How I Built a 7-Figure Business in Less than A Year with AI | Ben Tossell

We're talking about all these automated systems. We can't then be an automated system telling you about an automated system. I want to have some personality, some taste in what we're sharing. So we're a bit precious about any content being written by AI. I know people are going to be saying, well, you teach people how to build AI, like use AI to make content and all the rest of it, then why aren't you doing the same? I think that's important. AI is the co pilot. It's not the thing that takes it end to end. Supposed to assist you doing the work. So it assists us doing all of the work. It doesn't actually do the work. It's like it funnels down from this large task down to a smaller one. And we do the refining and the specific pieces of it rather than we just feed it into AI and have that come out with the output.

All right, well, welcome everyone. My guest today is Ben Tossel. Ben is the founder of Ben's Bytes, a seven figure business that has taught over 10,000 people how to use AI to save time and do better work. Previously, Ben sold his no code platform MakerPad to Zapier. Welcome, Ben.

Thanks for having me.

Great. So let's go back to a few years ago. First you had a daily AI newsletter. I think it had over 100,000 subscribers. What made you decide to move from a daily newsletter that's free to a paid plan?

They're mostly through things that I hate because I think dealing with advertisers, I don't think it's not many people's choice. And also for a reader, it's like if you asked a reader what would you want to see more of in a newsletter, they wouldn't say, more ads. If you ask them how would you want to pay for access to something like an education platform, they wouldn't say, oh, every month, bill me $50 or whatever. They just really want to and like doing things because they're just the right thing to do. And also I think that it's an easier thing to quantify in your mind where if I pay once I get access to all this content. And also we started with like 50 tutorials when we first launched and we're now over 320. I think as of today, it just gets more valuable over time. The whole point of this building it like this is we can just keep stuffing value in there. We started workshops recently with experts and it just gets better and better for everyone who joins. So like, who wouldn't want that as A user, I think, is kind of our point. And also there's so much like so many subscriptions to newsletters, Netflix, all the rest of it. Like, you don't need another one. And I think also with learning specifically, we'd have to hound you all the time to be like, are you learning something this month? Did you learn something next month? How about you take a challenge? How about you set yourself some goals? I don't want to bother people for them to want to learn something. You either want to learn something and there's something there that you maybe joined and you can dip into whenever you want. There's just some of some other, like, education platforms, they email you all the time to try and be like, don't forget your streak. Don't do this. And it's just like, pressure that people are busy. There's a lot of things going on. So we've just gone the simplest model possible. Like, it's not the best business model possible. But I'm not optimizing for that. It's optimizing for what's best for everyone who's going to come and be a member.

Yeah. If you do deliver, like, a ton of value to members, then, like, word of mouth is obviously one of the best channels to grow. And I think you've only been doing this for, like, the education portal for what, eight or nine months, right?

Yes, it's March. We launched it.

Yeah, got it. And then you tweeted that it's already bigger than your last startup. So I was wondering what have been some of the key drivers of this growth? Or did you just have a big subscriber audience from the free newsletter or what are some other drivers?

Yeah, I think so. The newsletter obviously is a big portion of where we got a lot of our early members, so that definitely helps. And it wasn't my plan the whole time to have a newsletter, then shift it into something else. It was just the newsletter. And also I think obviously AI. Everyone has heard about AI, or most people in sort of a professional capacity have heard about AI, whereas my previous company was around no code. So it's effectively teaching you how to build a website or a membership site without having to code. Now, there was a wave happening at the same sort of time. I was early on that as well. But that was something where you had to teach someone first, educate them about the fact that no code exists. And like, to a normal person, coding is this. So no, coding is the opposite of this. And also we teach you this thing by doing these things. It's like there's too many things where if you come onto our site and you see, here's how you use ChatGPT to generate B2B sales calls or whatever it is, you know what you've heard of ChatGPT, you might have tried it and like the market has educated the market on what AI is and why it's useful and why it should be used with everyone. I think companies are telling their employees like, what do we do about AI? Like what are you learning it? Like there's pull from everywhere else. So we can sort of sit here and like say, yeah, come on in.

We welcome home, just kind of meet the demand, right?

Yeah, yeah, ideally.

So are most of your members actually like not these people who are on Twitter all the time, like normal people or what's the profile, the typical member or is it very diverse?

It's pretty diverse actually. So it's difficult to pinpoint, but I think there's a big chunk of people in marketing, sales, customer support, I think all of those where AI is really sort of in their face, in their jobs. There's plenty of sort of founders and entrepreneurs that are looking to do more with their business or even if it's them saying to their team, hey look, we can do, I don't know, optimized SEO at scale by doing this, we can point our team to do that. And how about people do more with less? Because I think that kind of ultimately is where people have been talking about big companies built with small teams. And I think there's lots of these individual contributor roles that are going to come back out rather than like a big department of marketing with a manager and lots of like mini marketing managers. This is something I'm stealing from Sahil, from Gumroad who I had on yesterday who talked about this exact, this exact scenario where. And that workshop is for free, so anyone can go watch that. But it's interesting that founder mode whole post brought that out. It brought that out to be like, well it's with AI there's all like the power is in the hands of the people who are going to use the tools to get as much done as possible. And it might be that one person can do that role that four or five people would do before, but it just means that you just have to be the person that is doing the work. And like using AI to do that.

Yeah, it's like important to scale up because a lot of these jobs are going to get reduced and like you said, one person will be doing four persons work. So hopefully you can become that one person.

Yeah.

Okay, well, let's talk about, let's get into the details now. So let's talk about how you use AI to run your business at a high level. How would you break down your usage of AI tools in a typical day?

I really wish I could come on this podcast and say AI everything, like it's all AI, but it's not. And mostly the most boring answer, unfortunately, is it's a big split between Claude and ChatGPT. That's where I use most of, most of my AI usage is. So, yeah, it's probably about 80% Claude, 20% ChatGPT right now. It changes as new models come out, new features come out, whatever it is. Yeah, I really like Claude at the moment.

Have you been able to, like, dedicate any tasks entirely to AI? Like, you know, just like you? I use it to, like, basically clean up a bunch of data, summarize a bunch of, you know, server results, or like, maybe help me edit my content. But have you been able to, like, just like, hey, you go take care of this and I'm not going to worry about it anymore?

This is the funny thing I think about AI, where everyone's expecting it to do all like this huge grand thing that you don't have to touch anymore. But it was the same, the same paradigm as automation, where some of the time you need to be involved and you need to check. And with LLMs, obviously it changes responses and there's a lot of prompt engineering to test it and all the rest of it. But we have some things where when signups come in, it'll sort of pull out the domain of who signed up, if it's not Gmail or Outlook or something. And then is it a publicly traded company? Is it a big company with like checking? Is it over 500 people, those kind of things? And I label them effectively. We've tried to do it with notion AI with our tutorial content. So we have our backlog of. Here are all the tutorials we're going to be making. And then we have roles, categories and tags. So they're all like marketing manager in the category of marketing. And tags are ppc, event marketing, product marketing, whatever. Nifty. I is not very good, I don't think, because we wanted to just effectively look at this and make the next best guess of what is the right category and tag and they should all link. Right. So we haven't found that that's been very effective yet. But that's something we're trying to get better at and generally with things like that, we tend to Keshav on the team. He sort of is building tools to analyze the tutorials and do it that way and build it through using LLMs, but through a tool like using a replit but building a replic tool.

Got it. Let's go back to the email automation workflow. Is that using Zapier or. Yeah, okay. Yeah. So I never figured out how to use Zapier properly. So using Zapier and like ChatGPT API or something to kind of like. So maybe walk us through how that works like the steps like.

Yeah. And actually Sitqi Chen who runs a company called Runway, he's got a good workshop on this somewhere, I don't know where exactly. Maybe we can find it in the show notes where he talks about like a really, really in depth way that he has something similar. So it'll be a new member created through our authentication system. Then it's. You can do it a few different ways. But I think we've moved to prompt ChatGPT, put out the domain that this email is from and then either. I think you can do a web search now with, with the Zapier step where it'll say search the web for this company. Look at its LinkedIn or Crunchbase profile. Does it have over a thousand employees? If so, say high, high interest or like I can't remember what the names are. And yeah, just then put it into Slack. So it's kind of a monitoring system in that way. But it's. It's really three or four steps in. As a.

You mentioned that you have a team members now. So it's not quite a one person business anymore. Like how many people do you have now? And like how do you, like what things do you have to dedicate to people still versus using AI or it's like combination, I guess.

Yeah. So we do have a team. There's me and there's Shanice. So Shanice was my coo at MakePad and she's now COO at Ben Spites. And then we have Keshav who he's been writing the email for a long time now or helping me write it since about January last year. So he does a lot of the email stuff. And since working at Benspites he sort of picked up coding by using AI to learn how to code. And now he's doing a bunch of like internal tools for us. And Adam, who's my brother, he helped me build the site effectively. So he does all the design work and all that kind of stuff and then we both play around in webflow and things like that. So that's the core full time team. And then we, we have three content creators, Garrett Tanmay and Andrew, who are creating tutorials. Yeah. And we've got Ben who's a developer recently who's helping us like level up because we're on no code stack and we're banging our heads against the wall. So we're, we're trying to like add a bunch of these things. So all the tracking stuff that I talked about, the reason it's not possible is because we built on no code. So we're trying to make sure that a more proper company is supposed. Yeah. And yeah, the stuff that we want humans on. I've always been the same with the newsletters, never been automated because in if we're talk. I don't know if it's like a pride thing of. We're talking about all these automated systems. We can't then be an automated system telling you about an automated system. I want to have some personality and taste in how we communicate. So we're a bit precious about any content being written by AI. With tutorials specifically, we tried a bunch of ways in the early days when we were trying to. It was like me doing the tutorials. I was thinking how am I going to make 200 of these myself and do everything else? So I set up a bot using po. Like one was coming up with tutorial ideas, one was scoping out what the tutorial steps should be. I would then go and do the steps in ChatGPT, for example, and take the screenshots. Then I'd put that into a document and I'd use Lex, which I'm also an investor in. And then from there you can basically prompt Claude directly and say this is a tutorial. These are what perfect tutorials look like that are on our site already. Can you fill in the steps for what should be written in each of these and obviously look through all of their screenshots to make sure it's all cohesive. And it was always like it's okay, like it's fine. But it wasn't. This is a good tutorial that someone's going to find useful is very basic and wouldn't really go down like the reasons why you're doing this thing in a tutorial or like make sure you give it examples or make sure you do that. So we've kind of scrapped all of that. That was only really for us to get to launch. I don't think we even used them because we ended up editing them heavily. So, yeah, it's just. Yeah, people make tutorials and people make the content that we have that we put out. So I know people are going to be saying, well, you teach people how to build it, like use AI to like make content and all the rest of it. Why aren't you doing the same? I think it's just. That's the point. I think that's the point is the copilot, it's not the thing that takes it end to end, supposed to assist you doing the work. So it assists us doing all of the work. It doesn't actually do the work. It's like it funnels down from this large task down to a smaller one and we do the refining and the specific pieces of it rather than we just feed it into AI and have that come out with the output.

Yeah. But it's also good to kind of push the boundaries on what it can do or try to test the boundaries because maybe six months or a year from now it can actually do more of the work. But I totally agree. Right now it's very much a copilot.

Yeah. Well, like I said to you, I was testing O Preview yesterday and I prompted it with a tutorial that we have. So I gave it the title of the tutorial and I said, create this tutorial, like step by step tutorial. And it did a pretty good job of actually because it's thinking more at every step, it's thinking more like what is needed for this person to understand this thing at this point, like what will be needed then in the next step. So I think some of the extra thinking time, the reflection and all the rest of might be more helpful for writing tutorials. But again, it's like it's in chatgpt language. It's not always good for like, easier to read. Yeah. Who knows, maybe it'll one day be 100% all AI tutorials. And if that is the case, we've got to do something else.

Yeah. Just a quick question aside, like, I've had trouble finding help for my content stuff. Like, how do you find these? I mean, obviously your brother is your brother, but like, how do you find good content creators? Or like, do you use AI to try to recruit them or you just find them online?

Yeah, I mean, so we found. Or I found Keshav from. I tweeted because I was looking for some help with the newsletter because I knew I had to go on holiday at some point and I was like, how am I going to take a week off the newsletter and then he just came through that way. Everyone. My hiring process is the same for everyone. We sort of find someone, we just put them straight into the work and do like a two week paid test and say, let's try this. For what? If it's like a tutorial, send us one to two tutorials. We pay you your hourly, your hourly rate for those tutorials. If they're good, then we move you into like a longer paid trial and then if that's good, we'll carry on with working with you. And that's how we found our three content creators. Garrett was already in the community and this is actually what happened with MakerPad previously as well. We found our tutorial creators from the community. So like this, it just happens to happen like that with the kind of people where we're interested in, in hiring. And yeah, we, I think we put out like, oh, we're looking for two tutorial creators on the newsletter and on Twitter and stuff. Often they're already members or they followed the work and they're basically doing similar stuff. My, me and my brother thought of an idea to have. There should be a place to find people who can make educational content. Like that's. It's difficult to find a list of people who are good, who can do that.

I think, yeah, that might be a good startup idea because, yeah, the process you just talk through makes sense, but the people aren't good. It takes a long time, a lot of trial and error to find the right person. Has the community been key, do you think, to the education portal's growth? Because you're selling a lifetime plan. Right. So it's just like you join the community once.

Yeah. I think with makespad, for example, I think the community was one of the main drivers of why people would join in the first place. It was like surrounded by other people, other. No coders, figuring out how to build their business. This one has seemed less like that. But what I've really enjoyed seeing is the community. A percentage of the people who join join the community, not everyone, which is fine. Like there's enough going on in your life and then over time it's becoming gradually more engaging and more active. It's not like going into a discord if you've ever joined one in AI, it's just like, oh my God, there's so much stuff here. How can I get a word in edgeways? How can I speak to anyone, connect with anyone, understand what anyone's doing? I didn't want it to be that I didn't want to be like, let's get everyone into community. Let's really push people to engage. And I think it's. It's naturally become like a more organic community now from not trying too hard to make it a community, if that makes sense.

Yeah.

And I think it just. That's the best. The best way to. To do it. You sort of. For the community. I always said lead by example. So the soulful responses, it's thoughtful help. People are responding to questions all the time. People are helping one another. It just naturally follows that state. Instead of one channel being like bombarded with messages all the time and short messages from everyone, that's just going to quickly get overlooked by anyone who's actually looking to learn anything in this space.

Yeah. I think with communities, it's actually better to start small so you have a core group of people who actually engage.

Yeah.

Then to try to get like 5,000 people into Discord right away. Like, that doesn't work. Are you using Slack for your community?

Yeah, we're using Slack too.

Okay. Makes sense. I mean, you know, just a word of warning for people who want to do this. I mean, it is a lot of work, right, to engage in the community.

Yeah. Yeah. To build one is very difficult. And yeah, I think there's a few ways to do it and it's just gotta be intentional. I think. Like, think exactly what the person wants. Lead by example. Like, tell. Like, you got to remind people that it's there. Well, we have a member digest that goes out every Monday. This just talks about our new courses, tutorials. And then. So this is actually one thing completely automated with AI now from. Keshav built this. So we wanted a. We have like a. Here's what's happening. Interesting conversations in the community. So that's a section in the newsletter and it basically says like, oh, Ben asked about who's trying 01 preview. And like, whatever it is Keshav asked about, like recommended a tool that does this thing. So there's a bunch of conversations that happen in the community and people miss stuff all the time. So Keshav built a system where it pulls all the activity from Slack. It then effectively categorizes it, formats it exactly how you want, like links to the chats and then we put that into the members digest every. Every Monday. So.

Wow.

Definitely one thing that actually has done really well with. With AI that's forgot about that. Which is kind of the point. Right. You built it so you don't have to think about it again. And it just Works perfectly.

So is I use Zap Zapier again or is that.

No, no, he built it with code. So yes, it's a rep wall on replic and it's easy. Is Claude. Yeah.

That's incredible, man. Like this guy was just a blog writer before. Right now he's been upgrading stuff.

Yeah, well, we did a write up on our blog so you can read about how he did it and what he did. So we can put that in the notes if you want to.

What have been some of the most popular tutorials you have? 300, right. So what have been the ones that people love the most?

The most popular of all time is the Learn ChatGPT series, which I think.

Yeah, that's kind of like a big one. Yeah.

And then Claude Artifacts has been a big one. And then, I mean I can pull it up right now and we can actually have a live look at what they are. I should know off the top of my head, I suppose. But there's a lot like you said. So we have. Yeah, ChatGPT, actually the Claude one has just surpassed ChatGPT. And then there's how to write effective prompts. Creating interactive dashboards, a job search one. So accelerate your job search has become quite a big one. Market research. Yeah, things like that.

Got it. Job search, like get AI to send out like 100 different applications.

It's not quite like that, but it's. Yeah, it's how you can accelerate your job search using AI and to do a bunch of that stuff. I think that one only came out recently, the last couple of weeks.

So interesting. I gotta check that out. Yeah, let's. Before talk about like the ecosystem, let's talk about like writing great prompts. Like, I feel like after using the stuff every day, I had like a pattern that I use to write prompts using evolves. I try to put in a few examples. I try to try to think step by step and so on and so forth. Like, do you also write your prompts a certain way now?

I don't know if I should be like admitting this, but I don't do any of it. We teach the best way to write a prompt and we teach you how to do those kind of things that you mentioned. But also the way that we teach. It's not like this is the perfect prompt to do to accelerate your job search, for example, because these tools are chat interfaces. They're built to have a conversation. They're built then also to have this context of your conversation to then continue their output with the next request that you've asked for. So we teach it as a conversation because that's how it should be used. It's not like every prompt is going to do that. I will always put some context as like, I think these are the key parts, like you are an SEO professional and like, I'll put some of the things like that. I really like the way that X writes certain blog posts. If I was using it for that example, I would put some of the basics in like that.

Yeah.

But I wouldn't necessarily put in the perfect prompt, as you could share on Twitter and everyone would say, yeah, really good prompt. And I will often upload just like transcripts or like text, but copy loaded text and paste it into Claude because it sort of adds that as an attachment. So I will do a bunch of that because in the conversation that's where I'm stuffing in the extra context. And I'd rather sort of refine it that way because I'm never going to get the right answer perfectly out of the first. The first go. Because I'm usually trying to work through something and figure it out from start to finish and then I can go back and all that kind of thing. So I think there's two sides of prompt engineering. What is you're building a tool that, that needs to have a similar output every single time. So the prompt needs to be correct and that is the one you need to optimize to be the right context, the right length, the right examples, all the rest of it. And the other ones are using ChatGPT or something. You just don't need the perfect prompt there. But you need enough to actually guide your conversation the right way.

Yeah, like if you're just asking a bunch of. To help you solve a bunch of random problems, which I do all the time. Yeah, it's more just a chat conversation. But sometimes I have like these specific tasks like, you know, I wanted to clean up my transcripts, like fix all the grammar and make it more concise. And then I have these like really long prompts of examples that I just copy and paste from like a Google Doc.

Yeah, yeah, yeah. I've seen people who do that on Google Sheets. This. So Danceship has got a tool called Spiral. So this is kind of built for that kind of use case. So you can create your own spiral. You can give it a prompt and a bunch of examples. So if you're always doing transcript to blog post or transcript to tweet or LinkedIn post, whatever it is, you save each one as a spiral and it'll have like a consistent feel every single time you do it. So you just click on that one. Oh, I want to do this transcripted blog post today. You dump in your transcript and then it'll do all the things that you've said you've saved. So yeah, that's one thing to try.

And sometimes I also like, you know, after sometimes it doesn't get it right and I have to give it bunch more chat messages and then I will close by saying, okay, now go update my original prompt based on, based on this additional input. Yeah, that helps a lot too sometimes. Okay, cool. Let's close by having some fun. Let's kind of try to share our honest takes on popular and underrated AI tools. So I want to start with one that I think you have a hot take on, Perplexity. What are your thoughts on Perplexity?

Well, I knew this was coming and I'm sure you're going to be the last person to ask me. So obviously firstly, I don't wish any company like bad vibes or anything. I just never got good results with it. Like to me it feels like it's a different. And I had the pro, I had all the best models turned on and everything. And to me it's like a, it's a different function, it's a different job to be done to be searching on Perplexity than Google. So if I'm googling like I want to find the intercom homepage and I don't know what it is, I just type in the word intercom. The first thing comes up is intercom homepage. Click on it, I'm there. Perplexity of time in intercom starts telling me the story of how intercom started. Like there's no. There wasn't. There probably is now, but there wasn't at the time. A quick way to say, like, I'm just looking for this one thing. Give me that one link to that one thing. Because it's default to here are a bunch of sources and we cross reference those to give you the best answer. Sometimes I want to go to the actual page. So that's that bit I find odd. So then when it was something like, okay, find me reports that have been done about AI at the work in the workplace from the last eight months from companies like Accenture, Deloitte, LinkedIn, Microsoft. It can't do that either. I'm like, I don't. What is it that you're supposed to be doing there? If that's not the exact thing that I should be trying to use perplexity for. So I can't type that into Google Search and get great results without going through it all myself. But I can Write it in ChatGPT and I get way better results actually without even their search thing, which I haven't got access to.

Yeah.

So to me it's like the experience of using it wasn't the promise of what it is or what it should be able to do. And I think all these AI tools are a new way of computing. And I think if we're going to be sat in ChatGPT type systems a lot, doing a lot of our problems, usually the point where you want to search something is within that chat or within that workflow. Me to then go somewhere else search something and not get the right answer is like, how many times am I going to do that before I drop off? And I have churned as a customer, but how many times are the average person going to do that versus when search is available in Claude chatgpt properly, I'm just going to do it there and then in line with my conversation. When I'm doing that thing, it feels like a different thing, but it might be that it's a different use case.

Interesting. Yeah. So basically, I think like some of these big model providers will add a search Property or like ChatGPT has already added search and you think that will kind of disrupt perplexity?

Well, I just, I don't, don't mean to undermine what they've built because they've been building more since I left. I deleted my account, but it's not, it wasn't, it's hard for me to build, of course, but it's not hard technically to build. I think the founder of Julius built a clone of Propriety in like three or four days. There's others who have built open source versions. There's other companies like U.com, there's companies like Exa AI. They're really good. There's all these different ways to search. I feel like the better ones, they might not have their own specific place to search, or they will, but also they'll be integrated with the tool you use the most to do all the most general purpose things that you're going to do daily, because that's what ChatGPT is just like the tool that does all of your general tasks that you want to know about, but you need like the EXA plugin or whatever. But I know they're not doing plugins, but that's the kind of thing I'm thinking is it feels like it has to be within the general product for it to be huge scale. I don't know.

Let me defend them a little bit.

Yeah, I was going to say you try and convince me otherwise. I'd love to know what either of it.

I probably use it like 30% of the time and it's cloud 70% of the time. I actually completely turn from ChatGPT. Maybe I'll come back with this new model, but for perplexity, I think it's really good for doing research on current things. Like for example, I'm going to Japan next year. I want to know when is cherry blossom season and what the latest events in Tokyo is. Really good for doing that and giving me an itinerary of stuff that I should do. Because Google would just tells you maybe like one piece of information but doesn't give you like kind of like the whole itinerary or maybe you want to do some research on like what are some A B testing tools that I should use or something. Like it's pretty good that you do have to like kind of dig a little bit deeper to make sure it doesn't miss like clear gaps. But it's pretty good at like looking at a bunch of, bunch of current articles. You're trying to figure out what the latest and greatest is. I also use it. I used to search Reddit using Google like, you know, just like Reddit, what are like the best socks to buy for my kids or like the best clothes. But now I switch to Perplexity. You can, you can change it to focus on social sites and then it actually does a pretty. I, I feel like it does a pretty good job of summarizing a bunch of Reddit threads for like a particular question that I have. Yeah, that's not what I use it for.

Yeah, I can see that last one being useful, but I would. To me, when you said the first one, for example, looking at your trip to, to Japan, that's like one of the main use cases. People say ChatGPT, you can use that to write poems and plan trips and itineraries and stuff. Yeah, I just feel like it could probably do a similar job. And it's interesting you said perplexity versus I can't go and Google that easily. But you could use ChatGPT more easily to do a similar job and the outputs. Like there's two different things. I think there is Google versus Perplexity is perplexity can get better information than you can on Google and a more specific like I'm looking for this specifically, whereas Google is just like here's the links, you figure it out. So I understand that but I think there's obviously there seems to be two, like there's different types of I'm searching for something specific or I'm researching something, you figure it out. So complexity works for that. Google doesn't work for that. Yes, but then also it's like your comparison was between Perplexity and Google. And I agree, I don't like Google as a product either. But to me I'm looking at it as ChatGPT type products versus perplexity rather than Google versus perplexity. So yeah, it's strange and I just, I mean I hope I'm, I hope I'm wrong. I hope I've just come back and I pay for it and I love it. It's just when I was using it for that expect that reason, the report one was recent because everyone keeps on talking about you as well, like saying how great it is and I'm like, I just don't know if I missed it when I was saying all these things that I was like, I don't think it's going to be that great or it hasn't been that great for me so I'm going to churn. And then I tried it again and then people were like, no, you got to put the Pro mode on, you got to put the GPT4 one on. And so I did all those things. I changed the model, I did every version, none of it was great. And then like other people were like, oh, what did you do? I'll try it somewhere else. And like no one really got it right and the most right I got was was from ChatGPT. They were the most accurate results I had. And that's just their baked in search with Bing, I think. So interesting. I was like, yeah, it hasn't convinced me again but I'll keep doing these tests every so often. But that one didn't convince me. But we'll see what the next one is.

Yeah, you should do like a quality Perplexity test and tweet them out. I'm sure it'll go viral. Yeah, yeah.

At some point it'll be like everyone hoping, is this going to be the day that Ben changes his mind? And then one day it might be.

So what about this race between OpenAI Anthropic Cloud and maybe meta, like just like the foundational model race. Any thoughts on that? I feel like I don't have any real loyalty to any of them. I'll just Use whatever the best one is.

Yeah, so will I. And I think a lot of customers might do the same. I fear that we're all in the Twitter bubble and the Twitter bubble will do the same, but the everyday person might not. So I'm trying to, I'm trying to not bucket everyone in with me in like you will just use whatever's best, but like the mentor AI one, I don't know if it's because I'm in the uk, I haven't even experienced it, I don't even know what it's like. I just think no one cares. No one cares. Again, this is all generalizations, but no one cares. In terms of which model is it, who's providing that model? It's just, is the output good for what I put, for what I asked or what I put in? If it's good enough, if it does the job, that's great. And like these bells and whistles might be more for people like me and you who are like, I really want this kind of specific taste or vibe or whatever it is, or needed to think through a million steps beforehand, but I just think they, they all have big reaches. It kind of feels like the Apple vs Android thing. Yeah, people say it's like the war vs Apple vs Android, but everyone who's on Android doesn't care about Apple. Everyone's on Apple doesn't care about Android. You like the system that you use, use it for specific reasons. Like ultimately they do the same things really, and most of it is very similar. It's just more going to be on the, like, whether the most people use it or whether the most businesses use it for their AI, for powering their AI systems. Is it going to be open source and stuff? But ultimately I don't think anyone wins. I think they're all, they're all big models, they're all going to have loads of big capabilities. They're all going to be very slightly different and you'll probably just pick the one that you like or it's on the toolbox or it's either you've got an Apple so you use the Apple one, or you're on Facebook all day so you use the meta one. Like, I think it's just going to end up like that and it'll be like a nothing. Yeah, war. That is never really anyone at anyone's throat.

It does feel like it's kind of hard to, you know, make billions of dollars from, from the model itself, like just, just because, like, there's no real loyalty and meta's out there, like trying to open source everything. Like it's just kind of.

Yeah, yeah. I think. Yeah. Consumers. I don't think it matter. Like it doesn't matter or no one cares. And I think it's just like businesses just want to know who can do the best job for the cheapest amount of money. And they. They won't have loyalties if it's the cheapest and the best. It's just which. Which can do those two things. And I think Matt is obviously trying to do that, but OpenAI is trying to just make sure it is the best. The best. The best. But it will cost you. So. Yeah, so let them fight it. And we just will watch and enjoy all the extra features and capability.

Yeah, there you go. What about. What about like image generation or video generation? Like, I often want to make like infographics and I haven't found a really good AI tool to do that. Or like. Yeah. Have you found any useful tool there?

Infographics. So there's one I tweeted about the other day called Napkin AI, which has like a degenerative kind of flowcharty. Like it'll spin out. You know, when you see the. You see like illustrations on people's websites that will have like Ben spites and it'll have sort of arrows coming out and then have a bunch of different things like courses, workshops, whatever it is. You should have a look at that. I think that would be interesting for that use case. And then there was another one. I'll try to remember. Remember the name and send it to you. But that was quite similar as well. It's like these.

It's called Napkin AI.

Napkin AI. Yeah.

Okay, I'll go check it out.

Yeah. And then. And those people thought I built the tool when I tweeted about it, but I've got no affiliation whatsoever. I could just sort it and then. Video tools. I'm biased because I'm an investor in Pica Labs. So I think they're great. I just can't like, if you're such an insane output for such a small team.

Yeah.

Image tools. I really like Playground. I like. Feels like that's got taste again. There's two others ever Art and Glyph. I really like Glyph. I think that's quite a cool. A cool tool and helps you remix stuff. And they've got Chrome Extension and. Yeah, it just. They're all like sort of. They're all going in their own mini direction. Like Playground feels like a bit more Canva esque. Where it's like building all these different ways you can do it for social media images or blog posts, whatever it is. Glyph is more like fun and friendly and Ever Art. You can like upload a bunch of your stuff or you know what you can upload? This is what I tried to do with Ever Art before. Would you upload 10 images or five or 10 images of a style you like, then you generate something or you upload a picture of yourself and say, I want it in this style. So you can do a lot of like mini training models of images using Ever Art.

Oh, nice. Yeah, maybe I can generate some YouTube reaction faces.

Yeah, yeah, maybe. I think Photo AI is doing that as well now, so.

That's right.

Yeah, yeah.

Any other like underrated AI tools that people should check out?

Well done. I don't use that many because I'm not trying to. I'm in the point in my life when I'm not trying to over optimize everything. I'm like, yeah, back to Apple Notes. I'm back to like, let's just take everything off the cards. And I do like then upload Apple Notes to Claude and then ask for stuff there. So that's the AI version of that. But there's one tool I like, but I don't use it because I feel like I'm not, I'm not this person yet. But it's called Voice Notes, so it's built by Gigio, who is the guy behind Buy Me a Coffee, if you know that, like that tipping website where you can buy people a coffee. So he built this recently, him and his wife. And again it's just about like, they just have so much taste. They built such a nice looking, nice feeling product and that's effectively for like video, audio, journaling or you can text in there as well and they can tag it and then you can like look back at things like that. But I'm, I'm early in my journaling life. I started last week. I'm just writing in notes and yeah, I'm not much of an audio person. Like I'll do voice notes to people on WhatsApp, but like I've got the advanced ChatGPT voice stuff but I just don't, I don't walk around like chatting like that or giving, doing it that way. So yeah, yeah, I feel like I'm not, I'm not a good person for that. But maybe one day I'll be chatting.

You don't, you don't want to be like a tech bro in the uk just chatting to your phone.

Well, I'm just happy just sitting in silence to be honest. Just writing it down. So.

Yeah, yeah, yeah, maybe that's a good thing. Close on. So you have, you have twins, right? How old are they?

Yeah, 15 months.

That's crazy man. Like how do you. I guess you have help now but like how do you, how do you balance everything?

Yeah, luckily they sleep well. They have done since they're about four months with a sleep training so that was just the best thing ever. So I'll do a full night's sleep and have them like I said, for a while. Yeah. Generally I, again it comes back to the membership, the type of business that I have built. I build it to suit me and to suit the kind of people that I want to attract to be members. I'm not trying to optimize my whole entire life so every second is planned. Like I only do calls except for this one. I only do calls Monday to Wednesday in the afternoon. I just time block that and then I like play tennis in the mornings and on like half the week and the other half the week I'm getting up with the kids and feeding them and I'm always done at five to do like, what is it? Feeding, bath, bed and then I'll feed myself and then I'll just do a bit more sort of lower impact work. Sat in front of the TV for a few hours. But it's. I'm just trying to take it. I say take it easy but I kind of am trying to take it easy and just like go at my own pace. Not everything's not at the expense of life just like everything is stressing me out because it's all too hectic. Like I never want to be hectic and it's hectic enough in my house, isn't it? So I don't need work to be hectic as well. So. Yeah.

Yeah, I think it's like very underrated to build a nice business for yourself that, that doesn't have like a ton of VC funding. Like you don't have to go big or I kind of go, go home.

Yeah.

You have a good lifestyle. Yeah, I just.

Yeah, I've done a lot. That's probably another episode. Talk about like building businesses, but I think building a business, like when you're picking a business, a business picture you. Because the business comes with things out of the box that you probably can't control. If you start an agency you have to have clients, you have to have contracts, you have to have like project based work like that comes out of the box. So don't pick an agency if you don't want to do that stuff. If you want to do other things, if you want to do a tutorial company the way I've done it, you could pick. It's a one to many situation. You can charge lifetime if you want to. Like, you can do all of these things. There's nothing that is like, there's nothing that comes out of the box of the type of businesses I build that I'm not happy with. I'll only build a business if I'm happy with what comes in that box because it has to suit how I want to live my life. So that's kind of how it just happens. And this is the second time this happened. And we had a small amount of funding for Makefair. We haven't had any funding for Ben's Bytes. And yeah, it's doing much, much better. And who knows where it goes from here.

Awesome, Ben. So I guess it's pretty obvious, but where can people go find the tutorials and go learn more about you?

Yeah, you can sign up for free on BenSbytes.com we've got about 30% of our content is just freely accessible. And then I'm on Twitter or X as it is now with Ben Tossel.

Great, thanks so much. This is a super practical, very honest conversation. Cool.

Yeah, thanks so much.