Hank: I'm sure the ASMR people are going nuts right now. I actually would keep them and take them to a rave. To a rave? Yeah, it's fun to have them. Is this why you got to bring the nanny to Europe so you can go to a rave? Oh, also, do you want to see my swag from Laravel Live Japan?

Gonto: Yeah.

Hank: Okay, well, I got two items. Here's what we got today, guys. I got two items of swag. They're simple. That's quick. Then, some meaty topics. We've got a CMS that, like the CMS, Contentful Getting Bought. We've got discussion, debate, the look into the future on Loops from basically the two smartest guys when it comes to coding agents. And then there was this fun Mafia video that came out that we got to talk about. So first, my swag. Here's the coolest one. Ready? Very Japan. Come on, this is cool. Check out the, like, the little, you know, signature stamp is Laravel. This one says Kaizen, Continuous Improvement. And then I've got another one that says Artisan. I'm sure the ASMR people are going nuts right now.

Gonto: I actually would keep them and take them to a rave. To a rave? Yeah, it's fun to have them.

Hank: I'm not a rave guy, but I guess you are. Is this why you got to bring the nanny to Europe so you can go to a rave? Okay, the other thing is, you know, I love, like, we have these patches for our jackets. Oh, you can see the Denver one up here. You're watching the video patches. I mean, the patches are awesome. Like, you put this on your jacket and it's sick. And look how sick this one looks. If you're not watching a video right now on YouTube or Spotify, what are you doing? Even listening to us pull the car over and check it out. Okay, those are the swag. Those are the topics I brought today. Today you brought the meat because I'm still jet lagged. CMS got bought. This is like the biggest CMS by the biggest CRM. What's your take?

Gonto: Yeah, like, I'm curious for what Salesforce bought them. It's like, my question is more about the divide for the people or would they actually start selling more CMS with folks? Like, I don't know. We talked about it in other episodes. That's the dream of the CMS is fake. Where before, when you were told to use a CMS for your product pages, for your homepages and stuff like that didn't work because when a new feature came out, they wanted to add a new image or change the format or do a new animation or something. And with the CMS, you could just change the text or maybe add like a section with some already built-in things, which looked meh. So I don't know. Like, I think they did well on selling, even though they didn't sell well. Like they sold at half of the last round. So from acquisition side, I think they fucked it. They fucked up mostly because they probably weren't growing, weren't doing well. So I don't know. What's your take?

Hank: I think it's a weird, I mean, it's a sign that Salesforce is a little too clingy to the past. I think CMSs, they're just going to go out of style. Like, agentic-first CMSs are going to be the way to go. And I'm seeing this, like, I've got a client who's going all-in on zero CMS and just giving everybody, they're using Repl.it agent, and they're just giving everyone who needs access to change the site just access through Repl.it. And for me, I showed off, it would have been, I think, was it last episode or two episodes ago? I showed off my sites and I just have a skill now that I home-baked that is my CMS skill. So I have a special URL I go to on my sites and I enter, like, my authentication or password, and then I can edit anything on the site. And I love it because I'm not stuck. Like, part of the problem with the CMS is that the cure is also a new disease. Like, what you're curing is that you can't move fast enough. Basically, what you're curing is that marketers don't have an engineer at their beck and call. But now they do. And so now this cure of being able to iterate quickly and publish pages quickly, the new disease is you get no creativity, no innovation, and all your pages look the same. And OK, I said in our pre-show, I said I wasn't going to show my sites and self-promote today, but I want to show something that you can't do with a CMS. Let me pull this up. So if I go to my code-to-market.ai, first off, you've seen progress from last time. OK, this little speech bubble thing. In a CMS, you might get a banner, right? But you could never do something like this. OK, I got to hover it. Oh, I just moved. Oh, got him. And then there's that. Did you see it? Did you like it? Was it funny?

Gonto: I liked it.

Hank: I had seen it before, I'll be honest.

Gonto: But I thought it was good. I agree.

Hank: Like, this type of creativity is very hard with a CMS. And that's just like a dumb little thing that I threw in in the last minute before I published the site. Like, oh, I need a banner for the first people who get a discount.

Gonto: And I think part of the problem also of the CMS is that they focused on the wrong pages and the wrong things. Like, they focused on home product pages, solution pages. I just invested, for example, in a company, they're called Flint, where what they help you is actually build custom landing pages with your design using AI. But their focus is not your home or your product page. It's basically having one landing page for each ad that is basically custom done for that. It's creating landing pages for AI to find and have a better rating based off of that. Have special pages for each of the companies that you want to sell to and do some type of account-based marketing with AI. And like, that to me should have been the right focus for the CMS. It's the pages that are maybe not on the front and center of your site, but still you want them to look good. But they don't have to be home-level good. They have to be good, be your brand, but you don't want to focus anytime. You don't have engineers. So having something that actually helps you build those landing pages where you can just prompt them and edit if you want the copy, or if they add change that they automatically change, I think that's a better problem to solve and a space that I believe more in, which is why I invested in Flint, because I think that focus makes sense and not the CMS focus. And then for the focus that the CMS focus on, like home and product, you just custom build them now either with engineers or you bytecode them using an existing design system.

Hank: And I think engineers are also going to be more comfortable with marketers and content creators having direct access to the site because when you do it through an agent, the agent is going to be much safer and better at coding than, you know, marketers have been and are much less likely to break the site. Now, there are other risks, but we don't have to get into those. Now, speaking of agents and access and prompting them to update your site, our two boys, what are their names? Pete and Boris?

Gonto: Peter and Boris.

Hank: Yes. Peter and Boris, they're telling us we shouldn't even be prompting anymore. We should be building loops. Can you explain what the heck that means? Because I have thoughts.

Gonto: Yes. And before I explain a bit of what it means, I think not understanding what it is, which I think was what was happening on Twitter, is the opportunity for everybody. Like there was one day difference where there was a video of Boris saying, I just implement loops and they will prompt the agent. And then Peter did the same thing the next day, but they didn't say how that would be done. So there's two things that I think happened. One is one guy actually wrote a really good blog post about this. His name was Matt Van Horn, where he actually explains it and he promotes his open source library. This could have been, I think, a fantastic opportunity for any of the AI companies to actually explain this and how to do it and why. I actually texted my customer.

Hank: Yeah, if you see a wave, this is classic Ganto. If you see a wave starting to pick up, surf it.

Gonto: And specifically on this one, people were so fucking confused that I thought it was a great time to do so. I chatted, for example, today with Factory AI on like, look, we can explain how to do loops inside Droid and make it automatic for people so they understand that it works great. And I think a lot of companies missed that opportunity and part also like Peter, for example, replied to a random guy saying like two projects that he built that were examples of this. And first of all, most people didn't see a reply because it was a reply to a random dude. And I was just obsessed and reading everything. And then people could have used those to actually build an explanation or build something based on those two. So that I think is the important part of this. And overall, if I understood it correctly, I don't think it's rocket science. It's more like you have an agent orchestrator that will launch other agents to run things. And then when they come back, it will test the things that they do. And if they are meh, it would just either push them to do more or create another sub agent. But basically what you do is that loop is this orchestrator agent and you only implement that orchestrator agent to then call and use the others. And doing this meta prompt is different and it's a different thing that you need to do. So that's what I think about these loops. But the main thing I wanted to come up is these are the things where you have to reply really fast, even more when the two geniuses of AI coding are talking about it. What's your take, Hank?

Hank: I agree with all that. Like you should be catching the wave, like offering explanations and education because it just gets you the impressions and traffic. If I can offer what is potentially a controversial opinion on this topic of loops and whatever, and just where working with AI is going. Here's the controversial opinion. The most important skill of 2026 and 2027 is middle management. The thing that everybody has said is useless for years, which is basically having meetings and telling other people to do stuff. That's basically what you have to do now if you're managing agents. I actually disagree. I don't think... Well, let me explain a little more then. So the reason why I think it relates so well is like, okay, what is a loop if not a recurring meeting where you're just like, hey, are you doing your job? Hey, are you doing your job? Hey, are you doing your job? Please do this. Like, give me the data. Cool. I'll make a couple of decisions and now go back to work. Like that's what a middle manager does. And I think that's basically what we're being told to do with agents. And I think it's effective.

Gonto: Okay. Tell me why you hate that. Yeah, I disagree. Mostly because the main problem with middle managers is that they're not good at making decisions. And they are more about like, they hear from leadership what to do and why. And then they make sure that whatever leadership decided gets done through recursive meetings, pinging people and stuff like that. But I think with agents, it's actually more important to have the thought on where to go to, how to build it and why you're going that way more than anything else. And I don't think middle managers ever had that. That always came from leadership.

Hank: I suppose. Well, so in a way, I agree with you. But that's just what being a good middle manager is. And they do exist. But the problem is, you know, most middle managers aren't great. So the thing here is, is there a world like, how does that affect everything? Like, let's just think of marketing because that's what we're talking about here. Should there just be fewer marketers and like a marketing exec and director? Basically, the leaders should just be having these loops more because they're the ones who know how to judge the work and what work should be done and how it aligns to the C-suite and founders. Like, what's the point of like the very junior roles anymore? Are we about to have the same crisis as like junior engineers are having? I think so, personally.

Gonto: Like, I think it's not just exec, but you do have some ICs that are really good at the vision and the creative side. Like, I think they will do well. The others will not. I was talking to a friend, for example, and it was interesting because what he hated, it was the one thing that I liked. What he told me was that he preferred coding as before because there was a thinking time. But then once he had the idea and he knew what to do, he was basically just coding. And that was more relaxing. And he liked the mix. Now he says that everything is just thinking, which I actually think is more fun. I like it. But I understand his point on it. And I think that's going to be the future for us. It's like, if you enjoy thinking all the time, you're going to do well. And if you're assertive in some ways, if you don't enjoy the thinking and you want something else, I think you're going to lose your job.

Hank: There's a lot less zoning out. I mean, yes, I can't get over just the instant gratification I get now where I would move heaven and earth to get like a good engineer and a good designer just in my control for a project. And then I would do these loops. I would have a meeting and be like, here's the thing. Let me spit out some crazy ideas. You guys tell me what's possible, what's not. I argue with you about what's possible. And then I say, cool, we're going to meet again tomorrow would be like usually the fastest. Sometimes I could be like, we meet in the morning. I'd be like, you guys got to show me progress in the afternoon. You know, the double day meetings when you're like really in a crunch. But usually it's like one day later or two days later or sometimes when it's not so urgent, it's every week. And then you come back and the progress is done. And now I'm like, and this is really funny, actually, because Claude Code will build a plan with me and it'll give me these estimates. I'll be like, this is like three to four days of work. And I'm like, no, you have 20 minutes, dude. Like I'm going, I'm going to go take a call or I'm going to go eat lunch and I'm going to come back and it's going to be done and it's going to be tested three times over and all that. And you better have, you know, cogitated for 20 minutes. And that thing that usually took one day or two days, it's done in 20 minutes. And then I do it again. So I don't know, to your point, the people who like thinking and iterating and have a lot of focus for like ideas, I think they're going to thrive.

Gonto: I 100% agree. And yeah, that's I think where these loops is going to help with. And I think it's the only problem with loops now is that they are too fucking expensive still. Like it's prohibitive where these two folks can say this because they have infinite tokens. But that's honestly the main blocker now. But I think as we start improving this and we start having agents pick the best model for a task, I think it will do much better.

Hank: And like you can start small with stuff like Claude Code routines are awesome. And Codex has a similar thing. You can set up cron jobs other ways, but they don't have to be. I mean, one way to get started on this type of thinking is set up a daily report or I have a because I have a bunch of like repos now with a bunch of dependencies. I have a routine that just checks all the like alerts and clears out all the security crap every morning. That's a neat little thing that, you know, normally some engineer would have to think about that. But I don't have an engineer. I've got Claude Code. I don't know. So what do you think happens to the junior roles? Like, what are the junior roles in marketing supposed to be doing?

Gonto: That I don't know. Unless they are extremely creative, like there's junior people that are really creative that maybe can come because experienced people come with baggage. Junior people come with no baggage. So they can come with new ways and new thoughts on doing things. And maybe others do not. That I think can help. Other than that, I don't know.

Hank: The with baggage point is huge, by the way, because I have been part of and also been told about a lot of like CMO and VP marketing searches and all the like really experienced, like huge resume CMOs are flocking to AI. The majority of them haven't even opened up like Claude Code or Codex. And so they'll meet with these, like some of the CEOs are smart enough to actually ask them how they're using AI. And they're basically like a lot of these CMOs with baggage. They're just answering, oh, I use GPT. They have flowery answers. But most CEOs can see right through it if they think to ask. And then the ones who didn't think to ask, like they're firing these CMOs in like three or four months. And it's crazy. So if you're a founder listening, make sure you're asking how people are using like coding agents when you're interviewing them.

Gonto: Yeah, I think that's true for all roles. Like I interviewed recently a head of marketing ops and they didn't have really good examples on how they were building AI tools for the rest of the org or how they were using AI. And I think it's like when they say something, you just re-ask, re-ask, re-ask and double down until you get the specifics. And if there are no specifics, they really don't know. But I think you need an AI build for whatever role you hire at this point.

Hank: Otherwise, you're going to be stuck doing things the old way. I mean, and old tech is expensive and you're going to be buying CMSs and scoring platforms and like all this junk. I had a client who had a bad sales ops like interview exercise. And to prove it, I just popped it into Claude Code. And I said, do this exercise. And 10 minutes later, I had a better result than the candidate had produced. I was like, this isn't a good enough exercise. Like you're not testing thinking or anything. Like it was, it was like build a scoring and territory model. I'm like, that's table stakes now. You don't need an employee for that. I agree. Anyways, getting on a rant here. Yeah. Let's talk about, let's talk about the fun thing.

Gonto: I'll switch to our last topic. I think we can make it short. I don't know if you've seen the Mafia episode that Founders Fund has shipped, but it's just so good. Like in every fucking level, like I don't watch things that are 40 minutes. I watched this entire thing. And what I think is really good is they got really known people who retweeted their shit. So they got a lot of attention from that. They played a game that not only people in Silicon Valley play, but also you played as a kid. Like I was a big fan of the Mafia game. So I think you understand it from the get go. And number three, the Mafia game specifically is focused on manipulation, deceiving, and lying, and also strategy on being smart. So all of the people that went, you actually want to know how good they are at deceiving, at lying. And that's, I think they're fan of the episode. It's like, is Sam Altman really good at lying in this game? Like, does he do the same on the other one? Is Moxie really good at deceiving? Is he deceiving us when he has his private messaging signal network? And that's to me, what was really fun about it is you see known people that you want to understand how they behave on these things. And it was also a fun event. So it was so good in just so many levels to me.

Hank: I'm excited. I watched the first minute and I was like, I need to watch this. I haven't watched the whole thing. So no spoilers. I have my guesses on how some things will turn out. But yeah, like in terms of the marketing, yeah, we've talked about this before. Like if you can put on an event and record it, that's just interesting and just pure fun. Sometimes that's just a good thing to do. And this is perfect type of content from like a VC. There you go.

Gonto: With that, I think we're done for the day. Exactly. Thank you for joining. As usual, comment, tweet, I don't know, tweet or whatever. But let us know if you want any other topics or how you felt about this one. But thank you for listening. Ciao.