WEBVTT 00:00:00.000 --> 00:00:02.919 There is a trap that almost everyone falls into 00:00:02.919 --> 00:00:05.700 when they first start using these modern AI tools. 00:00:05.900 --> 00:00:08.099 I find myself doing it, even though I know better. 00:00:08.279 --> 00:00:12.080 It's the Google plus chat GPT trap. You treat 00:00:12.080 --> 00:00:14.640 the search bar like a slot machine. You pull 00:00:14.640 --> 00:00:17.260 the lever, you type in a generic question, and 00:00:17.260 --> 00:00:19.140 you just sort of hope the algorithm spits out 00:00:19.140 --> 00:00:21.640 gold. Right. It's the pull and pray method. Exactly. 00:00:21.820 --> 00:00:25.300 It's passive. And honestly, it is how we have 00:00:25.300 --> 00:00:27.940 been trained to use the Internet for 20 years. 00:00:28.280 --> 00:00:30.000 But according to the research we are looking 00:00:30.000 --> 00:00:32.500 at today, if that is how you are using perplexity 00:00:32.500 --> 00:00:35.719 AI, you are effectively driving a Ferrari in 00:00:35.719 --> 00:00:39.460 first gear. You are utilizing maybe. 10 of its 00:00:39.460 --> 00:00:41.479 actual power which is wild because the other 00:00:41.479 --> 00:00:44.159 90 isn't just about getting slightly smarter 00:00:44.159 --> 00:00:46.700 answers yeah it's about changing the fundamental 00:00:46.700 --> 00:00:49.159 relationship you have with the tool yeah it stops 00:00:49.159 --> 00:00:51.560 being a search engine you visit and starts becoming 00:00:51.560 --> 00:00:53.579 a research system that you design that is the 00:00:53.579 --> 00:00:55.640 phrase that stuck out to me intentional system 00:00:55.640 --> 00:00:59.560 design so welcome to the deep dive today we are 00:00:59.560 --> 00:01:02.039 unpacking precision research we're going to look 00:01:02.039 --> 00:01:04.900 at how to master the advanced workflows of perplexity 00:01:04.900 --> 00:01:08.799 ai and we are moving past the basics here Know 00:01:08.799 --> 00:01:11.700 how to sign up tutorials. This is for the person 00:01:11.700 --> 00:01:13.959 who wants to go from casual user to power user. 00:01:14.140 --> 00:01:16.859 We have a roadmap for this conversation. We are 00:01:16.859 --> 00:01:19.599 going to cover five specific shifts in workflow. 00:01:20.019 --> 00:01:22.260 First, we'll talk about controlling where the 00:01:22.260 --> 00:01:25.319 AI looks using search operators. Second, we'll 00:01:25.319 --> 00:01:28.359 look at automating the busy work with slash commands 00:01:28.359 --> 00:01:31.439 and recurring tasks. Which is a massive time 00:01:31.439 --> 00:01:33.819 saver for anyone who hates repetitive typing. 00:01:34.620 --> 00:01:37.099 Third, we'll discuss building a persistent brain 00:01:37.099 --> 00:01:40.260 using spaces. Fourth, selecting the right mode 00:01:40.260 --> 00:01:42.879 and model for the specific job because they aren't 00:01:42.879 --> 00:01:45.159 all the same. And finally, we'll get into advanced 00:01:45.159 --> 00:01:47.439 prompting techniques like chaining and perspectives. 00:01:47.739 --> 00:01:49.700 It's a full stack for information processing. 00:01:49.980 --> 00:01:51.819 It is. So let's start with the first concept, 00:01:52.019 --> 00:01:54.120 the art of constraint. The source material makes 00:01:54.120 --> 00:01:56.540 a point here that feels counterintuitive. It 00:01:56.540 --> 00:01:58.620 says that quality gains don't come from expanding 00:01:58.620 --> 00:02:01.739 your search, but from limiting it. Yeah. This 00:02:01.739 --> 00:02:04.230 is the biggest mental hurdle. We are trained 00:02:04.230 --> 00:02:08.189 to think that big data means more data. If I 00:02:08.189 --> 00:02:11.150 ask a question, I want the AI to scour the entire 00:02:11.150 --> 00:02:13.370 Internet, right? Why wouldn't I? Theoretically, 00:02:13.370 --> 00:02:15.650 yes. More sources should equal a better answer. 00:02:15.889 --> 00:02:19.030 But practically, no. Because the Internet is 00:02:19.030 --> 00:02:22.169 noisy. It is a dumpster fire of conflicting information. 00:02:22.629 --> 00:02:25.469 If you ask a technical question, let's say you're 00:02:25.469 --> 00:02:27.870 trying to figure out a specific nuance in a Python 00:02:27.870 --> 00:02:31.289 library or a tax law change, and you let the 00:02:31.289 --> 00:02:34.620 AI look everywhere. What happens? It pulls from 00:02:34.620 --> 00:02:36.780 everywhere. It stitches together an answer from 00:02:36.780 --> 00:02:39.500 the official documentation, sure. But it also 00:02:39.500 --> 00:02:42.639 pulls from three angry Reddit threads from 2019, 00:02:42.879 --> 00:02:45.620 a marketing blog trying to sell you a course, 00:02:45.740 --> 00:02:48.560 and maybe a YouTube comment section. It flattens 00:02:48.560 --> 00:02:50.780 the hierarchy of truth. A random comment weighs 00:02:50.780 --> 00:02:53.080 as much as the documentation. Exactly. It creates 00:02:53.080 --> 00:02:55.080 a Frankenstein answer. It might be plausible, 00:02:55.240 --> 00:02:57.520 but it's not precise. So the source highlights 00:02:57.520 --> 00:03:00.139 the power of the site operator. This is the most 00:03:00.139 --> 00:03:01.979 useful tool in your kit. Walk us through that. 00:03:02.000 --> 00:03:03.879 How does it work in practice? It's incredibly 00:03:03.879 --> 00:03:06.259 simple. You type your query, and then you add 00:03:06.259 --> 00:03:09.759 site, followed by the domain you trust. So site 00:03:09.759 --> 00:03:13.780 .nasa .gov or site .apple .com. So you're forcing 00:03:13.780 --> 00:03:16.379 the AI to web blinders. You're forcing it to 00:03:16.379 --> 00:03:18.870 ignore the noise. Let's say you want to understand 00:03:18.870 --> 00:03:21.469 how the new Stripe API handles recurring billing. 00:03:21.669 --> 00:03:24.289 If you just ask perplexity, you'll get a mix 00:03:24.289 --> 00:03:26.729 of opinions. But if you type your question and 00:03:26.729 --> 00:03:30.669 add site .docs .stripe .com, you are guaranteeing 00:03:30.669 --> 00:03:32.250 that the answer isn't coming from a confused 00:03:32.250 --> 00:03:34.729 developer on a forum. It eliminates opinions. 00:03:35.330 --> 00:03:38.129 We usually think more data is better, but here, 00:03:38.270 --> 00:03:41.310 less is actually more. Why is exclusion so powerful 00:03:41.310 --> 00:03:44.189 in this context? Because exclusion removes the 00:03:44.189 --> 00:03:47.400 noise. forcing the AI to focus only on trusted 00:03:47.400 --> 00:03:50.500 truth. The source also mentions time -based operators 00:03:50.500 --> 00:03:53.520 after and before. I usually ignore these in Google 00:03:53.520 --> 00:03:55.400 searches. Why are they critical here? Because 00:03:55.400 --> 00:03:58.060 of how fast the world moves, especially in tech 00:03:58.060 --> 00:04:00.840 or news. If you ask about the best open source 00:04:00.840 --> 00:04:03.539 LLM right now. And perplexity pulls an article 00:04:03.539 --> 00:04:06.219 from November 2023. That answer is worse than 00:04:06.219 --> 00:04:08.539 useless. It's actively misleading. The field 00:04:08.539 --> 00:04:10.960 moves too fast. Exactly. By adding after plot 00:04:10.960 --> 00:04:13.580 2024, you ensure freshness. You aren't reading 00:04:13.580 --> 00:04:16.220 history. You're reading news. And my personal 00:04:16.220 --> 00:04:19.759 favorite from the deep dive, filetype .pdf. Ah, 00:04:19.920 --> 00:04:22.819 the academic filter. It's the anti -SEO filter. 00:04:23.269 --> 00:04:25.769 If you want deep research on a medical topic 00:04:25.769 --> 00:04:28.389 or economic trends, you don't want the top 10 00:04:28.389 --> 00:04:31.329 Google results. Those are all SEO optimized marketing 00:04:31.329 --> 00:04:33.850 fluff. You want white papers. You want government 00:04:33.850 --> 00:04:37.470 filings. Adding filetype .pdf tells Perplexity, 00:04:37.529 --> 00:04:40.209 don't show me the blog posts. Show me the receipts. 00:04:40.389 --> 00:04:43.470 That brings us to our second segment, systemizing 00:04:43.470 --> 00:04:46.050 the workflow. This is about moving from a one 00:04:46.050 --> 00:04:49.649 -off tool to a daily assistant. Right. The goal 00:04:49.649 --> 00:04:52.250 is to stop typing the same things over and over 00:04:52.250 --> 00:04:54.790 again. The source talks about slash commands. 00:04:55.149 --> 00:04:58.029 Now, for people who use tools like Slack or Notion, 00:04:58.110 --> 00:05:00.670 this might feel familiar. But how does it apply 00:05:00.670 --> 00:05:03.329 to a search engine? Think of slash commands as 00:05:03.329 --> 00:05:06.250 saved prompt templates. We all have those tasks 00:05:06.250 --> 00:05:08.610 we do repeatedly. Maybe you take messy notes 00:05:08.610 --> 00:05:10.649 and you always want them formatted into a clean, 00:05:10.670 --> 00:05:13.389 bulleted list. Or you take a complex article 00:05:13.389 --> 00:05:15.189 and you always want a three -sentence summary. 00:05:15.430 --> 00:05:17.629 Right. And usually I type out, please take this 00:05:17.629 --> 00:05:19.970 text and summarize it into three sentences every 00:05:19.970 --> 00:05:23.300 single time. Exactly. And if you're tired, maybe 00:05:23.300 --> 00:05:25.699 you forget to say three sentences or you forget 00:05:25.699 --> 00:05:28.639 to say keep it professional. The output becomes 00:05:28.639 --> 00:05:31.480 inconsistent. With the slash command, you save 00:05:31.480 --> 00:05:34.600 that instruction once. You call it summary. So 00:05:34.600 --> 00:05:36.860 I just type slash summary. You type in the box, 00:05:36.939 --> 00:05:40.160 pick summary, paste your text, and hit enter. 00:05:40.799 --> 00:05:43.779 It executes the exact same prompt structure every 00:05:43.779 --> 00:05:46.740 time. It guarantees consistency. It removes the 00:05:46.740 --> 00:05:49.759 cognitive load of having to manage the AI. You 00:05:49.759 --> 00:05:51.839 just trigger the workflow. I have to admit, I 00:05:51.839 --> 00:05:53.839 still type the same prompt structure manually 00:05:53.839 --> 00:05:56.139 every morning. I haven't automated my own habits 00:05:56.139 --> 00:05:58.279 yet. It feels like one of those things where 00:05:58.279 --> 00:06:00.120 you know you should do it, but you just keep 00:06:00.120 --> 00:06:01.959 doing it the hard way because it feels faster 00:06:01.959 --> 00:06:04.300 in the moment. It's a classic, I'm too busy chopping 00:06:04.300 --> 00:06:07.160 wood to sharpen the axe problem. But the source 00:06:07.160 --> 00:06:08.939 points out that slash commands are currently 00:06:08.939 --> 00:06:11.759 a web and desktop feature. But once you set them 00:06:11.759 --> 00:06:14.660 up, it saves hours over a month. But slash commands 00:06:14.660 --> 00:06:17.379 still require you to be there typing. The source 00:06:17.379 --> 00:06:19.779 mentions a pro feature called recurring tasks. 00:06:20.319 --> 00:06:23.180 This sounded fascinating. This is where we cross 00:06:23.180 --> 00:06:27.019 the line from tool to agent. A recurring task 00:06:27.019 --> 00:06:29.980 is a search that runs itself. Without me. Completely 00:06:29.980 --> 00:06:31.720 without you. Yeah. Let's say you need to track 00:06:31.720 --> 00:06:33.860 a competitor's pricing or you want to know if 00:06:33.860 --> 00:06:35.560 there's any new news about a specific stock. 00:06:36.560 --> 00:06:39.699 You set up a prompt, you tell perplexity, run 00:06:39.699 --> 00:06:42.420 this every day at 8 a .m., and you walk away. 00:06:42.579 --> 00:06:45.480 So I don't open the app. No. You wake up, you 00:06:45.480 --> 00:06:47.379 check your notifications, and the briefing is 00:06:47.379 --> 00:06:49.779 there. It shifts to the dynamic. You aren't checking 00:06:49.779 --> 00:06:51.939 for updates. You are receiving them. So this 00:06:51.939 --> 00:06:55.019 essentially turns the AI from a tool you pick 00:06:55.019 --> 00:06:57.860 up into a worker that clocks in for you. Exactly. 00:06:57.939 --> 00:06:59.819 It works in the background, so you wake up to 00:06:59.819 --> 00:07:02.620 answers, not empty search bars. Let's move to 00:07:02.620 --> 00:07:04.620 the third segment. This is about memory and context. 00:07:05.199 --> 00:07:08.180 The source calls this spaces. This solves the 00:07:08.180 --> 00:07:10.519 amnesia problem. The amnesia problem. Standard 00:07:10.519 --> 00:07:13.079 AI chats are ephemeral. You have a great conversation, 00:07:13.220 --> 00:07:15.839 you close the tab, and it's gone. Or it's buried 00:07:15.839 --> 00:07:19.000 in history. If you start a new chat, the AI has 00:07:19.000 --> 00:07:21.519 forgotten everything you just discussed. It doesn't 00:07:21.519 --> 00:07:23.519 know who you are or what you're working on. So 00:07:23.519 --> 00:07:26.720 how do spaces fix that? A space is a container. 00:07:27.399 --> 00:07:30.019 It's a persistent environment. You can create 00:07:30.019 --> 00:07:32.839 a space called Project Alpha or Marketing Strategy. 00:07:33.519 --> 00:07:37.420 And here is the killer feature, knowledge injection. 00:07:37.779 --> 00:07:40.519 Knowledge injection? Sounds sci -fi? It kind 00:07:40.519 --> 00:07:43.839 of is. You can upload files, PDFs, spreadsheets, 00:07:43.860 --> 00:07:46.500 massive internal documents into that space. Right. 00:07:46.579 --> 00:07:48.420 And you can tell perplexity. Everything in this 00:07:48.420 --> 00:07:51.160 space must reference these files. So if I upload 00:07:51.160 --> 00:07:53.800 my company's brand guidelines into a space and 00:07:53.800 --> 00:07:55.980 then I ask it to write a blog post inside that 00:07:55.980 --> 00:07:57.939 space. It'll write that blog post using your 00:07:57.939 --> 00:08:00.160 brand voice, adhering to your guidelines without 00:08:00.160 --> 00:08:02.899 you having to remind it. It remembers. The source 00:08:02.899 --> 00:08:05.079 also mentioned something about isolation. This 00:08:05.079 --> 00:08:07.899 is huge for enterprise or deep research. You 00:08:07.899 --> 00:08:10.220 can toggle a setting in a space that says no 00:08:10.220 --> 00:08:13.360 web. No web, but it's a search engine. But sometimes 00:08:13.360 --> 00:08:16.060 you don't want the web. If you are analyzing 00:08:16.060 --> 00:08:18.680 confidential internal financial reports, you 00:08:18.680 --> 00:08:20.920 don't want the AI hallucinating numbers from 00:08:20.920 --> 00:08:23.560 a Yahoo Finance article. You want to look only 00:08:23.560 --> 00:08:25.879 at the files you uploaded. By turning off the 00:08:25.879 --> 00:08:28.100 web, you force it to ground its reality entirely 00:08:28.100 --> 00:08:31.319 in your data. Whoa. Imagine the scale of that. 00:08:31.519 --> 00:08:33.480 You're basically building a custom brain that 00:08:33.480 --> 00:08:36.580 only knows what you want it to know. You strip 00:08:36.580 --> 00:08:38.299 away the rest of the world and just focus on 00:08:38.299 --> 00:08:41.460 your signal. That's the power. It becomes a personalized 00:08:41.460 --> 00:08:44.259 knowledge base that actually remembers your specific 00:08:44.259 --> 00:08:47.200 context. Does this change the relationship with 00:08:47.200 --> 00:08:49.679 the AI from a search engine to something else? 00:08:50.159 --> 00:08:52.039 Yes, it becomes a personalized knowledge base 00:08:52.039 --> 00:08:54.240 that actually remembers your specific context. 00:08:54.559 --> 00:08:56.820 We have so much more to cover, including the 00:08:56.820 --> 00:08:59.100 different brains you can swap in and out and 00:08:59.100 --> 00:09:01.500 the Socratic method of prompting. But first, 00:09:01.659 --> 00:09:04.980 a quick break. We are back. We're deep diving 00:09:04.980 --> 00:09:08.600 into precision research with Perplexity AI. We've 00:09:08.600 --> 00:09:12.299 covered operators, automation, and spaces. Now 00:09:12.299 --> 00:09:14.440 let's talk about the engine under the hood, or 00:09:14.440 --> 00:09:17.080 rather, the engines. Right. This is segment four, 00:09:17.279 --> 00:09:19.980 the brains and the modes. The source makes a 00:09:19.980 --> 00:09:22.200 distinction here that I think confuses a lot 00:09:22.200 --> 00:09:25.720 of people. Perplexity isn't one single AI model. 00:09:25.879 --> 00:09:28.860 It's a wrapper. It's an interface layer. Underneath, 00:09:28.860 --> 00:09:30.879 you can actually choose which brain is doing 00:09:30.879 --> 00:09:33.259 the thinking. And this is critical because different 00:09:33.259 --> 00:09:35.179 models have different personalities. Let's run 00:09:35.179 --> 00:09:36.960 through them. What are the options and when should 00:09:36.960 --> 00:09:39.600 we use them? So you have the default model. It's 00:09:39.600 --> 00:09:42.419 balanced. It's fast. It's great for 80 % of things. 00:09:42.539 --> 00:09:45.320 But then you have Claude. From Anthropic. Right. 00:09:45.639 --> 00:09:47.950 Claude is generally calmer. more structured, 00:09:48.070 --> 00:09:50.570 and significantly better at writing. If you want 00:09:50.570 --> 00:09:53.610 a blog post or a nuanced explanation, you switch 00:09:53.610 --> 00:09:56.470 the model to Claude. And Gemini. That's Google's 00:09:56.470 --> 00:09:58.990 model. It shines with mixed inputs if you're 00:09:58.990 --> 00:10:01.909 analyzing images or need fast multimodal processing. 00:10:02.190 --> 00:10:05.809 And then, of course, GPT -4 from OpenAI. GPT 00:10:05.809 --> 00:10:08.950 is the logician. If you have a complex reasoning 00:10:08.950 --> 00:10:12.190 problem, a math puzzle, or a coding task, GPT 00:10:12.190 --> 00:10:14.029 tends to be better at following step -by -step 00:10:14.029 --> 00:10:16.470 instructions without getting creative in the 00:10:16.470 --> 00:10:19.100 wrong way. the source suggests a testing method 00:10:19.100 --> 00:10:20.779 for this yeah and it's something everyone should 00:10:20.779 --> 00:10:24.159 do at least once take a complex prompt run it 00:10:24.159 --> 00:10:26.759 through the default model then switch to quad 00:10:26.759 --> 00:10:30.559 and run the exact same prompt then gpt the answers 00:10:30.559 --> 00:10:33.419 will be different not just in style but in substance 00:10:33.419 --> 00:10:36.159 you'll start to develop an intuition for which 00:10:36.159 --> 00:10:39.039 brain fits your problem now alongside the model 00:10:39.039 --> 00:10:41.740 there is the mode the mode is the method how 00:10:41.740 --> 00:10:44.500 does the ai go about finding the answer The source 00:10:44.500 --> 00:10:47.379 lists three main ones. Search, research, and 00:10:47.379 --> 00:10:49.899 labs. Search mode is what we're used to. Speed 00:10:49.899 --> 00:10:52.279 over depth. What is the capital of France? Who 00:10:52.279 --> 00:10:55.200 won the game last night? It grabs a few sources, 00:10:55.379 --> 00:10:58.340 synthesizes them. Boom. Done in seconds. But 00:10:58.340 --> 00:11:00.240 then there is research mode. The source calls 00:11:00.240 --> 00:11:03.220 this the powerhouse. Research mode is fascinating. 00:11:03.440 --> 00:11:06.460 It's slower by design. When you ask a question 00:11:06.460 --> 00:11:09.340 in research mode, perplexity doesn't just answer 00:11:09.340 --> 00:11:12.259 it. It breaks your question down into sub -questions. 00:11:12.669 --> 00:11:14.669 Can you give us an example? Sure. If you ask, 00:11:14.769 --> 00:11:17.549 what is the impact of the new tax law on small 00:11:17.549 --> 00:11:20.350 businesses? In search mode, it finds a summary. 00:11:20.549 --> 00:11:22.789 In research mode, it acts like a project manager. 00:11:23.049 --> 00:11:25.669 It says, OK, I need to look up the tax law text. 00:11:25.870 --> 00:11:27.889 I need to look up analysis for retail businesses. 00:11:28.129 --> 00:11:29.909 I need to look up analysis for tech startups. 00:11:30.090 --> 00:11:32.049 I need to look at projected economic impacts. 00:11:32.730 --> 00:11:35.549 It runs four or five parallel searches, reads 00:11:35.549 --> 00:11:38.129 10 or 20 sources, and then writes a comprehensive 00:11:38.129 --> 00:11:40.909 report. So it synthesizes. It synthesizes. Yeah. 00:11:41.230 --> 00:11:43.110 It's for when you need a report, not just a fact. 00:11:43.269 --> 00:11:46.389 Labs is for artifacts. If you need it to write 00:11:46.389 --> 00:11:48.690 code, generate a chart, or create a specific 00:11:48.690 --> 00:11:51.730 document structure, labs acts more like an agent 00:11:51.730 --> 00:11:54.549 developer. It seems most frustration comes from 00:11:54.549 --> 00:11:57.570 asking a logic model to do creative work, or 00:11:57.570 --> 00:12:00.710 vice versa. How do we build the intuition for 00:12:00.710 --> 00:12:03.590 which to pick? Trial and error. Test the same 00:12:03.590 --> 00:12:05.950 prompt on two models. The winner becomes your 00:12:05.950 --> 00:12:08.470 default for that task. Let's move to our final 00:12:08.470 --> 00:12:11.159 segment, advanced reasoning. This is where we 00:12:11.159 --> 00:12:13.340 stop looking at features and start looking at 00:12:13.340 --> 00:12:16.159 how we talk to the machine. The source discusses 00:12:16.159 --> 00:12:18.899 chaining. This is the antidote to the God prompt. 00:12:18.980 --> 00:12:21.340 The God prompt. You know, the one you try to 00:12:21.340 --> 00:12:24.240 cram everything into one massive paragraph. Please 00:12:24.240 --> 00:12:26.299 explain quantum physics compared to string theory. 00:12:26.480 --> 00:12:28.600 Tell me which one is better for a sci -fi novel 00:12:28.600 --> 00:12:31.179 and write the opening chapter. Guilty. We all 00:12:31.179 --> 00:12:34.159 are. But the AI gets overwhelmed. It tried to 00:12:34.159 --> 00:12:35.899 do everything at once and gives you a shallow 00:12:35.899 --> 00:12:38.419 version of everything. Chaining is about breaking 00:12:38.419 --> 00:12:41.259 it down into a conversation. Step one, clarify. 00:12:41.799 --> 00:12:44.320 What are the main theories of quantum physics? 00:12:44.620 --> 00:12:47.899 Get the landscape. Step two, go deeper. Explain 00:12:47.899 --> 00:12:50.879 string theory in detail. Step three, compare. 00:12:51.320 --> 00:12:53.799 Create a table comparing quantum loop gravity 00:12:53.799 --> 00:12:57.820 and string theory. Step four, apply, which offers 00:12:57.820 --> 00:13:00.100 better plot devices for a time travel story. 00:13:00.419 --> 00:13:02.799 By the time you get to step four. The AI has 00:13:02.799 --> 00:13:05.200 all the context from steps one, two, and three. 00:13:05.559 --> 00:13:08.360 You aren't asking it to guess. You've built a 00:13:08.360 --> 00:13:10.570 foundation of facts together. The reasoning becomes 00:13:10.570 --> 00:13:12.490 much sharper because you forced it to show its 00:13:12.490 --> 00:13:14.690 work along the way. The source also mentions 00:13:14.690 --> 00:13:16.929 perspective prompts. This is a simple trick, 00:13:16.929 --> 00:13:19.610 but effective. Default AI answers are neutral. 00:13:19.809 --> 00:13:22.149 They try to be objective, which often means bland. 00:13:22.389 --> 00:13:24.070 On the one hand, on the other hand. Exactly. 00:13:24.210 --> 00:13:26.509 But if you say, answer this as a cynical SaaS 00:13:26.509 --> 00:13:28.830 founder or answer this as a cautious academic 00:13:28.830 --> 00:13:31.769 researcher, the output changes completely. How 00:13:31.769 --> 00:13:34.669 so? The SaaS founder will prioritize speed, profit, 00:13:34.870 --> 00:13:37.549 and growth. The academic will prioritize accuracy, 00:13:37.950 --> 00:13:40.320 citations, and caveats. The facts might be the 00:13:40.320 --> 00:13:42.580 same, but the wisdom is different. You get to 00:13:42.580 --> 00:13:45.259 view the data through a specific lens. Finally, 00:13:45.279 --> 00:13:47.799 there's a mention of Notebook LM as a companion 00:13:47.799 --> 00:13:50.379 tool. We've talked about Notebook LM on other 00:13:50.379 --> 00:13:53.840 deep dives, but how does it fit here? Perplexity 00:13:53.840 --> 00:13:56.460 is for hunting. It's for finding the information. 00:13:57.480 --> 00:14:00.179 Notebook LM is for gathering. You don't want 00:14:00.179 --> 00:14:02.500 to lose that great research you just did. So 00:14:02.500 --> 00:14:05.620 the workflow is hunt in perplexity, find the 00:14:05.620 --> 00:14:07.879 gold nuggets, and copy -paste them into Notebook 00:14:07.879 --> 00:14:10.200 LM. That becomes your permanent library that 00:14:10.200 --> 00:14:12.840 you can query later. This reminds me of the Socratic 00:14:12.840 --> 00:14:15.159 method. We aren't just demanding answers. We 00:14:15.159 --> 00:14:17.259 are teaching it how to think about the problem. 00:14:17.500 --> 00:14:19.500 Right. You're building a ladder of logic, stepping 00:14:19.500 --> 00:14:22.019 up from facts to wisdom. So we've covered a lot 00:14:22.019 --> 00:14:23.840 of ground today. Let's try to pull this all together 00:14:23.840 --> 00:14:27.039 in a Big Idea recap. Sure. The big idea is that 00:14:27.039 --> 00:14:29.279 perplexity is not just a better Google. It is 00:14:29.279 --> 00:14:31.519 a research engine that requires a pilot. If we 00:14:31.519 --> 00:14:34.399 break it down to the key takeaways. First, constraints. 00:14:35.500 --> 00:14:39.090 Use operators like sight. And after, to filter 00:14:39.090 --> 00:14:41.230 out the garbage. Quality comes from exclusion. 00:14:41.610 --> 00:14:45.049 Second, automation. Use slash commands for templates 00:14:45.049 --> 00:14:47.269 and recurring tasks for things that happen on 00:14:47.269 --> 00:14:49.710 a schedule. Make the machine work for you. Third, 00:14:49.929 --> 00:14:53.590 context. Use spaces to ground the AI in your 00:14:53.590 --> 00:14:56.429 own data and files. Give it a memory. Fourth, 00:14:56.669 --> 00:14:59.769 selection. Be intentional about the model, Claude 00:14:59.769 --> 00:15:02.809 versus GPT, and the mode, search versus research. 00:15:03.570 --> 00:15:06.639 Don't use a hammer to turn a screw. And finally, 00:15:06.799 --> 00:15:09.740 reasoning. Chain your questions. Don't ask for 00:15:09.740 --> 00:15:12.580 miracles in one prompt. Guide the thinking process. 00:15:13.200 --> 00:15:15.399 The source ends with a really provocative thought. 00:15:15.519 --> 00:15:17.879 It says that once you master these tools, the 00:15:17.879 --> 00:15:20.279 AI stops feeling impressive and starts feeling 00:15:20.279 --> 00:15:22.759 reliable. I love that. It implies that the wow 00:15:22.759 --> 00:15:25.340 factor is actually a distraction. It is. When 00:15:25.340 --> 00:15:28.059 technology is new. We want to be dazzled. Oh, 00:15:28.100 --> 00:15:29.940 look, it wrote a poem. But when you're actually 00:15:29.940 --> 00:15:31.940 working, you don't want to be dazzled. You want 00:15:31.940 --> 00:15:34.580 to be supported. You want the tool to disappear. 00:15:35.019 --> 00:15:37.399 Reliability is boring. Reliability is boring. 00:15:37.480 --> 00:15:40.460 And boring is productive. If you use these workflows, 00:15:40.659 --> 00:15:42.799 if you set up the tasks, if you build the spaces, 00:15:43.200 --> 00:15:46.019 perplexity becomes like electricity. You don't 00:15:46.019 --> 00:15:48.179 marvel at the light switch. You just expect the 00:15:48.179 --> 00:15:50.779 light to turn on. That is where we want to get 00:15:50.779 --> 00:15:53.100 to. That is the goal, to reach a point where 00:15:53.100 --> 00:15:55.059 you just have the knowledge you need when you 00:15:55.059 --> 00:15:57.059 need it. Thank you for listening to this deep 00:15:57.059 --> 00:16:00.159 dive. Go try out a recurring task this week. 00:16:00.440 --> 00:16:02.919 See if you can automate just one piece of your 00:16:02.919 --> 00:16:04.940 daily curiosity. It's worth it. We'll see you 00:16:04.940 --> 00:16:05.299 next time.