00:14:53.019 --> 00:14:55.940 that verified research from Proplexity and go 00:14:55.940 --> 00:14:58.379 to Claude, maybe. Why Claude? Because we want 00:14:58.379 --> 00:15:00.899 a warm, engaging tone, right? So you give Claude 00:15:00.899 --> 00:15:04.320 the role. You're a warm, empathetic, mental health 00:15:04.320 --> 00:15:06.679 blogger. Got it. Then you give it the research 00:15:06.679 --> 00:15:10.360 and the task. Write a 500 -word blog post based 00:15:10.360 --> 00:15:12.639 only on the provided research points about these 00:15:12.639 --> 00:15:14.919 five hobbies. Based only on the research. Key 00:15:14.919 --> 00:15:17.759 constraint. Crucial. So Claude writes the draft. 00:15:18.159 --> 00:15:20.960 Now step three. The visuals. The creator. Exactly. 00:15:21.039 --> 00:15:23.399 You go to ideogram, you ask it to create a custom 00:15:23.399 --> 00:15:25.179 image, maybe something cozy, showing someone 00:15:25.179 --> 00:15:27.820 watering plants or knitting, matching the blog's 00:15:27.820 --> 00:15:30.519 theme. Okay. So research, write, illustrate, 00:15:30.779 --> 00:15:33.440 using specialized tools for each step. And what's 00:15:33.440 --> 00:15:36.570 the result? in maybe, I don't know, 15, 20 minutes. 00:15:37.070 --> 00:15:38.929 You have a thoroughly researched, well -written 00:15:38.929 --> 00:15:41.409 blog post with custom visuals. The quality you 00:15:41.409 --> 00:15:44.610 get from that workflow, it's just leagues beyond 00:15:44.610 --> 00:15:48.509 asking one single AI to write a blog post about 00:15:48.509 --> 00:15:51.929 stress. Whoa. Yeah, I can see that. And imagine 00:15:51.929 --> 00:15:55.070 scaling that kind of researched quality workflow 00:15:55.070 --> 00:15:57.929 across tons of different... tasks or queries. 00:15:58.070 --> 00:16:00.129 That's the transformative potential right there. 00:16:00.309 --> 00:16:02.909 Real efficiency, but also high quality. So what's 00:16:02.909 --> 00:16:04.850 the biggest advantage of this manual stacking 00:16:04.850 --> 00:16:07.570 compared to just using one tool? Guaranteed quality 00:16:07.570 --> 00:16:10.169 by using each tool's proven strength for its 00:16:10.169 --> 00:16:13.470 specific job. Beat. Leverage a specialization 00:16:13.470 --> 00:16:16.190 for a better outcome. Excellent. Okay, that brings 00:16:16.190 --> 00:16:18.029 us to the final, and you said maybe the most 00:16:18.029 --> 00:16:20.710 crucial, skill. Skill number six, the human polish. 00:16:20.850 --> 00:16:24.049 Yes, the final touch. Because, let's be honest, 00:16:24.169 --> 00:16:27.559 a lot of raw AI content, it feels Cold. Generic. 00:16:27.679 --> 00:16:30.159 Corporate. Yeah, all of that. Yeah. Devoid of 00:16:30.159 --> 00:16:32.179 real personality. And the second a reader senses, 00:16:32.299 --> 00:16:34.720 oh, a robot wrote this. What happens? They disconnect. 00:16:34.840 --> 00:16:37.600 They stop trusting it. Exactly. Because AI, for 00:16:37.600 --> 00:16:40.480 all its power, lacks three key human elements, 00:16:40.840 --> 00:16:43.360 vision, taste, and care. And providing those, 00:16:43.399 --> 00:16:45.659 that's still our job. OK, I understand the goal 00:16:45.659 --> 00:16:48.100 making it feel human. But isn't spending time 00:16:48.100 --> 00:16:50.200 polishing the text after the AI did the heavy 00:16:50.200 --> 00:16:53.580 lifting kind of inefficient? If the goal is speed. 00:16:53.850 --> 00:16:55.889 Why slow down at the very end? That's a fair 00:16:55.889 --> 00:16:58.309 question. But think about the 80 -20 rule. Primitive 00:16:58.309 --> 00:17:01.210 principle. Right. Let the AI handle the 80 % 00:17:01.210 --> 00:17:03.090 of the initial draft, the research compilation, 00:17:03.230 --> 00:17:05.369 the outlining. That's where the massive time 00:17:05.369 --> 00:17:08.630 savings are. But that last 20%, that's the human 00:17:08.630 --> 00:17:11.289 polish. And that 20 % is where the value is added. 00:17:11.410 --> 00:17:14.630 That's where you turn a fine or OK draft into 00:17:14.630 --> 00:17:18.029 something genuinely great, something that connects, 00:17:18.170 --> 00:17:20.230 that people actually trust, remember, and share. 00:17:20.410 --> 00:17:22.289 It's the difference between competent and compelling. 00:17:22.539 --> 00:17:25.799 Okay, I am sold on the why. Now how? What does 00:17:25.799 --> 00:17:28.099 this human polish actually look like in practice? 00:17:28.500 --> 00:17:30.980 Technique one. Technique one. Add your voice. 00:17:31.240 --> 00:17:33.779 Inject your vision. AI gives you general truths. 00:17:34.240 --> 00:17:36.440 You replace them with specific stories, anecdotes, 00:17:36.920 --> 00:17:39.319 personality. Insanity. AI might write, consistency 00:17:39.319 --> 00:17:42.200 is key to achieving fitness goals. Which is true, 00:17:42.740 --> 00:17:47.170 but, eh. Your human polish adds a story. My old 00:17:47.170 --> 00:17:49.349 track coach used to hammer this into us. He'd 00:17:49.349 --> 00:17:51.849 say you don't get faster on the days You feel 00:17:51.849 --> 00:17:54.210 great and want to run you get faster on the days. 00:17:54.210 --> 00:17:56.450 It's raining You're tired and you run anyway 00:17:56.450 --> 00:18:00.269 much more memorable connects better. Okay technique 00:18:00.269 --> 00:18:05.250 to cut the AI words Exercise your taste AI loves 00:18:05.250 --> 00:18:10.930 empty formal jargon like what? Oh, you know In 00:18:10.930 --> 00:18:14.349 the context of furthermore a it is imperative 00:18:14.349 --> 00:18:18.500 that In today's fast -paced digital world, ugh. 00:18:19.099 --> 00:18:21.119 Yeah, those definitely scream AI -generated. 00:18:21.279 --> 00:18:23.539 Delete them. Simplify the language. Make it sound 00:18:23.539 --> 00:18:26.140 like a real person talking. Use shorter sentences. 00:18:26.299 --> 00:18:28.779 Use contractions. That's taste. Got it. And technique 00:18:28.779 --> 00:18:32.220 three. Add connection. Show some care. Insert 00:18:32.220 --> 00:18:35.039 empathy. AI is often very cold and instructional. 00:18:35.200 --> 00:18:38.029 Example of that. Yeah, might output. Users experiencing 00:18:38.029 --> 00:18:39.970 login difficulties should follow the prescribed 00:18:39.970 --> 00:18:42.589 password reset procedure. Super robotic. You're 00:18:42.589 --> 00:18:44.349 already detached. Your human polish changes it 00:18:44.349 --> 00:18:46.069 too. I know it's incredibly annoying when you 00:18:46.069 --> 00:18:48.109 get locked out of your account. Honestly, we've 00:18:48.109 --> 00:18:49.789 all been there. Here's the quickest way to reset 00:18:49.789 --> 00:18:52.789 your password. That little bit of empathy, we've 00:18:52.789 --> 00:18:55.230 all been there, totally changes the feeling. 00:18:55.690 --> 00:18:57.650 Completely. It shifts the dynamic instantly. 00:18:58.069 --> 00:19:00.430 How does adding personal stories and empathy 00:19:00.430 --> 00:19:02.990 fundamentally change the reader's relationship 00:19:02.990 --> 00:19:05.700 with the content? It moves it from machine -to 00:19:05.700 --> 00:19:08.140 -machine instruction to human -to -human connection. 00:19:08.799 --> 00:19:11.920 Beat. Builds trust and relatability instantly. 00:19:12.099 --> 00:19:13.940 Okay, let's pull this all together then. A recap 00:19:13.940 --> 00:19:16.460 of the process. Yeah, let's summarize. We covered 00:19:16.460 --> 00:19:19.400 the six core skills. First, choose your specialized 00:19:19.400 --> 00:19:23.000 tools. Build that AI team. Second, clarify your 00:19:23.000 --> 00:19:25.119 problem before you start typing. Know your goal. 00:19:25.589 --> 00:19:29.269 Third, communicate clearly used frameworks. Show, 00:19:29.390 --> 00:19:32.150 don't tell. Fourth, and this is huge, verify 00:19:32.150 --> 00:19:34.950 the facts. Always check the AI's work, especially 00:19:34.950 --> 00:19:37.809 with researchers. Fifth, combine tools through 00:19:37.809 --> 00:19:40.390 stacking used specialists for each step. And 00:19:40.390 --> 00:19:43.150 sixth, finish with that crucial human polish, 00:19:43.329 --> 00:19:45.910 add your voice, taste, and care. Essentially, 00:19:46.210 --> 00:19:48.390 AI is the incredibly capable doer. But you are 00:19:48.390 --> 00:19:50.630 the director. You're guiding the process, ensuring 00:19:50.630 --> 00:19:52.829 quality, adding the value. So for listeners, 00:19:52.970 --> 00:19:54.990 what's the first step? It feels like a lot to 00:19:54.990 --> 00:19:57.269 implement all at once. Oh, definitely don't try 00:19:57.269 --> 00:19:59.599 to master all six today. That's overwhelming. 00:19:59.980 --> 00:20:02.400 Just pick one skill to focus on this week. Seriously, 00:20:02.680 --> 00:20:05.359 just one. Like maybe. Maybe just really commit 00:20:05.359 --> 00:20:07.599 to using that six -part framework for your next 00:20:07.599 --> 00:20:10.740 important prompt. Or maybe just build the habit 00:20:10.740 --> 00:20:13.740 of always running AI facts through perplexity. 00:20:13.900 --> 00:20:16.319 Just one small change. Exactly. Even just doing 00:20:16.319 --> 00:20:19.619 the verification step consistently will dramatically 00:20:19.619 --> 00:20:22.079 improve your results and safety almost overnight. 00:20:22.460 --> 00:20:25.099 Okay. So here's a final thought to leave everyone 00:20:25.099 --> 00:20:29.059 with. Something to chew on. If AI can now reliably 00:20:29.059 --> 00:20:32.900 handle, say, 80 % of the grunt work, the research, 00:20:33.019 --> 00:20:35.539 the drafting, the summarizing, then maybe the 00:20:35.539 --> 00:20:38.079 new essential human skill in this era isn't really 00:20:38.079 --> 00:20:40.579 about becoming a master prompt engineer or knowing 00:20:40.579 --> 00:20:42.920 the tech inside out. Maybe it's shifting more 00:20:42.920 --> 00:20:44.960 towards cultivating taste, expressing empathy, 00:20:45.079 --> 00:20:48.369 and having a clear vision. Interesting. If basic 00:20:48.369 --> 00:20:50.390 content production is becoming commoditized, 00:20:50.490 --> 00:20:54.390 almost solved, what entirely new kinds of creative, 00:20:54.589 --> 00:20:57.410 strategic, or editorial roles might emerge that 00:20:57.410 --> 00:20:59.650 really lean into those uniquely human skills? 00:21:00.190 --> 00:21:02.250 What does work look like when the doing part 00:21:02.250 --> 00:21:04.630 is largely automated? That's a deep question. 00:21:04.869 --> 00:21:06.849 Something to really think about. Thanks for tuning 00:21:06.849 --> 00:21:09.970 in to the Deep Dives, everyone. OTero music.