- Platform: meet
- Source: Read.ai 01KME0HJMYD35PMDRTF9C8X8HD (via Ben Ornstein’s account, imported 2026-08-06)
Transcript
Ben Ornstein: Yo. Michael Houck: How’s it going? Ben Ornstein: Good. How are you? Michael Houck: Pretty good. Pretty good. Ben Ornstein: How was your weekend? Ben Ornstein: It was nice. Um, I went to actually back to New York for the weekend and, um, hung out with some friends, played some golf and then, um, hung out with my mom for a little bit. So that was good. Michael Houck: Oh, that’s cool. A little family time. Um, I gotta learn how to play golf. Everyone down here is, it’s like all they do. And I have the hand-eye coordination, I just don’t know the mechanics. Ben Ornstein: Apparently my high school buddies are big golf players. They say like tennis players, it translates really easily. And like when I first started playing, they were like, Dan, your stroke’s like already there. It’s kind of weird. Michael Houck: Oh, that’s cool. Well, hopefully I am as translatable as you, I suppose. We’ll find out. I got a lesson at some point next couple of weeks, so we’ll see. But do you mind if I go camera off I got to go through a bunch of Clothes that I brought down to my parents house. So I’m just gonna be kind of doing that as we go through if you don’t mind Yeah, that’s totally cool. UNKNOWN_SPEAKER: I’ll stay off as well. Michael Houck: Okay, cool. I Think we got some new studio clients coming in I just add you to two new channels and there’s like one or two more I’m expecting to convert in the next couple days and then Draper is Expanding to three more people on sort of the scope there, but it’s gonna be pretty exciting in the next week or so. I’m hoping to lock it down this month, but we’ll see. That’s huge. On the other side, though, Augment looks like they’re going to drop the social and just do the video stuff. They said they’ve been blown away by the quality of the video stuff, but that new guy, Paul, wants to bring the social writing in-house. That’s what they’re likely going to do. I’m talking to Sammy about it right now. Ben Ornstein: Got it. Yeah, that’s unfortunate. I mean, like even last week, we had a great week last week, like 300,000 impressions on a post or on like for the week. That’s unfortunate because we’re we’re doing just so well for him. Michael Houck: Yeah, I’m going to I’m going to shoot Paul a note and actually do that now. Ben Ornstein: I get it, like they’ve got the team in-house and then there’s like this one thing that’s kind of outside. I understand where they’re coming from, but I think there’s definitely an argument to be made that like it makes sense. And I know you could easily explain that better than I could, obviously. Michael Houck: Yeah, just checking in before we make the switch because the results We’re doing NOLs right now, not the company account, right? Correct. Ben Ornstein: Only NOLs. Michael Houck: Okay. Okay, cool. I just sent him a message. We’ll see what he says. But okay, I’m gonna start going through this stuff. But yeah, I guess just catch me up on on your side. What’s going on? Ben Ornstein: Yeah. So one thing is the Structify.ai. They’ve got their campaign on Wednesday. Their founder said Payment should be coming in. It’s uh, they did the same thing as last time where I gave them a bunch of options They only chose like the engagements for like like the megaphone engagements kind of things Rather than they picked one sponsored post for like a hundred bucks. So That was unfortunate, but so it’ll be like 800 bucks. Michael Houck: Yeah It already came through I saw it. Ben Ornstein: Oh It came through I didn’t see it It’s, it’s, they use ramp. Michael Houck: So it’s, it’s different. I get a notification before you do. Ben Ornstein: Got it. Okay. So yeah, so there’s that. And on that, I bring that up mostly because they selected your repost. So can I just repost from your LinkedIn? I can handle that myself since I have access. Michael Houck: What’s the post going to be? Ben Ornstein: It’s going to be about Structify. They’re launching basically What’s the best way to explain them They’re an AI platform that unifies Data sources to allow for lightning-fast revenue changing insights specifically targeting rev ops teams Um, yeah, that’s fine. Michael Houck: Yeah, let’s do it. Um, no big deal. Ben Ornstein: Okay, cool. Ben Ornstein: So yeah, I’ll handle that for you then. Um, and yeah, I mean, outside of that, it’s been all things, Claudia. So, um, I’ve just been, it’s been a hell of a process where it’s like, I build something, it temporarily works. Okay. I got to figure out if it fits into the whole workflow, if doesn’t figure that out, then I just take a step back, look at like the token burn, if I have to fix that, maybe I should fix that. All these like different pieces, but I’m starting to get a feel for how this all works now. It’s taken a long time to get here, but I mean, there’s some things like I’ve started to easily build, like that voice functionality now, totally built. Now when we make changes, he automatically updates on the architecture side, which is so crazy. Michael Houck: Yeah, I love that. Michael Houck: That’s gonna make it so easy to kind of track what what changes we’re making and see how they’re doing. It would be cool if like, if that site also had like version control, your version control is, I don’t know what that is now. It’s an engineering term that basically like, you know, it’s it logs when changes are made. And you can go back and see like, oh, you know, the the change that was made on March 20. Made these four or five different changes right so you can see like oh when bugs get introduced and things like that oh that would be super helpful okay I’m gonna make a note to add that that’s that’s very important cool okay you can tell it to take inspiration from how github shows version control and we can just have it do something similar. Ben Ornstein: Take inspiration from Okay, yeah, I made a note there. Yeah, no, I’m, it’s like there’s a lot of frustrating things where it’s like Blake’s reaching out and he’s like, I have no idea what Blake’s voice I was like, bro, I just spent so much time building all this stuff in and you can’t remember that it’s like It’s it’s frustrating, but I’m also proud of like some of the things that have been built like this architecture site I’m actually thrilled about that The quick switch to the voice functionality. That’s great. And now like I’m just going back and forth forth with him and DM saying like, okay explain I finally I’ve only gotten to a point where I can do this. It took a long time to get to like, okay, let me just go through these workflows now that I think the infrastructure is there and tell me what you know about workflow number one, for example. And then it goes through it and says, do I understand this right? Am I missing things? And I’d be like, ah, you’re missing this actually. In this scenario, you do this. And those are the things I need to continue building out. But that’s how I know we’re getting closer Now I’ve visually shown Claudio’s architecture. I know how he thinks in, of course, like a high level way. I don’t know the absolute mechanics, but I feel a lot better than I did. I’ll tell you that. Michael Houck: Yeah. I mean, look, I see the improvements and I know that you’re understanding how to identify and fix issues, and that’s really the main thing. It’s one thing to have the system work better, right? It’s another thing to know that when you run into roadblocks, you’ll be able to work through them, and it seems like you’re picking that up, which is the most important thing. What’s on your agenda for this week? If we consider that basically by the end of next week, we want it to be predominantly at least for Blake and I? What needs to happen for that to be true? Ben Ornstein: Yeah. At this point, I think it’s got to be testing and like just using it. And I say that, of course, as Blake tested this morning and it didn’t work, but I’ve since fixed and verified that it is fixed. I think we just need to keep doing that. And what I should do myself first is give myself, because I don’t have like a voice I made it one but like I don’t really have a voice profile so I don’t know if it’s actually like doing it or not. Have I given it a voice profile for me? UNKNOWN_SPEAKER: Do I need to do that? Ben Ornstein: So I kind of gave it one for you. What I did is I gave it your last 10 posts on LinkedIn as well as told it to look at founding journey to get your newsletter voice. So it has your newsletter voice and your LinkedIn voice. Obviously that means there are going to be a lot of improvements to make. So what you could do is ask him what, um, you know, what you’re thinking, like what you think. Um, and UNKNOWN_SPEAKER: Ask him what it thinks like about my voice. Ben Ornstein: Sorry, ask him, ask him what he knows about your voice. Um, and then we can change it from there. Cause yeah, Blake had that crazy markdown file that he gave. Um, so he, he’s loaded that, um, do you have something logged somewhere? Michael Houck: No, absolutely not. But what I’m going to do is I’m going to take Blake’s and I will like make a version of it that fits what I want more, you know? Cool. UNKNOWN_SPEAKER: Yeah, that works. Michael Houck: I think that the more we can templatize stuff like that, the better. So I trust that Blake has done a lot of that. So I’ll just do I’ll just use that as a template for your mind. Ben Ornstein: Cool, OK. Let me forward you his, just so you have it. OK, there you go. UNKNOWN_SPEAKER: Cool, thanks. Ben Ornstein: Yeah. So there’s all these playbooks. I’m going through, actively going through, making sure that he’s understanding what goes into each playbook. And then I want to verify how he gets to that playbook. I’m trying to go as far back in the flow of how he thinks to make sure each thing is connected properly. Just what Wojo said, right? It’s like each piece has to connect and make sure it connects. And why it connects. And so those are all the things I want to be able to understand and be able to verify. And so it’s taking time because I’m tackling pieces of things. So for example, right before our call, I was asking what it knows about memory. And if it reads a Slack thread, does it remember? Michael Houck: I think that one big difference right now that I’m seeing is when you’re doing this, you’re doing this through chatting with it. You’re like relying on it to tell you about itself rather than like you’re able to sort of explore the architecture itself, right? You’re like relying on it to provide that to you, right? UNKNOWN_SPEAKER: Yeah, yeah. Michael Houck: Yeah, I think it would be good if you could get it to a place where you didn’t need to, I mean, there’s multiple benefits. One is the cloud credits, right? And two is just know that you’re getting the actual ground truth of information from it. If you can like at least navigate, you don’t need to make any changes this way, but at least navigate the command line and the terminal in order to like see how it’s laid out. So do you think that’s something you’d be able to do? Ben Ornstein: Yeah, I can look into it. Like I can, yeah. I know, like, some of the key things to, like, open the terminal and things to, like, show me certain things, but maybe that’s the next step, so, because you’re right, like, I’m burning so many Claude credits by interacting with him. And he’s solving things, of course, when I do that, but sometimes it takes a long time to get there. Because you’re right, I should be able to navigate the terminal itself. Michael Houck: It’s not just the credits. Like, the credits are part of it, sure, but it’s also, like, you know, we’ve already seen that he doesn’t remember everything. And so if you’re relying on the system to tell you how the system is working, but we know that the system isn’t always remembering everything, you’re, you might not always be getting the full picture, like you might not be, quote, unquote, honest with you. Right? Yeah. You’re then making changes and decisions based on that. And then you’re relying on it those changes that way too, it’s just, there’s a lot of, there’s a lot of room for bugs to be introduced that way. And so the, the, the more directly connected we can get you to the actual infrastructure itself, uh, it’s, it’s worth the time investment to like learn how to do that. And so the first step is just learning how to navigate it. And then the second step, once you feel comfortable with that is learning how to make changes to it in the terminal and the command line. Ben Ornstein: Yep. Yep, for sure. That second part will be harder. I know like the basic things like LSCD, those kind of things, navigating through the files through that, but like, what things to run when that, you know, that part is very difficult. Michael Houck: Yeah, but like, again, like, that’s not the expectation, right? Right now, the expectation is just like, we can make changes to it. And we can. And then the next the next phase is what I’m saying, which is basically it’s like, know how to navigate and how to find ground truth. Because then if you’re navigating through the command line, then you can have the conversation with it in Slack. And you’re like, hey, Claudie, I noticed that this is your architecture. This is wrong in this way, please update it, right? Rather than relying on it to pull the architecture and be honest with you. And then also on the other side, update it correctly when you ask it to. So you’re just removing one of the two areas of potential bug introduction. Ben Ornstein: Yep. No, I totally see what you’re saying. Okay, that will be my next goal. Absolutely. Yeah. Michael Houck: And then I guess another thing is if we’re saying we have this two-week timeline where we want to make it functional, right? For Blake and I. Workflow-wise right now, is it basically just how we have it set up? Oh, sorry, I’m going to sneeze. Allergies down here. Right now. But is there any usability things we need to consider to build into this MVP release, basically? Does it schedule to type fully correctly? What’s the flow there? Or is it just conversations that I have with it in Slack, and then I go copy-paste the post? That one is better than the other. I mean, one is better than the other, but not the one that’s like needed. I just want to know, like, what are you scoping into the next? Because two weeks is a lot of time, right? We’ve already seen what you can do in two weeks so far. And it’s great. So like, if we put another two weeks on this, like, where do you realistically think we can get to if we continue to kind of push? Ben Ornstein: I think very, very realistically, he has 10 different workflows that he can do. And these are ones that I’ve built out for different reasons. Some of them would be cron jobs, that would be all if they’re fully active and have all the pieces of the workflow, like, integrated. But to answer your first question of, like, yes, it can post a draft to Typefully, and we can ask it to schedule it, but I think it has a hard bar of, like, only putting it at draft in Typefully. It has the ability to do that. And so, like, I can get it as close to, like, extreme production ready across the key workflows. The big one, of course, is the one that we cannot get wrong is when you or Blake write in the chat and be like, hey, I want a piece of content on X. It needs to go through the workflow seamlessly, given what you’re asking. And for that use case, in two weeks, are you saying like that’s the only workflow that you would like correct? You ask it something and it gives you, and it goes through the correct workflow, given what you give it? Michael Houck: I mean, I’m not saying anything. I don’t have the context that you do here. So I’m really asking you, I’m like, if you grind on this for the next two weeks, I’m going to be like, you know, don’t put a ton of other stuff on your plate. Like, you tell me, like, what are the non-negotiable features that you want to make perfect? And if so, what improvements are we making to get there? And then number two, sort of like, what were the nice to have features that we also think we can fit into the next two weeks. I’m just trying to think about like, you know, when I was at Uber and Airbnb, we would do these exercises every year, usually every quarter, where we’d have like all the things we could do listed out in a Google sheet. And then the product manager would go around and talk to the engineering team, the design team, everyone else and be like, all right, and like leadership and be like, all right, here’s the priorities. Here’s how much time each one’s going to take, right? We have roughly, you know, X amount of man days during that time, and each one’s going to take X amount of man days to do. So realistically, we have a certain amount of like, bandwidth. And therefore, we need to consider priority, and we need to consider size and scope, and impact, and then consider, okay, what goes sort of above the line of the things we think we can get done this quarter? And what is, you know, we want to do, but what’s realistic that’s basically not going to happen this quarter. And so I’m basically just asking for a small, rough version of that, where you’re like, hey, these next two weeks, we set this timeline for end of March, we’ll let it go a little bit into April. But what should I, as your user, stakeholder, exec, what should I expect, and why are we prioritizing these features, basically? That’s just what I wanna know. And that’s you as a sort of like engineer slash product manager, that’s you need to lead that. I can’t, I don’t have the same like low level context that you do. Ben Ornstein: No, of course, right. You couldn’t do that. Okay, I understand. I can give you a mini version of that now and then get it way more in depth after. Michael Houck: Yeah, that sounds good. Michael Houck: Let’s do it like that. Just give me the sort of rough off the cuff expectation now. Maybe spend like, you know, an hour or something like actually writing it out in the Google Sheet. And I can give you a template to use for that. Ben Ornstein: Okay, yeah, that’d be great. So the non-negotiable that will be done within that timeline is if you or Blake write something in the chat. I call this workflow the direct topic request. If you write something in the chat, I want to post about X, it clarifies who it’s asking for the content for. Is it going to be for you? It’s going to know if you mean X or LinkedIn, and if you don’t clarify, then it asks. What kind of post do you want it to be about? If it’s something that requires research, then it goes and does research. If not, if you provide your entire perspective in that first message, then it just goes with that. It doesn’t need to do research because it gives your perspective. And then you can, of course, follow up and say, oh, well, hold on. I want research on x. That is the direct topic request flow. It goes through. The entire workflow that it of course loads your voice, it looks through all of that, personalizes, humanizes, it goes through that humanizer skill and make sure that the content is as best it possibly can be for you and you uniquely. That’s one workflow and there’s more and more to that workflow. Michael Houck: Yeah, so I think that there’s a lot in there. I think overall, direct topic request is good as a p0 I’d like to see that when you I just sent you the link to the sheet I’d like to see you break that down more and so here we can do it like this actually we can do it like feature and workflow so like direct topic requests might have like four different things and each one might have like a different priority so like whatever the features are you can But P0 is like, must get done. P1 is like, hopefully get done. And P2 is like, probably won’t get done, but still something we want to do. And let me actually just make this easy to do here. So P0, make that red. P1, make that yellow. P2, make that blue. And then Scope, I’m going to do the same thing, Scope is going to be extra large, large, medium, small, extra small. And then impact is going to be just simply high, medium, low. And so, like, the sizing, Like for scope, XL is something that’ll take you like a full, let’s say two weeks. Large would be like something that’ll take like between a half week to a full week. Medium is like a half week. Small is like a day. And extra small is like less than a day, basically. Does that make sense? Ben Ornstein: Yeah. I don’t remember all of those exactly what they mean, but that’s why I have the Read AI guy in here, so. Michael Houck: Yeah, no, it’s just standard engineering like stuff. Like scope, like t-shirt sizing scope is what they call it. So anyway, yeah, you can use as many features as you want. The goal is like, if you like Ben, our Ben days, right? Ben days is realistically 10 days. So we need like 10 days worth of stuff over the next, in this sheet right now of like stuff that’s actually P0. Go ahead. Ben Ornstein: Okay. Michael Houck: So, okay, so we have the direct topic requests workflow. That’s one workflow. What other workflows are you considering prioritizing stuff in? Whether it’s, you know, as high priority as direct topic request or not. Just I’m curious what the full list is. Ben Ornstein: Yeah, so another one is the internal team batching, which is what Blake did, where he asked for like a bunch of different content in his voice, more than one piece of content, which is, you know, that matches his like niche, that matches things that he would talk about, but like different pieces of content that, like Blake said, short form or whatever. Michael Houck: I know what you’re referring to. Okay, so direct topic request is maybe more appropriately named like single post request. UNKNOWN_SPEAKER: Yeah, maybe I should I should change that actually. Michael Houck: And then the other one would be like, like batch request. Yeah, right. Okay. Ben Ornstein: Yeah. And the, for both of those, the trigger is just a message from the team internally. That’s, yeah, for those. Michael Houck: So like, yeah, I guess maybe even like workflow is probably not the right terminology here. It’s probably like, I don’t know what the right word would be. Cause it’s like kind of multiple things. But like, we could just call it like area for now. So like a single post request workflow, I know that’s like a mouthful and batch request workflow. Those are like areas and then another area would be like infra, right? And another one would be like integrations, right? With like other apps and stuff. So like there’s lots of different areas that you can have features in and they can sort of be different types of things. They can be a certain workflow for users. They can be an infrastructural thing. They can be, you know, yadda, yadda, yadda. I think you get the idea, basically. Ben Ornstein: Yeah. Okay. Okay. Yeah, man, this is a whole new world for me. Michael Houck: I know. There’s definitely a little bit of a learning curve here. That’s why I’m kind of like I’m not expecting it to be perfect, but I want to give you sort of the tools for the toolbox for you to sort of explore with, you know? Because I think that as this gets more complex and as this gets more integrated into our company, it’ll be important that we have like standard stuff set up. So like another area, for example, would even be like the editor agent. Right like that’s an area um so um yeah just like this is how we want to visualize it on like a like a roadmap sheet like this and then what we’ll do is after you go through and add all the features and the priorities and stuff as you see them then what I’ll do is I’ll go through and I’ll prioritize them based on like what I think we actually want to do in the next two weeks because I have broader business context like that’s our two roles here right your role role is you’re in the weeds, you’re building this thing, you understand how it’s working, you understand what the problems are better than I do. I have the broader business context as a sort of like stakeholder. And so I come in and I say, okay, here’s what we should prioritize based on what you’re telling me about priority scope and impact, basically. Ben Ornstein: Yep. Okay, that makes sense. Okay. There’s so those are the two work that are, those are like ones that are technically active that should in theory work. And the internal batching, that one’s not fully like 100% verified. I’m very, I’m pretty confident that direct topic request works, but I, again, I have to go through and test and verify. Then there’s, then there’s more in-depth ones. So workflow six, and these, these are a lot of these are more now for client like related ones that are definitely for clients. Michael Houck: So yeah, so also just remember, like, like, we can also include like the partnership stuff in here. Right? Like, this is like, broad is how we should think about this, like not just constrained to even just the content agent side of this. So like clients within the content agent is one we’ll want to do, but we’ll need to prioritize that against like expanding to the daily brief, right, and expanding to partnerships work. So that’s just what I’ll flag. Ben Ornstein: Got it. So this is outside of just, okay, okay. Michael Houck: Yeah, this is our prioritization roadmap for like all of our open claw work that you’re doing, basically. Ben Ornstein: Right, right. The briefing is, that’s, yeah, I guess I have that kind of, I don’t know why it’s listed here really shouldn’t be on the architect I have it there. So like the morning and afternoon briefing, I had it going. So that would be something that we need to build where it’s like 8 a.m. And 4 p.m., beginning of day and end of day. What needs to be linked, what things are important to know, how does it actually… Yeah, I haven’t done any work on that really. Michael Houck: No, I know. That’s been right to not do the work so far. But like, regardless of whether it’s these next two weeks, or whether it’s something we do in April, we’re going to want to do that, right? So I just want to get a sense of like, how big is that project actually? And like, then we’ll prioritize it from there. Ben Ornstein: Yep, yep. So yeah, and then there’s other, there’s other client ones that are like more proactive. 8 a.m. The one we’ve talked about 8 a.m. On Mondays it surfaces questions Proactively for the client or the user really it’s done that to me. Michael Houck: Actually, it’s kind of worked That’s like the source is what we call it, right the what? The the sorcerer the topic sorcerer, yeah. Ben Ornstein: Yeah the topic sorcerer So it’s like creating asking questions and like asking for the user’s perspective. Of course, these are not random questions. They’re supposed to be for, you know, specifically for a given client in all the context. So that’s one of the ones that I would want is like a cron job that actually worked. And I’ve seen it work. So like, I just need to make sure the workflow makes sense, but it has asked me relevant questions given what I’ve given it. So like that, it kind of works basically. Michael Houck: And there’s probably like multiple features within that that would make it production ready and better, right? So it’s just a matter of like identifying what those specific things are. And and scoping them, essentially. Ben Ornstein: Right, right. There’s another another like proactive one called breaking news content, where The easiest example I could give is Noel. A company raises a big private round and we don’t have content on it ready yet. So Ploy proactively reaches out to an internal channel. He says, hey Ben, OpenAd just raised a ton of money. Should Noel, can I send this to Noel to get his perspective on, or not even perspective, like should we create content on this? UNKNOWN_SPEAKER: Right. Ben Ornstein: So that’s like the breaking news. I have that as his own workflow right now. Michael Houck: That’s smart. I think that’s the right way to do it. UNKNOWN_SPEAKER: Yep. Michael Houck: And like, are we treating these as like, just to use consistent terminology, are we treating these as skills or like guides? Or like, what’s the relationship there between how we’re talking about it versus how it is in practice? Ben Ornstein: Yeah, I need to clarify if it’s technically a skill or a guide. I don’t really know. I’ll get back to you on that one as well. Michael Houck: Cool. Yeah, I would just say like we want to make sure that when we’re adding stuff that the structure fits within the overall framework that we’re setting up of like context equals guides and guides become skills when you combine them together in like a workflow, right? Skills are workflows that can be called by different agents. Different agents have different personalities. Therefore, they’re better suited to use different skills and they might use those skills in even different ways from each other, basically. Ben Ornstein: Right, right. See, these are the things like I just I don’t have clarified and like figured out. Michael Houck: Yeah, honestly, like I think that there’s, I think that there’s benefit to grinding away at like, you know, making something work, but I also think there’s benefit in you just like sitting with this and like even turning your computer off even for like a couple hours and just like writing this all down on like physical paper or in an ocean dock and just like architecting it a bit, you know? UNKNOWN_SPEAKER: Right, right. Michael Houck: It’s like, I don’t want you just like blindly hacking away at building this stuff. I’d rather we approach it with like a systematic strategy for each one. Like what’s it, what, what of this is our guides? What of this are skills? What of this needs to be in an agent’s personality and therefore might require a specified agent for this type of task area? UNKNOWN_SPEAKER: Right. Yeah. Michael Houck: Yeah. Ben Ornstein: That’s, yeah, I mean, that’s important. I need to think more like an engineer. My ops brain kicks in where it’s like, just fucking get it going and move. Michael Houck: Yeah, I mean, I would guess engineer, but also like product manager, right? Like each, like different areas of the business might be better owned by different agents. There’s no limit to the amount of agents we can have. Necessarily have like a million agents but like you need to find that balance between when do we create something new versus when is something we already have best equipped to do this right and just think about it like people like if you were the ceo then when would you hire someone new versus when would you um when would you you know give it your project to somebody who’s already on the team right and I think I think that it’s just like, if you think about it like that, treat the agents like people and then consider like what skills do they need to do the job I’m hiring them for? And okay, what sort of like guides do I need to give them in order for them to do those skills? Like kind of work backwards at it that way. Ben Ornstein: Okay, that makes sense. Yeah, so I think maybe I do need to just step back to get Because I’m way too in the weeds, I think. Michael Houck: Yeah, that’s my Read. My Read is that, yes, if you keep working at the altitude that you’re working at, yes, you will get things done, and yes, things will get better, but I actually think that it will save you time if you just, whether it’s the rest of the day today or whether it’s tomorrow, if you just step back and be like, all right, here’s everything I’ve learned about how this works, If I were going to design a system across all of Rye Valley that operates this way, and I know what the things we want to build are, you know, where do I draw the lines? Where do I hire quote-unquote people, aka agents? Where do the things they’re doing contain different skills? When do they contain different guides? You know what I mean? Just like map that out for yourself. And then when you go in to build it, you’ll save yourself a lot of time than just like trial and error, asking it to test. Ben Ornstein: Yeah, I can picture like what I need to do. Like manually draw. Michael Houck: I feel like you’ve learned like probably more than you even realize right now that you’ve learned. And so taking that time, I think will A, let you kind of see and feel that, but also will be like genuinely a time-saver for you because you’re just in a different kind of place now when you were when you started this yeah yeah for sure okay yeah I need to do that yeah it’s wild coming into this world that is just so foreign to me I know honestly this is like the most audacious project that I’ve given you since we’ve been working together and you know you’re just stepping up and you know diving in and I love it and I think that it position like this is literally what I said right when like you were worried like oh we’re are we gonna replace the whole team with AI like if these open call guys gonna come in like whatever and like the answer is no like do I want to do this stuff that you’re doing right now no absolutely not right like I can I can go there and do it but like do I want to know what I want is for me to be able to do marketing and growth and sales and for you to be the sort of like architect orchestrator switchboard guy who knows how the system works and can go in and add to it change it etc as the business requirements get updated that’s like a very valuable dynamic for the business and for me so like the more you dive into that you know the more benefit we’re going to see yeah for sure um and now what I’m now what I’m thinking about the architecture it’s like i’ve gone so deep on claudia’s content workflow but if I go deeper on and like well not deeper actually the opposite take a step and look at the holding company as a whole, and then build out every agent or possible skill or guide or new agent from there, I think that gets really interesting, actually. Yeah, and like, you know, remember, like Claudie is one agent, right? And Claudie might not even be the best agent to do half the things we’re asking it to do right now. Maybe there’s better agents we can build that are more specialized to do the different things like that dynamic where talking about between like content writer, editor, content ops person, right? Those are three different jobs. If that’s three different agents, you know, where is quality best equipped to do versus where would another agent be better equipped to step in? And what does that partnership between them look like, UNKNOWN_SPEAKER: Yeah. Ben Ornstein: And then the part that’s even more confusing for me is how do I get both agents to, I mean, how do I even start a new agent from scratch now that I’ve done all this? And like, how do they connect? Michael Houck: Yeah. That’s the classic engineering thing of like, why it’s good to do things that do scale, why it’s good to do things that are templatizable, because the best engineers, they’re not the ones who, you know, brute force through a million different Projects they’re the ones who build systems that can solve a lot of different problems in a simple and elegant way, right? So if you build quality in a way where you can just copy paste 90% of its infrastructure But just change some things in the soul file or the whatever the heart file or whatever. Yeah, and that’s entirely different personality But still valuable to the business and knows the business context Well, that’s fucking brilliant because then we can just spin up this new guy in 20 minutes and then give it some new skills and then all of a sudden it’s a total a different persona, you know? Ben Ornstein: Yeah, God, I wish I had approached it from that angle. But if I did, I wouldn’t have learned everything, so. Michael Houck: No, you approached it exactly how I think you should approach it the first time you do it. But my point is, like, now that it’s not the first time you do it, and now that you have sort of seen all this stuff, that’s why I want you to take a step back and sort of be like, all right, now let’s really map this out a little bit. I know how this shit works. Yes, there are bugs, whatever. But you know how it works. So take that step back, map it out, and then we can prioritize with this roadmap file what we actually wanna do every couple of weeks, every sprint, basically. Ben Ornstein: Yep, yep, yep. Makes sense. Yeah, that’s really important. So I will do that. Cool. UNKNOWN_SPEAKER: Yeah. Ben Ornstein: And the, I guess I just realized that two workflows are the same thing. Ben Ornstein: Research question. Ben Ornstein: Yeah, workflow one and seven are the same thing. Okay. Michael Houck: Yeah, and like when you realize things like that, look for the things that are generalizable. About them, but also look for the things that aren’t. Like what does really make them unique? Is it just one skill? Is it an attribute of the agent who should be doing them? Like note what that is so that as you’re designing the whole system, you know where to draw those lines of demarcation. Ben Ornstein: Yeah, for sure. And then there’s similar to the breaking news content, but different. Is like proactive content surfacing. So like, it just does research. And when it finds something interesting that may not necessarily be breaking news, but it is still relevant for a client to comment on. That’s, it’s like a proactive content surfacing thing. Michael Houck: So what would be an example of that that is not breaking news? Because in theory, like, wouldn’t anything proactive be like, I’m speaking from ignorance here, right? Wouldn’t Wouldn’t anything proactive be like something new that’s happening? Or are there things that like it could discover that are not new? And like, let’s give me an example of what that would be. Ben Ornstein: Well, I guess it would still be breaking news. Again, I always think from Knowles since for some reason that’s the one that goes to. So like private markets, there’s this new legislation that comes out, but I guess in a sense that’s still breaking news. So maybe there isn’t room for that. Michael Houck: Unless we can find an example, I would treat them as the same thing. Ben Ornstein: Yeah. And this is, again, the value of me taking a step back and actually doing all that. Right. And then the last workflow I came up with was content recycling. So this is specifically from something Jacob did where he’s like, hey, can you go through Brindis’ post from anything from August 2025 or before that did really well, can you like show me these, like put the URLs and let’s make something either the exact same or similar because we know it did well. And so like, it does like a content recycle where it brings back past posts and then does something with them. Michael Houck: Yeah, I would break that into two parts. Part one is like every act like, look back in the past and identify, you know, high-performing posts, and then analyze, like, what made them perform well, what patterns emerge. And then part B would be, like, okay, if you get the thumbs up from the user, the user here being a team member, probably, then actually use those to go create, like, a new batch of content. Because, like, Step one is understanding it and step two is then acting on it. We don’t just want it to act on it and then be like, yeah, I made these posts, right? We want us to have that intermediate step of like, we understand what about this account is doing well versus not. Ben Ornstein: Yeah, that makes sense. Cool. Um, in terms of like the workflows that I have somewhat defined, that that’s all of them on the content front. Um, Yeah, yeah. Michael Houck: Yeah, I think that’s totally fair. What about not on the content front? Because again, we don’t just want to think about this in terms of the content agent. Ben Ornstein: Yeah, so my focus has been on all things content so far. So I know you said this week you wanted me to focus back on things outside the content. So when you speak with the potential acquirers that we’re doing stuff outside content. So maybe, yeah. Michael Houck: I would like to see us treat Clawdee internally right now as like a product for Blake and I and for our team. I don’t want us to treat it like a client product right now. Because if we do, then they’re going to argue it should be included in the sales. Like when we talk about it and what we actually build for it should just be focused on the team for now. Ben Ornstein: Yeah, for sure. And so I think the two things would be Do I make the editor bot, or do I focus on the partnership side and figure out what are the most important things to build there? I think that’s pretty high leverage, unless we decide that Blake should start around and just keep doing what he’s doing. Michael Houck: Sorry, Blake Stewart. UNKNOWN_SPEAKER: Blake Stewart, right. Michael Houck: Yeah, I mean, I’d like to automate the partnership stuff. I think that there’s an argument to be made for that. The editor, I think, is something that for now Blake and I can continue to do for ourselves, like you, Mal, and I can continue to do for ourselves. So I would say, take the step back, show me this prioritization list, like tomorrow, and then I’ll do a final pass on what we should do. My guess is that I will want to prioritize non content stuff, but we’ll probably also be still a couple small content things and just like quality of life improvements for the existing functionality as it exists today. Ben Ornstein: Got it. Okay. Yeah, that makes sense. Yeah, that makes sense. Michael Houck: Okay. Ben Ornstein: Yeah, my brain is just filled with this stuff. I’m like constantly thinking about it all day. Oh, yeah, he’s doing that. Michael Houck: You’re doing a really good job of it, man. Like I continue to be impressed. You’re really like owning it, which is what I wanted. So keep it up. Ben Ornstein: Yeah, I know I’m not doing everything quickly or efficiently, but I’m trying, I promise. Michael Houck: It’s, you know, again, no expectation of perfection when you’re literally learning an entire new discipline here. It’s okay. Ben Ornstein: Yeah, when you put it that way, it makes me feel better. Michael Houck: Yeah, that’s for sure. Michael Houck: So yeah, just keep at it. And then I think I think the big thing is just get me that sheet so that I can kind of get visibility. And then I’ll be able to say with certainty, like, okay, here’s what we should prioritize for this. We should create it as sprints, right? So like, let’s say that like sprint one was the last two weeks, sprint two is this week and next week. All right, what are we gonna ship in the next two weeks, basically? Ben Ornstein: Yeah, yeah, yeah, that makes sense. I do like the more structure here, because I’m being like, oh, well, I guess I should fix that today. Oh, that shit still isn’t working. I guess I should try to attack that first. I’m kind of going blind. So I’m actually glad we talked about this, because I need a little structure for how to go forward here. Michael Houck: Any engineer will tell you that things would get really chaotic if we didn’t, we didn’t at least do a little bit of tracking. Obviously, there’s like, there’s a way to unbalance that in the other direction, too. And, you know, focus too much on, oh, well, the, you know, we’re doing things outside of the roadmap. But like, things change, we need to be adaptable. But just like every couple weeks, we want to reset and be like, Alright, what have we learned? What’s changed? What do we prioritize right now? You know? Ben Ornstein: Definitely. No, that’s, that’s going to be super helpful for someone who still doesn’t know what they’re doing. So cool, man. Michael Houck: All right. Well, um, I guess he has the next step. Just shoot me over the, um, the sheet when it’s ready for me to take a look at, and then I’ll do final prioritization. Um, And we’ll, we’ll go from there for, for what’s on the docket for this week. Ben Ornstein: Cool. Sounds good. Um, I’ve got, I’m almost done with, uh, Tuesday sand. I just have to add two more house picks and then I’ll, I’ll send a message letting you know it’s ready for review. Michael Houck: Okay. I did not get the newsletter out this weekend. Um, so if we have sponsors from there, just like let Blake, uh, no, we need to reschedule them or whatever. Okay. UNKNOWN_SPEAKER: I’ll message him. Ben Ornstein: All right, dude. UNKNOWN_SPEAKER: Thank you. All right. Talk soon. UNKNOWN_SPEAKER: See ya. UNKNOWN_SPEAKER: That’s great.