Ranking 22 Of The Most Popular Ai Tools (Q4 2024 Ai Tier List)

0:00:04 – Model agnostic, ship fast, solve a specific problem.
0:00:06 When people are making the AI startup,
0:00:08 they should be looking at cursor and perplexity.
0:00:09 – It’s pretty damn impressive.
0:00:11 Where would you put that one?
0:00:12 – S tier.
0:00:13 – That’s one of the few that we can totally agree,
0:00:16 like S tier, that’s one of our main go-tos.
0:00:19 One of the most useful AI tools available out there
0:00:20 right now in my opinion.
0:00:23 (upbeat music)
0:00:25 Hey, welcome back to the Next Wave Podcast.
0:00:27 I’m Matt Wolf, I’m here with Nathan Lanz.
0:00:31 And today we’re gonna do a super comprehensive
0:00:33 AI tool breakdown.
0:00:36 In fact, we’re gonna do an AI tools tier list
0:00:39 where Nathan and I breakdown kind of all the AI tools
0:00:42 that we’ve used in the past are using right now,
0:00:44 the most talked about AI tools.
0:00:47 And we’re gonna put them on a tier list from S to F
0:00:50 and really rank where we would put them.
0:00:53 And also talk a little bit about how we’re using each one
0:00:57 and give a little context as to why we’re ranking them
0:00:58 that way.
0:00:59 So you’re probably gonna listen to this.
0:01:01 You’re probably gonna completely disagree
0:01:02 with some of our rankings.
0:01:03 You’re probably gonna be frustrated
0:01:05 that we excluded some of those tools.
0:01:07 Let us know all that stuff in the comments.
0:01:07 We wanna hear it.
0:01:09 We wanna know where you would rank them.
0:01:11 We wanna know what tools we’re missing.
0:01:12 And if you liked this episode,
0:01:13 we’ll probably do more like it.
0:01:15 So let us know that as well.
0:01:17 But without any further ado,
0:01:20 let’s go ahead and just break down this tool list here.
0:01:22 (upbeat music)
0:01:25 – When all your marketing team does is put out fires,
0:01:26 they burn out.
0:01:27 But with HubSpot,
0:01:30 they can achieve their best results without the stress.
0:01:33 Tap into HubSpot’s collection of AI tools,
0:01:36 Breeze to pinpoint leads, capture attention,
0:01:39 and access all your data in one place.
0:01:40 Keep your marketers cool
0:01:42 and your campaign results hotter than ever.
0:01:46 Visit hubspot.com/marketers to learn more.
0:01:48 (upbeat music)
0:01:53 – So if you’re listening to like the audio version
0:01:54 of this podcast,
0:01:57 this might be one that you might appreciate
0:01:59 the video version if you actually wanna see the tier list.
0:02:02 We’ll do our best to talk about the tool
0:02:05 and tell you where we’re putting it on the tier list.
0:02:07 But this is sort of a visual episode.
0:02:08 I’m gonna start all the way on the left.
0:02:11 Let’s start with Grock here.
0:02:14 Where would you put Grock/XAI?
0:02:18 – Yeah, I mean, I hate to not give it like a super high rank
0:02:19 ’cause like obviously I’m a fan of Elon Musk.
0:02:22 I think I’m one of the few people that I know
0:02:25 who thinks that long-term that Grock and XAI
0:02:27 is going to be the main competitor to OpenAI.
0:02:29 Like I think they’re gonna be bigger than Anthropic
0:02:32 is my current opinion, which seems to be controversial.
0:02:34 But right now, I would probably put it as a B.
0:02:35 It keeps getting better.
0:02:38 Like the version two is a lot better
0:02:39 than most people realize.
0:02:42 It’s not as good as ChatGBT or Claude,
0:02:44 but it is nice that you can talk to it
0:02:45 about pretty much anything you want to.
0:02:49 It is less censored and the image generation is pretty good.
0:02:51 The image generation is a lot more free
0:02:55 to make whatever crazy meme you want or anything.
0:02:58 So I think Grock is like a solid B right now.
0:03:01 I think long-term it could be an S.
0:03:05 – Yeah, so for me, just based on like how often I use Grock,
0:03:07 I’d probably put it like down in D.
0:03:10 Like that’s not me saying it’s a bad product.
0:03:11 – Yeah.
0:03:13 – But for me, it’s saying that like,
0:03:16 all right, if I need to go use a Chatbot right now
0:03:18 to get an answer to something,
0:03:20 Grock isn’t gonna be my first choice.
0:03:23 It’s not even gonna be like my top three, right?
0:03:25 Like ChatGBT, Claude, perplexity,
0:03:26 you’re probably gonna be my top three.
0:03:29 Grock would probably fall number four for me
0:03:31 for actually going and using it
0:03:33 to ask questions right now.
0:03:35 – One thing that I’ve noticed that’s kind of surprised me
0:03:37 is I thought it would be a little bit better
0:03:39 at getting real-time data from X somehow.
0:03:40 – Yeah.
0:03:41 – And I don’t think they really nailed that yet.
0:03:43 So that’s one thing that kind of sucks is like,
0:03:44 I went there several times like thinking like,
0:03:47 oh, I can ask it some question about an X user
0:03:50 about my own profile or whatever for some kind of insights.
0:03:53 And it kind of is not that great at it.
0:03:56 So I don’t know, maybe a C, maybe a green C.
0:03:59 – I don’t think I’m gonna keep it in D for myself
0:04:01 because I think that like the biggest redeeming factor
0:04:06 of Grock for me is that it uses the Flux 1.1 Pro model.
0:04:09 I think it’s Flux 1.1.
0:04:12 But, you know, Flux has three models, right?
0:04:13 The AI image generator Flux.
0:04:17 It’s got Flux Snell, it’s got Flux Dev,
0:04:19 and it’s got Flux Pro.
0:04:22 Flux Snell, I believe, is like the free version
0:04:23 that you can use.
0:04:25 Flux Dev is a little bit better model
0:04:28 that you can use with like the API
0:04:31 and it’s improved a little bit above the Snell model.
0:04:33 And then you’ve got the Flux Pro model,
0:04:36 which is actually kind of expensive to use,
0:04:39 but it creates really, really good AI images.
0:04:44 And Grock is actually using the Flux Pro model in there.
0:04:47 So by using Grock, you actually get access
0:04:50 to like one of the better AI image-generated models
0:04:51 is out there.
0:04:55 So for me, I think I’d probably solidly put it as a C,
0:04:57 just because it’s not my main go-to,
0:05:00 but it’s got a great AI image generator in it.
0:05:02 – I do see a lot of memes and stuff on X
0:05:04 that are definitely being generated with Grock.
0:05:06 – Oh yeah.
0:05:09 Anytime you see like an image that has like blood
0:05:12 or like an actual, like famous person in it,
0:05:13 like a lot of the Kamala and Donald Trump
0:05:15 and Elon Musk memes that you’ve seen,
0:05:18 most likely created with Flux,
0:05:20 most likely done inside of Grock.
0:05:23 All right, so next up, let’s go chat GPT.
0:05:26 Where do you, where would you put chat GPT?
0:05:27 – Yes.
0:05:29 – You put chat GPT as S.
0:05:32 – Like I still use chat GPT more than anything else.
0:05:34 – So you’re going to chat GPT more than Claude,
0:05:35 more than Perplexity?
0:05:36 – Yes, yeah, I still am.
0:05:39 At first some of, yeah, I felt like Claude was getting better
0:05:41 in terms of like the projects were better
0:05:43 than custom GPTs and it still is slightly,
0:05:46 but like OpenAI’s already caught up there too,
0:05:49 where like there’s more, like you can put multiple files
0:05:51 into the custom GPT and things like that.
0:05:55 So that’s, I use it daily like for translating with my wife.
0:05:59 I use the voice daily now.
0:06:00 So I use it that way.
0:06:02 I still have a custom GPT that I use,
0:06:06 that I created for helping me write threads on X.
0:06:08 I’ve been using that daily.
0:06:11 And also I have a thing that I use for tracking my,
0:06:13 the calories I eat and the protein I intake every day
0:06:14 for working out.
0:06:15 – Oh yeah.
0:06:16 – I use that every day.
0:06:18 So like, and like I’m sure I could do that stuff in Claude too,
0:06:20 but I’m just finding like the combination of features
0:06:24 that it has, it still has just, it’s my main go-to.
0:06:28 – So for me, it’s not my, it’s not my first go-to.
0:06:30 I’ll probably agree and leave it in S tier
0:06:33 because I do think that chat GPT is like
0:06:36 constantly setting the bar for everything else, right?
0:06:39 Like I feel like I probably go to Claude in perplexity
0:06:42 a little bit more than I go to chat GPT.
0:06:45 But, you know, they’ve got their 01 model,
0:06:48 which is the best reasoning logic model
0:06:51 that exists right now, according to pretty much
0:06:55 any sort of benchmark you look at GPT 4.0
0:06:57 is kind of leading the pack.
0:07:01 You know, almost anytime you see a new model come out,
0:07:04 you know, Google Gemini or Claude model or Lama model
0:07:07 or any model really, who do they compare it to?
0:07:10 They’re always like, oh, it’s almost as good as GPT 4.
0:07:13 It’s right up there on par with the benchmarks of GPT 4.
0:07:16 So I almost feel like you gotta leave it in S tier
0:07:19 just for the reasons that it is like setting the bar
0:07:22 that every other company is trying to get to.
0:07:24 – Yeah. And, you know, I mean,
0:07:26 we’re talking specifically about chat to be T here,
0:07:27 but like if we were talking about OpenAI as a company,
0:07:31 definitely S because like it sounds like they have more
0:07:33 stuff in the works that, you know, with the agents,
0:07:35 I think they’re ahead in agents.
0:07:36 And like you said, with the reasoning model,
0:07:38 they’re definitely ahead there.
0:07:40 And I might not use that every day,
0:07:43 but like I am seeing like tweets from like real scientists
0:07:45 saying like, I’m using this every day now.
0:07:49 And holy crap, if 01 is better as people are saying,
0:07:51 they’re super pumped about it.
0:07:53 And it’s like what one scientist said that it’s already,
0:07:56 01 preview that a paper he was working on,
0:07:58 it helped him get it done in two days,
0:08:00 what normally would have taken him a month.
0:08:02 And so yeah, I would definitely put them as tier.
0:08:03 – Cool. Yeah. I’m not going to,
0:08:05 I’m not going to fight for that one too hard.
0:08:08 It’s not my first go to, but all right.
0:08:11 So next up is Claude for me.
0:08:13 I put Claude in S tier because it’s probably the one
0:08:14 I go to the most.
0:08:18 I have tons of custom projects that I’ve built out in Claude.
0:08:20 You know, like I’ve got a project that helps me write scripts.
0:08:22 I’ve got projects that help me with my notes
0:08:24 for making YouTube videos.
0:08:27 I’ve got all of these various projects that I’ve made.
0:08:28 So when I need to make a script,
0:08:31 I can literally drag and drop an article in there.
0:08:33 And it’s sort of rough draft script for me,
0:08:37 or I can pull in all of the URLs from like a YouTube video
0:08:38 that I’m putting in.
0:08:40 And it creates like a list of resource links for me,
0:08:43 really clean and easy that I use in my description.
0:08:48 So like Claude is probably my second most used tool
0:08:51 behind perplexity.
0:08:53 So for me, that one’s S tier.
0:08:55 And I also really, really love the artifacts, right?
0:08:58 Like I love that I can have it generate code.
0:09:00 And then it will actually show me what that code looks like
0:09:02 over on the right sidebar.
0:09:05 That to me is like really, really valuable.
0:09:07 So for me, Claude is an S tier.
0:09:08 – Yeah, I mean, I can agree on S tier.
0:09:11 I guess if it was my own tier list, I’d probably put it A.
0:09:13 I’m just, I’ve always felt weird about the fact
0:09:15 that it’s like basically an employee from OpenAI
0:09:16 who left and tried to do his own thing.
0:09:19 And like, and for the first year,
0:09:21 it was like basically all copies of JVT.
0:09:23 So like, it’s like, there’s just a weird feeling there.
0:09:25 And like, and it’s called Claude.
0:09:26 I don’t know, I hate the name.
0:09:31 And also, I’ve noticed recently, like it, like,
0:09:32 people said it was better at writing.
0:09:34 I’m just not seeing that lately.
0:09:36 Like I see it’s like pretty much the same as Chatch PT.
0:09:39 Maybe on code, it’s better, you know, it seems to be.
0:09:41 I think that’ll change quickly.
0:09:43 Artifacts, definitely give that to them.
0:09:44 Like artifacts is amazing.
0:09:47 Like that’s like the one invention that they’ve created
0:09:49 that’s like actually really, really cool is artifacts.
0:09:52 – Yeah, well, they also showed off the computer use
0:09:55 recently as well, where you can basically give it
0:09:57 instructions and it will actually click around
0:09:58 on your screen and stuff.
0:10:00 – Yeah, which no one seems to be using that.
0:10:02 And my understanding from like hearing, you know,
0:10:04 from friends is like that basically OpenAI
0:10:06 already had something like that, like out for a while.
0:10:09 And it’s just like, I think Claude tried to beat them to it.
0:10:11 But then like, they’re also probably behind there.
0:10:14 And it’s like, yeah, it’s just not useful yet.
0:10:15 – So what do you think?
0:10:16 Do you think it should stay in S tier?
0:10:19 Or do you think it should bump down to eight here?
0:10:22 Cause it’s constantly playing catch up.
0:10:24 – I mean, I do think it’s eight here.
0:10:25 I really do.
0:10:28 And the other thing too, I’ve noticed is people used to say
0:10:31 that, you know, chat to PT was super woke
0:10:33 and it was super left-wing and maybe just slightly,
0:10:35 but I feel like they’ve gotten a lot better at that.
0:10:36 Like where you can talk to it.
0:10:40 Like, like I had it edit a newsletter issue
0:10:42 that I put out the other day that had some talk
0:10:44 about the Trump election that just happened,
0:10:46 like Trump winning and, you know,
0:10:48 and I understand that’s controversial people.
0:10:49 Some people hate Trump,
0:10:51 but I thought that like why it’s good for AI.
0:10:55 Claude refused to edit the newsletter issue, right?
0:10:57 I was like, that sucks.
0:10:59 And then I went to chat to PT and it not only did I edit it,
0:11:00 but I felt like it did it as good of a job
0:11:02 as Claude would have or better.
0:11:05 And so the fact that it’s like that you can’t,
0:11:07 that there’s certain things that just won’t help you with,
0:11:09 like long-term if they keep doing that,
0:11:10 it’s like, I don’t know how I can rely on it
0:11:15 as like my main assistant for my life if they do that.
0:11:17 – So to add to what you’re saying
0:11:20 and to actually sort of agree with bumping it down to eight
0:11:21 here, I do use it a lot.
0:11:23 It is one of my more favorite tools,
0:11:24 but here’s like another argument
0:11:26 for pushing it down to eight here.
0:11:29 They jumped to bump the price on Haiku, right?
0:11:31 Because they said, well, it’s smarter.
0:11:33 So we should cost more, right?
0:11:34 So they bumped the price on Haiku.
0:11:36 And the other thing about Claude,
0:11:38 like the biggest complaint I hear the most
0:11:40 is getting rate limited.
0:11:42 I don’t really hear people talking about how like
0:11:45 in chat, GPT, they get to a rate limit very often.
0:11:47 But in Claude, everybody seems to hit the rate limits.
0:11:50 I don’t really find I hit the rate limits very often
0:11:51 unless I’m writing code.
0:11:53 If I’m just like having conversations
0:11:55 about like a YouTube script or something like that,
0:11:57 I never run into rate limits.
0:11:59 But I do know that’s like a big complaint people have
0:12:00 is that for whatever reason,
0:12:03 they run into rate limits a lot with Claude.
0:12:04 – Yeah, I mean, I think, I mean,
0:12:07 they have to be more cash strapped than opening eye for sure.
0:12:10 Like no one talks about that, but they have to be like,
0:12:12 like if you’re the number two player in a space,
0:12:14 you’re always going to have a way harder time
0:12:16 raising capital.
0:12:17 – Right.
0:12:19 – But they also have the backing of Amazon
0:12:20 and Google, right?
0:12:22 Like Google is like huge investor in Claude
0:12:24 and Amazon basically said,
0:12:28 we’re going to use Claude as the future Alexa, right?
0:12:30 So it’s like, they do have that Amazon backing.
0:12:32 Like Amazon could be to anthropic
0:12:35 what Microsoft is to chat GPT, you know?
0:12:36 – Yeah.
0:12:37 – True, true, true.
0:12:38 – So we’ll see how that plays out.
0:12:40 But I think you’ve talked to me
0:12:43 into sort of conceding it down to an A.
0:12:45 All right, so next up is Gemini.
0:12:48 The thing about Gemini is we have Gemini
0:12:49 and we have notebook LM on the list
0:12:51 and notebook LM is powered by Gemini.
0:12:55 So if we’re just looking at like Gemini on its own,
0:12:57 like ignore the fact that notebook LM is powered
0:13:00 by Gemini, we’ll rank that one separately.
0:13:02 Think about like the Gemini chat
0:13:07 and like maybe let’s include like Google AI search, right?
0:13:10 Like the AI overviews that Google does,
0:13:12 we’ll consider that part of Gemini as well.
0:13:15 So considering the Gemini chat
0:13:17 and considering Google’s AI overviews,
0:13:19 where would you put Gemini?
0:13:20 – F tier.
0:13:21 – You put Gemini and F tier.
0:13:23 (laughing)
0:13:25 – I mean, I feel bad.
0:13:27 Like Logan is gonna hate me for saying that.
0:13:31 And I will concede, I have not used it recently
0:13:33 and I know everyone said it’s gotten a lot better
0:13:34 and so I should use it again.
0:13:37 And I just, I feel like, I haven’t seen a good,
0:13:39 like notebook LM is a great product.
0:13:41 So like that’s actually the first great product
0:13:43 I’ve seen from Google in a long time.
0:13:45 I feel like the Gemini release,
0:13:47 I feel like it’s been like a totally botched launch
0:13:49 and I don’t know anyone who personally uses it.
0:13:51 So that’s why I put it, put it F.
0:13:52 (laughing)
0:13:54 – I’m in the boat of like,
0:13:57 I hate to put anything in the F tier.
0:13:59 Like, ’cause you know,
0:14:03 I actually like a lot of people at Google.
0:14:06 I, the thing about Gemini that there are redeeming factors
0:14:06 about Gemini, right?
0:14:09 Like Gemini’s got the largest context window
0:14:11 of any platform, right?
0:14:13 So if you want to upload the entire Lord of the Rings
0:14:16 book series or the entire Game of Thrones book series
0:14:18 and have discussions about it,
0:14:20 Gemini is pretty much the only game in town
0:14:24 that’s gonna let you upload that large of a context window.
0:14:26 So I think that’s sort of a redeeming factor
0:14:27 for them as well.
0:14:31 Gemini, I believe also has the ability
0:14:33 to sort of read videos now too, right?
0:14:35 So you can put a video in there
0:14:36 and it understands what’s going on in the video
0:14:40 and can tell you about what it sees inside of that video.
0:14:43 I don’t know if any other models can do that.
0:14:46 Gemini also powers notebook LM,
0:14:48 which, you know, we’re kind of trying to keep the two separate
0:14:52 at the moment, but like notebook LM is really, really useful.
0:14:55 So I have a hard time like throwing it as like a solid F,
0:14:57 but the only reason I would put it sort of lower
0:15:00 on the tier list is for the same reasons you mentioned,
0:15:02 it’s not really one of my go-tos.
0:15:03 It’s not something where I’m like,
0:15:05 okay, I need to go use AI for this.
0:15:07 Let’s go pull up Gemini.
0:15:09 No, I’m usually gonna go to Claude, ChatGPT or Perplexity,
0:15:10 right? – Yeah.
0:15:14 – So for me, it’s just not the top of mind tool yet
0:15:17 that I would go and use when I need AI’s help.
0:15:19 – Yeah, it’s got great capabilities.
0:15:21 Like if I was ranking it purely on capabilities,
0:15:23 I might put it A or B.
0:15:26 – And I’m not a huge fan of the AI overviews yet.
0:15:29 Like you mentioned, it kind of didn’t know the difference
0:15:31 between real news and like memes, right?
0:15:33 Like that’s why I was telling people to put glue on pizza
0:15:35 and eat 10 rocks a day or whatever, right?
0:15:37 Like it just, it didn’t know
0:15:39 that like this wasn’t factual information.
0:15:42 It was just information it found on the internet, right?
0:15:45 But I think they’ve improved a lot of that stuff as well.
0:15:46 But it’s gotten to a point now
0:15:49 where I don’t really use Google or see AI overviews too often
0:15:51 ’cause I go to Perplexity for that kind of stuff now.
0:15:53 So I don’t know, personally,
0:15:55 I have a hard time throwing it as an F
0:15:57 because it is a capable model.
0:16:00 It’s just not one of the models it’s like my go-to.
0:16:03 I think I’d probably maybe put it at like a C or a D.
0:16:06 – D, let’s do D, something has to be lower.
0:16:09 Like otherwise we’re gonna have like all ABs and Cs
0:16:10 – That’s true, that’s true.
0:16:11 And if I’m looking at the two,
0:16:15 I probably still would go to Grock before Gemini as well.
0:16:16 – Yeah, I would, yeah.
0:16:18 (upbeat music)
0:16:19 We’ll be right back.
0:16:21 But first I wanna tell you about another great podcast
0:16:22 you’re gonna wanna listen to.
0:16:26 It’s called Science of Scaling hosted by Mark Roberge
0:16:29 and it’s brought to you by the HubSpot Podcast Network,
0:16:32 the audio destination for business professionals.
0:16:34 Each week hosts Mark Roberge,
0:16:37 founding chief revenue officer at HubSpot,
0:16:39 senior lecturer at Harvard Business School
0:16:41 and co-founder of Stage Two Capital,
0:16:44 sits down with the most successful sales leaders in tech
0:16:47 to learn the secrets, strategies, and tactics
0:16:49 to scaling your company’s growth.
0:16:51 He recently did a great episode called
0:16:55 How Do You Sol For A Siloed Marketing in Sales?
0:16:57 And I personally learned a lot from it.
0:16:58 You’re gonna wanna check out the podcast,
0:17:02 listen to Science of Scaling wherever you get your podcasts.
0:17:05 (upbeat music)
0:17:06 – Let’s do perplexity next.
0:17:09 So perplexity for me, it’s all an S tier.
0:17:10 I use it every single day.
0:17:11 – S tier.
0:17:12 (laughing)
0:17:13 Yeah, yeah.
0:17:15 Like right now I’m like chat, speak to you in perplexity.
0:17:16 That’s like daily for me.
0:17:18 And they just, you know,
0:17:20 I mean, Ervin was one of our first guests who came on.
0:17:22 Like, you know, I’ve talked with him personally.
0:17:25 He’s like, he’s a great guy and they ship so fast.
0:17:26 You know, and like,
0:17:28 and coming from like Silicon Valley and doing startups,
0:17:31 I know like how fast you ship things is so important.
0:17:33 I would say they’re shipping faster than anyone.
0:17:35 Like they’re doing better than open AI in that regard.
0:17:37 – Well, and they’re more like Gnostic, right?
0:17:40 So like what you can like perplexity
0:17:42 powers part of future tools, right?
0:17:45 And on future tools, it uses the,
0:17:47 it uses llama with perplexity.
0:17:49 So it does a search on the web,
0:17:49 fives the information,
0:17:53 uses llama to sort of like write the summaries and stuff.
0:17:56 When I use perplexity like on my desktop,
0:17:57 not within the API,
0:18:00 but just like I’m gonna use it to look up some information.
0:18:03 I’m using it with Claude 3.5 sonnet.
0:18:04 They even added a new feature
0:18:09 where you can use the GPTO1 model or the open AI O1 model
0:18:13 and actually have it do more like logical thinking through
0:18:16 with what it found in its search results, right?
0:18:19 So like it actually allows you to use
0:18:22 like whatever model is your favorite
0:18:24 plus the web search results, you know?
0:18:26 – Yeah, that’s awesome.
0:18:28 – That’s one of the few that we can totally agree.
0:18:30 Like S tier, that’s one of our main go-tos.
0:18:33 One of the most useful AI tools available out there
0:18:34 right now in my opinion.
0:18:35 – That is interesting.
0:18:37 So it kind of has replaced not only Google
0:18:39 but Wikipedia as well in a way,
0:18:41 which they probably use Wikipedia as a source,
0:18:42 but that is interesting.
0:18:43 – Yeah, yeah, yeah.
0:18:45 And I’ve done a lot of like research.
0:18:48 Like the other day I was looking for a new like headset
0:18:50 with a microphone on it, right?
0:18:52 Like headphones with a microphone.
0:18:54 And I asked it to basically like look
0:18:57 for the most recommended headset microphones
0:19:01 by YouTubers and rank them like top 10.
0:19:03 And it went and looked for all of this
0:19:05 and it linked me up to the YouTube videos
0:19:07 and it ranked them for me and it gave me the source.
0:19:10 And then I went, okay, cool, this is how it ranked them.
0:19:12 Now I’m gonna start watching some of these YouTube videos.
0:19:13 And I started clicking into the videos
0:19:16 to see what the people were actually saying about them,
0:19:18 what the sound quality was like and all that kind of stuff.
0:19:20 So I was actually starting to like click
0:19:23 into all the resources after it did all the research for me.
0:19:25 And like, man, it’s just, it’s, for me,
0:19:27 it’s been really, really helpful.
0:19:30 – It’s probably the most useful AI tool.
0:19:33 – Cool, let’s do Lama next.
0:19:35 I’m curious your thoughts,
0:19:37 ’cause I don’t think you probably use Lama very often, but–
0:19:38 – Yeah, that’s what I would say.
0:19:39 I have a hard time ranking.
0:19:41 I’m gonna like trust your rank a little bit better than mine.
0:19:43 I’ve used it one time.
0:19:46 – So Lama, like I mentioned, Lama is actually the AI
0:19:49 that powers what’s going on at Future Tools.
0:19:51 So all the like descriptions and things like that
0:19:54 that are on the website are a combination of perplexity
0:19:56 and then Lama, right?
0:19:57 ‘Cause Lama is really kind of the cheapest model
0:19:59 that you can use right now.
0:20:00 – Okay. (laughs)
0:20:02 – ‘Cause it’s open source, right?
0:20:03 – Yeah, yeah.
0:20:06 – So like, I think the fact that it’s open source,
0:20:09 you get, it’s gotta have some like brownie points for that,
0:20:09 right?
0:20:11 Like that, you know, gives it some cred.
0:20:13 Because so many of the open source models
0:20:16 that people are using now, they started with Lama
0:20:19 and then sort of modified them, forked them,
0:20:20 whatever you wanna call it
0:20:21 and like built their own models,
0:20:24 but we’re based on Lama originally, right?
0:20:27 So I think that’s like an important factor.
0:20:30 I would probably put it into like B tier.
0:20:35 Like I would use Lama probably before I would use Grock
0:20:38 or like getting a text response
0:20:40 just because it is that open source model
0:20:43 and it is insanely inexpensive to use
0:20:45 as a large language model.
0:20:47 When I use my Meta Ray-Ban glasses,
0:20:48 they’ve got Lama built into them.
0:20:51 So if I rarely ask my Ray-Ban glasses questions,
0:20:54 but when I do, it’s answering with Lama.
0:20:55 For me, it’s gotten to a point
0:20:58 where like most of the large language models
0:21:02 are like as good as each other for my use cases, right?
0:21:05 Like a lot of people that use these large language models
0:21:08 that need them for like the logic reasons
0:21:13 or more like really in-depth scientific complex topics,
0:21:16 they’re probably a little bit more sensitive
0:21:18 to which model they use.
0:21:20 For me, if I’m just like asking questions
0:21:23 that’s like general knowledge that’s out in the world
0:21:26 or if I’m trying to get it to help me like write a story
0:21:29 or if I’m getting it to try to summarize something for me,
0:21:31 most of those use cases,
0:21:34 like all of the models do pretty well at this point.
0:21:36 And so the fact that we’ve got an open source version
0:21:40 that does what ChatGPT and what Claw does
0:21:43 and you know, what all these other models do
0:21:46 really decently for most people’s use cases,
0:21:48 that’s why I think there’s value there, right?
0:21:52 Like most people are using something like ChatGPT
0:21:53 to help them write their emails better
0:21:57 or to cheat on their social studies homework
0:22:01 or to do fairly simple things
0:22:03 that pretty much all the large language models
0:22:05 are pretty profession at now.
0:22:08 And so, you know, Claw was the open source version of that.
0:22:09 – Right.
0:22:11 Yeah, I mean, since I haven’t used it much,
0:22:12 I’m not gonna like argue on it.
0:22:14 I probably would have put it the same as GROC
0:22:16 if it was just my own ranking.
0:22:19 But since the fact that you’re actually using it
0:22:21 and like neither one of us daily use GROC
0:22:24 on a daily basis, like, yeah, it should be a B then.
0:22:26 – Let’s move on now to,
0:22:28 let’s talk about some of these like coding platforms.
0:22:31 So we got GitHub Co-Pilot, we got Hurster.
0:22:34 – Replet, AI agent would be another one.
0:22:35 – Yes, we got Replet.
0:22:38 – There’s another one that people
0:22:39 are probably gonna be angry that we’ve put on here.
0:22:41 I can’t remember the name of it.
0:22:41 – Is it Bolt?
0:22:42 Are you talking about Bolt?
0:22:44 – Oh, is that, yeah, it is Bolt, isn’t it?
0:22:45 – Bolt is pretty good.
0:22:46 I’ve only tried it for a couple of minutes.
0:22:48 So I don’t think we should put it on this tier list
0:22:51 because like I haven’t played with it enough myself.
0:22:51 – Yeah.
0:22:53 – But it was impressive.
0:22:54 Like I literally told it, like,
0:22:57 I was watching a video earlier today from Tiff and Tech,
0:22:59 who’s a YouTuber who does tech videos.
0:23:03 And she actually asked it to make me a clone of Spotify.
0:23:04 And it did.
0:23:07 And I was like, like, literally that was the prompt.
0:23:09 Go make me a clone of Spotify.
0:23:10 And like when I was using it,
0:23:13 I asked it to make me like an Evernote type tool
0:23:16 where I can sort of save bookmarks from around the web in it.
0:23:19 And it made it pretty quickly.
0:23:22 Like, it’s not like cursor where it looks like a IDE
0:23:24 where you’re, you know, you can see the code
0:23:25 and all that kind of stuff.
0:23:27 It looks like a chat bot.
0:23:28 You ask it to do something
0:23:30 and it just sort of runs through the code.
0:23:31 And it’s like, all right, here’s your app.
0:23:33 I don’t do a lot of code.
0:23:35 So I’m gonna have to lean on you a little bit
0:23:36 for these code ones.
0:23:38 – Yeah, and disclaimer, I don’t code every day.
0:23:40 Like, you know, I’ve done three tech startups.
0:23:42 I’ve coded on and off most of my life.
0:23:47 You know, GitHub Copilot, originally I loved it.
0:23:49 It’s actually how I got involved in AI
0:23:51 or at least more in like the current way of AI.
0:23:52 I’d actually been doing some stuff with AI
0:23:54 in my previous startup.
0:23:56 We’d actually had a computer vision department.
0:23:58 In this current way, that’s how I got interested.
0:24:00 It was like trying out GitHub Copilot
0:24:01 and seeing how amazing it was.
0:24:04 I was like, I was actually going back and like just for fun.
0:24:08 I was like, oh, I should like pick up C++ again.
0:24:11 And maybe I can make like a cool little game demo
0:24:14 like something I was doing like a Zelda kind of game
0:24:16 but it was kind of like Dark Souls theme
0:24:17 just like messing around.
0:24:20 And it was amazing how fast I was able to like learn C++
0:24:22 again just using GitHub Copilot.
0:24:25 I was like, oh, ’cause I know basic like coding logic
0:24:25 and stuff and it was like,
0:24:28 it was just like helping teach me things so fast.
0:24:30 But I haven’t used it recently
0:24:32 and it seems like less people are using it.
0:24:35 It feels like maybe they fell behind a little bit
0:24:37 compared to cursor and other ones.
0:24:39 So I think I’d probably put it like in a solid,
0:24:43 I don’t know, B tier would be where I’d put it, yeah.
0:24:43 – Okay, yeah.
0:24:46 I mean, I’ve never used Copilot like not even once.
0:24:48 So like I really don’t have any sort of rebuttal
0:24:49 or argument to that.
0:24:51 So B tier it is.
0:24:52 – Yeah, I mean, it’s mainly really good
0:24:54 for like inline suggestions of like,
0:24:55 oh, you’re writing something
0:24:57 and this is what I think you’re wanting to write
0:24:59 and give you a great suggestion.
0:25:00 You press tab to complete it.
0:25:02 It’s like, that’s how they got started.
0:25:04 I think they have some other more advanced features
0:25:06 but I haven’t heard of anyone actually using them so.
0:25:08 – Yeah, yeah.
0:25:09 And then we have cursor, right?
0:25:12 And cursor, we actually did an episode with Riley Brown
0:25:15 where we used cursor and built a whole app
0:25:17 like within the course of that episode.
0:25:20 It’s pretty damn impressive.
0:25:22 Where would you put that one?
0:25:24 – So I mean, I would put cursor at S tier
0:25:26 ’cause I think it’s actually having a major impact
0:25:27 on software development.
0:25:29 Like it’s like, there’s lots of AI stuff
0:25:30 that we talk about where it’s like,
0:25:32 oh, it’s cool and it has promise in the future.
0:25:36 I’m hearing from lots of friends in Silicon Valley
0:25:39 that like lots of teams are using cursor now.
0:25:41 And it’s the fact of, yeah,
0:25:42 it can give you the inline suggestions
0:25:44 like GitHub co-pilot does,
0:25:47 but also it can like look at your entire code base
0:25:50 and it can have context of the entire code base.
0:25:52 It can, it can, it’s helping people a lot
0:25:54 for like debugging issues, you know?
0:25:56 ‘Cause like debugging issues is like one of the things
0:25:57 when you’re in software development,
0:25:58 it’s like the thing where you’re like,
0:26:00 damn, why am I doing this job?
0:26:03 Like you feel like the job is amazing
0:26:04 when you’re like, you’re having a concept
0:26:06 and you start, you know, you’re putting it together
0:26:08 and you get to see a demo of it work.
0:26:09 But as soon as you start debugging,
0:26:11 it’s just like, I hate this job.
0:26:14 – Yeah, yeah, I’ll keep it up there and S tier, I agree.
0:26:17 I, you know, my sort of main experience using cursor
0:26:20 has been watching Riley go back and forth with us
0:26:22 on an episode.
0:26:25 And within what, the course of 45 minutes we went from,
0:26:26 like we didn’t have an idea.
0:26:29 Like we used AI to come up with the idea on that episode.
0:26:31 And then by the end of the episode, we had a working app.
0:26:35 And I was like, this is a 45 hour long episode
0:26:37 and we have software now.
0:26:39 Like that is blowing my mind.
0:26:41 – Yeah, something else it’s gonna do that’s amazing
0:26:44 that I haven’t heard a lot of people talk about is you can,
0:26:45 like a lot of times when you’re coding,
0:26:47 you might be working with a certain API
0:26:48 or some kind of, you know,
0:26:50 some kind of library that you’re using
0:26:53 that you don’t fully know how to use it.
0:26:55 And you end up having to go back and back to the docs
0:26:56 over and over and over.
0:26:58 It’s like, oh, look, how do you do this?
0:26:59 How do you do that?
0:27:01 You can like just like link to docs
0:27:05 and you can like, like kind of like upload it to a cursor.
0:27:07 And like you could say, here’s whatever API.
0:27:09 And then you can just like at tag that
0:27:12 and talking to cursor and it knows all of the docs.
0:27:13 And then you can just ask it any question
0:27:15 and get the answer to how you do that.
0:27:16 – Oh yeah.
0:27:17 – That’s like so much better
0:27:19 than like go like opening up another window
0:27:20 and like trying to have to search through like,
0:27:21 oh, how do you do this?
0:27:23 And like learn it all again.
0:27:25 They even have stuff too for like the command line
0:27:27 where it’ll help you like doing command line stuff.
0:27:28 It gets really tedious,
0:27:30 especially if you don’t do it every day.
0:27:32 If you do it every day, you learn the commands.
0:27:33 It’s not bad.
0:27:34 But like, if you do it like once a month,
0:27:36 like me, like you kind of forget some stuff.
0:27:37 – For sure.
0:27:39 – It can just like, oh, here’s what you’re trying to do.
0:27:40 And here’s the command.
0:27:43 – Every tool out there has like an API
0:27:46 and every API is sort of slightly different
0:27:47 from every other API.
0:27:48 – Yeah.
0:27:50 – Yet somehow when you use cursor,
0:27:52 it seems to know how to implement the API
0:27:54 for the tool you’re trying to implement, right?
0:27:56 Like if I’m trying to implement
0:28:00 the whisper speech to text feature in an app, right?
0:28:03 It knows how to just use the open AI API.
0:28:08 When I was trying to get it to do like a whizzy wig editor,
0:28:09 right, where I can like, you know,
0:28:12 make fonts bold and stuff in my text,
0:28:15 it just automatically knew how to pull in the proper APIs
0:28:16 for that kind of stuff, right?
0:28:19 So that to me is also really, really impressive.
0:28:22 It’s like, you don’t need to go to
0:28:23 whatever tool you’re trying to integrate with
0:28:26 and learn all about how that API works.
0:28:29 It seems like cursor will go and do that work for you
0:28:31 to figure out how that API works.
0:28:33 – Yeah. And I think this is an example too of like,
0:28:34 when people are making the AI startup,
0:28:37 they should be looking at cursor and perplexity.
0:28:42 Like model agnostic, ship fast, solve a specific problem.
0:28:44 – Yeah. And then we got replet,
0:28:45 which I have used replet a little bit,
0:28:50 but to me, replet was more of just like that IDE, right?
0:28:51 It was more of like a replacement
0:28:53 for visual studio code for me,
0:28:55 but it was saving it all to the cloud
0:28:57 instead of saving it all to my computer.
0:28:58 – Well, yeah, on that side it’s like,
0:29:00 it’s like on the, it’s for like deployment, right?
0:29:01 Like they’re kind of a competitor
0:29:03 with like a VERSEL and some of these other startups
0:29:05 that help you easily deploy websites.
0:29:08 But I’m specifically wanting to rank the replet AI agent,
0:29:10 which is like, is a newer thing that they have.
0:29:14 So I probably put it in like a C tier right now
0:29:16 with the potential to be an S tier.
0:29:18 Like I tried it, it’s cool.
0:29:20 I mean, at least you go from like start to finish.
0:29:23 Like you can literally, I sat down with my son
0:29:24 and he had a ton of fun.
0:29:26 We basically just sat down,
0:29:28 we gave it some ideas of what we wanted to build,
0:29:29 like a website.
0:29:32 He wanted to do something that was a Minecraft
0:29:33 and League of Legends.
0:29:36 And it was like a guide for these two or something, you know?
0:29:38 And it went from beginning to end
0:29:40 and it shows you the code too.
0:29:41 And so you can dive in more
0:29:43 and see the actual code it’s generating as well if you want.
0:29:45 It takes a long time.
0:29:46 Like a lot of times it was taking
0:29:48 like three to five minutes to generate the stuff.
0:29:51 But it helps you create the entire website
0:29:52 and then deploy it.
0:29:54 So it’s like from beginning to end,
0:29:57 the quality was not as good as if I wouldn’t,
0:29:59 if I would have just used my own template
0:30:00 and then probably used cursor to help me and stuff,
0:30:02 I probably would have got something better out.
0:30:04 But you could see longterm where
0:30:06 there will be a class of people who,
0:30:08 if you have an idea for a startup,
0:30:10 probably the best way in like a year from now
0:30:13 will be to go to something like the Replay AI agent
0:30:15 and say, make me this landing page.
0:30:17 It’s gonna make you a beautiful landing page.
0:30:18 It’ll deploy it for you.
0:30:21 The whole thing will be done in 10 minutes.
0:30:22 Yeah, yeah, yeah.
0:30:22 Right, right.
0:30:25 That’s gonna be probably possible in the next year.
0:30:26 And then it’ll be S tier.
0:30:27 Or you can just sketch something out on a piece of paper,
0:30:30 like scribble out like a rough layout of what you want
0:30:32 and then upload that image and say,
0:30:34 make me a landing page.
0:30:35 Here’s a sort of quick template.
0:30:37 And then it’ll make a beautiful version
0:30:39 of what you, you know, you sketched out.
0:30:41 Something that we saw, you know,
0:30:43 back at the GPT for demo,
0:30:46 but it’s never really panned out
0:30:49 to be like what we’ve seen in the demos yet.
0:30:50 Right.
0:30:51 Okay, I’ll put it there
0:30:53 because I can’t, or you otherwise.
0:30:58 So let’s, let’s, let’s do like the AI image generators now.
0:31:01 Let’s start with probably the most well-known
0:31:03 with mid-journey.
0:31:06 Where would you rank mid-journey?
0:31:07 F tier?
0:31:08 (laughs)
0:31:10 Yeah, totally.
0:31:12 Some of our friends are gonna kill me if I do that.
0:31:13 God, I have a hard time ranking it.
0:31:15 Like there’s like, part of me that wants to say S
0:31:16 and there’s part of me that wants to say A.
0:31:17 I guess that’s kind of like-
0:31:19 Honestly, I’ll be honest with you.
0:31:20 I’m gonna struggle with the AI art tools
0:31:24 ’cause like I actually use them all for different reasons.
0:31:25 Yeah.
0:31:26 I mean, I used to, you know,
0:31:29 when I first started dabbling with AI art,
0:31:30 I used stable diffusion
0:31:32 ’cause I could run it locally on my machine.
0:31:34 And I just thought that was so cool.
0:31:36 And then I started using mid-journey
0:31:38 and I was like, oh yeah, you get, you know,
0:31:41 it’s easier to create beautiful things with it.
0:31:43 I hated that you had to use it on Discord.
0:31:44 It took them forever to get the web version.
0:31:46 I thought that was so dumb.
0:31:47 Well, I would think mid-journey,
0:31:50 I think mid-journey kind of argued that like,
0:31:51 they were more of like a research company
0:31:53 building up a really good art tool.
0:31:55 And they didn’t, they were less concerned
0:31:58 about the user experience at that point, right?
0:31:59 They were just concerned
0:32:02 with making the best image model they could.
0:32:03 This is our user experience.
0:32:05 Live with it or go somewhere else.
0:32:06 We don’t really care.
0:32:08 We’re gonna make the best image generator there is.
0:32:11 And only recently have they started putting like effort
0:32:13 into user experience.
0:32:15 Yeah, I’m kind of convinced that the image generation part
0:32:16 is gonna become a commodity
0:32:18 and the user experience is a thing.
0:32:21 So I think I would put them at A.
0:32:22 I think a lot of people would put them at S,
0:32:24 but I think I’d put them at A.
0:32:25 Just ’cause I think-
0:32:26 – I would agree with A, I think.
0:32:29 – Yeah, I’m not using it as much as I used to.
0:32:30 There’s great alternatives.
0:32:33 I think they ship the website
0:32:35 and the new user experience way too slow,
0:32:38 which makes me more pessimistic on them long-term.
0:32:39 – Yeah, I think, you know,
0:32:40 if I’m going back to how I was sort of
0:32:43 ranking things earlier on in this episode,
0:32:45 I was kind of ranking them by like,
0:32:47 how often I would go to that tool
0:32:50 opposed to the other tools, right?
0:32:52 Mid Journey isn’t really one that I go to
0:32:56 is like one of my top, you know, two or three anymore.
0:32:59 Maybe it’s top three, but it’s not top two, right?
0:33:02 Top two for me would be Leonardo and Stable Diffusion
0:33:04 are the two that I use the most.
0:33:07 I don’t know, I would probably put Mid Journey as B tier.
0:33:09 – Oh, really? Okay. Yeah.
0:33:11 – But I could probably be convinced otherwise.
0:33:13 The thing about Mid Journey is it’s like,
0:33:15 it’s the most well-known.
0:33:18 It’s the one that I feel like at least for a long time,
0:33:21 like ChatGPT was setting the bar with AI Image Generation.
0:33:22 Like everybody was going,
0:33:24 oh, this one’s almost Mid Journey tier.
0:33:26 Oh, have you seen the new version of Stable Diffusion?
0:33:28 It’s almost as good as Mid Journey.
0:33:30 Have you seen the new ideogram?
0:33:32 It’s almost as good as Mid Journey, right?
0:33:37 Like, so like to me, it gains some points as being that one
0:33:39 that sort of bent everybody’s benchmark
0:33:41 of like trying to reach,
0:33:46 but I don’t really find myself using it as often anymore.
0:33:48 And I think it’s just because,
0:33:50 not because I don’t think Mid Journey is very good.
0:33:52 It’s just because I think some of the other models
0:33:55 have gotten better than Mid Journey at this point.
0:33:57 – Yeah, and people are saying that they’re going to move
0:34:00 into like AI video and that’s what’s going on.
0:34:02 And also they’re going to move into,
0:34:05 like you’ve created some kind of character persona
0:34:07 on Mid Journey and that can be consistent
0:34:08 across different things.
0:34:09 So I guess you can make like comics and stuff.
0:34:11 So maybe they will,
0:34:13 maybe they are ahead in that area and we don’t know,
0:34:16 but also recently Sam Altman kind of like had this grin
0:34:19 on his face when he was talking about how much better AI art
0:34:21 in video is going to get soon.
0:34:24 So I have a feeling that behind the scenes too,
0:34:26 like all the other stuff that OpenAI is building
0:34:28 is probably going to give them some edge there too
0:34:29 that people are not going to expect.
0:34:31 Just like when we first all saw it, it’s like,
0:34:33 holy crap, that’s amazing.
0:34:34 They probably have other things that are amazing
0:34:37 that have not been revealed yet, if I had to guess.
0:34:39 – The other sort of like negative against Mid Journey
0:34:41 is of the image generators.
0:34:44 It’s probably the most expensive to use as well.
0:34:45 – Yeah.
0:34:46 I keep canceling it and then re-subscribing.
0:34:48 I’m like, I am not using it and I’m like, it’s expensive.
0:34:49 Then I cancel it.
0:34:50 I’m like, I want to use it again.
0:34:52 I subscribe, but this happened so many times now.
0:34:53 – Yeah, yeah.
0:34:55 So like, you know, for me,
0:34:57 because it’s on the more expensive end
0:34:59 of the AI image generator models,
0:35:03 I personally don’t find myself going to it as much anymore.
0:35:03 That’s why I’d put it there.
0:35:06 I feel like flux might have passed it
0:35:10 or at least caught up with it as terms of realism.
0:35:11 Right?
0:35:14 And when it comes to like artuny, like colorful,
0:35:18 high contrast, like aesthetically pleasing images.
0:35:20 For me, I tend to go to Leonardo a little bit more.
0:35:23 Now I will say that I am an advisor for Leonardo.
0:35:24 Like I feel like-
0:35:26 – So S tier, of course, Leonardo’s S tier, right?
0:35:27 – I need to disclaim that that like,
0:35:30 I do have equity in Leonardo.
0:35:33 But I legitimately do go and use it more
0:35:35 than I use Mid Journey, right?
0:35:39 Like I do go and I don’t think I’m going to put it in S tier.
0:35:40 I’ll probably put it in A tier
0:35:43 just ’cause it’s the one that I use the most.
0:35:45 – Wait, are we finishing up Mid Journey though?
0:35:46 So let’s-
0:35:47 – Oh yeah, okay.
0:35:49 – So I agree with, I had said S or A,
0:35:52 but I think you kind of convinced me of B, honestly.
0:35:53 I think that’s like a good spot
0:35:57 ’cause like they were S tier for a while, now they’re B.
0:35:59 Yeah, that makes sense to me.
0:36:00 – Yeah, and then, you know,
0:36:02 so I was sort of sort of rolling into Leonardo
0:36:04 just because I was like referencing it
0:36:06 in comparison to Mid Journey.
0:36:08 For me, Leonardo is like,
0:36:10 it still has some of the like issues
0:36:11 you might get with stable diffusion
0:36:13 where you get some of the like seven finger hands
0:36:18 or like, you know, like that bird flying up there
0:36:19 looks a little weird.
0:36:20 Why does it have three eyes or whatever?
0:36:22 Like you get some of that weird stuff
0:36:24 that you get with stable diffusion,
0:36:27 but aesthetically like the Leonardo Phoenix model,
0:36:28 which is their proprietary model.
0:36:31 It’s not like one of the stable diffusion models.
0:36:33 I think they’ve gotten, it’s gotten really, really good.
0:36:36 It’s like, if you look at my YouTube thumbnails,
0:36:37 the way I make those thumbnails
0:36:41 is I usually generate an image in Leonardo first
0:36:42 and then I take that image
0:36:44 and then I pull it into stable diffusion
0:36:46 that I have locally installed on my computer
0:36:49 and I do a face swap with stable diffusion.
0:36:52 So it’s like the image was made in Leonardo
0:36:54 and then stable diffusion is what I use
0:36:56 to sort of mask out my face
0:37:01 and then put my real face on the image that was generated.
0:37:03 So I’m using that one constantly
0:37:05 ’cause it’s what all the thumbnails are made with.
0:37:06 But I don’t know.
0:37:08 Have you, have you used Leonardo before?
0:37:10 – Yeah, I have like twice.
0:37:11 Honestly, it’s been a while
0:37:13 so I definitely haven’t tried the Phoenix model.
0:37:15 My impression was like when I tried it,
0:37:17 I was like surprised how good it was to be honest with you
0:37:19 ’cause I felt like I hadn’t heard much about it
0:37:21 except from you honestly.
0:37:23 Like I had heard about it a little bit on the X
0:37:26 but like I was like nowhere near as much as mid-journey.
0:37:27 And when I tried it, when I,
0:37:29 this was probably a year ago now.
0:37:31 So it’s probably way better since then.
0:37:35 I felt like it was 95% as good as mid-journey
0:37:37 in terms of quality, which surprised me.
0:37:40 I was like, oh, this is pretty close to mid-journey
0:37:42 and I liked the interface a lot better.
0:37:46 Like I thought it was a cooler app to use honestly.
0:37:47 I think I just had the mid-journey subscription
0:37:49 so I just ended up not going back to Leonardo.
0:37:51 That’s honestly like kind of what happened, but.
0:37:52 – When it comes to realism,
0:37:55 I think mid-journey still has Leonardo beat.
0:37:56 You know, I think flux in mid-journey
0:37:59 are better at realism than Leonardo.
0:38:00 But when it, like, do you remember back
0:38:03 when like mid-journey was on like version three
0:38:05 and they had these like really like vibrant,
0:38:08 like contrasting colors and every image you looked
0:38:10 had like this mid-journey aesthetic
0:38:12 that just had like this cool look to it, right?
0:38:15 That’s how I feel about the Leonardo Phoenix model
0:38:19 is it’s like, it’s got this high contrast HDR
0:38:22 sort of like really, you know, the colors really pop.
0:38:24 It’s got like that aesthetic to it,
0:38:26 which to me is really, really pleasing.
0:38:28 I feel like with mid-journey,
0:38:30 like each version that’s come out of mid-journey,
0:38:33 they’ve sort of moved away from that like mid-journey aesthetic.
0:38:35 And now they just kind of look like, you know,
0:38:37 whatever you’re prompting,
0:38:39 like whatever style you’re prompting,
0:38:42 Leonardo to me still has like a style, I guess.
0:38:45 – Yeah, so you are using it for your thumbnails
0:38:47 on a daily basis, is that, that’s right?
0:38:48 – Well, I don’t put out a new video every single day.
0:38:50 – Not daily, not daily, yeah, but yeah.
0:38:51 – Every video I put out, yeah,
0:38:54 I use Leonardo for the thumbnail, yeah.
0:38:57 – Yeah, I’m kind of like, I feel like I could see A,
0:38:59 but also like people are gonna freak out about that
0:39:01 and saying Leonardo over mid-journey.
0:39:02 And also since you’re an advisor,
0:39:05 I’m like, damn, I should probably put it as a B.
0:39:08 – I’ll put it as a B, just as a precautionary measure.
0:39:11 I, you know, so here,
0:39:12 just kind of thinking it through here, right?
0:39:14 I think mid-journey and Leonardo
0:39:17 should be kind of on the same level.
0:39:19 Maybe like, if we’re right and I’m like this way too,
0:39:22 so like this is a higher B maybe,
0:39:25 but like they’re probably about as good as each other.
0:39:26 Like I would go to mid-journey
0:39:28 if I need something a little bit more realistic,
0:39:29 I would go to Leonardo.
0:39:31 If I’m looking for that sort of like aesthetic,
0:39:32 the color’s gonna pop,
0:39:34 it’s gonna make a good thumbnail kind of thing.
0:39:35 – Yeah.
0:39:36 – But they’re probably on par with each other.
0:39:38 – Okay, I could see that.
0:39:41 – All right, so I put stability AI here,
0:39:43 but this is meant to be for stable diffusion.
0:39:46 Stable diffusion itself doesn’t really have a logo
0:39:49 and everybody kind of like associates stability AI
0:39:50 with stable diffusion.
0:39:52 I mean, so for me, stable diffusion,
0:39:54 I’m probably just gonna put it right alongside
0:39:57 these other ones ’cause I use it just as much
0:39:58 as I use Leonardo, right?
0:40:01 Like that’s what I use for like my face swapping.
0:40:02 Maybe I put it as a C tier actually,
0:40:05 just because like I don’t use it to generate images,
0:40:08 like almost ever I use it to do like in painting.
0:40:12 So I generate the image with a mid-journey or a Leonardo
0:40:14 or like one of those tools first.
0:40:16 And then the only thing that I’m using stable diffusion
0:40:19 for at this point is to go in, mask out my face
0:40:23 and then like superimpose my face using AI.
0:40:24 So I can’t, I don’t know,
0:40:26 I don’t think I can put it on the same level
0:40:28 as these other ones, but-
0:40:30 – I was gonna say F, which is like probably getting me killed
0:40:31 by like everybody who-
0:40:33 (both laughing)
0:40:36 But I mean, I feel like they dropped the ball so much.
0:40:38 Like early on-
0:40:39 – Yeah, but I think you’re referring to stability AI,
0:40:41 the company, right?
0:40:42 – Yeah, okay.
0:40:43 Yeah, true.
0:40:45 – The stable diffusion is like an open source
0:40:47 image generation model.
0:40:50 The ability AI has made a handful of the weights
0:40:54 that have been made available like SDXL, SD 2.0, SD 3.0.
0:40:56 Like those were made by stability AI,
0:40:59 but stable diffusion itself was an open source model
0:41:01 that was built before stability AI even existed.
0:41:05 – Right, which was super, like I feel like
0:41:06 there wasn’t a whole lot of transparency around that.
0:41:07 So yeah, I’m probably judging the company,
0:41:10 not the model itself.
0:41:13 Cause like I have a lot of feelings about the company
0:41:16 in terms of the statements that were made to investors
0:41:17 and other things.
0:41:19 – When SD, was it 3.0 came out
0:41:21 and everybody was generating the images
0:41:22 of people like laying in the grass
0:41:24 and it would have like these mutated people
0:41:27 that had like only legs and no head
0:41:29 and you know, random stuff like that.
0:41:30 – Yeah, yeah, yeah.
0:41:31 I guess I was also like, you know,
0:41:34 early on I was using stable diffusion a lot.
0:41:36 And I just, I personally don’t use it anymore,
0:41:38 but it sounds like you do actually have a friend
0:41:39 who has a company as well.
0:41:41 And I know he, like he has like a discord bot
0:41:42 or something like that.
0:41:43 I know he uses it.
0:41:45 So it sounds like people are using it.
0:41:47 So probably, yeah, it’s probably not an F.
0:41:48 – Well, I guess if we’re looking at that as a whole, right?
0:41:51 Like most of the models that people are gonna use
0:41:54 for stable diffusion today are the models
0:41:56 that were generated by stability AI.
0:41:59 So I guess like, you can kind of lump it in, you know?
0:42:00 Well, just as we’ve been talking about this,
0:42:03 I’ve already knocked it down two tiers.
0:42:04 – Okay, okay.
0:42:06 Yeah, DC, I think DC seems good.
0:42:07 Like it’s gonna be controversial.
0:42:08 Some people are gonna say it should be like an A.
0:42:10 I don’t think anyone’s gonna say S.
0:42:13 – Well, again, my logic on it is like,
0:42:15 how often do I find myself going to it
0:42:16 when I need to generate stuff?
0:42:18 And I do use it often,
0:42:20 but I don’t use it to generate the images.
0:42:21 Like the only value I’m getting
0:42:26 from stable diffusion right now is that face swapping feature.
0:42:28 – Okay, D, let’s leave it at D.
0:42:29 We have to have something that’s lower.
0:42:30 I mean, like.
0:42:33 – But looking at what’s left, what would go lower?
0:42:34 – Yeah, yeah, that’s true.
0:42:35 That’s true.
0:42:37 We may not get anything at F.
0:42:39 Like we’ll just like the people who make it.
0:42:40 – We’re just too nice.
0:42:41 Yeah, we don’t want to burn any bridges
0:42:43 with any companies completely.
0:42:46 So it’s like hard to throw anything below a D.
0:42:47 All right, but stable diffusion itself,
0:42:51 let’s leave it in D because it is open source.
0:42:54 You can literally generate anything
0:42:55 if you’re running into like roadblocks
0:42:59 with censorship or whatever with any of the other AI models.
0:43:03 All right, so this next one here, this is Playground AI.
0:43:05 Quite honestly, I haven’t used it a ton,
0:43:07 but I did use it recently.
0:43:10 They’ve sort of rebranded a little bit
0:43:14 to being like the AI generation tool
0:43:16 for like graphic designers, right?
0:43:18 Like they’ve made it really good
0:43:21 at creating almost like vector style art.
0:43:22 Like it’s not actually generating vectors,
0:43:25 but it’s generating like vector style art,
0:43:28 like logo type stuff and stuff that would be good
0:43:31 on like t-shirts and mugs and stuff like that.
0:43:34 It seems like they’ve kind of gone more that direction
0:43:36 and sort of niched into it, right?
0:43:39 Like you’ve got like scenario.gg, right?
0:43:42 They’ve really, really like leaned into like game assets
0:43:46 where Playground, they’ve really seemed to lean into like
0:43:47 where what you go to if you want to create logos
0:43:51 or graphics for like your company assets
0:43:53 and icons and things like that.
0:43:57 And for that use case, it seems pretty dang good,
0:43:58 but I haven’t used it a ton.
0:44:00 Like that was me playing around with it
0:44:02 for like 30 minutes or so.
0:44:04 – Having a hard time ranking this one
0:44:05 for like multiple reasons.
0:44:06 Like I haven’t tried it recently.
0:44:09 I tried it before, haven’t tried it recently.
0:44:09 Next time we do this,
0:44:12 I probably should try everything before we do this.
0:44:15 And you know, the founder, Sue Hale,
0:44:18 I have a lot of mutual friends with him in Silicon Valley.
0:44:19 He created Mixpanel back in the day,
0:44:22 which was like, was for a while considered
0:44:23 like the coolest analytics tool.
0:44:25 Like instead of using Google Analytics,
0:44:27 you use Mixpanel if you wanted more details
0:44:29 and the stuff that mattered to you.
0:44:32 So, you know, he’s definitely a great founder,
0:44:34 but when I did try it,
0:44:37 like the user experience and interface was like really good.
0:44:42 It was like that and Leonardo were like two of the ones
0:44:43 I thought were pretty cool.
0:44:47 – I mean, just kind of going back on my sort of fallback
0:44:51 like logic with how I ranked everything else.
0:44:53 I don’t really go to that one very often.
0:44:56 I mean, I use stable diffusion more than I use playground.
0:44:58 But I also think that this should probably still rank
0:45:00 above stable diffusion.
0:45:03 ‘Cause as far as like people listening to this episode,
0:45:07 it’s a lot more user friendly than stable diffusion.
0:45:08 Like you’re probably more likely to go,
0:45:10 like unless you’re using it for a very niche use case,
0:45:12 like I am with stable diffusion
0:45:15 where I’m going and face swapping for YouTube thumbnails,
0:45:17 you’re probably not gonna use stable diffusion, right?
0:45:19 You’re probably gonna prefer something like playground,
0:45:22 which has stable diffusion models built into it.
0:45:24 So you could just select the model
0:45:25 and get playground to generate the image
0:45:27 with stable diffusion for you, right?
0:45:31 Also, it’s similar to like the AI version
0:45:35 or the AI art version of like a perplexity or a cursor
0:45:37 where you can go into playground.
0:45:40 You can use playground model V3,
0:45:41 but you can also choose Dolly three
0:45:46 or you can also choose stable diffusion SDXL or SD 3.5.
0:45:49 Like you could go and pick whatever model you want to use
0:45:52 or you can use their own like proprietary built-in model.
0:45:56 So I think like that gives it more value
0:45:57 than stable diffusion right off the bat
0:46:00 ’cause it does what stable diffusion does plus, you know?
0:46:02 – Yeah, and anyone could listen to this
0:46:03 and just go instantly try it
0:46:05 and get the gist of it in like a few minutes.
0:46:08 – But as far as like the images that it generates,
0:46:10 I don’t think I’d put them on the same level
0:46:12 as Leonardo’s mid-journey.
0:46:15 I like, I don’t think it’s up to par with those quite.
0:46:18 Like it’s close, but it’s not quite as good as those
0:46:20 like with the quality of image.
0:46:23 – Yeah, the last time I saw images from it,
0:46:26 it seemed like kind of closer to like stable diffusion level.
0:46:27 – Yeah, yeah.
0:46:29 And then we’ve got, just moving on,
0:46:30 we’ve got Adobe Firefly.
0:46:33 This might be my first E tier.
0:46:36 Here’s my problem with Adobe Firefly.
0:46:38 Out of all the image generation models,
0:46:41 it is the absolute hands down most censored, right?
0:46:43 Like I tried to get it to generate an image
0:46:44 of the Eiffel Tower,
0:46:47 and it said, “No, we can’t generate that for you.”
0:46:50 And I went and looked it up and apparently images
0:46:53 of the Eiffel Tower are technically trademarked
0:46:54 or copyrighted or something like that.
0:46:57 Like it wouldn’t generate an image of the Eiffel Tower.
0:47:00 When I get it to generate images of actual people,
0:47:01 it’s not great.
0:47:03 Like the realism isn’t quite there.
0:47:06 When it comes to like aesthetics and generating images
0:47:09 that like I just think are aesthetically pleasing
0:47:10 with like the color palette and the contrast
0:47:13 and stuff like that, never very impressed.
0:47:16 Like every time I’ve gone to Adobe Firefly to go try it
0:47:19 to see if like, okay, has this one gotten better
0:47:21 than it used to be?
0:47:24 It doesn’t really, like the UI has gotten a lot better.
0:47:27 Like the UI, I have no beef with the UI.
0:47:28 I think it’s a decent UI.
0:47:31 I just really, really don’t like the images it generates.
0:47:33 And it is so insanely censored
0:47:35 that like I’m always sitting here banging on my desk.
0:47:37 I was so frustrated that some of the images
0:47:38 it wouldn’t generate.
0:47:39 – Yeah.
0:47:41 So I only saw examples of it then.
0:47:42 I haven’t tried.
0:47:45 I used generative fill and some of the other Photoshop features
0:47:47 they rolled out, which are pretty cool.
0:47:48 But yeah, it sounds pretty bad.
0:47:50 Like I didn’t see any examples.
0:47:52 They looked great either, like on Twitter or anything.
0:47:54 Like there was nothing that was like, oh, that’s awesome.
0:47:59 And to think about like Photoshop for, it’s a creative tool.
0:48:02 It’s like, how can you put such a restriction
0:48:05 on a tool that creators use on a daily basis?
0:48:07 Like there’s no logic there for that.
0:48:10 Cause like, maybe they’re doing it for legal reasons,
0:48:12 but like, it’s like saying that you’ve got a hammer
0:48:13 and you can’t use a hammer in certain ways.
0:48:16 Like, well, that’s, why would I buy that hammer?
0:48:19 It’s like, you know, it’s a, like you can use Photoshop
0:48:21 to draw anything, right?
0:48:22 Like so.
0:48:24 – Well, you’re bringing up some points here
0:48:25 that actually want to touch on.
0:48:29 Like Adobe claims to be like the only AI,
0:48:31 like ethical image model, right?
0:48:33 Because they’ve only trained on like images
0:48:35 that are inside of Adobe stock
0:48:38 and like public domain photos, right?
0:48:43 So they claim that like, you’re not generating on a model
0:48:46 that was trained on like everybody else’s artwork
0:48:49 because they had the license to use all of the images
0:48:51 in their training data.
0:48:53 So the people that are really, really concerned about the,
0:48:56 like the ethics of how the models were trained
0:48:58 would probably go to Adobe over the others
0:49:02 because of that sort of like ethical, you know,
0:49:04 piece that they put on there.
0:49:08 Also, you mentioned that Firefly in Photoshop,
0:49:09 that has actually been useful.
0:49:11 Like, I kind of forgot about the fact
0:49:12 that it’s actually using Firefly
0:49:14 when you’re using Photoshop.
0:49:16 And I do every once in a while use Firefly
0:49:17 to like fix stuff, right?
0:49:20 Like I’ll generate an image with Leonardo
0:49:21 or Majorna or something like that.
0:49:23 And there’ll be like a weird artifact in it.
0:49:26 And I’ll pull it into Photoshop,
0:49:28 circle that little weird artifact,
0:49:30 click the generate, generate a fill button
0:49:32 and it’ll like remove that for me, right?
0:49:34 And it works at that really, really well.
0:49:35 But even saying that,
0:49:37 the censorship is so bizarre sometimes.
0:49:40 – I honestly think this should be our first F tier
0:49:43 because like a creative tool that doesn’t allow you
0:49:44 to create things you wanna create.
0:49:46 And also I haven’t seen anyone sharing good examples
0:49:47 of using it.
0:49:49 I haven’t heard of anyone actually using it.
0:49:51 – Well, everybody uses it in generative fill though.
0:49:54 Like if you’re using Photoshop with generative fill,
0:49:56 that’s where you’re most likely gonna use it, right?
0:49:57 – Yeah.
0:49:58 – And that’s where I find myself using it
0:49:59 from time to time.
0:50:01 But even there, like I’ve had times where like,
0:50:03 it would be my face in an image
0:50:05 and I would circle my eyes and say like,
0:50:06 put sunglasses on me.
0:50:07 And it would say like,
0:50:08 oh, our guidelines don’t allow that.
0:50:12 Like sometimes the guidelines just present false positives
0:50:13 or false negatives
0:50:15 or you know, whatever direction that should go.
0:50:18 It thinks you’re trying to generate something
0:50:20 that you’re not actually trying to generate, right?
0:50:22 And sometimes it’s just like really weird like that.
0:50:25 Or like, like I’ve had issues where like,
0:50:27 I had like a six fingers on one of the hands
0:50:28 that it generated, right?
0:50:30 And I was like, circle one of the fingers
0:50:31 and it’ll be like, oh, sorry.
0:50:32 This is against our guidelines.
0:50:34 We can’t change this image
0:50:36 ’cause it obviously thinks the finger is like,
0:50:38 you know, a different appendage or something.
0:50:41 And so like, you run into that kind of issue with it.
0:50:42 But I have a hard time putting in F tier
0:50:44 because I do find it useful.
0:50:46 – But we should treat it as like a standalone
0:50:47 versus a generative fill.
0:50:49 Like if we had generative fill,
0:50:50 maybe we’re talking like a B tier or something
0:50:53 like that feature in Photoshop.
0:50:56 But for like a standalone image generator,
0:50:59 I think we have to judge it that way
0:51:00 to get like an accurate judgment on it.
0:51:02 And I do think that that would be an F.
0:51:03 – Yeah.
0:51:05 You know, and another negative for Adobe too
0:51:08 is like their pricing model
0:51:10 has always spelt a little predatory, right?
0:51:11 Like where they do the whole thing
0:51:13 where it’s like, it’s this much per month,
0:51:15 but you have to agree to a year long contract
0:51:17 to get that at this much per month.
0:51:18 And then people don’t use the tool anymore
0:51:19 and they can’t get out of there.
0:51:21 It’s just, that to me feels very icky.
0:51:23 And I’ve always hated that about Adobe.
0:51:24 – Yeah.
0:51:27 There’s been once or twice where like unintentionally
0:51:29 I end up paying them like a 500 bucks or something.
0:51:30 – Yeah.
0:51:31 – And it’s like, that sucks.
0:51:33 – All right, so next up is Magnific.
0:51:36 I haven’t actually used Magnific to generate images,
0:51:38 but I’ve used their upscaler quite a bit.
0:51:39 And I think it’s pretty cool.
0:51:41 But I think the big complaint about Magnific
0:51:42 is the cost, right?
0:51:46 A lot of people say it’s like for an upscaler,
0:51:48 it’s like kind of pricey for that.
0:51:50 But what are your thoughts on it?
0:51:52 – Yeah, I got pretty crazy.
0:51:55 That’s my problem as well.
0:51:57 Is they did give me some credits
0:51:58 so I could play around with it.
0:52:00 But like, whenever I’ve talked about it
0:52:01 on a YouTube video or something like that,
0:52:05 people are always like, Magnific, prices is crazy, you know?
0:52:06 – Right.
0:52:07 I haven’t used it recently.
0:52:09 So I’m gonna have a hard time giving it
0:52:11 like a really high ranking.
0:52:13 But when I used it, it seemed great.
0:52:14 Like upscaling was always one of the issues.
0:52:16 Like you’d use different AI art tools.
0:52:18 And then like there was, it was crazy
0:52:20 that there was no good upscalers.
0:52:21 It was like, why is it?
0:52:23 How’s this still an issue?
0:52:24 And then the fact that,
0:52:25 and then it was kind of cool
0:52:26 how it would actually make them look better.
0:52:27 Like some stuff you put in there
0:52:29 and all of a sudden look nicer.
0:52:30 Which sometimes that was bad too.
0:52:32 ‘Cause sometimes it would actually change
0:52:34 the essence of the image as well.
0:52:36 So I had a hard time kind of like playing with the tweaks
0:52:38 on that and stuff like, how do you get that right?
0:52:40 – It upscales, but it also hallucinates.
0:52:42 Like it hallucinates on purpose, right?
0:52:43 Like it will add extra stuff
0:52:45 that wasn’t in the original image.
0:52:48 But it creates this like cool effect.
0:52:49 Like I’ve always thought it looked really cool.
0:52:52 Like we’ve seen those things that circle on Twitter
0:52:54 and stuff where it’s like Laura Croft
0:52:55 from like the original Tomb Raider game
0:52:57 and somebody upscaled it with Magnific.
0:53:01 And then it looks like a triple A game graphics.
0:53:02 I wouldn’t say it looks like ultra realistic,
0:53:04 but it looks like modern day game graphics
0:53:06 with the upscale done to it.
0:53:09 Like some of that kind of stuff is really, really cool.
0:53:12 But to me, it falls into the category of like novelty.
0:53:14 Like I don’t use it a ton.
0:53:16 – Yeah, I thought you were gonna say like the new Lara Croft
0:53:18 and then it changed it to more like the original Lara Croft.
0:53:20 – Well, I’m sure you can go that direction
0:53:20 if you want it as well.
0:53:24 (all laughing)
0:53:24 – Yeah, I don’t know.
0:53:28 I’m feeling like probably, geez, where would we put it?
0:53:30 Should we put it like C or B?
0:53:31 – See, I’m going B or C.
0:53:33 I’m probably leaning more towards C
0:53:36 ’cause if I’m looking at like the value of the tools in B,
0:53:39 I don’t think I put Magnific at like the same level
0:53:41 of value of these tools.
0:53:43 – And it’s not to like disclose.
0:53:45 We probably both have talked to, what’s his name?
0:53:47 Javi, is that how you say it?
0:53:47 – Javi, yeah, yeah.
0:53:52 I’ve talked to him several times on Twitter, nice guy.
0:53:54 I was kind of shocked that he built this
0:53:56 ’cause like I just knew him as like this guy on Twitter.
0:53:58 And it was like, awesome that he built this amazing startup.
0:54:00 So it’s like props to him,
0:54:01 but he’s already sold to the company now.
0:54:03 – It’s got the FreePik, right?
0:54:04 So now it’s owned by FreePik,
0:54:06 which FreePik is an AI image generator
0:54:09 that also has Magnific built into it.
0:54:10 – Yeah, so we can talk more crap about it
0:54:11 since he’s already sold it.
0:54:12 (all laughing)
0:54:15 No, so I would say C ’cause it’s expensive
0:54:20 and just long-term, I don’t see the value long-term
0:54:22 just because it seems like, you know,
0:54:24 and I think I may have even said this to him.
0:54:25 “Yeah, you should sell it, like if you can.”
0:54:27 (all laughing)
0:54:29 Because like it took off like crazy fast
0:54:30 and it’s like, but like long-term,
0:54:33 how does that become its own company or own product?
0:54:36 Like it’s an upscaler, like a cool-upscaler.
0:54:37 – Yeah.
0:54:39 – That should be part of mid-journey or Leonardo.
0:54:41 – We’ve kind of covered the AI art.
0:54:43 Now we’re sort of shifting into AI video here.
0:54:46 We’ve got Runway, Pika, Kling and Luma,
0:54:48 ’cause Luma’s got Dream Machine
0:54:50 and Dream Machine was pretty damn groundbreaking
0:54:51 when they showed it off.
0:54:53 But let’s start with like, let’s start with Runway.
0:54:57 For me, Runway, for me, Runway is an S tier.
0:55:02 And my reasoning for that is like, Runway just ships, man.
0:55:04 In the last couple of weeks,
0:55:05 we saw Runway ship act one,
0:55:07 which was like the sort of lip sync.
0:55:10 You can, and it follows the emotions of a character
0:55:12 so I can make a video where I’m talking
0:55:15 and I’m moving my eyes around and every time I blink,
0:55:17 the animation on the video blinks, it does all that.
0:55:22 And then they released the multiple camera angles
0:55:25 where you can make a video from like any angle
0:55:27 of an image that you upload in there.
0:55:30 And they, and that’s really cool.
0:55:32 And then they’ve got like frame interpolation
0:55:34 for like images where you can take one image
0:55:36 and another image and have it like frame interpolate
0:55:38 between them and do this like morphing effect.
0:55:40 And they’ve got the ability to go
0:55:42 and like remove the background of any video
0:55:44 and turn it into like a green screen video.
0:55:47 And like to me, when it comes to AI video,
0:55:51 like every week, Runway’s got some sort of cool new feature.
0:55:53 They just, they just ship and they ship and they ship.
0:55:55 And the generated videos that come out of Runway
0:55:58 are really impressive to me.
0:56:00 When I’m going to generate up an AI generated video,
0:56:03 I’m either using Runway or Luma most of the time.
0:56:06 And for me, I’d put it in S tier.
0:56:08 – Yeah, I get the S tier.
0:56:10 – I just don’t think it’s on the same level
0:56:12 as those other three companies
0:56:16 in terms of like actual value that can be created with it.
0:56:18 Like as of right now.
0:56:19 – That makes sense.
0:56:21 And I’m biased ’cause I do a lot with video, right?
0:56:23 So like I’m playing with video tools a lot.
0:56:25 I’m sure a lot of people that listen or watch this,
0:56:29 you know, this podcast don’t do AI video as much as I do.
0:56:33 But when it comes to the AI video platforms
0:56:35 that are out there, to me, Runway always seems
0:56:37 to be like one step ahead.
0:56:39 Again, we don’t know where open AI is with video.
0:56:42 Like they haven’t released it to the public.
0:56:44 You know, we talked to Don Allen, Stevenson,
0:56:45 the third on our podcast,
0:56:47 who is somebody who actually got to use Sora.
0:56:49 And basically he was like,
0:56:50 I can’t give you more details,
0:56:52 but it’s better than you think, right?
0:56:57 So like there’s good stuff coming out of open AI
0:57:00 in that video front at some point.
0:57:02 But as far as tools we actually have access to
0:57:03 and can play with right now,
0:57:05 I think Runway sort of leading the pack
0:57:07 with AI video at the moment.
0:57:08 So I think A makes sense.
0:57:10 All right, so next up you got Pika.
0:57:12 There’s no way I’m putting Pika in the same
0:57:14 like realm as Runway.
0:57:18 Pika generates AI generated videos as well.
0:57:20 The videos are not great.
0:57:23 Like I don’t really feel like they’ve got great realism,
0:57:27 but their new sort of thing is like, you know,
0:57:29 you see the videos where like there’s a giant
0:57:32 like compressor that compresses somebody
0:57:34 or like the is it cake kind of thing
0:57:36 where a knife comes in and like slices,
0:57:37 whatever you’re looking at,
0:57:38 it makes it look like it’s cake
0:57:40 or like they got that like balloon feature
0:57:42 where it’ll blow up really big and then like float away.
0:57:45 And it’s got like all those kind of like gimmicky
0:57:49 like like effects that you can put on videos.
0:57:50 And that’s kind of fun to play with.
0:57:53 There’s not a whole lot of use out of it
0:57:58 out of like out of like memes, but it’s kind of fun.
0:57:58 Yeah, yeah.
0:58:01 And when they first came out, I got, I got like a demo.
0:58:03 I’m not sure if you did, but like a few like AI influencers
0:58:04 got a demo before other people.
0:58:08 It was at that moment, like for whatever reason,
0:58:10 like cartoon kind of characters, it seemed pretty good.
0:58:12 It seemed like better than some of the other models
0:58:14 for like that kind of art, that kind of style.
0:58:17 But then it seems like there’s been so many other things.
0:58:18 There’ve been all these models from China,
0:58:21 Kling and others that are like probably better looking
0:58:22 for that kind of stuff.
0:58:24 The special effects look cool.
0:58:25 Maybe that’s what they’ll keep doing.
0:58:27 Maybe it’ll be like the special effects
0:58:29 of viral memes kind of company,
0:58:30 but it seems like that’s expensive.
0:58:33 Like are people gonna really pay for that?
0:58:35 Like, you know.
0:58:37 – That’s where I run into like a little bit of struggle
0:58:39 with like ranking some of this stuff is it’s like,
0:58:43 I think it’s really cool and it’s fun to play with,
0:58:45 but I don’t know if I see the real world use case
0:58:47 where like, if you’re paying for it,
0:58:49 you’re gonna get ROI on it.
0:58:49 You know what I mean?
0:58:51 – Yeah, it does seem like that kind of puts it
0:58:52 in like the sea though.
0:58:55 – I think I agree with sea.
0:58:58 I think, I think it’s kind of like a novelty and, you know,
0:59:03 like realistically, most tools should fall under a sea, right?
0:59:05 Like that’s, that should be like our average.
0:59:06 If it’s like a bell curve, right?
0:59:07 So.
0:59:08 – Yeah, it’s cool.
0:59:09 It’s cool.
0:59:10 It’s fun to use.
0:59:14 It’s not the best one out there, but it’s not bad.
0:59:16 I am skeptical about the company longterm
0:59:18 just because I don’t know how they build a business
0:59:20 around that because like, well, the runway, I can see it.
0:59:23 Like if you get baked into Hollywood and stuff,
0:59:25 obviously like, like I’ve told you,
0:59:26 I dealt with Hollywood a little bit.
0:59:28 Like they spend so much money on special effects.
0:59:30 Like some of these films is like $100 million
0:59:31 on special effects.
0:59:33 Like, well, if you could cut that down to a million dollars
0:59:36 or whatever, you know, that’s huge.
0:59:39 – And runway was founded by like legit like AI pioneers.
0:59:42 Like one of the founders, I don’t remember exactly
0:59:44 like names here, but like one of the founders of runway
0:59:46 was one of the original developers of stable diffusion.
0:59:47 – Yeah, yeah.
0:59:49 So I think sea makes sense.
0:59:50 – And then you’ve got Kling,
0:59:53 which I haven’t really used that much.
0:59:55 It’s one of the few AI video tools
0:59:57 that I personally haven’t played with a ton.
0:59:58 – I played with it a little bit.
1:00:00 I have noticed like, you know,
1:00:01 I used to do those AI video threads.
1:00:04 And I have noticed that it seems like right now though,
1:00:07 most of the videos that people are putting out on X,
1:00:09 a lot of them are Kling actually.
1:00:12 So it seems like in terms of visual quality,
1:00:15 they are in the realm of runway is what it seems like.
1:00:17 It seems like they’re in the realm of runway,
1:00:18 but with like the editing tools
1:00:20 not being as good as runway.
1:00:22 – So maybe like a beat here.
1:00:22 – It seems like a beat
1:00:25 ’cause it seems like they’re putting out better videos
1:00:27 than Pica.
1:00:29 – Well, it also seems to me like Kling
1:00:32 is a lot more uncensored, right?
1:00:34 Like, if you went to runway and said,
1:00:36 generate a video of Will Smith eating spaghetti, right?
1:00:38 Like the very classic meme
1:00:39 that’s been given around for years now,
1:00:41 it’ll say we can’t do that
1:00:43 ’cause we’re not gonna make videos of Will Smith.
1:00:44 I believe Kling will.
1:00:45 I think Kling is like,
1:00:46 we’ll make whatever you want to generate.
1:00:47 – Kling’s from China.
1:00:48 They’ll do whatever you want.
1:00:50 (laughing)
1:00:52 And that is one thing I’ve seen people kind of talk about
1:00:55 online recently, like why is,
1:00:56 and there’s another,
1:00:57 there’s one or two other Chinese models too.
1:00:59 We probably should be listing them,
1:01:01 but they’re also pretty good in AI video.
1:01:03 And it’s like, why are they so good at it?
1:01:04 – Is it like how do AI or something?
1:01:05 I didn’t figure out how to pronounce that one.
1:01:06 – Yeah, exactly.
1:01:08 Yeah, that’s why it’s so hard to say that.
1:01:09 I shouldn’t be able to say it.
1:01:10 But those are all pretty good.
1:01:13 Like those are all probably better than Pica right now
1:01:15 or in the ballpark of it.
1:01:16 Pica’s got the cool special effects,
1:01:18 but like they’re in the ballpark of being,
1:01:20 somewhere in between runway and Pica,
1:01:21 which is pretty impressive.
1:01:22 They may end up being like,
1:01:25 having some of the best models out there long term.
1:01:26 – Yeah, I agree.
1:01:28 I think B is the right place for them.
1:01:30 I think they live a little bit above Pica,
1:01:31 but not quite runway level.
1:01:34 So then you’ve got Luma and Luma has their dream machine,
1:01:39 which is probably about on par with what runway does
1:01:42 as far as like the video generation.
1:01:44 Not nearly as many features as runway has,
1:01:47 but Luma’s, the cool thing about Luma dream machine
1:01:50 that I really like is that you can give it two images.
1:01:53 You can give it a starting frame and an ending frame.
1:01:55 And then it figures out how to make a video
1:01:57 that goes between those two frames.
1:02:01 So for example, if I put an image of myself
1:02:03 as the starting image and an image
1:02:06 of like a wolf howling at the moon as like the other image,
1:02:08 and I want it to like animate as me morphing
1:02:11 into a wolf howling at the moon, it can do that.
1:02:14 And like that’s where I feel like Luma really stands out
1:02:17 is the start frame end frame created animation
1:02:19 between the two frames feature.
1:02:21 I really, really think that feature is cool.
1:02:23 It doesn’t put it on par with runway,
1:02:27 but I think it puts it a step ahead of Pica maybe.
1:02:29 – Yeah, probably a B then.
1:02:30 Yeah, I play with it one time.
1:02:31 I liked it better than Pica.
1:02:33 – Well, another thing about Luma is Luma was really
1:02:35 one of the companies that led the way with
1:02:38 nerfs and Gaussian splatting and stuff like that, right?
1:02:41 Like if you have the Luma app on your phone,
1:02:44 you could go like take an object in your room
1:02:45 and like scan it in with your phone
1:02:47 and create a 3D version with the Luma app.
1:02:50 So not only does it have a video generator,
1:02:53 it also can scan in real world objects
1:02:55 and convert them into nerfs and Gaussian splats
1:02:56 and stuff like that,
1:02:59 which I know is getting really technical in the weeds,
1:03:01 but if you’re really into like creating 3D assets
1:03:04 and stuff, Luma is also capable of that.
1:03:06 – Right, yeah, I think B makes sense.
1:03:07 ‘Cause it seems like in a way they’re gonna be competing
1:03:09 probably directly with Pica then too,
1:03:10 like on special effects and things like that.
1:03:13 And it seems like they’re probably already ahead of them.
1:03:15 – Yeah, yeah.
1:03:16 All right, so next we got Descript.
1:03:18 And I know you haven’t really played with this one,
1:03:20 so I’ll just talk about this real quick.
1:03:22 But Descript is a tool where you can upload
1:03:25 any audio or any video file.
1:03:29 It will automatically use AI to transcribe that whole file.
1:03:32 And then you can edit the audio or video file
1:03:33 by editing the text, right?
1:03:35 So if there’s like a sentence that I wanna remove
1:03:37 from an audio or a sentence
1:03:38 that I wanna remove from a video,
1:03:41 I just delete that sentence in the transcript
1:03:43 and it automatically edits the audio file
1:03:47 or the video file to make that edit for me.
1:03:50 And I found it really handy for,
1:03:53 I use it as like my main transcription tool.
1:03:55 So like whenever I release a short form video
1:03:58 on YouTube shorts or Instagram reels or a place like that,
1:04:02 I always record the video, pull it into Descript,
1:04:04 get the transcription and then use that transcription
1:04:06 for the subtitles on the video.
1:04:07 There might be better ways to do it.
1:04:09 In fact, I’m sure there are better ways to do it.
1:04:11 But for me, it’s just like,
1:04:12 it’s become a really, really simple way
1:04:14 to transcribe every video that I make
1:04:16 and then make captions on that video.
1:04:18 And it uses AI to do that all.
1:04:22 Is it like something that’s life changing?
1:04:24 I don’t know.
1:04:24 I think it’s really, really helpful
1:04:27 if you make a lot of video or audio content.
1:04:30 But I’d probably put it like middle of the pack.
1:04:33 – You can edit the video with it as well or no?
1:04:37 Okay, but you just don’t use that feature.
1:04:38 – I don’t use that feature, no, no.
1:04:41 I mean, ’cause I have editors that I hire.
1:04:42 – Yeah.
1:04:43 – And I use a tool called TimeBolt,
1:04:46 which isn’t technically an AI tool,
1:04:50 to do a lot of the cutting out the gaps in the audio.
1:04:52 But Descript will do that as well.
1:04:54 To cut out us and ums, I also use TimeBolt for that,
1:04:56 but Descript will do that as well.
1:04:59 I use Descript a lot for just getting transcriptions.
1:05:01 Another thing, Descript was like,
1:05:03 even before 11 Labs came out,
1:05:05 Descript was the first tool that I ever saw
1:05:07 that can clone your voice.
1:05:12 So if you had a podcast and you misspoke during the podcast,
1:05:15 and you wanted to go and change one sentence,
1:05:18 it actually, you can train it on your voice,
1:05:20 go and type in the new sentence you wanted it to say,
1:05:22 and it would replace the original sentence
1:05:26 in the audio version of that with you saying it.
1:05:29 Now 11 Labs does it so much better.
1:05:32 The Descript version is still fairly mechanical AI sounding
1:05:35 when you do it, but it was the first tool I ever remember
1:05:38 seeing that had like voice cloning built into it.
1:05:40 – Yeah, it’s, I guess, I can’t argue with that.
1:05:44 I guess, see, it like the editing sounds way more interesting
1:05:45 to make it like the transcript thing.
1:05:47 It seems like that’ll just be like,
1:05:50 so you’ll be able to do it with like anything in the future.
1:05:51 Like I can imagine even YouTube
1:05:52 will probably just have all that.
1:05:54 – Well, YouTube does have a transcription built in too,
1:05:56 but it’s like, you got to upload the video first,
1:05:58 and I need the transcription before the video’s uploaded.
1:06:00 – Right, got it.
1:06:02 – But yeah, I mean, I think like transcription
1:06:04 is going to be very commoditized.
1:06:06 You’ve got the Whisper API from OpenAI.
1:06:07 It’s really cheap to use.
1:06:10 So it’s really easy for tools to just build in transcription.
1:06:13 – But it’s somewhat useful for you today.
1:06:13 So like seed makes.
1:06:15 – It’s somewhat useful for me today,
1:06:17 but I also think like, if I was,
1:06:19 if I didn’t already have like workflows in place
1:06:22 and systems in place for creating and editing my videos,
1:06:23 I would probably use Descript a lot more
1:06:25 for editing the videos.
1:06:27 It’s just like, I haven’t really wanted to change
1:06:28 the workflows that I have.
1:06:30 So I think anybody that’s trying to get into podcasting
1:06:31 and they’re going to edit their own stuff
1:06:34 or anybody that’s trying to get into like creating video
1:06:36 content and they want to edit their own stuff,
1:06:38 I really think Descript is worthwhile
1:06:40 for like speeding up the editing process as well.
1:06:41 All right, we got three more.
1:06:44 Now we’re getting into like the audio apps.
1:06:46 Yeah, so we got Soono first.
1:06:48 I love Soono.
1:06:50 Like Soono is so much fun to me.
1:06:51 – Yeah.
1:06:53 – I can’t put it in last year though.
1:06:55 It’s too much of a novelty.
1:06:56 – Yeah, yeah, I love it.
1:06:58 I’ve only used it twice.
1:07:01 So I think you use it more than me, right?
1:07:04 – I’ve made hundreds of songs with it.
1:07:05 – Oh, that’s awesome.
1:07:07 Yeah, and you like, you play music or you used to, right?
1:07:08 So that makes sense.
1:07:10 – Yeah, yeah, I’m a musician.
1:07:12 So I’ll use songs that I make in Soono
1:07:13 instead of YouTube videos.
1:07:15 Like if you want background music
1:07:16 and you don’t want to worry about copyright,
1:07:20 you can make Soono make you a song without any lyrics
1:07:21 and use that as your background music.
1:07:24 So you can say, I need some like calm emotional music
1:07:25 for this part of the video
1:07:27 and then I need upbeat, exciting music
1:07:28 for this part of the video.
1:07:30 And you can tell it to not generate lyrics
1:07:32 and it will generate like a sound bed
1:07:33 for you for your videos.
1:07:35 I’ve also made like montages
1:07:36 where I have like a whole bunch of clips
1:07:38 from an event that I went to.
1:07:40 And then I actually typed in the lyrics that I wanted.
1:07:44 Like, oh, now I’m looking at, you know,
1:07:46 now I’m looking at the Insta360 cameras
1:07:47 on the wall over here.
1:07:48 Those are pretty cool.
1:07:51 Look over here, it’s a really cool futuristic car
1:07:53 and look at that, it’s a flying car.
1:07:55 And like I would put all these like random things
1:07:58 that I was like, basically what I’m seeing in the video,
1:08:00 I would put that as the song lyrics,
1:08:04 make a song and then make the montage video
1:08:07 like sync up to the song where the lyrics of the song
1:08:09 are actually what I was looking at in the video.
1:08:10 And you can do like really fun stuff like that
1:08:13 that makes for really engaging videos.
1:08:15 So it’s fun and I use it.
1:08:17 And most of the time I’m using it,
1:08:20 I’m using it just to screw around and just like play with it.
1:08:21 – It’s so much fun.
1:08:26 I used it to make a song to my wife when she was my fiance.
1:08:29 And it was just like a love song to her.
1:08:31 And like, I even like, I wrote in like our like story
1:08:32 of like meeting in Kyoto and stuff
1:08:34 and it made this awesome song, you know?
1:08:35 – Yeah, yeah.
1:08:38 – So it seems like it’s like, it’s amazing technology.
1:08:39 It’s a novelty.
1:08:41 It’s kind of like, it’s just kind of fun to use,
1:08:43 but you do get some real use out of it.
1:08:45 So it seems like that’s probably like what,
1:08:47 A or B, something like that?
1:08:49 – Yeah, I’d probably have to put it in a B,
1:08:52 just like going back to like how often I use it
1:08:55 for like actionable practical use cases.
1:08:56 – Yeah.
1:08:57 – And not very often.
1:08:59 I’m using it for fun and screwing around more often
1:09:00 than anything.
1:09:03 Would I put it on the same par with Claude?
1:09:04 That would be tough, right?
1:09:06 If we’re saying Claude’s an A,
1:09:08 like I don’t think I can put Sudo at the same-
1:09:09 – Yeah, B makes sense.
1:09:10 – As Claude.
1:09:14 All right, and then you got 11 labs.
1:09:16 So anybody who doesn’t know what 11 labs is,
1:09:17 it’s a voice cloning tool.
1:09:18 You can train it on your own voice.
1:09:19 You can train it on anybody’s voice.
1:09:21 You type in words and it will speak out
1:09:23 in a very realistic tone.
1:09:25 It’s pretty valuable.
1:09:27 And they keep on adding like features, right?
1:09:29 Like they’ve got the 11 labs reader now
1:09:31 where you can load in blog posts or PDFs
1:09:33 and it will like read it in your voice
1:09:35 or it’ll read it in whatever voice you choose.
1:09:36 And so you could go out for a walk
1:09:40 and turn like a PDF or a news article
1:09:42 or something like that into almost like a podcast
1:09:44 that you listen to, but it sounds like a real human
1:09:47 and not like an AI robot talking to you.
1:09:48 So that’s, you know, that’s useful.
1:09:50 It also creates sound effects now too.
1:09:52 So you can say like, I need a clang sound effect
1:09:55 or I need like a dog barking sound effect
1:09:58 or a car crash sound effect or whatever, right?
1:10:00 And it will create that sound effect for you.
1:10:02 So it’ll create talking audio.
1:10:03 It’ll create sound effect audio.
1:10:07 It can be used to like make audio from text files.
1:10:10 Like there’s a lot that you can do with it.
1:10:11 – Were they the ones too
1:10:12 who created the whole persona thing?
1:10:14 Or was that Suno?
1:10:15 I feel like there was one where
1:10:17 that you can like create like a personality now
1:10:19 and like you kind of like save that personality.
1:10:19 – That’s Suno.
1:10:21 Suno just created the Personas feature.
1:10:23 – Oh, yeah, I was getting confused.
1:10:23 Okay.
1:10:25 I was like, oh, that’s great.
1:10:27 – With 11 labs, you can train your own voice into it.
1:10:29 And then they even have a marketplace
1:10:31 where you can sell access to your voice if you want.
1:10:32 In fact, Matt Vid Pro,
1:10:34 he actually put his voice in the marketplace
1:10:37 and then started seeing ads online
1:10:39 with his own voice promoting a product.
1:10:42 And he’s like, I should take my voice off the marketplace.
1:10:43 – Right, it’s probably worth more money
1:10:44 than that too, like longterm.
1:10:46 But yeah, especially if it’s like promoting the wrong thing.
1:10:47 Right?
1:10:48 – Yeah, yeah, yeah.
1:10:50 – Here are these pills that help you, you know.
1:10:51 – I’m Matt Vid Pro.
1:10:53 Here’s why you should buy blue chew.
1:10:53 Yeah.
1:10:57 – Rhino pills or whatever.
1:10:57 – Yeah, yeah.
1:11:02 Here’s my gut feeling, and we can debate it a little bit.
1:11:04 For me, 11 labs is eight here.
1:11:05 – I was gonna say A is what I was gonna say,
1:11:07 ’cause like it’s actually being used by people.
1:11:09 Like there’s people in certain industries
1:11:11 are actually using it right now.
1:11:12 – Yes, like a lot, a lot.
1:11:15 Like people are using it for doing voiceovers
1:11:16 on like sales videos.
1:11:19 People are using it for the faceless channels.
1:11:21 Like it’s really, really popular
1:11:23 among like the faceless YouTubers, right?
1:11:25 They just put out content.
1:11:28 You can create voices that never existed before.
1:11:30 You could say I want it to sound like a, you know,
1:11:33 an 80 year old grandpa who, you know,
1:11:35 stepped on attack or whatever.
1:11:38 Like it’s just like random prompt.
1:11:39 – Do we wanna reorder these at all?
1:11:41 Like do we wanna say like, oh, it should be ahead of runway
1:11:43 ’cause people are actually using it?
1:11:44 Or does that not matter?
1:11:46 Do we not, don’t worry about it.
1:11:48 – Yeah, maybe it does go ahead of runway.
1:11:51 And also when it comes to like voice AI,
1:11:54 so I actually interviewed one of the founders of 11 labs
1:11:58 for like an NVIDIA panel that we did a while ago.
1:12:02 And when it comes to like the voice, the speech AI,
1:12:04 like 11 labs is leading the way, right?
1:12:06 Like there’s a lot of other companies out there
1:12:10 that are like voice clone, like AI voice tools.
1:12:13 But like 11 labs is the chat GPT of voice tools, right?
1:12:15 It’s the one that everybody’s trying to catch up with.
1:12:17 It’s the one that everybody says
1:12:19 we’re almost as good as 11 labs, right?
1:12:24 Like they’re the one everybody’s, they set the bar, right?
1:12:27 – Right, I mean, the fact that so many people
1:12:29 are using it in facelift channels, that’s like awesome.
1:12:31 That’s like an AI tool that’s actually being used
1:12:33 that changed an industry, right?
1:12:35 So definitely makes it say.
1:12:37 – And then the last one,
1:12:38 I’m not sure if this is the proper icon or not,
1:12:41 but it’ll have to do the trick for this one is a notebook LM,
1:12:43 which as we talked about at the beginning of this
1:12:45 uses Gemini underneath,
1:12:48 but we’re more specifically talking about like the feature
1:12:51 where you can upload a whole bunch of documents
1:12:55 or YouTube transcripts or links to articles.
1:12:57 And then you can ask the questions
1:12:59 and like actually have a chat with all those documents
1:13:03 and it will also create like an audio narrated podcast
1:13:05 where it actually sounds like two hosts
1:13:07 talking back and forth about the topic,
1:13:10 which I think is really cool as well.
1:13:11 – I have mixed feelings.
1:13:12 So like– – I know, this one’s gonna be tough,
1:13:13 I think.
1:13:16 – I think it’s an awesome technology.
1:13:17 It’s a great product.
1:13:20 I can see so much potential for it,
1:13:22 especially like in like learning and things like that.
1:13:24 Like, you know, you want to learn a subject,
1:13:26 you can make a little podcast and you can listen to it.
1:13:27 That’s awesome.
1:13:28 It’s entertaining.
1:13:29 My son was like,
1:13:31 I don’t think I’ve ever seen him laugh as much
1:13:33 as like when we created a podcast talking about,
1:13:36 oh, by the way, the earth has just been invaded
1:13:39 and like, you know, one of you is actually probably an alien.
1:13:42 My son just like thought this was so hilarious
1:13:44 listening to them, kind of talk that out.
1:13:46 Like, by the way, maybe it’s you, you know, kind of thing
1:13:48 or like, this is weird.
1:13:50 I’m just really skeptical about Google’s ability
1:13:52 to make it mainstream or like,
1:13:55 to release new products in general
1:13:56 and have them become popular.
1:13:59 I’m skeptical of that because I haven’t seen it.
1:14:01 So, I don’t know, I have a,
1:14:03 it’s probably A or B though, I guess.
1:14:04 Some, something like that.
1:14:05 – Yeah.
1:14:07 – It’s one of the most amazing like things
1:14:09 I’ve seen in the last, in the last six months.
1:14:10 – Definitely.
1:14:11 I think it’s the most impressive thing
1:14:14 that has come out of like Google in a while, right?
1:14:15 – Yeah.
1:14:18 – Like, Gemini’s always felt like it was playing catch up
1:14:20 with chat GPT, right?
1:14:22 It has the bigger context window,
1:14:23 but other than that, it kind of felt
1:14:25 like it’s playing catch up.
1:14:27 But then when notebook LM came out,
1:14:29 it was like using Gemini underneath
1:14:31 and everybody went, this is like,
1:14:32 this is a cool use case.
1:14:33 This is a really, really cool way
1:14:36 to like consume information.
1:14:38 And I’m using it quite a bit, right?
1:14:40 Like, I’ll pull in a whole bunch of like news articles
1:14:43 to try to understand what’s going on with something
1:14:45 and have them explain it to me.
1:14:46 Dang that.
1:14:49 I do know it has some issues.
1:14:52 Like, it will get confused about things sometimes
1:14:55 where if you give it like two different news articles,
1:14:58 it will say like, I’m having a hard time
1:15:00 thinking of an example,
1:15:03 but it will like say something happened
1:15:06 in one of the news articles that didn’t actually happen
1:15:08 because it got confused with the wording
1:15:10 between the two news articles.
1:15:12 So sometimes I know it can get confused
1:15:15 and the speakers on the podcast
1:15:17 actually give the wrong information
1:15:19 because of its confusion
1:15:22 between all the documents you uploaded.
1:15:23 Yeah, I do wonder.
1:15:26 There was some recent research.
1:15:27 I think it came from, what’s his name?
1:15:30 That guy, Ethan, a professor at Wharton.
1:15:31 I think he shared it on X.
1:15:33 I think he shared it showing that the more,
1:15:35 that there’s new data out,
1:15:37 showing that the more restricted a model is,
1:15:39 like the more restrictions they put on it
1:15:41 in terms of being more censored and things like that,
1:15:44 the less intelligent the model is,
1:15:46 because you had all these different layers of complexity
1:15:49 there that just makes it less intelligent.
1:15:50 So I do wonder if that’s like one of the problems
1:15:52 they’re having, like Google’s having behind the scenes
1:15:55 in general, like just because of the culture of Google
1:15:58 or whatever is that it’s all more restricted.
1:16:00 So it just, it’s ended up being less intelligent,
1:16:01 even though it’s got great technology
1:16:04 and a huge context window and all that.
1:16:06 And I do wonder, like it feels like, you know,
1:16:07 the last great product that Google released
1:16:11 that I can think of is like Gmail, right?
1:16:14 And then you got Google Analytics,
1:16:16 which they killed, basically.
1:16:18 – Yeah, yeah.
1:16:20 Are you talking about like products that Google built?
1:16:23 ‘Cause I mean, they’ve acquired some good companies.
1:16:25 – Oh, YouTube, yeah, of course they acquired, yeah.
1:16:27 – I mean, YouTube, you’ve got Google, you know,
1:16:30 the whole Google Drive, Google Docs, Google Sheets,
1:16:31 all that like whole suite of tools.
1:16:33 I use those constantly.
1:16:35 – Man, Jim and I just feels like Google Plus to me.
1:16:37 And so like I just feel like, you know,
1:16:38 where they just copied another company
1:16:41 and no book L.M. is not that it’s like it’s unique.
1:16:44 So I wanted to like get a decent ranking.
1:16:46 I just feel like they, it’s not there yet.
1:16:48 Like it’s like has so much potential
1:16:49 and they’ve shown the potential.
1:16:51 And I’m not sure they’re gonna be the one
1:16:54 to actually capture the value from that potential.
1:16:56 – Yeah, I also ran into some issues
1:16:58 when I was trying to create some content.
1:16:59 Like they just added a new feature
1:17:00 that I was really excited about
1:17:01 where you can give some additional context
1:17:04 and sort of steer the conversation, right?
1:17:06 Like there’s a new like customize button.
1:17:08 And I pressed the customize button and I said,
1:17:11 hey, you’re a podcast called,
1:17:12 I don’t remember where I called it,
1:17:15 but you know, you’re a podcast called the next wave.
1:17:17 When your episode starts, say,
1:17:19 hey, welcome to the next wave
1:17:21 and then move into the rest of the conversation, right?
1:17:23 That wasn’t what I was calling it.
1:17:25 That’s just for example purposes.
1:17:27 And it ignored my instructions completely, right?
1:17:29 I also gave it instructions one time
1:17:30 where I gave it like transcripts
1:17:32 from like three different videos
1:17:35 that were interviews with Casey Neistat, right?
1:17:37 And I said, in the custom instructions,
1:17:40 pull out what Casey Neistat does different
1:17:41 than other YouTubers
1:17:44 and make a podcast about what makes him different
1:17:46 and you know, so much more successful
1:17:48 than so many other YouTubers.
1:17:50 Totally ignored my instructions
1:17:52 and just like talked about Casey
1:17:53 without specifically honing in
1:17:55 on the things I asked it to hone in on.
1:17:57 I almost felt like the customize button
1:18:01 that they put into notebook LM is pointless.
1:18:03 Like right now, like for me, I couldn’t get it to work.
1:18:06 It like, it didn’t follow my extra instructions.
1:18:09 So that was interesting to me as well.
1:18:10 – Yeah, it does make me wonder
1:18:13 that that’s like possibly a bad product decision.
1:18:15 So that, and again, that’s like my criticism
1:18:16 of Google in general.
1:18:18 So like, maybe this is like some great research
1:18:20 that turned into like a really simple product
1:18:22 but can they take it to anything beyond that?
1:18:25 – Yeah, well like it was originally Project Tailwind, right?
1:18:28 And Project Tailwind was designed to like
1:18:30 allow you to chat with all of your documents
1:18:32 that you put inside of like a specific folder
1:18:33 inside of Google Drive or whatever, right?
1:18:35 And that’s what this became.
1:18:37 The podcast thing was, I think,
1:18:39 just like an extra fun little like novelty feature
1:18:40 that they’re like, oh, this is kind of cool.
1:18:42 Let’s put that on top of it.
1:18:43 And it turned out to be the feature
1:18:45 that people like the most from it.
1:18:47 And so I think that’s kind of how that played out.
1:18:48 – Yeah, so in terms of like,
1:18:50 it’s like one of the most impressive things
1:18:51 I’ve seen the last six months,
1:18:53 there was like part of me that wanted to put it like as a,
1:18:55 but I do feel like maybe it’s B or C
1:18:56 just because of, I just don’t know
1:18:57 what they’re gonna do with it.
1:18:58 Like, I don’t know if it’ll actually
1:18:59 become a big thing or not.
1:19:01 Like it, it feels like maybe it’s already like,
1:19:03 it’s got that novelty, it’s been cool.
1:19:04 Like especially some of the videos
1:19:06 where like they were surprised where people told it like,
1:19:09 oh, you are AI by the way and it actually freaked out about it.
1:19:11 Like that was an amazing viral clip,
1:19:14 but what they do with it beyond that, I’m not sure.
1:19:15 Like I’m not sure if they’ll actually improve it
1:19:16 and make it a big.
1:19:19 – Yeah, I’m leaning towards B, I think.
1:19:21 I have a hard time putting an A tier.
1:19:23 Like it’s not on the same level as Claude 11 lives
1:19:24 and runway for me.
1:19:27 – Yeah, C seems to mean cause it is a pretty cool.
1:19:28 – Yeah, like I’m definitely using it
1:19:31 more than Grock or Replit or, you know,
1:19:32 any of the tools that we’ve got down there.
1:19:35 So I don’t know, B feels right to me.
1:19:37 B would have potential for S.
1:19:39 So like if they actually like make it better
1:19:41 and like do a good job with the product,
1:19:43 they should like, now it’s kind of hidden on the side.
1:19:44 It’s like so hard to even find.
1:19:46 Like they need to like launch it
1:19:49 as like a standalone product
1:19:51 and like really nail some kind of use case,
1:19:53 maybe education or something like this.
1:19:55 Like you’ve got a knowledge base, hit your company,
1:19:56 now make some stuff that explains stuff
1:19:58 about your company to people, something.
1:20:00 They need to nail some kind of use case
1:20:01 and launch a product.
1:20:02 – Here’s what I can see.
1:20:04 And I’m just sort of shooting from the hip here,
1:20:07 but like a combination of what perplexity does
1:20:09 and what notebook LM does,
1:20:13 but like in like a conversational format, right?
1:20:15 Like let’s say I want to go for a walk
1:20:17 with my dog or something.
1:20:19 It would be really cool is to open up a tool
1:20:22 that’s like this notebook LM perplexity hybrid
1:20:25 and just say, hey, today I want to learn about quantum physics.
1:20:27 Give me a podcast about quantum physics.
1:20:30 It goes and does the like sort of perplexity thing
1:20:33 of hunting down all the articles and information you need
1:20:36 and then pulls them into the notebook LM side of things
1:20:38 and turns it into that podcast.
1:20:39 So now all I’m doing is giving it an audio command
1:20:42 of like, here’s what I want to learn about.
1:20:43 And then it gives me a whole podcast
1:20:45 after doing the research for me.
1:20:46 – Yeah.
1:20:48 – That seems like it should exist.
1:20:49 – But that’s why perplexity is S tier
1:20:51 ’cause they’re more likely to do that than Google is.
1:20:53 – Yeah, I agree, I agree.
1:20:54 (laughing)
1:20:54 And you know, all right.
1:20:56 So this is like, that’s all the tools we came up with.
1:20:58 I know there’s tools that we’re missing.
1:21:01 I know like there’s probably a lot of people that disagree.
1:21:03 Like I’m actually looking forward to seeing the comments
1:21:05 on this episode ’cause I want to see people go,
1:21:08 I can’t believe we put that in D tier or whatever.
1:21:10 I think that’s gonna be really fun to see and debate.
1:21:12 But at the end of the day,
1:21:15 I think every single tool that we put on this list
1:21:17 has the potential to jump to S tier at some point.
1:21:20 Also has the potential to fall to F tier at some point.
1:21:22 I think, you know, like this isn’t anything personal
1:21:23 about the companies.
1:21:25 Like I actually, for the most part,
1:21:28 like all of these companies and have messed with the tools
1:21:30 and think what they’re doing is really, really cool.
1:21:31 So there’s like, you know,
1:21:34 I’m trying to save face here a little bit and be like,
1:21:36 “Hey, don’t hate us to put a D in this tier.”
1:21:39 But, you know, I do think all of these tools
1:21:41 have the ability to like move up and down.
1:21:43 This is sort of more of a ranking of like
1:21:45 how useful they are to us in our lives,
1:21:47 which is obviously very subjective right now.
1:21:52 So, you know, just want to kind of disclaim that
1:21:53 before we wrap this one up.
1:21:55 But I don’t know, this was kind of fun.
1:21:57 I think this is gonna be our longest episode
1:21:59 we’ve ever put out, but it was like really fun
1:22:02 to sort of like rip and talk about all these tools.
1:22:03 – Yeah, it was more fun than I expected.
1:22:04 I think if people like it,
1:22:06 I think we should like go into like different categories
1:22:08 ’cause obviously with the AI video,
1:22:10 there was like two other companies we could have listed
1:22:12 and like with AIR, there was one or two others.
1:22:14 We could do like a different categories
1:22:16 and like actually spend time to like
1:22:18 make sure we’ve tried the latest version of them too
1:22:19 ’cause some of these, you know,
1:22:20 haven’t tried them recently.
1:22:21 – For sure.
1:22:22 Well, cool.
1:22:24 I think on that note,
1:22:26 we should probably go ahead and wrap this one up.
1:22:28 Thanks everybody for tuning in.
1:22:30 If you enjoy this type of content
1:22:33 and you wanna learn about the latest AI news and tools
1:22:35 and keep your finger on the pulse
1:22:37 and learn actionable strategies to use AI,
1:22:39 make sure you subscribe to us on YouTube
1:22:42 or follow us wherever you follow podcasts.
1:22:44 And thank you so much for tuning in.
1:22:45 Really, really appreciate you
1:22:48 and hopefully we’ll see you in the next episode.
1:22:49 – Thank you all.
1:22:51 (upbeat music)
1:22:54 (upbeat music)
1:22:56 (upbeat music)
1:22:59 (upbeat music)
1:23:02 (upbeat music)
1:23:04 (upbeat music)
1:23:06 you