Maximize Efficiency with AI (Without Losing Your Voice) with Trey Griffin from Raptive

Disclaimer: This transcript was generated using AI.

Bjork Ostrom: What if your content could earn more and do more for your business audience and your future? You might know Raptive as the ad management platform behind thousands of the world’s top creators, including Pinch of Yum, but today, Raptive is so much more than ads. They’re a true business partner for creators, helping you grow your traffic, increase your revenue, and protect your content in an AI-dRaptiveen world. Unlike one size fits all platforms, Raptive customizes strategies for each creator, whether you’re growing a niche food blog or running a multi-site business, they offer expert support in SEO, email and monetization strategy and they’re leading the charge on AI advocacy to protect the future of creator owned content. And the best part, Raptive supports creators at every stage from Rise, their entry level program for growing sites to their top tier Luminary level, their offering scale with you so you can get the right support when you need it the most. Apply now at Raptive.com to get a personalized growth strategy and join a creator community that’s shaping the future of the open web. Thanks again to Raptive for sponsoring this episode.

Ann Morrissey: Hey everyone. Welcome back to the Food Blogger Pro podcast. This is the second episode of our Raptive series. Can’t Be Automated where we talk about scaling, standing out, and the importance of authenticity as a content creator. This week Bjork is sitting down with Trey Griffin, who is the Senior Manager of AI industry Analysis at Raptive to talk about how AI is reshaping the content creation sphere. The reality is AI is here and it’s getting better every day. It’s bringing a huge wave of both opportunities and threats to the creator space. And if you’re not thinking about it, you’re already behind. This episode is your playbook for how to approach AI, not as a replacement for your work, but as a powerful assistant. You’ll hear Bjork and Trey talk through how you can use tools like Custom GPTs to boost your efficiency, letting you take on bigger creative projects and focus on your core strategy. The big secret they unpack is that as digital noise gets louder, your personal brand and unique human touch are actually becoming your most valuable assets. You’ll definitely want to stick around for this episode. If you’re looking to learn more about how to leverage AI for growth while making sure your content remains uniquely yours, be sure to tune in next week for the final episode of this series with Raptive. And now without further ado, I’ll pass it over to Bjork.

Bjork Ostrom: Trey, welcome to the podcast.

Trey Griffin: Hey, thanks for having me.

Bjork Ostrom: I’m going to start with something that I saw recently on LinkedIn. So OpenAI, they have this big presentation, it’s a Developer Day. They release all this information and I think about stuff like this when Apple does a big release, so occasionally they’ll show here’s the update on iTunes and they’ll show artists and it will be a certain artist that they show for this open AI one. They were like, here’s the power users of ChatGPT, I think it’s ChatGPT.

Trey Griffin: Is the API

Bjork Ostrom: Is the API, the API. Your name was on the website, presented as somebody who had used a hundred billion tokens of usage.

Trey Griffin: I was actually just the person that was there for Raptive. I didn’t use all 100 billion of us to, I think I could maybe only account for 90% of ’em,

Bjork Ostrom: Which is amazing and I feel like it is a great little piece of evidence as to the amount of time and energy and focus that you are putting into the world of AI specifically around creators. And if I remember right, I saw the post that it was actually Paul Bannister who’s been on the podcast before, mentioned it, but it sounds like that was related to a project that you had worked on with Raptive to help ingest a bunch of data and sort through things.

Trey Griffin: Just a way to more quickly expand our leads really who are good creators that could be part of the Raptive network. We took the amount of time that it takes to do that and decreased that from about nine months to a week, and then we took the cost and reduced that by 97%.

Bjork Ostrom: It’s because my guess is now instead of an individual doing it, instead of people hours, it’s a computer that can look through stuff a thousand times faster, 10,000 times faster, a large set of data, great use of AI within the context of Raptive. One of the things that we’re going to talk about today as we’re doing this series with Raptive is this idea of, okay, that is obvious, that has made the business better. There are a lot of examples for creators on how we can make our businesses better by using AI. Also, there’s a lot of things with AI that are really scary for a creator who has a business. An example is, and maybe we can start by talking about this, I spent maybe 15 minutes on Sora last night. I finally got an invite and got in. It is bonkers.

Trey Griffin: It’s quite good.

Bjork Ostrom: It’s really good. And the thing that was weird to me was that that was 15 minutes, 30 minutes. Now the app is on my phone. There is at this point an equal likelihood that I would open Sora and scroll through. It is all AI. It’s not even randomly come across an Instagram reel and be like, oh, that’s AI.

Trey Griffin: You can’t even upload a video that’s not AI-generated.

Bjork Ostrom: It is strictly AI created content and my time otherwise would’ve been used in that instance on Instagram or TikTok.

Trey Griffin: Yeah, it’s fun, isn’t it though?

Bjork Ostrom: It’s fun but it’s also scary. I think what’s not like fear-inducing scary, actually feel it physically scary. It’s scary from a business model perspective because for somebody it’s an opportunity. So that’s attention. People are going to learn how to create in a way that gets attention on a platform like Sora and they’ll be able to create a business out of it. But for somebody who’s not used to that model and would need to reinvent ourselves in order to build a business on there, that’s the part that I think is not even so scary, but it just feels exhausting potentially. Tell me, have you used Sora a little bit and what has your experience been with it?

Trey Griffin: I think Sora is pretty fun. I think that would be of my, one thing that I kind of find with it is that after about 30 minutes, I’m kind of over it. I don’t fall into those. If you’ve ever had a rough day and you’re sitting in your bed at the end of the day and you just scroll Instagram reels or TikTok for two hours, soa definitely the application anyways is not engagement maxing the same way that I feel like an Instagram or a TikTok does. So I’m not sure actually to change in the future, but I’m not sure actually if there will be. It will be a big deal, but I don’t know if there will be many businesses that are built on top of Sora where the business, the whole thing is built off of it. Instagram influencers and stuff like that. I doubt you’re going to have many accounts that are able to turn it into, I make money with this now simply put, it’s really expensive computationally for OpenAI to make those videos. So it’s a great attention seeking moment for them, but the reason they didn’t roll it out to everybody is it would melt their servers and they wouldn’t be able to put anything else out there at the time

Bjork Ostrom: Specifically, you put a prompt in and the prompt creates a video that is six seconds long or whatever you’re saying that is a super expensive process, super expensive.

Trey Griffin: So in their AI, the cost to do 12 seconds of video with the highest end version of Sora, which is not what you get on the app costs $6. So for 12 seconds of video, it’s insanely expensive for them if they’re charging us that much because typically the AI game is get the price pretty much as low as you can get it because you can’t really guarantee your product will be the best for very long. And we actually saw Google respond to Sora today with version 3.1, which is their video model.

Bjork Ostrom: Interesting, but point being things are changing very quickly within the world of video. One of the threats as a creator is as a creator that exists in are the current forms that we are creating today is there inevitably will be a shift to where not only will content creation become easier because you can do it from a prompt, but also people will be consuming potentially more AI content. Again, there’s opportunities there, but there’s also a threat. The other piece is this idea that these LLMs are being trained on creator content and so within the context of like, Hey, is AI a friend? Is it a foe, is it a frenemy? What is it? We have all of these different instances, let’s look at the category of foe for a little bit. So not only is it potentially something that is going to be taking attention away, but also it’s trained on creator data. Can you talk a little bit about that and how we should be thinking about that within the context of creators who are publishing stuff to the open web.

Trey Griffin: So I think there’s the business thing and then there’s almost a sort of spiritual part of it, which is a weird thing to come out of my mouth, but I guess that’s what we’re going to roll with today. So the business aspect of it is you’ve got sort of two angles to it. The first is the SGE, AIO perplexity chat, GPT search, AI-powered search tools, so like SGE and AI or Google,

Bjork Ostrom: SGE, meaning a search generative experience,

Trey Griffin: A search generative experience, and now it’s AI overviews. So

Bjork Ostrom: Same thing. So you do a search on Google and it pops up, gives top with

Trey Griffin: An answer. Here you go. And we thought that was going to have a huge, huge impact. The impact of that’s been kind of muted compared to our expectations, but then you look over at some other stuff like Pinterest where you have a pretty significant flood of content. So this is the other business side of it, which is that it has already become really easy to make content. Pinterest kind of tends to reward stuff that’s newer. So if you can just jam a million articles out there, Pinterest likes that moves fully AI generated stuff towards the top of a lot of Pinterest keywords that are really important for our food creators.

Bjork Ostrom: People are just crushing AI images, AI text, the whole thing is the whole pipeline. They just

Trey Griffin: Just

Bjork Ostrom: Click button stuff isn’t super expensive.

Trey Griffin: No,

Bjork Ostrom: And so then people are able to do it in mass.

Trey Griffin: Yes, exactly. I mean if you don’t really care about the quality of the content you’re putting out into the world, I mean you can generate hundreds or tens of thousands of articles and hours relatively inexpensively with the imagery and everything associated with it. It’s not very hard. So those are the big foe points that I would say. And then when we talk about training it who is in the early days of OpenAI was one of the key research members And he’s a big believer in that you can do what these LLMS do under the hood as a next token predictor. If you could predict the next token, you could do almost anything is basically his idea is if you know the ending of a sentence, you necessarily understand all the component parts. We know that LLMs are just predicting the next token just like your keyboard in your phone. But it turns out that when you scale that thing up to a relatively large size, it does begin to understand what it’s actually saying. We know that’s true. It’s not really that up for it depends on how you can get a really philosophical argument, but if it can predict the next token with a hundred percent accuracy, and we’re probably nowhere close to that with modern alms, but you may not need to get to a hundred percent before it starts to become a problem. Then if you start writing an article in your style, in your voice, it’s trained in all of your content, it should be able to finish that article in your exact voice if it’s able to predict the next token, it’s almost like a sort of technological impersonation That these LLMs are going to be able to push out. In fact, one of the things that’s really different about GPT–5 is they’ve spent OpenAI in particular spent a lot of time basically hammering its personality to death with reinforcement learning specifically so that it wouldn’t do things just like you or right, just like this person. And there’s a variety of hypothesis as to why OpenAI would do that, but there’s sort of the business things where it’s like, okay, it’s going to take eyeballs away from us, attention away from us. I’m going to have these more competitive moments and then there’s the sort of fear that I have that you could have a hundred little robots running around impersonating your exec speaking style all over the internet and nobody would be able to tell the difference. And I think that we’re actually maybe closer to that second thing than a lot of people are ready to realize.

Bjork Ostrom: That was the thing for me with Sora, which I know is different with this but maybe not exactly was they had the Jake Paul did the partnership with Sora so you could use his likeness and it’s like, I don’t know Jake Paul that well, but when I watched that’s pretty good videos of Jake Paul and that’s Jake Paul, and it was the first time that I thought, oh, we are now at a point or the Queen of England. We are now at a point where it’s going to be hard for people to distinguish the difference.

Trey Griffin: I think the voice cloning technology, which OpenAI doesn’t let their models clone people’s voices. You can’t go but ElevenLabs, which another will let you do it. They don’t care. The voice cloning models are good enough to where with 10 or 12 seconds of audio from this podcast, somebody could go out there and clone both of our voices and have us say pretty much whatever they want and it would be if you’ve used one of the higher quality providers, it’s very hard, very hard to tell that it’s AI generated right off the bat. So

Bjork Ostrom: Especially if it’s not somebody super well,

Trey Griffin: Yeah, so quite soon you will have AI models that aren’t from people’s. We like to give OpenAI a lot of guff and I think that they deserve it, but there are a lot of people out there that are way less concerned about this kind of stuff than OpenAI is and

Bjork Ostrom: Probably people who are incentivized for these to perform. Well, if you are working on soa, you’re probably not as concerned somebody who is a content creator who makes their money from videos.

Trey Griffin: When you’ve got your OpenAI stock and it’s like a 500 billion valuation or wherever they’re right now, I think then you don’t really care whether or not one particular product at your company, you just got to stay there long enough for the stock investment. But I’m thinking more like foreign actors. If you look at the Chinese AI market, there’s basically no censorship. There’s no limitations of what rely with their models. They’re normally six months to a year behind the US but they always catch up. So there will be a Chinese version of Sora that will run for a fraction of the cost that will let you completely clone Bjork and Trey having a podcast from a 15-second clip and then put it up. It’s actually you and there’s

Bjork Ostrom: In there is not going to be copyright because it’s, yeah,

Trey Griffin: It’s impossible, but really our brains don’t have a defense for it. If you see an AI-generated clip of the two of us talking and it looks just like this,

Bjork Ostrom: Yeah, you don’t

Trey Griffin: Know. Your brain goes, that looks real.

Bjork Ostrom: So as somebody who is a content creator who has built a business that is based on content, how do you suggest we approach this? What is the mindset? What have you seen people talk about? Because it is changing in a really significant way and there are always opportunities which we’re going to talk about some of the things that are really great about AI and how they can be used in super helpful ways within our business

Trey Griffin: Definitely.

Bjork Ostrom: But just looking at the category that maybe feels like there’s a threat there, how would you recommend we categorize that as we process it in our brains? What does that look like?

Trey Griffin: I think that it’s very case by case, which is not super helpful. Some creators will find immediate wonderful uses for AI that they feel very good about. So if you’re trying to do some basic fact verification where you kind of know what it’s supposed to be right and what’s wrong and you could send the AI out as a second pair of eyes, that’s fine. That could be something that’s really beneficial to you, especially as newer models are much less prone to make things up. It’ll still happen, but it’s nowhere near the rate that it used to be at even six months ago. So that’s great. But then there’s other sorts of content creation in particular food where it’s like, are you just going to generate a recipe on a whole cloth without even bothering to cook it? No, probably not, right? I mean it’s probably not a good idea to do that anyways because AI doesn’t have a tongue and doesn’t know what tastes good and there’s all sorts of issues associated with that. I think that there, it’s kind of a tricky little bit of a minefield where you have to think about your own risk tolerance around it. My sort of viewpoint when it comes to using AI in your work is if I took something that I used AI to help me create and I put that in front of somebody, could I stand behind it and be like, I’m proud of this work, I’m proud of what I’ve put out into the world. And if you’re standing behind a piece of AI generated content or content where you had the assistance of an AI and you’re like, I’m going to put it on my website and if we feel like you would be embarrassed if somebody called you out on it, maybe don’t do that.

Bjork Ostrom: Sure, yeah. It’s like a filter that you have as a creator around what is helpful and what isn’t. So that would be in the category of let’s say, okay, you’re using it as a brainstorming tool or maybe it’s like, hey, this is completely different industry, completely different company. It was a real estate company working on core values and the person who did them ran ’em by me. I took what he had drafted and then said to ChatGPT, hey, here’s the intent behind this core value, here’s the current statement around this core value. And I think what I said was just like, can you clean this up and just to see what it came back with and it essentially cleaned it up, it just made it easier to read, made it a little bit more simple. Then it was like, Hey, we actually want to include something related to the importance of trustworthiness within it. What would that look like? So it was kind of shaping and that was one an example where I sent that back to them and I said, Hey, just so you know, my refinement process with this was intentionally putting it through chat bt to see what it gave me back. But that’s kind of in the category of like, okay, how do we as creators use it positively? How about just, I guess my question is do you have any predictions on the future of content consumption and that shifting, I

Trey Griffin: Mean in the predictions game, but I’m happy to make one. I just want to have the caveat that I have no idea if it’ll be accurate. I’ll say this, I’m pretty good on predictions about where the technology will, I said we’d have video models that were good enough to do at least some cinematography work and this year I feel pretty good about that one. I didn’t feel that great about it two months ago, but I feel pretty good about it now. The first useful AI agents, I was like, I think we’re going to get those year, if you know anything about the coding agents that have come around, those have become incredibly valuable and useful as to where content consumption is going. I don’t think that they have a plan these AI companies at all for how content consumption is going to work in a world where you can just flood the internet with not just content but content that has been highly optimized for aesthetic beauty interestingness capturing your attention. I mean, you think that Instagram reels and tiktoks are bad for you. You ain’t seen nothing yet. You have sort of this approach from Facebook which is quite literally, we are going to train the AI to make Facebook more addictive. So that seems bad. It seems like there’s not a whole lot we can do about it though, so it’s not that productive to talk about. And then you have on the other end of the spectrum open AI who’s like, we are going to release this thing and we’re going to try to make it not that engaging. We’re worried about what might happen if it’s too engaging. And then you have folks philanthropic over here on the side, which are the claw people and they’re just not even interested in this kind of social media building AI around content creation at all. They’re just like, I need more tokens to make more code because that’s my primary client. So they don’t have a plan and until they sort of lay out what they think the future will be like, then I think you’re talking about taking these AI generated systems like these AI generated content, putting ’em into the existing engagement algorithms is what we have in the near term. So if you’ve scrolled Instagram reels, I’m sure you’ve seen a bunch of videos at this point with the store watermark bouncing around. The great thing about that is there is a so of watermark bouncing around so you can kind of tell The question is are people going to start having a preference for that content and what does that mean for content creators? If a market comes out here and says, we want AI generated content, that is what makes us happy, that’s capitalism, baby. We’re going to have to adapt. We’re going to have to change change. Right now all the evidence points to there’s a slight favoritism towards human content, but it’s actually quite narrow. It’s quite narrow, and a lot of the times you have to tell them which is which first.

Bjork Ostrom: Sure.

Trey Griffin: Soon they will not know in

Bjork Ostrom: Order and soon you won’t know. And I think that’s the piece that is really fascinating to me, and to your point, if what we start to see is that AI-created content is more engaging on these platforms, then these platforms are probably going to favor AI-generated content. And it’s a little bit morbid, but I think of this the first, the guy who learned how to jailbreak an iPhone, I don’t remember what his name is

Trey Griffin: George Hotz.

Bjork Ostrom: Okay. Yeah. I was listening to an interview that he was doing, and this was four or five years ago, so was before I knew about OpenAI or any of these platforms, he had this prediction that he said, I think there’s going to be a time when these algorithms are going to get so good and these algorithms will be able to create their own content. The thing that changes from the algorithms today, we talk about algorithms as being like, gosh, algorithms are so good and they keep us addicted, but the one thing they don’t have today is the ability to create content. They have to use other people’s content in order to create a sticky algorithm. But just imagine if you had an algorithm that not only could optimize for what content to show you, but then also

Trey Griffin: Content made for you

Bjork Ostrom: Also create the content

Trey Griffin: For you. That’s like, that’s my nightmare scenario. People worried about the meteor’s going to hit the earth. I am worried about everybody’s brain being vacuumed up by opening on TikTok

Bjork Ostrom: And it was kind of one of these wally moments where it’s like everybody has those screens and they’re just floating around. His prediction was the algorithms will get so good that you’ll start to look at a phone and you won’t be able to look away

Trey Griffin: And you die of starvation. Yeah, it’s one of the nightmare scenarios. Totally. You optimize the AI’s ability to entertain you so much that it overrides the brain, it produces so much dopamine and it overrides all of your brain’s other survival functions basically accidentally hack yourself. You ever seen the Dark Mirror episode the most recent season where it’s like he sticks the thing up against the wall, we’re treading into science fiction territory too much, but

Bjork Ostrom: But I think the point that it comes back to is that the world of content creation as a business is inevitably changing. And so if your business is content creation, that means you have a lot of questions that are probably floating around right now. What I would say as somebody who creates content within the content of Food Blogger Pro, the podcast courses forum conversations within the context of Pinch of Yum. It’s food and recipe content. I would say the question for us then becomes is what does that mean for our business? And I think I’m going to put something on the table and would love to hear what you have to say about it. The options really are do nothing and continue to see how that plays out. There will always be some place for people who are creating content, food related content to show up to serve people, and maybe the business actually looks really similar today in five years than it does today. I don’t know. I kind of doubt it. So that’s one scenario. Another scenario is you go super deep on ai. You say, I’m going to become an AI forward creator, and that’s what your focus is like. You are becoming an expert on AI tools. You’re creating AI content. That’s your kind of specialty and expertise I would say for people who listen to this podcast probably doesn’t make sense, but I think there will be those people who create really entertaining content that it is somehow a business for them. I don’t know what that looks like yet, but I think that will exist. The third piece is this kind of hybrid piece. How do you use and leverage these tools in a way that allows you to do what you’re trying to do better? And I think you have to be a little bit indifferent as to how you get there. The purpose just has to be what am I trying to do and how do I do that as best as possible for an audience of people? And I think that’s the one that’s most fascinating to me is just what’s our mission? What’s our goal? What are we trying to do? What are the tools that we have at our disposal to do that and how do we do that as best as possible? That’s the one that I think is most interesting to me. I’d be curious your response to that and then we can dive deeper into it.

Trey Griffin: Yeah, so I’ll give you an example and we’ll pause it like a hypothetical workflow. This would require a little bit of work to pull together, but this is within the realm of possibilities. The technology exists today. You just might have to actually spend a little bit of time write some software, but let’s imagine for a second you’re creating a new recipe with Pinch of Yum. Instead of going through the process of taking a bunch of notes in your recipe, you just kind of have a scratch pad that you’ve written down the ingredients on and you’re just going to narrate to a phone why you do the recipe and then it takes that narration and it transforms it into an article written in your style and you take a couple of photos and then you touch those photos up in Photoshop or whatever your process is, and then you use an AI filter to go through them, makes them more appealing looking. And then let’s say maybe you don’t do video, but you’d like to do some animated gifs of the food being picked up and eaten or whatever the case may be. But the process for taking those pictures, the food’s not very appetizing by the end of it, so you may not have a chance to actually take any images or any video of people eating it, enjoying it. So what if you took a picture of that food, you’d plop it into something like Sora or VO3 and you just AI generate somebody eating and enjoying the food. There are different levels of tolerance that people will have to allowing AI to involve itself in each step. So the most aggressive is you’re taking the dictated notes and then it is like you have some bespoke fine tuned model that just writes the whole article and then you’re just basically placing it somewhere. I think that’s maybe a little too extreme for most people right now. Even if you felt like this AI actually does write like me, it does represent me in some way. Even if you feel that way, I think that that’s a bridge too far for most people and I understand that. But some of this other stuff, I could take a picture and I could clean it up with AI as opposed to having to learn Photoshop. That’s really doable. Now if you haven’t played with Gemini 2.5 flash image, also known as Nano Banana, play with it. It’s super fun and you can just edit up a photo in the image tool in two seconds. You can just completely swap the material of the bowl, but while preserving everything else in the image, it’s great. It’s super fun to play with. And then if you turn that into a moving video that played at the top of your channel, I think that’s probably okay too. I think most people would be like, oh, that’s a nice, another way to grab attention and eyeballs. Like look, there’s some video content that’s associated with this. Even if I’m not a big video producer can help you branch out with maybe some areas that you haven’t done before, could help with velocity a little bit, but I don’t think that we’re yet at the point to where it’s going to be this thing where everything is handed over to the ai. You’re going to have to sort of pick your pieces of that beginning to end flow that you feel comfortable with and focus there and then move your way sort of up that ladder of comfort.

Bjork Ostrom: That’s great. I love the idea of having specific examples that you’re sharing, and I think the core piece that’s important within what you just talked about was the recipe is still something we’re talking about the world of food. Obviously this is different as you get into other content categories, But the recipe is still something that you are in the kitchen and you are developing. That is the core component of what is the important piece of a recipe post is the recipe itself. Everything around that is how do you help people have the most success in creating that recipe? And if you can have an image that helps because suddenly there’s ingredients information on it or it’s explaining a step how to do a step that’s kind of hard. All of that can be additive. And if an AI tool allows you to get there quicker to create something that is additive, that will increase the success, great, all the better. Another thing that I think about is what are the things that you are currently outsourcing to another human to do that you aren’t doing? And would there be an opportunity instead to outsource that same thing if you’re trying to get from point A to point B, would there be an opportunity for some tool to do it instead of a human to do it? One of the ways that you’re starting to see that manifest itself is within the writing. I’m curious to know, and this may not be your department, but for Raptive, I know Raptive is outspoken around. Hey, we don’t want a bunch of AI slop sites that are really intentional to make sure it’s human-forward content. Where’s that line? And I know there isn’t one

Trey Griffin: That is not my department. You’re going to want to talk to Jasmine. She’s the person in charge of policy. I would say this, I mean this is not Raptive’s opinion. My opinion is that if you can stand behind it, a piece of content that you could create and you feel comfortable and confident about it, then that’s fantastic. However you got there is cool with us. What wrap lift doesn’t want to see is like, Hey, I generated 10,000 new articles today and there’s no way that a human being read all of this and I posted it all in one day and I’m expanding the Google results. Absolutely not. We don’t want to see that.

Bjork Ostrom: That makes sense. And so for you as an artistic, I had this same conversation with a friend who’s a musician, what do you think of this idea that I could go to one of these sites and say, write a folk song about traveling to Lake Superior in Minnesota in the fall and seeing the Leafs and it’s like, oh, it can do that really quick. And he was like, honestly, I don’t love it. I was like, okay, that makes sense. He said, but AI has been really cool for sampling certain sounds that I want to put into a song. And I feel like that’s a little bit of what we’re getting at right now, which is there are ways that you can build component parts, but it’s different building the entire thing. And I’ve seen a handful of these sites, I can think of three or four on one day on Pinterest and I looked at the site map and it’s like these were all published on June 11th. All 700 of these posts were published on June 11th, 2025.

Trey Griffin: They’re out of the box tools that you can get on GitHub that will just do this

Bjork Ostrom: And it just populates the site.

Trey Griffin: Yeah, just populates the site and host it and everything. You just got to go slap your credit card in or whatever. There’s people out there that will do it for you. So I mean I think a portion of this is very the sort of AI spammers, these technical people who are like, I saw on x.com that I can make a quick buck doing this. I’m going to take a swing at it. I don’t think they stick around super long. I think the people that are just persistent are people who are maybe not that technical but understand the game of running a website and they are using third party services. I think, I actually won’t say the name of it, but there’s one that I know that’s out there that allows you to do stuff like this,

Bjork Ostrom: Spin up a site,

Trey Griffin: Spin up a site, and then it gets blacklisted from Pinterest or whatever and who cares? You could spin up three more and if they all make $10 and you’re only spending two bucks to make ’em a 5X profit margin, I think almost anybody would take that.

Bjork Ostrom: And I think the thing that’s interesting for me is, okay, that’s inevitable. You can try and figure out how to combat it, but it’s like a genie out of the bottle, bend doors, box. There’s no way to combat it. So then what do you do as a business owner? And I think the answer is you differentiate. How do you show up in a way that is different, that is unique, that’s novel. It’s a little bit of what the past 15 years have been for us. You have to continually find a way to show up that is novel and unique considering the current reality of the landscape. I’m curious to know ways that you have seen that happen potentially. How can we show up and be novel? And part of it might just be leaning into our own story, our own humanity.

Trey Griffin: So I think that the way that many of our top creators differentiate themselves is they have a personal brand. It’s not just like this is website.com and we do recipes. It’s like we are a Pinch of Yum and we are these two people and we do recipes. There’s more to it than that. They put a little bit more of themselves in it. So from that perspective, the perspective of the human perspective, what people actually want to read is they want to read stuff that’s written by person or at least that’s had a lot of care put into it by human being at a minimum. But I think that from a business perspective, the specific tactics that I can tell you to do, I went up here in Pinterest more. I think we’re all kind of figuring that out together right now.

Trey Griffin: Yeah, totally. We’ve got stuff I can’t get into a ton of details about, but we’re definitely looking at how do we get Raptive creators back at the top of interest like we used to be or we’re interested in how we, if AI, the search results have grown from ChatGPT in the last year by 800%, how do we make sure that you guys show up in the search results? And a lot of it comes down to kind of simple stuff, at least initially make sure your site it’s really well structured and just stuff that you’ve been told to do forever, right? Make sure this, it’s like best practice, it’s website hygiene. A lot of it’s really basic stuff, but I think we’re going to probably over the next six to 12 months to start to learn some more interesting stuff. I’ll say I don’t think that there’s anybody on our side anyways that is actually trying to do any real original research around this. And a lot of the original research that’s been done around the AI’s impact on the internet has been not very good. You’ll hear of generative engine optimization, which you have people hold up as if it’s a lot of SEO people will say this is the future meaning, and then when you actually go meaning that how do you show up in generative AI

Bjork Ostrom: As a link when somebody’s searching for best snickerdoodle cookie recipe?

Trey Griffin: Yeah, exactly right. So it’s super funny that has come. I see that everywhere and every time I see it, my eyes roll almost completely out of my head. That paper was written in late 2023, early 2024. It tests GPT–3 0.5. Any advice it’s giving you is correct only by accident. It might be correct, but it’s just by chance that it still applies. Yes. Because you’re talking about the Gulf between a GPT–3.0.5 and GPT –5 as far as its ability to accurately find information on the internet is an ocean.

Bjork Ostrom: Yeah, well even within, so for those who aren’t familiar and there’s the releases of different ChatGPT versions 3.5 was

Trey Griffin: It was the very first one ever released,

Bjork Ostrom: So that was the one people were used to. I think even at that point it was like it had data up to September of 2021 or something.

Trey Griffin: They can, 2022, but it wasn’t much more than that.

Bjork Ostrom: You couldn’t query something that was like,

Trey Griffin: Yeah, no internet access

Bjork Ostrom: Time and the current version that most people would be using when they’re just using ChatGPT would be 5.0. So it’s these different releases, similar idea to you get an iPhone and you upgrade it to the next os. It’s like a new version that does things differently. And so your point is the paper that people often reference around GEO, generative experience optimization versus SEO, which would be the 10 blue links, is on an outdated piece of information. And so your point is people don’t really know how to show up within these yet.

Trey Griffin: We have some ideas, but as far as here is our 100% bulletproof, we could publish it in an Academic Journal Evidence with capital E. I think that a lot of that basic research has been hard to do because this where the people wrote the geo people had all the best intentions, the people who actually publish it, all the best intentions in the world, but there’s never been geo two, they don’t have geo two where it’s like now let’s look at the models as they are today, which would give totally different advice. And because there’s not that much out there about what can we do to show up in AI search results, a lot of the conversation has been based around this paper and a lot of who do for better lack of a word, just wish thinking.

Bjork Ostrom: Yeah, that makes sense. And point being, it’s hard to know if you are trying to optimize to show up in these chat interfaces, you don’t know if it’s possible or not. So one of the things to go back to this idea of so what do we do as creators? It’s leaning into the things that differentiate ourselves. Personal brand. One of the things that I think about is speaking very broadly, the more transactional your content is, the easier it is to replace. And so how long do you boil an egg? How long does raw chicken last for? Yeah.

Trey Griffin: If you’ve been the top expert on how long I can keep chicken in my freezer for the last 10 years? It may be time to think about updating that

Bjork Ostrom: Article totally. But the more that you become somebody that people want to follow along with, and I think that would be even an interesting prompt for ourselves to think about who am I interested in checking in on? Who am I interested in following? Am I interested in hearing from? Chances are part of that is they’re creating a thing. It might be transactional, but part of it is probably because you want to follow that person. You have a curiosity around, in our world right now we’re doing a kitchen edition, a kitchen remodel. That’s not something we’re going to rank for on Google, but it is something that I think people are interested in following along with. And that’s story, there’s a human component to that where we’re naturally curious about that type of stuff. And so what feels like it’s kind of difficult right now as a creator is that we’re walking this line between what has worked, which is like

Trey Griffin: It continues to work in a lot of cases and continues to work.

Bjork Ostrom: Totally. It’s like you create a piece of content, it ranks, people go to it because they use a keyword or they find it on Pinterest, and that is generally transactional. You created a thing, you rank for it, people discover it, but it feels like less and less that’s going to happen and more and more people will go to these LLMs, ChatGPT, Claude, even a generative experience or a AI overview within Google. The thing is we just don’t know how quickly that will happen and to what degree that will happen. And so we’re kind of in this middle ground where we’re seeing it kind of happen a little bit, but not to the point where it’s like, okay, we need to completely rebuild this

Trey Griffin: There’s a couple of things that have been sort of holding back LLMs from being the dominant search paradigm. The first thing that has holding them back is this idea of context rot. And basically this was originally called the lost in the middle problem, that it’s like they came up with a better term that gives the idea a little bit more easily, which is context rot, which is basically when you’re typing to a model, you put the words in and the model isn’t reading the words. The model is reading a string of numbers that represent tokens in latent space, which you don’t need to know what any of that means, but what you need to know is that it’s

Bjork Ostrom: Predictive

Trey Griffin: Around exactly. So it translates those into words. It splits them up in a specific way so it’s more efficient for it to read. So you only have a certain number of those splits that you’re allowed to have and those are called tokens. As you add more tokens to your prompt, the less the model is able to accurately remember what the prior token was because it’s not able to accurately recall it from its attention mechanism. So as a mechanism for what parts of the sentence are actually important. So the dog ran to the beach, dog ran beach, that’s all the AI probably actually cares about. All those words in between that link things together, it can kind of guess that those are probably there. So with search, you go and you put in your query, well, I can’t just take your query. That’s not the only tokens that go in. It also has to go and grab, go and look at every website that might be relevant to your query. So you have to have some system to prioritize that, and then it has to go and pull all that information in. It has to read all that information, and then it has to take the actual pieces of information from those sites that would answer your question and construct an answer that is then given to you as the end user. The problem was if it makes any mistake during that whole process where it’s doing a lot of different things, there’s a lot of points of failure. You can get back the famous, here’s how much glue you should put on your cheese pizza

Trey Griffin: Answer from AI overviews, right? And basically the context rot problem is pretty close to solved. So that was a big thing that’s holding it back. Other thing that’s holding it back from being more relevant is the speed and cost. Right now it’s just more expensive than giving you 10 blue links. It’s not that much more expensive anymore, but especially a year ago when we first all started thinking about that, it was really expensive,

Bjork Ostrom: And this is expensive for the context within the company itself. Open AI versus Google, it’s expensive

Trey Griffin: For Google, so they got to make money somehow. They can’t just burn money all the time.

Bjork Ostrom: And so you’re saying there would be a disincentive for them to lean too heavily into this because of the cost?

Trey Griffin: I wouldn’t say disincentive like Google if anybody can burn money, but if they don’t eventually get to a point to where they’ve reduced the cost significantly, they will stop doing it because it won’t make them any money. It doesn’t matter how brilliantly you present the information to somebody. If it costs you more money to do that brilliant presentation, then if you just left things as they were and people Google searched, then you’re cooked. I think that the first thing, the context fraud is pretty much solved. You have to use, there’s one model where it’s solved, but it’s with no open eyes Model GPT–5, if you’re using the thinking mode, so there’s three settings for it. You have GPT–5 Auto, which just picks a model and then gives you an answer. And then you have GPT–5 thinking, which always uses the highest tier GPT–5 model except for not the pro tier, which is,

Bjork Ostrom: Yep, takes longer, but it does a better job thinking through it.

Trey Griffin: So it basically over its context window is very close to a hundred percent accurate. It will still make mistakes, but the frequency of those mistakes is very rapidly approaching 1% of the time. And I imagine a year from now it will be well under 1% of the time that it makes mistakes. It’s slow. So that’s a problem against it. So it also needs to provide the answer quickly, but we know that Google has pretty quick models already. If OpenAI has figured out the context route problem, Google will figure it out. And the pricing of these models collapses every year. It’s genuinely hilarious when you see how good the existing models are versus how much they cost. And then you go look at these outrageously expensive models from three, four or five months ago and you’re like, this is so much better for half the price. So I think that you can expect the AI powered search stuff, long story short is going to impact us and soon because it’s going to be very good relatively soon.

Bjork Ostrom: That makes sense. So those two specific problems will be solved. Number one, the internal costs or number, I’ll do ’em in order. Number one, the ability to accurately interpret the prompt. Is it within the context of the prompt that you were giving

Trey Griffin: It? Yeah, you could sort of think of it this way. I want you to imagine for a second, I gave you 10 website articles and you have to remember every single detail from all 10 of those articles correctly. In other words, you get an electric shock.

Bjork Ostrom: Sure.

Trey Griffin: Would you be able to do that?

Trey Griffin: No. And neither can the AI, but they’re getting very close to where it will be able to do that relatively soon. I would say. I would argue GPT five can already do this. You can go load in 10 articles into its context window, even on your ChatGPT Plus plan or whatever, you can just load ’em all in, ask a question, it’s probably going to get it right. More often than not,

Bjork Ostrom: And I’ve done that even with PDFs, like, okay, Please thoroughly review you give it prompts to cause it to tip into thinking mode. If you don’t manually select it thoroughly review these documents and then you can ask prompt, prompt it based on those documents. So that’s getting better. That makes sense. Cost is going down. Your point within that is we’ve kind of been in this in-between of when does this really impact us in what significant way? And what you’re saying is it requires some of these things to potentially get to that point where it does tip over people start to use it a little bit more.

Trey Griffin: Yeah, you have to have, the customer has to have trust that whatever the AI is telling it is probably correct. The first time you get a wrong answer from a robot, you’re going to be, it’s so funny because I can give you a wrong answer five times during this podcast and nobody,

Bjork Ostrom: You’re a human, not a big deal.

Trey Griffin: AI Gets it wrong. Once we’re all up in arms, it has to be perfect. So they have to get pretty close, I think. And I think Google, if they’re not there, it’ll be soon. I would imagine we will know within the next eight days.

Bjork Ostrom: Yeah, that makes sense. And so even more so as your information that you’re creating, if it is transactional information that is a vulnerability within the context of LLMs. And so as we talk about this idea of being human or leaning into story, one of the reasons why is if your business is content content creation and it was purely transactional before, that’s going to be huge vulnerability. But the more that you can tell your story that you can have a personal brand that people are following you, they aren’t just getting information, the better. I think that’s the general point. As we move into this next decade of being creators, that’s going to be something that I think will ring true, especially within the context of search, which for this industry is especially important different for fashion. Fashion is visual. You want to see people wearing clothes, interacting. There’s still the ability to

Trey Griffin: How much longer though?

Bjork Ostrom: And it’s all vulnerable.

Trey Griffin: So if that image editor I was telling you about, you can go take a picture from amazon.com of a dress, you can take a picture of your wife and you can say, put the woman in the dress and yeah, I’ve seen snatch them together instantly and it’s exactly what you would look like in this dress or at least what it thinks you’d look like. So how much longer is it going to be before fashion starts to get pretty majorly disrupted? I mean, if you even look at Pinterest right now, so much of what’s on Pinterest is AI-generated, AI generated fake clothes that don’t exist,

Bjork Ostrom: And then you order them and they take two weeks to get shipped over it. They’re like seven sizes too small. Perfect. Yeah. A plus. Yep. That’s speaking from experience. Let’s talk about some of the uses of the tools that we can be thinking about as creators. And let’s start with ChatGPT. I’ve used both projects and custom GPTs. Can you talk about the difference between a project and a custom GPT?

Trey Griffin: I would say project’s a way of organizing a bunch of different chats or a bunch of different documents that you’re going to be coming back to on a pretty regular basis. That’s kind of just for you. If it’s that’s interested in this stuff, then projects are great because they’re really easy to manage and you just click in ’em and you go, right. It definitely feels like of the two features, it’s the one that OpenAI wants to push a little bit more and definitely gets updates a little bit more frequently. Custom GPTs are like if you have a team of six or seven people and you’re like, I have this great prompt that helps me do this one thing and I want to make sure that everybody else is using this great prompt, you can kind of set up a custom GPT super quickly, throw those instructions in the configuration, test it a few times to make sure it’s doing what you want it to do, and then you can quickly share it with all the members of your team. And they all have that exact prompt set up exactly the way you want with all the contextual information that you need. So it’s super great for that. We have one at Raptive that helps rewrite stuff and the Raptive corporate communication style that our comms officer brand came up with and

Bjork Ostrom: Fantastic. So it looks through and it’s like if Raptive isn’t capitalized and it’s like capitalized Raptive, or if there’s a word, if you use influencer, it might say, Hey, we use creator internally.

Trey Griffin: Exactly. Right. Yeah.

Bjork Ostrom: And so that would be for anybody who works with the team, you want to have them all have access to that projects would be something more internal. An example of a project I have is I have this idea of doing a family road trip where we go to national parks, and so it’s like I’m kind of brainstorming within that project. One of the things that I’ve done, and I’d be interested if you have some use cases of either custom GPT or projects that creators could use, is I’ve created one for sponsor content to help me quickly analyze and come up with some ideas for not the actual what this branded partnership will be, but just to help get quicker to a conversation that I’ve had with a brand and a conversation that I’ve had with Lindsay. And then how we work with brands to say, here’s what the potential pairing of those could be. And if I have data for all of those, this is a great video moment. This is a cat cameo for you. I love this. You need to do a quick explanation of who your cat is.

Trey Griffin: This is Wally. He has swimmer syndrome, so his back legs aren’t very good, and so he like to hang on your soldier and then you hold his legs up for him. It’s the ultimate

Bjork Ostrom: AI. And in cats, this could not get more internet based than talking about AI and cats

Trey Griffin: If overly online was one picture.

Bjork Ostrom: So anyways, that’s one of the ways that we’re using it. I remember a conversation that I had, my friend Alex from A Couple of cooks, he talked about building one kind of similar to the comms example that you talked about, which reviews a recipe to make sure, okay, are we talking about the ingredient in the same place as instructions? So let’s say the ingredient was the last one listed, but then it’s the first thing in the instructions. Okay. That should probably be looked at, but do you have other examples for ChatGPT, like projects or custom GPT that you could be thinking about?

Trey Griffin: It’s pretty good with SEO advice actually. Shockingly quite good with SEO advice and obviously SEO is kind of always in flux and it doesn’t know what it doesn’t know, so you should just bring your own knowledge to the table, of course, with anything around this. But I found that it’s pretty decent, especially if you can give it a little feedback and guidance to go the direction you want it to go. I think generally speaking though, right now a lot of the best use cases are allowed to take huge swaths of unstructured information and pulling structure out of it, which is kind of hard to do from ChatGPT, I mean, unless you really are patient, you have a lot of free time.

Bjork Ostrom: Hard to do.

Trey Griffin: In what way? The way that you would want to do it is you’d want to orchestrate it programmatically via an API, right? You’d want to print some code to do it, whereas if you try to do it through ChatGPT, you’re like, Alright, let me grab this recipe and this recipe and this recipe and we’ll do these three, and then we’ll do the next three in another chat window. And

Bjork Ostrom: There’s a lot of, it’s pretty

Trey Griffin: Manual. It’s pretty manual, and then you’re like, or am I actually saving time at this point or actually gaining that much? I think that a lot of the stuff, when it comes to custom GPT, that is actually super useful inside of ChatGPT is stuff where it’s pretty bespoke to what you’re doing. The way that I sort of look at it is this: if you’re going to ChatGPT and you’re writing a prompt to do something more than three times in a month, you should probably be making a custom GPT for it so that you can save that time of writing that prompt. And if you’re uploading the same six files every time,

Bjork Ostrom: Put that into project,

Trey Griffin: You should probably throw that in a project so that you can just ask questions about those files. It’s more about time savings rather than custom GPT giving you any sort of superpower or being all that much more powerful than just writing your question in a prompt. It’s about if I execute this thing 5,000 times a month, I want to make sure that I don’t have to go and write the whole prompt to do it,

Bjork Ostrom: Or my team does. Or

Trey Griffin: My team does. Yeah,

Bjork Ostrom: That’s great.

Trey Griffin: It’s also a good way of diffusing knowledge. If you have somebody who’s really good with these tools in your organization and they have a bunch of great custom gpt that help them do stuff that other people in your organization could benefit from, they can share that stuff with those people and then they can kind of drift off of this one individual’s AI skills and you don’t have to worry about everybody developing a bunch of great AI skills because this one person’s going to kind carry a team,

Bjork Ostrom: Has created the tool that then other people can use simply by prompting. Yep. Yeah, I mean we have hundreds of custom GPT at Raptive, hundreds, maybe thousands that people have created for their teams for, there might be three people that are working on a proposal document and it’s like how do you create a custom GPT to accelerate that process or,

Trey Griffin: Yeah, exactly. I mean there’s a lot of just double checking work. I mean, just getting a second pair of eyes on stuff can be enormously advantageous and AI is pretty good at offering good suggestions and advice and feedback, just custom gpt to review a specific kind of document that you did. Everything is an enormous time saver, but I think it’s hard to just be like, you use custom GPS for this. I mean, you can kind of use it for whatever you want, so maybe that’s not super helpful.

Bjork Ostrom: No, that’s great. I think that the challenge for anybody listening, if there’s a takeaway, it would be to just go in and create one. That’s what it was for me. I heard people talking about it, I heard conversations around it, but it wasn’t until I went and created one where I was like, oh, cool, I can see how this would be really helpful. And I can see some other instances where I would probably use this, but it wasn’t until I actually did it. That was the hard part is just going in and doing it,

Trey Griffin: Getting in there and doing it. I mean that’s thing learned for everybody I feel right now. And as far as technology come, new technology, this isn’t the cloud. You don’t have to have a cloud engineer to get in there and get value out of it. This is one of the most approachable new pieces technology. If you can construct a sentence in almost any language on earth, you can get value out of these models. So it’s really just get in there and get your feet wet. And I think that is probably the hardest hurdle to overcome. I think once people do a lot of the fear goes away, you start to see the brittleness a little bit. These models have weaknesses. The Terminator is not around the corner probably.

Bjork Ostrom: Yeah probably.

Trey Griffin: I noticed that you weren’t a hundred percent definitive with that probably. Yeah, well it’s one of those things where not to get too far off track though. I mean we, I guess we’re getting off track time. It’s now, right?

Bjork Ostrom: We can go there if you have time, we can go there. Yeah, absolutely.

Trey Griffin: I blocked out extra time before and after, so

Bjork Ostrom: Great.

Trey Griffin: These conversations always go long. Whatever the person wants to do, if they want to do an hour, it’s like, alright, let’s book two

Bjork Ostrom: Block an hour.

Trey Griffin: So I think that there is a compelling case to be made that what a lot of these AI CEOs like utopian scenarios could happen. Sure.

Bjork Ostrom: AI that is all knowing essentially.

Trey Griffin: Yeah, yeah. Basically you create a caretaker robot that takes care of the entire human race and keeps us from smacking each other in the head with flies, waters or whatever. I feel like that’s Sure why not. I also feel like in the same breath, if you’re a listener of the Ezra Klein show, people like Eliezer Yudkowsky are not totally crazy.

Bjork Ostrom: A little bit of background. Explain.

Trey Griffin: Yes. So Eliezer Yudkowsky is probably, he’s been doing this in 2003, is the AI dor. If you’re looking for somebody to give you nightmare scenarios that keep you up at night, nightmare Fuel, he’s a great resource for that. His most recent book, number one

Bjork Ostrom: Provider of Nightmare Fuel.

Trey Griffin: His most recent book that he published, which he was promoting on this podcast to listen to was If anyone builds it, everyone dies. So that gives you the idea sort of where he is. I would say that I am not anywhere near as pessimistic about the technology as he is.

Bjork Ostrom: Isn’t there a site where it lists every expert’s opinion on AI opinion on when they think the end of the world will be?

Trey Griffin: Oh yeah. The probability of doom p-doom, which you start to hear some people are like I listened to and they’re like, well below 1% as of right now, and then you have Eliezer Yudkowsky. where he is. It is 100% certain we will all die if we continue to pursuing this technology. But the median sort of response I see from people is between 15 and 20%, which is crazy high. If I told you there was a 15 or 20% chance that if you turned on your computer today that it was going to explode and kill you, you’d wait till tomorrow.

Bjork Ostrom: And that’s what a lot of these people are saying.

Trey Griffin: Yeah, a lot of they’re essentially saying one day you’ll turn on the computer and it will kill you, and they’re still turning the computer on, which is a kind of insane thing if you believe that’s true. Now there’s some evidence that we will skedaddle into science fiction fantasy land. I wouldn’t say the evidence is incontrovertible or overwhelming, but GPT–5 Pro solves about 20% of the questions correctly. On the epoch AI Frontier mathematics benchmark.

Bjork Ostrom: Which are highest of the high.

Trey Griffin: Yeah. Would take a PhD to solve even a single problem, a PhD in mathematics working as a tenured professor about an entire day.

Bjork Ostrom:: Sure.

Trey Griffin: So getting 20% of these questions correct means that functionally for the average person, like most language models at this point that are released by the top three labs, which is philanthropic, Google open AI, are better at math than you. They just are. So it’s already mastered this critically important domain for scientific advancement. But not only that, we’ve seen with one of Google’s systems called Alpha Evolve, it is come up with original new algorithms that saved Google 2% I think on their global compute costs, which for Google, that’s an enormous sum of money. And the AI basically did this by throwing spaghetti up against the wall until something eventually stuck. But that’s evidence that it could produce a new piece of novel information. It’s outside of its training distribution and wasn’t in the training data. So this argument that, oh, it’s just regurgitating stuff. It’s not capable of real creativity. Well, maybe, but also, and maybe this is just luck, right? You basically rolled the random number generator until you found a better way to do matrix multiplication, which is what they did. But maybe it’s not and maybe it’s actually

Bjork Ostrom: Figuring it out. Yeah,

Trey Griffin: Yeah. It’s actually maybe getting to a point to where language models start becoming more and more significant. And when you talk to professional mathematicians, Terrence Tao is probably the most well-cited example, and he’s considered by a lot of people to be the smartest person in the world. They’re using these models to help them do high level research in mathematics. So could we get to a point to where these models are so good at writing code and doing math that largely what the AI researchers are doing is standing over the shoulder of these models and double checking their work every once in a while maybe. I mean, it seems well within the realm of things that could happen,

Bjork Ostrom: Especially when pull it out 20 years.

Trey Griffin: I mean, I think that most people when they’re talking about this are not, if you say within 20 years, if you’re going to pull it out that far, almost certainly we will get there by accident within 20 years, even if you didn’t have just the amount of investment and data centers and energy and the resources needed to build these more complex AI systems is enough to where we should just get there by accident with 20 years. I’m talking like five, right? And what are the odds of this happening within the next five years? I think that’s where most of the industries mine is. And definitely when you’re looking at the valuation of these AI companies, you do not value a company with 700 employees like OpenAI at 500 billion unless you think that what they’re going to achieve, the thing they say they’re going to achieve.

Bjork Ostrom: That makes sense.

Trey Griffin: So you have all of that in one corner. This is my optimistic case for why we’ll achieve a generally intelligent system at some point in the relatively near term. My pessimistic case I think is much more drawn from a worldview where you don’t think about the technology very much and you just kind of think about what these actors are doing. If you’re trying to build a self-improving artificial intelligence system that will be smarter than any person, you probably don’t waste money on ai, TikTok. If you are trying to build an artificial intelligence system that is smarter than any human, you probably don’t focus on how well it can write. Cloud 4.5 seems to pretty tuned to be a great communicator and writer, but if you really believe that this was possible, all you would care about is math and code math and code math and code math and code math and code. That’s all you care about. Why do we have video models? Why do we have audio models? Why are all these, if you really think you’ve got a pathway there, you would behave differently than these companies are behaving.

Bjork Ostrom: Your point is such that if somebody’s worried about these becoming all-knowing and getting the point where they can communicate better than us, and one of the indicators would be the focus for these companies and what you’re saying is if they release soa, like TikTok ai, like you said,

Trey Griffin: That

Bjork Ostrom: Maybe is an indicator that they’re not putting all their resources towards,

Trey Griffin: They’re not putting all their eggs in the AGI basket.

Bjork Ostrom: That’s for sure.

Trey Griffin: I think that OpenAI will be around 10 years from now regardless of whether or not they have AGI within the next five. I think they’ll still be around because

Bjork Ostrom: Meaning artificial general intelligence.

Trey Griffin: Yes. Even if they don’t achieve that, I think they’ll still be around. But that’s largely because I see ’em putting eggs in other baskets, and I think that this is a little bit of OpenAI’s behavior in particular has been hedging a little bit

Bjork Ostrom: Because your point is, as a company, if you did think that you could create AGI, you’d want to be the first one to do it. You’d want to do it as quickly as possible, most important thing because of how valuable that would be for a company, and you probably wouldn’t be spending time partnering with Jake Paul. So

Trey Griffin: Just take it from a purely cynical perspective. If you have a machine that is smarter than any person, what is the number one determinative value of how powerful military is on this planet right now? It’s their technological prowess. Yeah, sure, yeah. How smart is the, it’s not how many soldiers you have. The US military is not the most powerful military in the world because they have the most soldiers. It’s because we have 11 aircraft carriers powered by nuclear reactors that nobody else knows how to make. So if you take it from a purely cynical perspective, I mean if you put this model, this supposed model that they say they’re going for in somebody’s hands, that person becomes one of the most powerful influential people on the planet overnight. Why would you bother with this

Bjork Ostrom: Stuff?

Trey Griffin: Because that’s what you think you’re headed towards.

Bjork Ostrom: Interesting.

Trey Griffin: So that’s sort of my real politic answer as to why I think that AGI is probably not right around the corner. But I also think that in order to have a drastically different world, you do not need to get anywhere close to a general intelligence. You just have to have an intelligence that is superhuman in a certain few key categories, Whether that’s drawing connections in data or whether that is drug discovery. I mean, think about AlphaFold, which is, if you’re not familiar with that, one of Google’s sort of AI dings that they were working on way before language models became their primary focus, which was able to discover all folded proteins. They just solved the protein folding problem. They’ve discovered every protein that’s underneath there, and it’s going to drastically accelerate the field of biology, but it is useless for anything except for that, but it’s going to change drug discovery completely. You’ll probably have Ozempic level discovery in drugs every year for the rest of our lives

Bjork Ostrom: Because of a tool that is able to do it with

Trey Griffin: Super quickly. Just basically it did a billion years of PhD level protein folding work in a year. So it’s completely upended the field of life sciences and biology and yeah, go

Bjork Ostrom: Ahead. I think the interesting thing within that is all of these, this is within the field of science as an example. You also have that within the field of video. We talk about Sora, so this is apart from the a GI stuff, which we can do a different podcast interview where we can go super deep on a GI. It’s one of my favorite things to talk about and one of Lindsay’s least favorite things to talk about, so I need somebody to talk to about it. Perfect.

Trey Griffin: Just let me know, happy to come back. But the other thing is what all of this is, and part of it is what is it today and what it is today is these are tools that allow us to do really incredible things in a way that we never could have done from an efficiency standpoint before. And so I think for us, if our job is we are creating content, we are monetizing that content with a brand partnership, we’re monetizing that, potentially partnering with Raptive, running ads against that content. We have our own product that we’re selling, e-commerce, whatever it might be, a digital product. We now have the opportunity to use a tool that is going to allow us to do something more efficiently. What do we do then once we have created those efficiencies, I think we do more of the things that we want to do and that we’re excited about and less of the things that we don’t want to do and that we’re not excited about. You can also take more risks.

Bjork Ostrom: You can take more risks that you have more bandwidth.

Trey Griffin: You have more time. Because one of the things with AI, one of the reasons I have this job is that I was able with AI to take more risks with my day-to-day life. I was able to go out and try a startup because I could get started on it at night. The help of an AI and then that startup eventually leveraged at this job. Time is something that you’re always risking. You’re always risk gambling your time, and this is a tool that can help you reclaim a lot of that time so that you can take risks on, okay, I don’t have to spend as much time managing my backlinks because I’ve got an AI that helps me with that now, but that means I have this additional five hours a week that I was using to backlink all my stuff. Now I can use those additional five hours, and I can go and try and kick off a video sort of content creation business that I’ve always wanted to do, but I haven’t had the time to do. Or I can get into a completely different vertical and start a new site in that vertical and go after that. Do a cooking class or everybody shows up and you’re doing it together.

Bjork Ostrom: Yeah,

Trey Griffin: Whatever it is that you want to do, right? AI is going to allow you to gamble time on bets that you maybe wouldn’t have wanted to gamble your time away on because you know that if I’m making content for my site, that’s going to make money, but I don’t know if I go and do this cooking class if that’s going to make me any money, but now you can have your cake and eat it too. I think that’s maybe the best way of thinking about it.

Bjork Ostrom: I think that’s great and I think maybe a final kind of thing that’ll leave people with to think about and then would be interested if you have anything else too is always ask yourself the question. I wonder if, so if you’re doing a thing, you get into a thing, you’re working on a thing, you start to get into it. Just ask a question. I wonder if there could be a way whether using chatt, Claude, Google, another tool that exists out there, I wonder if there’s a way that I could do this more efficiently where I’m doing less of the thing I don’t want to do in order to allow me to do more of the thing I do want to do, and I think that’s the thing that is most exciting to me about all things AI is that it opens up that world and that to me is I feel like to end on a, not like AI takes over the world and takes all of our jobs and shuts us all off, which is also again, very interesting path to go down, but today I know that to be true. I think that’s the thing that’s really exciting for me is finding those opportunities and efficiencies, and one of the things we’ll continue to do on the podcast is talk about what those look like. It’d be fun to have you back on at some point to talk through a tool session. Hey, what are these tools that we can use as creators?

Trey Griffin: I think there’s a ton of stuff out there too. I mean, even just get out of ChatGPT Land, there’s more than just ChatGPT out there as far as language models, I would say Claude 4.5 is an excellent writer compared to ChatGPT these days.

Bjork Ostrom: Yeah, I use both. It’s interesting. You pay for the premium plan for both, and spend a lot of time in each, and we’ll almost have them fact-check each other. I’ll do a prompt in one and then do the same prompt in the other. We could go on and on. I want to be respectful of your time. If people want to connect with you, I know that you’ll occasionally write content for Tive. I know that I mentioned LinkedIn at the beginning. You have a LinkedIn account, I’ve got one, and then there’s obviously Raptive as well to follow along with anything else.

Trey Griffin: Yeah, Raptive and LinkedIn are great. I think I’ve got a Twitter, but I couldn’t tell you what the handle is.

Bjork Ostrom: It’s there, though.

Trey Griffin: It’s out there. You can find me on Twitter, but really LinkedIn if you want to get in contact, that’s where you probably want to go, but you could always just drop in and say hi at an interactive event. I’m always down to say hello because they take me to all of them, so I’ll always be there.

Bjork Ostrom: The AI guy, you need to be there.

Trey Griffin: Got to be there.

Bjork Ostrom: Yep. Trey, thanks so much for coming on.

Trey Griffin: And thanks for having me. Take it easy.

Emily Walker: Hello, Emily here from the Food Blogger Pro team. I wanted to pop in today and thank you for tuning into this episode of the Food Blogger Pro podcast. We are so grateful for you for listening. Before we sign off, I wanted to talk a little bit about the Food Blogger Pro Forum. In case you didn’t know how it works, if you are a Food Blogger Pro member, you get access to our amazing forum. It’s one of my favorite places on Food Blogger Pro. I spend a lot of time there myself, and on the forum, we have tons of different topics for you to explore. We have a Building Traffic section, a Photography section. We have an Essential Tools section. We chat about generating income and essential plugins, all sorts of areas for you to ask questions and chat with your fellow Food Blogger Pro members. It’s a great place to connect with fellow members, troubleshoot any issues you’re having and brainstorm together. Our industry experts are always popping into the forum to help members with their questions. Casey Markee and Andrew Wilder are always popping in and so is Danielle Liss our legal expert. It’s a really great place to get access to these experts and have them help you with your concerns. The Forum is also just a fantastic place to find a community in this food blogging space as you’re working to grow your site and your business. If you’re ready to join Food Blogger Pro and get access to our wonderful forum, head to foodbloggerpro.com/join to learn more about our membership. We really hope you enjoy this episode and can’t wait to see you next week for another great episode. Have an amazing week.