Episode Transcript
[00:00:05] Welcome to Prompting Curiosity, a podcast for the AI curious. No coding background required. I'm your host, Dr. Shantae Cofield, also known as the Maestro, and I created this show to explore what these AI tools actually are. Really though, are the files in the computer, how to use them, and what they might mean for how we think, work, create, and move through life. Whether you're skeptical, intrigued, or already experimenting, you're in the right place. All that I ask is that you, you stay curious. All right, let's get into it.
[00:00:38] Hello, hello, hello, my curious people, and welcome to episode 54, the Big Five Four, of prompting curiosity. I am your grateful host, the Maestro, and today we are talking about Perplexity, the quote unquote answer engine. So real talk, right off the bat, I am making this episode because I've heard the name Perplexity for quite some time now, and I straight up don't know what it, what it is. I didn't know what it was. Uh, so I decided to get a little curious and do a little learning. And so here we are. So, based on what I found, this is definitely not an I'm talking about this AI because I recommend this tool kind of episode. Um, I've yet to really use Perplexity outside of a teeny bit of tinkering I did so that I could create the episode. Um, and I don't, you know, based on what I found, I don't have any use cases for it. Off the top of my head, I am also not convinced that it's any better than Claude for my specific needs, but tbd.
[00:01:34] Lastly, I'm also not sure if Perplexity is on its way out the door. Right. There's lots of negative talk on Reddit, uh, and we know what happens when companies stray from their original service, but we are curious people, so let's dig in. So what is Perplexity per its own website? Perplexity is an AI powered answer engine that provides accurate, trusted and real time answers to any question in the form of conversational summaries. And this is the important part. Backed by real time web citations, the translation, the main point that I want to highlight, it provides citations. They are links with its answers. All right, so it's just like a chatbot, but at the end of each of the answers it gives you, it will have links to websites and such that it found that information worth naming up front. Perplexity isn't its own actual foundation. Foundation model. Like, it's not its own LLM. It is what's called a harness and that harness that, let's think about a harness like that you would use on a horse. The horse is the engine, the horse is comparable to the LLM, the harness, the reins. That is what you know, allows you to have the horse jump and go do what you wanted to do. I'm not again, as I'm saying this like I am not for animal cruelty or anything like that, but that's the best way to understand this. Then it's called a harness. Right. So that is what Plexi is, It is the harness.
[00:02:57] Right. And it orchestrates other, other companies models. So it uses GPT, uses Claude Gemini Grok, there are some in house ones. Um, but it uses all of the big frontier models and then it places that on top of Live search. Right. So the user can select which model they want to use, whether it's GPT, Claude Gemini, whatever, and then you just input your prompt as usual.
[00:03:21] Worth noting though, only users on the plate on the paid plan can actually select the model they want to use. The free tier users, any free tier users, which is me, um, they must use whatever Perplexity has set as the default model. When I used it, it was using Claude Sonnet 5, I want to say, which is fine. That's what I met my data driver anyway. Um, but just, just worth noting that it does have access to all of the frontier models. But, but you can only choose to which one you want to use if you were on a paid plan. So Perplexity offers two plans just like the regular pricing for everything else. There's Per Perplexity Pro which is $20 a month if you're going with monthly billing, a little bit cheaper if you do it annually. And then there's Perplexity max, which is $200 a month and that is if you're paying monthly. All right, so there is no in between planner that I saw. Um, unlike, you know, Claude does have like the a hundred dollar a month plan.
[00:04:12] There is none of that for perplexity. It's 20 or 200 from what I saw.
[00:04:17] So who is this for? Who was Perplexity for? Who's probably the biggest users of this? We're looking at researchers, consultants, educators, analysts. Anyone who's deliverable has to, you know, like survive someone asking where'd you get information from? Right. It is way less useful for creative, uh, for creative writing, um, way less useful for any kind of coding, um, or huge single document deep dive. So I. Nothing like that.
[00:04:44] Perplexity has been around for a bit. It actually went live on December 7th in 2022. That's seven days after ChatGPT launched. ChatGPT launched at the beginning of December in 2022. And, uh, then a week later, here came Perplexity. So, like, I've heard this name for a bit and I'm like, what the fuck is it? Um, and it differentiated itself with those citations that I mentioned earlier. So I can't help but wonder if the timing was actually fortuitous, if it was actually helpful because we had chat GPT, you know, making up, being cool. It's like, I can write you a poem, but then you, like, ask it for information. You're like, this seems wrong. And perplexity comes out at the same time as like, hey, I'm, um, you know, here's the citation. Here's where I got this information from. Here's, here's why this answer is, you know, quote unquote, true, you know, and allows. This allows Perplexity to establish itself in the market as a trustworthy source, which is helpful for it.
[00:05:34] All right? But let's note that that same format that it uses, this, grab it from the web and reformat the answer with citations. That's all it does, right? It goes, it searches the webs, the web finds the answer, paraphrases it, and then gives you the citation that has landed Perplexity in a few active lawsuits with BBC, New York Times, Reddit, and multiple Japanese newspapers that are, uh, all alleging copyright infringement. So something to be said about that.
[00:06:04] So what can this perplexity actually do? I'm going to go over the big rocks here. If you're really interested in this thing and you want the minutia, as per always go try it out. So in general, you ask for plexity, something that's just like you would any other LLM, any other chatbot. And by default, it will run a live search and then it gives you a synthesized answer with clickable sources.
[00:06:27] Maybe you're sitting here listening and you're like, so does chatgpt, so does Claude. And that is correct. And we're going to talk about what that means for, for business in a second.
[00:06:36] Perplexity also has a deep research mode that you can access via the prompt bar.
[00:06:41] Um, and that is for you use that if you want to explore a topic and then make like a report from it, um, from all that web material. It's built for synthesis, citations, and answering what's out there type questions.
[00:06:54] I somewhat strategically did this episode on the heels of the episode I did about deep research, because we see Claude has this and GPT has had this built in where you can go and ask it something and it will make you a report. That is what Perplexity does. That is at the heart of what it does.
[00:07:16] An important note, this deep research mode, uh, is only available to folks that are on a paid plan.
[00:07:24] Perplexity can also create things like mini apps, dashboards, files and other project like outputs rather than just a written research report in order to prevent its fightings findings. This is known as a labs task. Again yes, Claude can absolutely do this and that competition. The fact that these other LLMs can do the same thing is what I want to talk about next. So given that we had Perplexity, it started out, it was like hey, we provide citations for our information.
[00:07:55] Everybody else started doing that and um, seemingly Perplexity was like yo, we gotta do something else. So earlier this year, February 2026 to be specific, perplexity ventured into the agent space and launched what they have named Computer. These fucking people are the worst at naming. I've read this very early on in my like dive into AI and then I read AI that basically people that are in tech are terrible at naming. Why would you call something computer? It's so confusing. You're like wait, I'm using the computer. But it's called computer. So their agentic AI, they named it Computer. So remember agents are simply AI programs that execute multi step tasks.
[00:08:38] That means that they actually do things, they don't just answer questions and they do this autonomously. Right. Uh, it will go make files for you, it will move things, it can send things, it can delete things.
[00:08:48] Right without you, it can do it on its own and multiple steps. So if you want the deep dive on agents, I just did an episode on this, this was episode 52. I will link that. You can go and listen to that or read the um, read the, the blog post, the computer's companion about agents. So the best way that I can explain what Computer is, right, Computer is Perplexities. AI Agent Agent is that Computer is analogous to Claude Cowork.
[00:09:19] Uh, Computer is an age. I hate that name. Computer is an agent that function that functions as an autonomous digital worker and it's capable of using various frontier AI models to execute multi step workflows. So the difference between this and Claude Cowork is that Claude will use, it can use the various models within that anthropic owns, whereas Computer can use the, the different frontier models from all the different companies. Whether they're there, the, whether they're the Exact same or not I think is tbd. And my guess is that they're probably not the exact same. And also I don't really understand how these companies, how a company like Perplexity can actually what the margins look like because now that means they have to also pay to access that model. Right? Because it's not their own in house model. Um, but that is what computer can do, right? It's an autonomous digital worker. It can use uh, any of those various frontier AI models and it will execute those multi step workflows as of July 2026. So this month, um, it is expanding into or I guess by time this episode launches. Will it still be July? Let's see, this episode will drop on the last day of July. So yes, um, but as of July 2026 perplexity, their computer is expanding into Microsoft 365. So it will be in Word, Excel, PowerPoint, Outlooks and Teams. Um, worth noting, Access again of course is gated to a paid plan. Right. Um, so the same way that cloud cowork can be all up in drive and Google Docs and things like that, we have computer being able to do things inside of the Microsoft ecosystem.
[00:10:59] Two other names that you will come across if you venture into, or terms I guess I could call them if you venture into the Perplexity ecosystem are Brain and Comment. Again, again worse with naming Brain and Comment Brain that launched in last month, that's June 2026 and that's a memory system for computer. Remember, Computer is Perplexity's uh, AI agent.
[00:11:23] What makes it different?
[00:11:25] Uh, what makes this memory system different than others is that it remembers the agent's own work, AKA what succeeded, what failed. It doesn't remember or it's not specific to your personal preferences. So Perplexity's own numbers, right? They, they measure these things and like hey, look how good this thing is. It claims accuracy gains, but you know that's a first party figure, so take that with a grain of salt.
[00:11:49] The second name or term is Comet and Comet is Perplexity's own web browser with AI baked into it. It can summarize pages, it can answer questions about what you're looking at and it can carry out multi step tasks directly inside the browser.
[00:12:07] Fun little tidbit and a glimpse into the impending, what we're going to call the impending agent wars. Amazon is suing Perplexities.
[00:12:16] Uh, Comet Breath is suing over Perplex. Wow, I cannot speak. Amazon is suing over Perplexities. Comet browser acting as a shopping agent on Users behalf. Why is this a problem? Well, they're saying that Comet logs into a user's Amazon account without the user permission. Excuse me. Wow. Comet logs into a user's Amazon account with the user's permission. But Amazon never authorized that. So it's like yeah, we, the user said you could do this, but Amazon hasn't said it. And this very much sounds like Amazon is just cranky pants because they want, they have their own AI inside of Amazon. And that in my opinion is very likely why they are suing. Uh, because they want to have the ad revenue as well as, you know, access to the user and the user using their own um, AI.
[00:13:11] Right. Because if, if you have a, an AI agent that is doing the shopping for you, that means the human is not seeing any ads and that is bad for Amazon's bottom line. So whenever someone's suing it's always like, well how is this affecting you? Probably your, your pocketbook.
[00:13:27] So Comet, that is Perplexity's web browser akin to Atlas. Remember Atlas was, and I say was because it's getting shut down. That was ChatGPT's, um, that was ChatGPT's web browser that they, that they came out with.
[00:13:44] All right, so uh, that is, those are just two terms to, to know. So should you even bother using this thing, you know, as per always go try it out yourself and you make the final call for yourself.
[00:13:57] Um, the free tier is definitely a perfectly fine place to start. Um, um, and clearly the best use case for Perplexity is questions where you want a source attached. But again Cloud can do that. So I actually messaged my friend K and I was like, do you still use Perplexity? And he was like yeah, I use it in a, like instead of Google, um, for clean search. He likes the search results. And I was like okay, interesting.
[00:14:23] I um, personally have found myself uh, with certain things I use Cloud all the time. But um, with certain things I do really like when Google has the AI generated summary at the top.
[00:14:35] So you know, will I be using Perplexity? Probably not. Um, but by all means give it a shot and see. I wanted to introduce you to it. Um, I clearly, I don't pay for Perplexity. I haven't used it outside of the little bit of poking around I did in order to click this episode. Um, so to that end I cannot say if it's worth it or not to pay and you know, get those pay gated features which would be just access to the other models largely. Um, and then access to Its computer, its agent. Um, but if you're like, hey, this sounds like something that may be helpful for me, give it a try and see if it's worth it for you.
[00:15:11] All right, last part. Last How I used AI this week. If you're new here, welcome. Each episode I share a quick example of how I used AI that past week. So this time I used my French assistant, Claude to build a custom content creation tracker for the members of my newest program that I'm running called Crew 52. So I have spoken in the past about how I've used Claude and specifically Claude code, um, to create a custom dashboard.
[00:15:40] And I use that dashboard to manage my messaging clients and I use the exact same scaffolding for this. The difference this time is that it's going to be client facing and thus needs to be able to withstand the stress and the stress testing of other, you know, external users, not just me. So far, so good. And the thing that I want to highlight is how much faster it was this time. Despite, you know, there was definitely increased complexity because I needed external users. I needed different dashboards. They have to have one. I have to have like a master, uh, what do we want to call it? A, uh, an admin dashboard. And so I had to have it create those things.
[00:16:16] This, um, took me about two afternoons, really. It was like one big afternoon, then a little bit of the next day. Um, but that's a really fast build. It took me building my dashboard the first time. It took me a long time. Um, um, and it wasn't much faster this time because I'm familiar with literally every single, just about every single part of the, of the build. There were way less errors and way less issues to deal with. Not that I had a ton last time, but, um, I also like, knew what to expect and knew what to, I knew what to do. Connecting things, connecting the, um, what is it called, connecting the database, um, having it backed, uh, up to GitHub, we call it pushing it to GitHub, um, having it deployed on Vercel, all these things. I have done these things multiple times.
[00:16:59] Um, and I think one of the biggest things, one of the biggest reasons, aside from this, the familiarity, um, that I had less errors and less things to, to deal with, uh, was that I planned it out with Claude. So as I was building my Crew 52 program, I was doing it all in a, In a, In a. Basically a single message with a single chat with Claude. But then I had it, you know, it was remembering the things that I needed and then I Had it build out the instructions that I used with Claude code, and I had it frame it so that we built it in five phases instead of just trying to one shot it of like, I want to make this dashboard, and it's gonna have this and this, and you just, like, have it do this thing. And then, you know, when that happens, the output you get, you're like, oh, but I want to change this and this. And suddenly you have to make all these changes that then break something else. Whereas if you just do it one phase at a time and you let Claude determine what those phases should be. Right? So I had in the plan, I was like, break it up, make it into phases that make the most sense, and then it goes through one phase at a time. And then you look and you're like, okay, this was made. Okay, I want to change this. I want the colors different. Oh, it would be helpful to have this kind of connection for this.
[00:18:07] And then when that's all done, then you go to the next phase. It is way more helpful, way more efficient to do it that way. Um, worth highlighting here is the reason. The.
[00:18:19] Probably the main reason that it took two days instead of, like, one was that I hit the usage limits and I had to wait for it to reset. So I am still on the 20amonth tier. And I used Sonnet 5 to build this. I actually asked Claude and I was like, which model should I use? And it was like, this is perfect for Sonnet 5. You don't need to use Opus. I was kind of nervous, um, about whether or not it was going to say was, because OPUS is very hungry, very token hungry. Um, and it was like, nope, Sonnet 5 is totally fine for this. And I was like, oh, great. So it did hit the limits twice. And, um, I had to wait for it to reset, which is fine. Not a problem. 20 for $20 a month. I will wait.
[00:18:54] Um, but the big takeaway here is that repetition is your friend, right? It will make this faster and it will make the process way more familiar. Right? So the translation, the tldr, the takeaway gets a building, and it will get easier. All right, that, my friends, is all for today. Hopefully you found this episode helpful. And if you did, maybe consider leaving a little rating or review. I have forgot to go look and see if you have any new ones, but if we do, amazing. I'll check it out for next time, but I would love to hear from. From y'. All. It really does. I can't say that it helps people to find the podcast. Uh, maybe a lot of people say that. I don't know. Um, but I enjoy it. I enjoy reading them, so that's why I ask for them. And if you've already done that, thank you. All right. Don't forget, I have a companion newsletter and blog called the Curious Companion that drops every Thursday and is basically and by basically I mean exactly, the podcast episode in text format. So if you prefer to read or you just want a written record of the things, join the newsletter or blog fam. You can head to pompnicuriosity.com forward/newsletter or forward slash blog. Or you can just take the easy route and check out the link in the the show notes.
[00:20:03] As always, endlessly, endlessly, endlessly appreciative for every single one of you. Until we chat again next Thursday, stay curious.
