Latest / The Koerner Office - Business Ideas and Deep Dives with Chris Koerner / What AI Tools Are Worth Your Time in 2025⏐Holdco Bros. Ep. #158
Transcript
- 0:00Over the last two weeks, I've talked to 30 entrepreneurs who
- 0:03are using AI in their business. Because of that, I now
- 0:05understand the use cases that they're using it for ChatGPT as
- 0:09a general thought partner, Claude, coding partner,
- 0:11perplexity research. Everyone I spoke to who actually
- 0:14uses these models, they're like, it's as good as deep research
- 0:16and it's real time Gemini workflows within the Google
- 0:18suite. There's a ton of good ideas, not
- 0:20a lot of good execution. Stay close to the money.
- 0:22Build something that people can use to directly make more money
- 0:26that will have the highest likelihood of succeeding.
- 0:28The number one use case for it right now.
- 0:33You know what I love about ChatGPT?
- 0:35I just realized this and you might have had a different
- 0:38experience and it may change because incentives aren't
- 0:41aligned. It is not biased.
- 0:43Like if I ask it about Republicans or Democrats or the
- 0:47Mormon church or the Catholic Church or seed oils or health or
- 0:50carbs versus fats versus proteins, I really feel like
- 0:53it's unbiased, like it is a superpower.
- 0:56So here's one example. I saw this Instagram real and do
- 1:00you know what ghee is? GH EE No, it's like a, a
- 1:04clarified butter. It's like a different version of
- 1:06butter that we've been using for hundreds of years.
- 1:08It's whole food. It's healthy.
- 1:11It's it's like an alternative to seed oils or canola oil or
- 1:14butter. So I saw this stupid Instagram
- 1:17reel that was saying basically ghee is unhealthy actually and
- 1:21canola oil is healthier than ghee.
- 1:23And I'm like, Oh my gosh, what are we doing?
- 1:25Like there's so much info out there that like literally no one
- 1:28knows what to eat because you don't know who to believe
- 1:30because you don't know who has the incentive to say whatever.
- 1:33So I went to ChatGPT 4 O and I said who paid for this study
- 1:37that says replacing ghee with canola oil is better.
- 1:41So this is what I said who? Who paid for this, right?
- 1:45And it's like great question. Here's the deal with most of
- 1:47those. Canola is better than ghee
- 1:49studies. Many are funded by big food or
- 1:50AG companies. Example studies in the past that
- 1:53have been backed by groups like the Canola Council of Canada,
- 1:56American Heart Association or Unilever.
- 1:59Like literally like Coke and Pepsi are funding the American
- 2:02Heart Association, right? So then they put out these
- 2:05studies that are like Coke is actually good for you.
- 2:06It's like, but like ChatGPT has no bias because there's nothing
- 2:11to click the incentive is there. Whereas if I Google the same
- 2:15thing, like the incentives are for me to click on stuff.
- 2:19So I I still can't believe what I Google, right?
- 2:21As optimistic as I am, I don't think the ChatGPT will stay like
- 2:24this forever because money, right?
- 2:27But I love that it's like this for now.
- 2:29Have you had the same experience?
- 2:31I have but a little bit different.
- 2:33So I think it's pretty unbiased when I'm asking it semi factual
- 2:37questions. But I do think it's biased when
- 2:40I'm asking it for feedback. And this is either it's bias or
- 2:43I'm insecure because whenever I ask her for feedback on
- 2:46something that I'm doing or about me, I feel like it's it's
- 2:51overly praising me if that makes sense.
- 2:55Where I'm like hey take a look at this post I wrote.
- 2:57What do you think this post is Fire fire emoji by the way, open
- 3:00AI. Can we cut the emojis?
- 3:02Like what's happening? There's so many freaking emojis.
- 3:05I hate them. Anyways, this post is fire.
- 3:07And I'm like, how would you rate it?
- 3:08I would rate it a 9.5 out of 10. And it's like, OK.
- 3:12And then I have to go back to it.
- 3:13I'm like, you're a neutral third party.
- 3:15Your job is, you know, content. Managing Seeing.
- 3:19Everything. It wants you to hear.
- 3:21Right, right, right. So right.
- 3:23To me, that's where I am seeing bias is like it says what it
- 3:27thinks that I want to hear. And there are times where like
- 3:30it's a good point. I want it to be totally unbiased
- 3:32and like give me an actual opinion of what it thinks of my
- 3:34business plan or this post or the argument that I'm making
- 3:37with my lawyer. You know what I mean?
- 3:39So anyways that that's the only piece that I would say I'm I do
- 3:42see it biased. It's funny because I'm I'm going
- 3:44to directly contradict myself and it's like it's not that it's
- 3:47unbiased, it's just biased in a different way there.
- 3:49There's always incentives. It's impossible to remove
- 3:51incentives. Opening eyes incentive is to
- 3:54keep us using jet GPT, right? So it's like the bias is not
- 3:58clicked on this study. It's like make sure Nick never
- 4:01churns, right? We want to keep Nick happy, you
- 4:03know? Right, I have a custom GPT where
- 4:07in the GPT prompt I'm like, you're a hard nosed, no nonsense
- 4:12content specialist. I need you to give me direct
- 4:14advice because bad advice is terrible.
- 4:17I I give an analogy about being on the war field, war field.
- 4:21I'm so stupid being on the being on the battlefield.
- 4:25And you know, field generals need to have not only real time
- 4:30information, but accurate information to make good
- 4:31decisions. Otherwise bad things happen.
- 4:33So like I give it that analogy like This is why I need you to
- 4:36give me good information because I need to make good decisions.
- 4:39And so that GPT is great because it gives me sort of unfiltered
- 4:43responses, but it's, it's when I'm just like trying to ask it a
- 4:46general like question that it's definitely biased.
- 4:49I'm kind of debating if I should say this part next part or not
- 4:53Say it. Say I don't think I want to, I
- 4:57think I want to. Relationship advice?
- 5:00Have you like with your spouse? Have you used it for that?
- 5:04No, but like what have you used it for, dude?
- 5:10Like I'm going to say this. You and I have talked about we
- 5:15each have superpowers. When you have a superpower, you
- 5:19have to know when to use it and when not to use it.
- 5:22And you also have to recognize that you can use your superpower
- 5:25for evil. To me, the reasoning that chat
- 5:29TPT has like let's say you have an argument with your spouse and
- 5:31you're trying to get your way where it can help you frame an
- 5:35argument is is out of this world.
- 5:40And then it's a very unfair fight because.
- 5:43You you're not citing to your wife, ChatGPT, according to No,
- 5:47you're not. Doing no, you're not citing.
- 5:48You're like, well, that's an interesting point, honey,
- 5:50because I would think that you would want to do XY and Z based
- 5:53on your previous, you know what I mean?
- 5:54And she's like, well, that's yeah, that's a really good
- 5:56point. And meanwhile, you're like
- 5:58looking down at your phone. You're like, she said it was a
- 6:00good point. What do I do now?
- 6:03What? This has never happened.
- 6:04She she said I'm smart. What's happening?
- 6:07Dude, I have been very impressed with how Chad CPT is handled
- 6:13like relationship advice with my spouse and I've, I've never,
- 6:17we've never seen like a marriage counselor or a therapist or
- 6:21really talk to anyone like we've got a good marriage, right?
- 6:24But I've just been asking it some things and it has been so
- 6:28helpful and it I feel like it's advice that like a very
- 6:31expensive, very, very good marriage counselor would give.
- 6:35And it's literally made our marriage better.
- 6:38And she has no idea. She has no idea.
- 6:42But it's the perfect use case for it, right?
- 6:43And if you're like, if you're trying to come at it from a
- 6:46genuine way and you're like, look, I'm going to try to give
- 6:48you all the facts that I can help me understand it, then
- 6:51you're getting like this impartial third party.
- 6:53And you're like, well, it's AI. So like if AI thinks I'm being a
- 6:56prick, yeah, I guess I am being a prick about that, right?
- 6:59Or vice versa. And it's, it's pretty incredible
- 7:03with the way if you give it enough context, right?
- 7:05Because you and I might have a disagreement, I might give it
- 7:08all the context in the world and just be like, yeah, Nick, you're
- 7:10right. And then you might have the same
- 7:12disagreement. But.
- 7:12It doesn't want Nick to churn. It doesn't want Chris to churn
- 7:15well. It doesn't want, it doesn't want
- 7:16me to turn. But also I'm just giving it my
- 7:18context. Yeah.
- 7:19And then you do the same thing. Your side of your story, right?
- 7:22And like, if you feel it biased info, it's going to give you
- 7:24biased info. Yeah.
- 7:26The trick is I would like to start using freaking out.
- 7:29A way to use this for mediation in business disputes, I think
- 7:35would be really fascinating if everybody just had if everybody
- 7:38just had the the. Limitless.
- 7:42Which I ordered by the way. And it's like, it's like, OK,
- 7:46upload all your limitless data. Now you'd have to have two
- 7:49parties who are like, OK with everything being uploaded, but
- 7:53that in and of itself would be a tell right?
- 7:54Where you're like, oh, this guy doesn't want it uploaded.
- 7:56What's going on? But it's like, all right, take
- 7:58all of the side of their story, enter all the information.
- 8:01Because that's essentially what an arbitrator is doing anyways.
- 8:04Yeah, that's really good. You just create a a fresh
- 8:08ChatGPT account and the two people could sit side by side.
- 8:11So here's my side of the story. Here's my side of the story.
- 8:14And then they both agree like what the prompt is, right?
- 8:17And it's it's neither one of their accounts.
- 8:19And you just hit enter and see what happens.
- 8:21Now the hard part would probably be agreeing what the prompt is.
- 8:25Right. Yeah.
- 8:26And just like any other mediation, the hard part would
- 8:27be actually agreeing on the output, because even if we're in
- 8:30a mediation and it sides with you, I'm going to be like, what?
- 8:33Well, I didn't. It's stupid, doesn't know what
- 8:34it's talking about. Yeah.
- 8:36But I do think if you're a mediator and your business is
- 8:39mediation, for those types of businesses, your bottleneck is
- 8:42you. It's your time essentially
- 8:44because you're an agency and you can only bill as much as you can
- 8:47actually go out and meet with clients.
- 8:48If you could use chat TPT to build a wrapper or to build a
- 8:52prompting system to get that information quicker so that you
- 8:56save an hour per mediation like then all of a sudden you can
- 8:59take on 20 more clients a month like you just you just grew your
- 9:03business in a way that's AI enabled.
- 9:05That would be a really, really cool use case.
- 9:07That is cool. You and I talked about this
- 9:09idea, I don't know, a couple weeks ago where it's just like
- 9:11buy limitless pendant and give it to somebody almost like a
- 9:15lead magnet to get into their business.
- 9:17And you're already doing this because you're recording
- 9:18everything. At some point, the ability of
- 9:22these models to digest more data is going to happen.
- 9:26So what's what becomes the limiting factor and how well
- 9:29these models will get to know you and be your own personal
- 9:32assistant. The limiting factor will be
- 9:34data. Contacts window.
- 9:35Power, yeah, or, or the ability to process or, you know what I
- 9:40mean? Or the ability, because what's
- 9:42happening right now is it may have this massive data set, but
- 9:44it's, it's optimizing for speed. And so it's only looking for
- 9:47certain whatever. At some point it's going to get
- 9:50way better and the the limiting factor is going to be the amount
- 9:52of data that you have to input it into the system for it to
- 9:55then tailor itself to you. So I don't have a business model
- 9:59off of this. I just think using ChatGPT,
- 10:02using a limitless pendant where you're recording every
- 10:04conversation that you have, it just creates more and more and
- 10:07more data so that when it's there, it's fed in and you're
- 10:11accelerated even faster. Yeah.
- 10:13Can we barbell this for a minute and go offline?
- 10:15Yeah. So you see my tweet about the
- 10:18podcast Billboard. Yeah, putting up.
- 10:22So that's been a lot of fun. It won't go live for like two or
- 10:25three weeks, but I used AI to generate a bunch of phrases,
- 10:30right? So you saw that I shared my
- 10:31screen. I put the podcast art on the
- 10:33left, Kerner office in the middle, icons to YouTube,
- 10:36Spotify, Apple podcast, and then like a catchy slogan at the top.
- 10:40So I had Simon make 34 of these. They all look the exact same
- 10:44except for the catchy slogan at the top is all different, 2 to 5
- 10:47words. Like you only have so much time
- 10:49to read a billboard, right? And so the catchy slogans that
- 10:53I'm testing are your dream is dying.
- 10:58Sell stuff, not hours 9 to fives.
- 11:00Kill dreams. Own your time.
- 11:02Never job again jobs or jail. Delete LinkedIn work sucks.
- 11:06Question mark escape the grind fire your boss still rat racing
- 11:11question mark. And what was the last one?
- 11:14But then there was Best Business Podcast and then Quit your job
- 11:19with an exclamation point. And then there's a bunch of
- 11:21others. But I started Facebook ads that
- 11:24are all the exact same. It's just like the Kerner Office
- 11:27Podcast title, text description, They all say the same thing.
- 11:31The Kerner Office podcast, the audience, the four counties in
- 11:35and around where the Billboard will be.
- 11:37And then I exclude anyone that's ever interacted with my content,
- 11:41so they don't follow me. They've never seen my video, and
- 11:43they live in those counties because that's bias.
- 11:45The people driving by don't know who I am.
- 11:47Right, right, right, right. And then I just run the ads on
- 11:51any Meta platform. Meta thinks messenger,
- 11:55marketplace, Instagram, it chooses.
- 11:57And then I just look at the clicks lowest cost per click
- 12:01wins. And then I get because it's 950
- 12:03bucks to print that billboard. The image that I'm using on on
- 12:07the Facebook ad is literally a picture of the Billboard with
- 12:10like trees in the background instead of just the image, you
- 12:13know? And it's 18 plus any gender in
- 12:16those counties that don't know who I am.
- 12:18What do you think? I think it's amazing.
- 12:20I have an idea for it as well. I know we're offline, but I'm
- 12:23going to take it back to AII. Used AI to generate most of
- 12:27those phrases by the way. What I like to do with AI when
- 12:30I'm like creating a title for a video is I'll be like generate
- 12:33me the 10 best titles and rank them, and then I'll like iterate
- 12:37from there. I would take all of your
- 12:39pictures and I would go to chat, TPT, Grok, Claude, Gemini, four
- 12:46of them, upload them and say rank these from what you think
- 12:51would be most to least effective.
- 12:54See what the four models spit out and then compare it to the
- 12:56actual results from Facebook. Yeah, I'm.
- 12:59Just going to deal with the phrases because all the pictures
- 13:01are the same. Oh, OK.
- 13:03Yeah, yeah. So like, I think that would be a
- 13:06really, really cool case study of like, all right, I'm, I'm
- 13:11trying it with real humans. This AI has been trained on the
- 13:14totality of human whatever experience, which one of them
- 13:18wins or which, how accurate is it?
- 13:20Because then in the future, if it's like, oh, they were, they
- 13:23all said the same 5. And those were actually the five
- 13:26that one, it might save you time, you know, running Facebook
- 13:30ads in the future. Yeah, I like that.
- 13:33It's interesting. I I didn't exclude my audience
- 13:35at 1st and I started running the ads and the winning one was Best
- 13:39Business Podcast. And that surprised me.
- 13:42And I was like, I wonder if that's because like, they
- 13:44recognize me, they see me right? And I don't yet know.
- 13:47It's still too early, but it just made me think like how much
- 13:50power there is in marketing to call your thing the best.
- 13:53I had Greg Eisenberg on a couple weeks ago.
- 13:55His story was he was walking around Miami.
- 13:57He had just moved there. He walks by a dentist and he
- 14:00sees a sign on the window that says voted best dentist by who?
- 14:04What? I don't know, I don't care.
- 14:06He walks in because he saw that sign, became a customer, been a
- 14:11customer for eight years, given them 10s of thousands of
- 14:13dollars, referred 20 friends there because he saw the sign.
- 14:16So his idea was like, there needs to be some sort of like a
- 14:18marketplace where you could get like certified by whoever.
- 14:22It's a racket. It doesn't matter.
- 14:23You pay 200 bucks, you get a sign.
- 14:26And so it's like, I just put up this billboard.
- 14:27Yeah, I think it's the best. My mom thinks it's the best
- 14:30whatever, Best business podcast, according to Chris's mom.
- 14:33Fine print. Asterisk, right?
- 14:35It doesn't matter. So maybe people clicked that the
- 14:38most because they knew who I was and I'm going to find out that's
- 14:40actually bad data. Maybe not.
- 14:42Maybe it's the winner again and people just it's confirmation
- 14:46bias. They're like, I believe you,
- 14:48Internet stranger. I will now click and listen
- 14:50because you said it's the best. Hearing you talk about Greg
- 14:53Eisenberg makes me think of that meme that's like Internet guy
- 14:55discovers rating agency. It's like these these these
- 14:59things exist, right? Like best business, like you can
- 15:02talk. They're already businesses in it
- 15:04of themselves. But yeah, this cracks me up.
- 15:06Well, there needs to be like for SAS companies or for AI
- 15:10wrappers, like you could do a rating agency for any number of
- 15:12things, right? Yeah, I.
- 15:14Mean that's what no BS crypto was when you were launching it
- 15:16right, it's like, hey, this is you were a rating agency.
- 15:20I don't know looks like we made it.
- 15:25I just like that idea a lot because you're like, you're
- 15:28currently competing. All of the tools, all of the
- 15:31advice. If you go and look at hey, how
- 15:32do I grow a podcast is do it on someone else's newsletter.
- 15:35Go do paid ads. Go to.
- 15:37Here on other podcasts. Brokers go to other podcasts,
- 15:39right? It's like, OK, well, everybody
- 15:40else is doing the same thing. They're all doing the same exact
- 15:43thing. Yep.
- 15:44And that we know that's a strategy, Zig, while everybody's
- 15:46acting. Yeah, So like, either this is
- 15:49going to be a total waste of money and I'm prepared for that
- 15:51to happen because I don't see other podcasts, billboards
- 15:53probably for a reason, or I'm going to unlock something really
- 15:56cool. Right?
- 15:57Right. Aren't you going to ask me how
- 15:58I'm going to track if I'm getting new downloads from this?
- 16:00Nick? How?
- 16:01Are you going to track? Softball me here set me up.
- 16:04Chris, let me ask you a question.
- 16:05Like, I get it, but how do you freaking track that?
- 16:08It's on a. Billboard probably like AQR
- 16:11code, right? No, QR codes are not allowed by
- 16:14Clear Channel, the company I'm using for this.
- 16:17OK, OK, so Buzzsprout is the hosting platform that we use.
- 16:22You can export your downloads by location, by every little city.
- 16:27So like when I export mine, there's like 8500 cities that
- 16:30have listened to my podcast sometimes once, sometimes 1000
- 16:32times. Right now, what you don't know
- 16:34is how many unique listeners there are per city.
- 16:37You just know how many downloads.
- 16:38So 25 downloads could mean one person downloading your podcast.
- 16:4225 episodes of it could mean 25 different people once each,
- 16:45right? You don't know.
- 16:47So what I did was I went to the Billboard will be in Tomball,
- 16:50TX, which is a city of 10,000 just north of Houston.
- 16:53I took Tomball, The Woodlands Spring, all the cities and
- 16:57Houston itself in and around Tomball and I exported every
- 17:02city by download. And then I found a percentage,
- 17:04like a ratio. How many downloads have I had
- 17:07all time that were in Tomball and or Houston Point O1, 2% of
- 17:11my downloads. And that's been consistent.
- 17:14And so I'm just going to run that export every week and see
- 17:17the raw number and the percentage as a ratio and see if
- 17:20it's climbing or not. Do you see any any fault in
- 17:23that? I do see a couple of faults,
- 17:25like you don't necessarily know where it is as a percentage over
- 17:28time that as it's mapped. Well, I'm going to run the
- 17:32export every week because that's the really the only way with
- 17:35Buzz Sprout you can do it. The other problem would be it's
- 17:39just all attribution related, right?
- 17:40Like you're looking at this geography or did you capture the
- 17:42right cities? What I would do, this is just a
- 17:45suggestion, is I would ask chat TPT the best way to track this
- 17:49and the best way to analyze it. You can throw data into chat TPT
- 17:54and then like ask it to run a regression model and the
- 17:57regression model will tell you whether or not that's
- 18:00statistically significant or not.
- 18:02Because you could just have, you could just blow up over the next
- 18:06month. Let's say you have like a couple
- 18:07episodes that are amazing and it's like everybody saw a bump,
- 18:10right? So of course, Turnbull would see
- 18:12a bump. It would be proportional in that
- 18:14sense, right? In that instance I would have to
- 18:16look at the percentage, not the the raw number.
- 18:19You have so many downloads, you have so much data that it would
- 18:22be easy to say with confidence whether or not it's
- 18:26statistically significant. And you just I would just ask
- 18:28chat TPT. No matter what I do, it won't be
- 18:30exact. But no.
- 18:31But I think it's. Amazing.
- 18:32It would be too coincidental if a disproportionate amount of
- 18:35people in that part of Houston suddenly started downloading my
- 18:38podcast, right? Right.
- 18:39What I what I wish I knew is how many?
- 18:42Because you just don't know how many raw people it is.
- 18:45How many loyal listeners? Because there could be 100
- 18:47people that listen to 5 seconds of it once and then bounce and
- 18:51it could look really good but not actually be good.
- 18:54Why didn't you do the here's the website and then have like one
- 18:58letter off? Right.
- 19:01I just don't want to mess with like URLs.
- 19:03I, I kind of want to go more for brand awareness.
- 19:05That's why I put the image of the podcast and one of them has
- 19:08TKO pod.com in it. I just want people to search for
- 19:11it. Another thing you can do is you
- 19:13can put up Facebook ads based on a latitude and longitude point
- 19:16of the exact billboard and it will show your ad to people that
- 19:19drive by that billboard. And so then you get double
- 19:23exposure, triple exposure, the Billboard even more effective,
- 19:27but it will make it even harder to find the attribution because
- 19:30you're adding another variable, right?
- 19:32At that point, it's like, OK, I'll spend $780 for a billboard
- 19:36and I'll run $1000 with the Facebook and, you know, Facebook
- 19:39ads. Who cares about the attribution?
- 19:41I know it works. You know what I mean?
- 19:42Yeah, cuz any Joe Schmo can run a Facebook ad, but you don't see
- 19:45a lot of billboards. It just adds another layer of
- 19:47legitimacy to it, right? Yeah, and that feeling of like,
- 19:49dude, I'm seeing this guy everywhere.
- 19:51Right. I kind of had the thought, like,
- 19:53wouldn't it be interesting if one of us, anyone, ran a test
- 19:56where they took like a random city in the middle of nowhere
- 20:00like Sioux Falls, Iowa? And they ran Facebook ads like
- 20:05saturated for your podcast, your brand, whatever, just in that
- 20:09city. And like you just became a
- 20:11freaking rock star in that one city.
- 20:14And you just put them everywhere.
- 20:15Instagram ads, TikTok ads everywhere, only Geo fenced to
- 20:18that city and just, I don't know, go speak have an event
- 20:22there. Like it could be an interesting
- 20:24experiment. Like no one else in the world
- 20:25knows who you are, except for Sioux Falls.
- 20:27What if you went to chat TBT and you said, I want you to run a
- 20:31contagion analysis or model only of cities less than 1,000,000
- 20:37people in the United States and tell me which city would have
- 20:41the highest likelihood of spreading this contagion the
- 20:44fastest. So it's like, you know, it's
- 20:47going to go look at Des Moines like, well, Des Moines actually
- 20:49has the most connecting flights of any airport in the United
- 20:53States, so. Contain them, right?
- 20:55I'm just like, yeah, yeah. Yeah, it's going to.
- 20:56And then it's like, OK, cool, I'm going to go target Grand
- 20:59Junction Co because it says that in of any city under a million
- 21:02people, that's the one that would have the like worst
- 21:05contagion, and that's the city I'm going to go target.
- 21:07So they're like Vegas, like a city, Orlando, where people are
- 21:10coming in and out of all the time, right?
- 21:12Like New York City has 20 million people anyway, right?
- 21:15People are coming in and out, but proportionally speaking,
- 21:18there's just a lot of people there.
- 21:20But for New York City, it's like you're or LA or San Francisco,
- 21:24there's just a lot of competitors.
- 21:26So standing out is going to be hard, generally speaking, right?
- 21:29But a smaller market like Orlando, I don't know, you
- 21:32probably have a better chance of standing out.
- 21:34That'd be a cool idea. So video is cool, but you know
- 21:37what's better? Long form audio via podcast and
- 21:40my newsletter, TKO pod.com. Go there to subscribe for free
- 21:44to my newsletter. It's one e-mail a week, very
- 21:46tactical. And then go to my audio podcast,
- 21:483 episodes a week. Stuff like this.
- 21:49You're going to love it all free, no sleazy sales bitch.
- 21:52TKO pod.com. So over the last two weeks, I've
- 21:55talked to 30 entrepreneurs who are using AI in their business
- 22:00because I feel like you and I have talked about this.
- 22:02There's a ton of good ideas, not a lot of good execution like
- 22:05Replit. How many things have you made on
- 22:07Replit that you haven't actually finished?
- 22:097. So it's like, oh, this is cool,
- 22:11I can build an app, this is great.
- 22:12And then you get to the final point.
- 22:13You're like, oh, how do I actually complete it?
- 22:16It's like, it's that last 10%. So I've been having
- 22:18conversations with people and it's been awesome.
- 22:22I've learned so much. I like it.
- 22:23My, my brain hurts every single day because I'm learning so
- 22:26much. But there are some really cool
- 22:28insights that I've gotten out of this.
- 22:30So I'm going to share some of my findings with you because I had
- 22:32chat TPT take all my findings and put them in the summary for
- 22:35me. Question #1 Chris, what do you
- 22:39think is the most popular LLM for people to use as their
- 22:45primary LLM? ChatGPT.
- 22:48It is. What do you think is the most
- 22:50popular use case? Like what are they using ChatGPT
- 22:53for? Writing emails like rewording
- 22:56emails. Mainly they're using it as a
- 22:58thought partner. So it could be rewriting emails
- 23:01generally not that it's usually like, hey, this is I have an
- 23:03idea for this. How would I do XY and Z thing?
- 23:05And it's like just to get them started, which was fascinating
- 23:09to me. Second most used LLM.
- 23:11What do you think it? Is Gemini.
- 23:13No. Grok.
- 23:15Nope. Perplexity.
- 23:17Nope. Claude.
- 23:18Claude. Dang it, Claude.
- 23:20And the surprising thing about Claude is it's actually used
- 23:23more in aggregate because I asked them two questions, like
- 23:25what's the primary? Like what's your primary
- 23:27favorite? And then what are the other LLMS
- 23:30you use? And I've mapped it all out.
- 23:32Clod is actually used more by more people than ChatGPT is, but
- 23:37ChatGPT is the most preferred, if that makes sense.
- 23:40OK, what do you think that they're using Clod for?
- 23:44I have no idea. That's like the only one I
- 23:46really hardly ever use. So I have it broken down in
- 23:50three tiers. non-technical, technical, intermediate, and
- 23:55then expert. For chat TPT the majority of the
- 23:59distribution is in the non-technical and intermediate.
- 24:02For Claude, the majority of the distribution is an expert and
- 24:06intermediate. And the main use case that I see
- 24:08people use it for is coding. It's the best coding tool like
- 24:15to help them actually write code.
- 24:17And then the second best thing. This one blew my mind.
- 24:21It's organization structure is fantastic.
- 24:23So you know how ChatGPT has projects clod you can set up
- 24:27projects, but it you have two fields, 1 is what is this
- 24:31project about? And so you can write a prompt to
- 24:33say like this is my content project.
- 24:36And then it has files that you can upload for context.
- 24:40How's that different from ChatGPT?
- 24:42From what I've seen, ChatGPT doesn't have the same
- 24:45functionality with the uploading of files and the reference for
- 24:49each one of the threads. Maybe I'm just not using chat
- 24:52TPT. That's how I use it.
- 24:54So you think you upload instructions?
- 24:56They call it files and instructions.
- 24:58And so you use individual files for each one of your projects.
- 25:02Yeah, I don't always use files, but it can.
- 25:05One of the people I spoke to, she's a writer and she's like,
- 25:08Claude does a much better. It's a much better writer,
- 25:11apparently. And she has this whole setup
- 25:13where it's like, this is my brand voice here, recent
- 25:17projects that I've done. And so she's like constantly
- 25:19changing the context so that it's up to date.
- 25:22But the interesting thing is, I didn't know this.
- 25:25You can only upload however big the context window is.
- 25:29Does that make sense what I'm saying?
- 25:30Yeah. So that was interesting.
- 25:32What was the least mentioned tool?
- 25:36Well, I would say grok, but I think that a lot of these calls
- 25:39came from people on Twitter, so maybe Gemini?
- 25:43Gemini, yeah, it was the least mentioned and everybody like
- 25:47dogged on it. They're like Gemini is not good.
- 25:50So that was that was just really interesting to me.
- 25:52That's interesting. Are they dogging on it because
- 25:54they're they've actually given it a fair shake and they don't
- 25:57like it or they're just it kind of has that brand around it
- 26:01because I dog on it, but I don't really use it.
- 26:03So it could be great. They're dogging on it.
- 26:05Not that it sucks. They're dogging on it because
- 26:08it's not as good as what else they use, if that makes sense.
- 26:11They're like the Claude model, ChatGPT, and I think ChatGPT
- 26:16just has a head start because so many people were using it.
- 26:19The two use cases that I found out about Gemini are if you're
- 26:23trying to create workflows within the Google suite.
- 26:27Gemini's the best because it just knows Google so well.
- 26:30Makes sense? I thought that was that was
- 26:32really interesting. And then the second thing that I
- 26:33thought was really interesting is the context window.
- 26:35I didn't realize this, but do you know how big the context
- 26:37window is for Gemini? I looked this up a while ago but
- 26:40I can't remember. I think Claude was the biggest
- 26:42that I saw. Let's back up.
- 26:43The context window is essentially how much information
- 26:45you can give the model for context of a specific question
- 26:49that you're asking it. For ChatGPT, it's 120,000
- 26:51tokens. And with that is it's like a 300
- 26:53page book, OK? For Claude, it's 200,000 tokens,
- 26:58so it's a lot bigger than ChatGPT.
- 27:00For Gemini, it's a million. Wow.
- 27:03Yeah. So it's a much larger context
- 27:05window. I thought that was just really,
- 27:07really interesting. Next finding what do you think
- 27:10was the most used tool that that interfaced with the LLMS that
- 27:16they work with? Like feature of the specific
- 27:19LLM. No, let me ask it a different
- 27:21way. What do you think is the most
- 27:23used AI tool that was not an LLM?
- 27:27Something that I would have heard of.
- 27:28Yeah. I have no idea.
- 27:30Cursor OK cursor and from several of them that like
- 27:34multiple people said, I think cursor is going to be bigger
- 27:36than Google. They were like, so, yeah, they
- 27:40they were just like, so in love. Used to talk to some coders
- 27:42then. I did.
- 27:43I did. I talked to a couple of coders.
- 27:45It blew my mind though. Like on the call with them, they
- 27:48were on Cursor and they were showing me you can natural
- 27:52language prompt into building an actual website.
- 27:55All of them who had like actual coding knowledge crapped on
- 27:58Replit. Yeah, which was really
- 28:00interesting to me. They're like Replit.
- 28:03Well, but I'm like, yeah, in order to use Cursor, you
- 28:07actually kind of have to have some coding experience.
- 28:09Like you can still you can use it at using natural language,
- 28:12but it would help if you had coding experience.
- 28:14Here's my push pull framework for for this Friday usability
- 28:18and functionality. That's the push pull of today.
- 28:23Repla is more approachable to more people, but it's less
- 28:26functional as like a coding assistant.
- 28:29Whereas cursor is harder to use for more people, but it's a more
- 28:32functional coding assistant. And the magic happens like the
- 28:36Unicorn status happens when you have like an Airbnb with
- 28:41beautiful design, beautiful user interface, like easy for anyone
- 28:45on the planet to use, yet so freaking functional.
- 28:49It just works. Everything just works.
- 28:51And it looks simple. Like from the outside looking
- 28:53in, you assume that it's very simply built, but on the back
- 28:56end, it is just a hairy mess of complexity and it's just really
- 29:00good engineering and design, right?
- 29:02But most companies, 99% of companies have one or the other.
- 29:05They have good UI or they have good functionality, not both.
- 29:09I agree. So I'm going to go through
- 29:11generally, broadly speaking, this is what I found to be the
- 29:14use cases. ChatGPT as a general thought
- 29:18partner, like someone they're collaborating with and like
- 29:20trying to figure out ideas. Clawed coding in particular,
- 29:24It's like that's their coding partner perplexity research.
- 29:28Everyone I spoke to who actually uses these models, they're like,
- 29:31it's as good as deep research and it's real time.
- 29:33So it can go and crawl the web in real time, which fascinating
- 29:37in and of itself. Grok.
- 29:39There was no like best use case, but it just seemed like
- 29:43generally speaking, everybody was impressed with Grok, the
- 29:46speed and responsiveness. I don't know if you've messed
- 29:47with Grok. Like it spits answers out very
- 29:51quickly. They're very good, has access to
- 29:53real time data. It's everybody's like, it's
- 29:56amazing what they've they've built a a top tier model in such
- 29:59a short period of time. And then Gemini is just, hey,
- 30:03it's really good for the Google suite.
- 30:04Those were like the general use cases that I'm seeing people
- 30:07actually use it. And I thought that was those
- 30:10things were fascinating. Can I give you my personal
- 30:12takeaways from what you just said?
- 30:15I need to be thinking of using Gemini more since I'm so
- 30:19embedded in the Google workspace, right?
- 30:21I just don't think of it. We've talked about notebook LMA
- 30:23bunch. I like the contact window size.
- 30:26I just don't like what it spits out.
- 30:28I don't like what it gives me right?
- 30:30So who cares how big the contacts window is if it's not
- 30:33working with it? Well, in my opinion, I'm using
- 30:36Perplexity more and more every day.
- 30:38As of today, I don't see a need to keep my $200 a month Pro
- 30:42subscription because Perplexity does deep research, which is the
- 30:45only thing I'm using Pro for now.
- 30:47I still like a second layer of redundancy.
- 30:49I still take my prompts, my deep research prompts and I put them
- 30:53in deep Research and Perplexity at the same time because it does
- 30:57give me different answers and I I still kind of like deep
- 31:01research a little better. I think it's more thorough.
- 31:04It gives me a longer answer with good relevant info.
- 31:07Have you tried Grok's Deep Research?
- 31:09I've barely even tested it. I can't really speak to it.
- 31:12It's. Very, very good.
- 31:14Very good. Surprisingly good.
- 31:16Actually I want to make a a YouTube video where I I take the
- 31:19same prompts and put them in all three for specific for deep
- 31:22research. So curse you think you need to
- 31:24use cursor more. I need to use cursor more and I
- 31:27need to use Rapplett less just because I'm I'm not able to wrap
- 31:30up these projects. This is one of the things that
- 31:33been fascinating to me is like, I wanted to learn how to do
- 31:35these things. So I did a project and have you
- 31:40ever heard of, I hate bringing these things up to you because I
- 31:43feel like a high schooler who's in chemistry one O 1 or whatever
- 31:47talking to a PhD chemist and it's like, have you ever heard
- 31:50of gravity? And the chemist?
- 31:53You did ask me what a lead magnet was a while back.
- 31:55That was a fence. Yeah, his offense.
- 31:57Have you ever worked in Google collab?
- 31:59Nick, you, it's a good thing you ask, because no, I haven't, OK,
- 32:02I haven't. I want to be like no I don't.
- 32:05OK, genuine. So I have all of these
- 32:07transcripts from my podcasts and specifically I just wanted to
- 32:11focus on the AI podcasts. What I wanted to do, and you
- 32:14know how this works, is I want to take all the conversations,
- 32:17aggregate them, and then be able to pull out information.
- 32:20OK, but how do you do that? The context windows on only
- 32:23120,000 tokens and like four of those transcripts are more than
- 32:28120,000. It's like it doesn't work.
- 32:31So I needed to clean the transcripts, get rid of all the
- 32:35Dead Space, oz timestamps, etcetera, and then get them into
- 32:39a format where they would all be combined and I had them all
- 32:42saved. Yeah, what about just do the
- 32:45transcripts identify who's speaking?
- 32:47Yes, I kept the speaker. OK.
- 32:49I did keep the speaker because I wanted to know by person and
- 32:52I'll show you the output that I got, but I wanted to know by
- 32:55speaker who's talking and what are they saying and why?
- 32:59Anyways, I don't know what to do and what do I do?
- 33:02I literally just go to chat TPT and I'm like, hey, this is what
- 33:04I want to do. I want to take all these
- 33:06transcripts. I want to combine them.
- 33:07I don't want to clean them. And it's like, OK, where are
- 33:09they stored other than saying Google?
- 33:11OK, perfect. You're going to go to Google
- 33:12collab. Great, Open a new project,
- 33:15attach your your drive. Great, I did that and run this
- 33:18script. So Google collab is an
- 33:20environment where you can run Python script for it to actually
- 33:22execute a function. So I'm like, OK, so it gives me
- 33:26the Python script. First couple times it doesn't
- 33:27work. Then I go to Claude.
- 33:28Claude fixes my Python script and it starts working.
- 33:32The first iteration suck, second iteration suck.
- 33:36But by the, you know, a couple hours later, I've got some
- 33:41really, really good data. In fact, I'm going to share my
- 33:43screen with you again. It took time for me to get to
- 33:47this place. Once it was spit out, I took it
- 33:50back to chat TPT and I was like hey please clean up this data
- 33:55for me, make it look presentable, etcetera and this
- 34:00is what it spit out. So now I have the episode with
- 34:02the person OK interview name, their business type, the
- 34:07industry, their technical knowledge, what their primary
- 34:10LLM was that they used. Oh, this is cool.
- 34:13The main tasks in use cases. This looks like in lead magnet
- 34:16to me. Nick, it's.
- 34:19On that motivation for using these LLMS, Strengths,
- 34:22weaknesses, the ideal user profile, and then the key
- 34:25insights. And like the key insight, I was
- 34:27like, hey, write me a 500 character summary of this
- 34:31conversation. But then I had to do a couple of
- 34:34other things. Then I had to actually go to
- 34:37Zapier and create a ZAP so that it would interface with open AI.
- 34:41And then I learned I could actually just attach the API.
- 34:44I'm going like real technical here, but whatever the API key
- 34:47into Google collab. So I didn't need Zapier anymore.
- 34:49But because of that, I now have this amazing spreadsheet with
- 34:54tons of conversations with entrepreneurs who are actually
- 34:56using AI and I understand the use cases that they're using it
- 35:00for. So now I can go to this and just
- 35:02ask it questions or I don't know, make a quiz that says or
- 35:07tell Chatty BT, which I've been working on, create a quiz for
- 35:10me. I'm a business owner and I want
- 35:12to know which one of these models I should use.
- 35:14So create a quiz that I can decide which model to use.
- 35:16And it uses this information to create a quiz for somebody so
- 35:19they know which model to use. Dude, it's been just so
- 35:21fascinating to learn from these people and see how much
- 35:25application there actually is, even in the limited state that
- 35:28we currently see it. That's amazing.
- 35:30I love that you did that because we know that I hate helping on
- 35:33calls with people. So you're able to do it and
- 35:35enjoy it and then relay what you learn.
- 35:37Yeah. The number one use case for it
- 35:40right now, sales. For AI in general, was there a
- 35:44selection bias? Did you happen to talk to a
- 35:46bunch of sales guys or? No, they were all either owners
- 35:49or coders and they were like, I'm using it for my sales team
- 35:54to follow up with them. I'm using it to track leads.
- 35:56I'm using it as an after hours phone service.
- 35:58I'm using it as a chat bot for customers.
- 36:02I'm using it for my sales team to prepare for calls.
- 36:04Like it was just all sales related.
- 36:06I thought that was really interesting.
- 36:07You know what framework that is? No what?
- 36:10When you're launching products or services, stay close to the
- 36:13money, right? Stay close to the things that
- 36:15directly generate revenue and you can tie it to revenue
- 36:18easily, right? So let's say you want to start a
- 36:21SAS AB to B SAS app. It's like, oh, let's do HR
- 36:25should we do, you know, administrative assistant?
- 36:27It's like no like build something that people can use to
- 36:31directly make more money and they know that your thing helped
- 36:34them generate more money that will have the highest likelihood
- 36:37of succeeding. Yeah, Yeah.
- 36:39It's a great point and that's exactly what I found.
- 36:41We're talking to these people. It's all sales related all.
- 36:44Right, what do you think? Please share it with a friend
- 36:45and we'll see you next time on the Kerner office.