Gonto: So I use Hermes for all of my coding, for everything.

Hank: Not to be confused with the high fashion brand Hermès, spelled the same. It's like OpenClaw. It's basically the same thing.

Gonto: It did the math looking at how many tokens I used, and with my $200 OpenAI plan, I spent $7,600 these past months.

Hank: That's how much it would have cost?

Gonto: It's how much it would have cost if the $200 OpenAI plan wouldn't be subsidized.

Hank: Wow. And we're back. Okay, in a month and a half from when this is published, we're going to have a Code to Market meetup in San Francisco.

Gonto: Call it Code to Market Summit. It sounds much better. It's not a meetup, it's a summit.

Hank: Nah, this isn't a summit though, you know?

Gonto: That is true.

Hank: Yeah, we do have the summit in February, which is, you know, we go and we will ski and have workshops and there's a spa for those who don't ski. It's all awesome. And you might not even make it for the second year in a row.

Gonto: I might not because I'm having a second kid. The second kid is coming. They're going to be born January 30, which is basically one day before the trip. So I don't think I'm going to make it.

Hank: Unless you're aiming for a divorce in the same week.

Gonto: Yeah, but we're doing a meetup. We'll both be in SF. We actually planned this. We'll be in SF for the same week. We're going to a conference. It's also a Supabase conference, which we'll both attend as well. So you're welcome to come and we'll do a meetup for developer marketers.

Hank: Yes, so that'll be September 30. We'll have a Luma. Luma's in the description. I think we have a sick office lined up from a big fan of the pod. Thank you. I won't reveal them yet. I think we've got a sponsor. So we'll have some like food and drinks. And we're not just going to do like, it's not just going to be like, come and chat, whatever. It's not just going to be like, come and listen to a panel. We're going to have some sort of game show. So that's our first topic here is like, what would be interesting to do? We could do like, maybe we get a panel of judges and the panel of judges, judges, people's hot takes. Maybe you got to come up and say something interesting you're doing with an agent. I don't know. What are we thinking?

Gonto: What I like the idea is like, I don't like panels because it's a very few people that are talking and then everybody else is just listening. And I think we are going to get so many incredible developer marketers that to me, the number one priority is everybody has to participate. So maybe we get everybody to talk for a minute or something like that. I like the idea of talking about context. I like the idea, for example, of real shit they've done with agents. Like, how are they using an agent to do a better job of developer marketing or what thing that they've done that they think is creative and nobody else has done recently. And I think we can have a small stage where everybody can come and talk maybe for a minute and we'll bring on some judges, including us, who can basically give some prizes to the winners sponsored. What do you think?

Hank: Yeah, we can come up with some prizes. Yeah, I like that too. We'll come up with something. We'd love some ideas. Leave it in the comments, DM us, something.

Gonto: Tweet at us, whatever. I like the idea of like agents on the wild or something like that. Like, OK, how are you really using them? Which will talk actually a bit about us, how we're doing it. But we want to hear from others to learn as well. But again, set a date. It's going to be September 30. It's going to be an awesome office in SF that will share the details soon. But book your time. We'll share the Luma soon as well. And yeah, we're excited to see you all there.

Hank: By the way, and we've done enough conference stuff, but I'll mention this because it's related to the idea. But at Laracon last week, instead of doing a panel, we did Ship or Die. And so we had seven people go up in front of a panel and they would just pitch their idea of a thing they've built. And then the panel, the four judges would say, oh, that's great. Ship it. Or they wouldn't say like die, but they'd be like, I'd want to see this before you ship it. So no. And then each panelist that they got to say yes, they got credits on the platform or something. So there's a cool prize. But I like this instead of just having panel people speak or lecture or whatever, get more people participating.

Gonto: Exactly.

Hank: That's been enough.

Gonto: And we're going to talk a bit about how we are using agents for marketing and what things we've done recently with it. Who should start?

Hank: I think you're doing way more than me. So I'm going to start with a simple thing. So something I did today was I had an advisory call and they just wanted one hour on email, which is an hour I have done just ad libbed many times. I can talk your ear off about email. I could do a full day workshop on email, but they wanted an hour. And what I did that I've never done before for prep is I just told Claude Code, hey, will you read through all the docs I've ever shared with you, which is me, read all the Granola notes that I've had with this company before, and then read all my Granola notes from other companies, everything I've said about email, and then just put together a structure for this call that you think I would like and all my tips. And it was really cool because it came up with like, OK, this was the funniest part that I can share, which it just gave me a bunch of things that I've said. It's like your own lines ready to deploy. And these are just quotes, things that I've already said. So it's like, hey, you can just repeat these and they'll like it. And like the first email is pure resource. It's pure give like it takes longer to write a shorter email. It takes humility to make the email about them and not you like, you know, here's a slight against marketing ops people, of which I am one. A marketing ops person is someone that you Slack and then an email happens. This is cool. Like, and I just had it. I always tell Claude Code to just give me stuff in HTML because it can write it really quickly. And as you can see, it gives it to me like pretty and easy to read. But that was just fun. It made the prep way easier for this advisory call. And it was a little less of me like thinking and humming and hawing. And I got more value delivered. So there you go. There's my simple thing. I think you're doing cooler, more advanced stuff. What are you doing on this?

Gonto: First, I do something similar, but it's done for all my goals. Like one thing I've implemented now is my agent at 2 a.m. every morning reads all of my Slacks, all of my emails, all of my WhatsApp and all of my recordings and updates my internal CRM and knowledge base daily. So then when it knows that I'm doing, for example, a call for emails, it will automatically read all of my internal like articles and knowledge and CRM and preview the markdown just for that call. And if it thinks it's something I'm going to use more in the future, like the email one, it will create like a meta concept page that is created. I use a skill called llm-wiki, which is basically based on what Karpathy shared, this idea of like building a wiki that a human reads, but it's only updated, touched and created by an LLM. And in this case, basically what it will do is it will have concepts that are abstract created from all of the others and it will automatically create them every time I have a new call. At the same time that it does that, it will also create to-dos for me. It will create to-dos that it can do for me, like sending an email or sending a WhatsApp or sending a Slack. And it will also keep track of what people owe me. I have a list that's called owed to me. And then if people promise me that they will do something by a certain date, it will put the date. And then one day after that date, if I didn't get that delivered, it will automatically create a draft for me, either in WhatsApp, Slack or email saying like, hey, you promised you would give me this for this date. You haven't, what's going on? So now basically I've implemented the GontoPush as an owed to me thing.

Hank: I love that. The GontoPush.

Gonto: I have an app where I review some of these drafts and stuff like that. And it's literally called the GontoPush. That's the name of the app where I can go and check it all out. But one thing I wanted to chat about is as a marketer, I've been building a lot of apps. I was a developer in the past, but I haven't coded manually in a long time. And this weekend, for example, I built a new app. I bought this thing called the Index 01. And it has a button there where if you click on the button with this finger, it has a mic and it will record whatever I say. And what I did was I built a small webhooking app that will automatically get whatever I say here and create a new chat in Hermes, like a new topic with whatever I say and process it. So it can send an email for me. It can tell my wife a WhatsApp. It can do whatever. And there's two skills that I've used to build it. And I think every marketer who is trying to build an app should use. Skill number one is called Grill Me. It's from Matt Pocock. And the skill name is Grill Me because when you're not a developer, you don't think of all of the edge cases. You don't think of all of the things that you need to build. What Grill Me does is it forces you to first start with an idea like, hey, I want to build this app. I wanted to do this, this and that. And then it will ask you between 15 to 40 questions on things that you forgot to think about, like edge cases or things that you need to ask. And it's called Grill Me because it will literally grill you for an hour. But once you do that, it will create a PRD for you. And then the agent is so much better at implementing by using that. So that's one thing I've used a lot lately. I think when you're not a developer, you don't think of edge cases. And this Grill Me forces you to think of that. And I think helps you build not just better apps, but for example, I've built recently an email automation and it was just better because I missed one edge case, which was that I already built another email automation last week. And it could happen that two emails get sent within an hour. And I didn't want that to happen. And the Grill Me realized of that instead of me. So I think Grill Me is one. And then the other skill that I've used for this, also from Matt Pocock, it's called to-prd, which basically after it grills me, it will create a markdown with my definitions, the answers, the questions, et cetera. And it's a full doc, like multiple pages docs on what it needs to implement. And the good thing about it is in a lot of cases, when you tell an agent to do something, it will do something, but then stop. And then you have to ping him like, hey, but this is missing, or this is missing, or this is missing. Now with this PRD, if you do slash goal in Codex, in Claude, all of them have it, and say, continue implementing it until this PRD is done, it will not stop until it's fully implemented. I've had it, for example, run for 20 hours straight the other day, just implementing it. And every time it was stopping, the slash goal is a meta agent that will check when the first agent says it's done to say like, no, you're not done, continue. I mean, he spent 20 hours and the result was good. I didn't have to be in the middle. So I think those are two skills that if you're building automation with your agents, you should definitely use.

Hank: What are you building that takes 20 hours? Come on.

Gonto: My 20 hour build, it was this GontoPush app. Like it's the, it's the cycle of the morning. It's the draft that it writes daily. It's like on every draft, I can snooze it, discard it or change it.

Hank: I can see that because for GontoPush, you want it to be very thorough. It needs to check lots of cases. It needs to check up against. Okay. What's already been like, how do I verify if they already gave it to me?

Gonto: Exactly.

Hank: Like, how do I check all these sources and there's styling? You know, how do you make it sound like Gonto? Blah, blah, blah. Very interesting.

Gonto: I actually checked yesterday. So I use Hermes for all of my coding for everything. It logs how many tokens.

Hank: For those who don't know Hermes, is it? I haven't heard it out loud. It's Hermes or Hermès or Hermet.

Gonto: It's Hermes. It's Hermes because I'm sorry, but it's Hermes.

Hank: Yeah. Not to be confused with the high fashion brand Hermès, spelled the same. It's like OpenClaw. It's basically the same thing. And just a bit more stable. You don't need to know more than that. Yeah.

Gonto: So I've asked it how many tokens I've used this month and it did the math looking at how many tokens I used given what would be the pricing if I use the API. And with my $200 OpenAI plan, I spent $7,600 this past month.

Hank: That's how much it would have cost?

Gonto: It's how much it would have cost if the $200 OpenAI plan wouldn't be subsidized.

Hank: Wow. That's crazy. Huh. I should ask what I've been using. I've been doing a lot. I need to do more. Just when I think I'm catching up on like my like personal assistant type of workflows, you're pulling ahead. So, and this is the type of stuff that like, yeah, these aren't necessarily marketing use cases, but the people who are listening to us, we know you, a lot of you are executives and leaders and time is everything. Like this is how you buy, this is how you buy time now and how you get really good at the follow-up and execution. So I got to get better at this. I've only just started building out my personal CRM and to-do trackers. So I've got to catch up.

Gonto: My wife says that my AI gave itself the name Faro, which is like a lighthouse in Spanish. And my wife says that Faro is my new girlfriend. And I like her more than my wife because I'm obviously long.

Hank: I wanted to circle back on that. How does your wife feel about all the AI and knowing that some of her texts are from AI and stuff like that?

Gonto: She's, I would say she doesn't care that some of the messages are from AI. What she doesn't like is that I chat so long with it. And that sometimes I just like, I have a WhatsApp group with her and the AI and my wife's mom left some chairs here for months. And I wanted her to take them out. So Faro started pinging her every day, take the chairs out, take the chairs out. She started to say, stop pushing me. And Faro was saying, I can only cancel the notification if Gonto tells me. If you tell me, I will, I need to double down. That's what Gonto said. That was what I set up. Every time she complained, it would ping her twice. After a week, she had the chairs out. So at least it works.

Hank: Wow. I don't know. Short-term gain, my friend, may translate into long-term losses. Tread lightly, everyone. We're good at giving some marketing advice. I don't know if you should take Gonto's relationship advice. Oh, geez.

Gonto: You were saying that you are using your agent for newsletter as well. Like, how are you using it for your newsletters?

Hank: Well, that was one of the things on the call. So I'm not using it, though I should. We should get better about like our Code to Market newsletter and stuff. We have a bunch of people who signed up. Though we're very, if you sign up for our newsletter, which you give us your email at ctm.fm, we're very judicious about when we send emails. Only for like events. So we'll send one out about the meetup and the summit. And when we have like a banger episode, we'll send it out with a summary of some other ones. So what we talked about on this advisory call I had today was they're not doing any sort of newsletter. It's like, guys, sometimes people spend too much time writing their newsletter and sending it up and doing it every month. And this company already does a changelog, multiple changelogs a month. They do lots of LinkedIn posts and they have events every month. So I'm like, boom, you have three perfect things for newsletters. Someone will find any of those useful every time. So you just set up a routine at the start of every month, look at all the changelogs, write a summary and give the links. On top of that in the email, put a, I was like, you need to keep a doc of like your best LinkedIn posts. And just like whenever one hits a certain number of likes or engagement, just drop it at the top of a doc. And then part of your routine will be, hey, check that doc for the latest LinkedIn banger. Put that in there with a link to it. And then put the best events that you're doing this month at the top. I was like, and you can automate that completely. The LLM knows enough how to summarize it. That's content people aren't going to unsubscribe from because they're always going to think, well, even if it's not useful this time, there might be something for me next time. And that's great, like easy. And there's so many things like that you could probably set up that people find useful, that your team finds useful. Just do it.

Gonto: How I've implemented newsletters lately is I want the founders. I want engineers. I want everybody to basically be inputting thoughts. Like if they see a really good article, they should use it. So what we've done now is we built an internal Slack channel where when people really like an article, they just post a link there and they can add some thoughts on to what they liked and why they liked it. And then we have an agent that basically goes into that Slack every month and it knows it has to do 20% content of the company and 80% content of in general generic. And it will read the Slack and it will also see using analytics, which were our most popular posts from that month from signups and unique visitors. And it will automatically build a newsletter just using the popular blog posts from us. And then the links that were sent by people randomly on that Slack. And that's how we built it. And it's very easy for people because they can just paste a link or paste a link plus a few words. And that's it. That automatically sends the newsletter every month.

Hank: One thing that's interesting here, I'll do a little plug for me and Mitchell's workshop here. But there's a lot of people, they hear us talking about this stuff and they just still haven't opened up the dang programs and gotten started. And, you know, there's so many little things to trip you up like, oh, you haven't installed Git or Node or whatever. And that's why me and Mitchell started this workshop, codetomarket.ai. And you'll find the Claude Code for Marketers workshop. Somebody who went to our first cohort of this, by the way, she's built an entire Italian language app. This is just personal software she built for herself after. So she's got an app and like she lives in Italy now. Her name is Abby. She worked with me at Laravel. And it's like an amazing app. Now I poke fun at her because I'm like, so wait, you had to get an Apple developer license or whatever. So isn't this like a hundred dollar app now? She's like, yeah, but it has everything I want. And, you know, I'm learning or whatever, but I don't know the barrier. Once you break through the barriers, it's kind of amazing what you can do as a non-technical person. Anyways, I agree. There you go. I think that's the pod today. I'm excited about what we've got planned. The meetup is going to be really fun. We hope to see a lot of you there. September 30. Save your time.

Gonto: Well, book your time, I would say. We're going to be in SF September 30 at night and bring your crazy idea, crazy agent usage so you can speak with us. Thank you all.