Sean is back from two weeks of Spanish immersion in Puebla, and the conversation picks up right where it always does. We get into MetaMonster's 100-demo sprint with their new co-founder, why AI keeps producing bad content for founders (and what Sean shipped to fix it), and somehow end up deep in a conversation about outcome-based agency pricing and why it is a finance nightmare in practice.

Links:

For more information about the podcast, check out https://www.smalleffortspod.com/.

Transcript:
00:00.41
Sean
Welcome back to Mexico City.

00:02.26
Andrew
What a

00:06.90
Sean
And let's get

00:07.11
Andrew
up? I still can't remember how, I still don't know how to say, can never remember how to say like, good to see you. I think it's like, mucho gusto verte, or like, con gusto.

00:20.02
Andrew
I think con gusto is like the the simplest way to say it, but... um

00:24.79
Sean
con gusto my friend um me amigo um

00:30.58
Andrew
Asking Claude, how do I say nice to see you in Mexico?

00:37.71
Andrew
I've been using Claude a bunch for, did i have I talked about this in a past episode?

00:40.96
Sean
yeah yeah yeah but no it's it's it's good it's like um

00:42.08
Andrew
Okay, my bad.

00:46.01
Andrew
Que bueno verte, me da gusto verte. Que gusto verte, que gusto verte is the most natural way. Literally, what a pleasure to see you. Que gusto verte.

00:58.77
Sean
I feel like that's not a thing people say. I feel like.

01:02.84
Andrew
I don't know how often how often people say it yeah

01:08.67
Sean
Yeah. Interesting. um how was How was your trip?

01:14.97
Andrew
It was awesome. um Just going from using Spanish maybe anywhere from half an hour to like two or three hours a day to using Spanish from 9 nine am until like five every day and then sometimes for another couple hours at night.

01:34.68
Andrew
depending on what plans I had with friends and stuff, um was definitely like much needed. Like I felt my vocabulary improving. i felt myself building some confidence, speaking Spanish, getting a little bit more comfortable.

01:50.14
Andrew
um Also underscored how much I still don't know and like how far I am from where I actually want to be, which is like kind confusing. good and frustrating at the same time.

02:03.32
Andrew
um But overall, really fun and also just nice to see a new part of Mexico. Like, Puebla is really beautiful. um It's a smaller city, a little bit more chill.

02:16.06
Andrew
um but has a really beautiful historic center with porticos that reminded me of like the porticos in Bologna in Italy. um And really good food.

02:29.66
Andrew
um I ate something and like fucked up my stomach for like a week. So that was a bummer, but also just part of being a gringo living in Mexico. um So it comes with the territory.

02:44.76
Sean
You were going for, I want to say it was two weeks, but it feels like it's been a very long time since we've talked.

02:47.10
Andrew
Just two weeks.

02:50.35
Sean
yeah, I miss you as well.

02:50.68
Andrew
Oh, I miss you too. Nah, it was just two weeks. Yeah, but super fun.

02:58.52
Sean
Good. Good, good, good. Yeah. That was a good picture from your group.

03:03.93
Andrew
Oh, of the of the class.

03:05.53
Sean
Yeah, yeah. That's good.

03:06.56
Andrew
Yeah, the school does a really good job of like, so you take classes from nine to one every day, and then you have lunch with the other students. um And you're strongly encouraged to speak only in Spanish while you're at the school.

03:21.82
Andrew
um So you're you're speaking to other students who are at different levels, but speaking in Spanish.

03:21.96
Sean
Nice.

03:28.48
Andrew
And then in the afternoons, Monday through Wednesday, you're paired one-on-one with a guide who's just a local. um And you walk around Puebla and go visit museums and tour different parts of the city with your guide. um ah The second Monday I was there, i got two of the other students and their guides to come with me to a board game store and we just played board games and talked in Spanish and it was super fun.

03:57.64
Andrew
Muy lindo, very cute.

03:59.72
Sean
Nice.

03:59.77
Andrew
um And, ah but yeah, so you get like a lot of one-on-one time with your guide. And then on Thursday, you do like a bigger expedition with the group to a city, you know, a small town or city nearby to Puebla. Yeah.

04:15.04
Andrew
And then, ah yeah, Friday afternoons you get off.

04:15.40
Sean
Nice.

04:18.78
Andrew
So,

04:19.82
Sean
It's like a full immersion type of thing. I like it.

04:22.43
Andrew
yeah, exactly.

04:23.14
Sean
Yeah. Cool.

04:24.44
Andrew
um Yeah, super fun.

04:25.43
Sean
You feel like you are drastically better at speaking the language now?

04:31.52
Andrew
No, I would not say drastically better. um I would say...

04:36.90
Andrew
marginally better. Like definitely, like i like I said, like definitely improved my vocab and my confidence, but like still two weeks just isn't enough to like for a drastic change.

04:46.47
Sean
Right, right.

04:47.76
Andrew
I, um yeah, i I need, I wish I could have spent

04:49.21
Sean
Gotcha.

04:55.55
Andrew
you know, two months there, because I think that would have been the kind of time I need to, like, really take a step forward. um Although I found this funny thing. i've I've spent some time on the language learning subreddit over the last, like, six months or whatever.

05:11.77
Andrew
And I found this funny joke where like every level thinks they're six months away from being fluent. Like every level of speaker is like, yeah, with like six more months of study, I'll be fluent.

05:21.25
Sean
Yeah.

05:23.68
Andrew
And it's like, you also just realize the more you learn, like, and Fluent is not easy to define and is kind of a shifting goalpost. um There's like true native speaker fluency. And then there's like, you know, understanding scientific terms. And um so like, you know, I think my goal is just conversational fluency, like be able to meet people who speak with a slightly different accent and not panic and like ah be able to like hang out in a crowd of like, you know, locals and like carry a conversation without being super stressed or watch a movie in Spanish and understand it without subtitles.

06:12.37
Sean
Cool.

06:15.19
Sean
Cool.

06:19.13
Sean
My brain's going into, like I forget what it's called, but there's like, a I don't know this is like i don't know if this is for like all languages, but like there's something like from what I understand, but Japanese, there's like an N1 or N2.

06:31.96
Andrew
Yeah.

06:32.28
Sean
that Is that it just like a language proficient sheet proficiency thing in general?

06:32.40
Andrew
So there's

06:35.93
Andrew
Yeah, I think so.

06:36.14
Sean
I

06:36.73
Andrew
um there's there's this There's some sort of like international organization that has created tried to create these levels um for language learners, um and they have courses and exams and stuff.

06:42.02
Sean
see.

06:51.33
Andrew
And so you can pay to take an official exam and it'll give you your level. And that level can be accepted as like certification for jobs and things like that.

07:00.17
Sean
Gotcha.

07:01.58
Andrew
and so

07:01.93
Sean
Right.

07:03.05
Andrew
It's like beginner is like A1, A2. Intermediate is like B1, B2. um Advanced is like C1, C2. And then I think beyond that, it's like native. It might be like N1, N2, or like just native or something. I'm not sure.

07:19.26
Sean
Gotcha.

07:19.26
Andrew
um

07:20.02
Sean
Gotcha.

07:20.06
Andrew
i took like a free online exam and I was B1, which kind of makes sense. It's like... like just beyond beginner, but still like can carry a conversation in the right conditions with the right person, but like still not, you know, anywhere near fluent.

07:34.28
Sean
Nice.

07:38.66
Sean
Right, Cool.

07:39.38
Andrew
I think like, I think it also is one of those things where it's not like a linear scale. So like getting from getting to be one doesn't take like it can take way more time to get from B1 to B2 than it takes to get from like a one to b one

07:53.84
Sean
Totally, totally. Yeah, I mean, I feel like um I feel like especially like as you get like as you become able to exist in a a society with it, even if it's not like perfect fluency, it like continues to like stagger. Yeah.

08:10.13
Sean
Yeah. But cool. Nice.

08:12.62
Andrew
Yeah.

08:13.18
Sean
I'll ask you again in six months and then we'll

08:15.12
Andrew
Oh, that'll be interesting. Yeah. um

08:18.32
Sean
see.

08:18.38
Andrew
six months ah Six months from now, I will have been here for 10 months. So like hopefully by then I'm getting more confident and comfortable. um I feel like my friends who have been here for, know,

08:32.88
Andrew
like I have a friend who's been here for four years and he also spoke French already because he is half French or something. um

08:41.55
Sean
Right.

08:42.51
Andrew
And so that helps for sure. But, you know, he speaks what to me comes across as fluent. I'm sure he's not like native level speaker, but to me, he seems fluent. He seems like totally comfortable speaking. And then like,

08:59.44
Andrew
my my teacher's boyfriend, Tom, who's no who I spent a bunch of time with last week, um he's from Germany, has also lived here for like four years and seems like kind of at that fluency, at like my definition of like conversational fluency level.

09:13.85
Sean
Right, right.

09:17.35
Sean
Right.

09:17.33
Andrew
So maybe four years would do it.

09:18.47
Sean
Yeah,

09:22.57
Sean
yeah it's not that long. ah You blink in it and it happens. Cool. um and Did you work MetaMonster stuff while you were there, or are you full?

09:34.03
Andrew
Barely at all. Yeah.

09:35.19
Sean
Yes.

09:35.55
Andrew
Going back and forth with lawyer on some legal paperwork, we're trying to get signed. But like um my plan was to do a bunch of metamonster work. I told myself like, all right, you're going to take classes during the day and then at night you'll have the nights free.

09:49.40
Andrew
And so you'll work on metamonster stuff. And one, my brain did not want to function after nine hours of trying to speak in ah another language.

09:59.02
Sean
Makes sense.

09:59.09
Andrew
um I was just like completely brain dead every day when I got home.

10:02.72
Sean
Makes sense.

10:02.86
Andrew
um And then, too, I ended up just, like, having more plans in the afternoon. And, um like, my teacher at this point, we've talked to each other every week for almost three years.

10:13.13
Andrew
And so, like, we've now become good friends. And so, like, I hung out with her and her boyfriend and her family a bunch um and wanted to, like, take advantage of being in Puebla and getting to see them for the first time.

10:15.85
Sean
Right.

10:25.26
Andrew
So I ended up just, like, it's more more of a vacation workcation.

10:25.73
Sean
Nice.

10:33.66
Sean
Nice, nice.

10:38.45
Sean
I think that's good. you know I feel like feel like it's good to like like fully you know be in the immersive experience. not like I feel like i feel like give you if you let yourself then do metamonster things, I feel like it impacts the later on like the later days where like you start weighing the, you know so I think it's good.

11:00.63
Sean
so

11:01.03
Andrew
Yeah, potentially, yeah. um And I think I mentioned we have a new, a third co-founder now. So like work kept happening while I was gone, which is great.

11:11.86
Sean
Hell yeah.

11:12.91
Andrew
um She's absolutely crushing it.

11:13.03
Sean
Good.

11:15.34
Andrew
ah we We like jokingly set ah a goal of 100 demos by the end of May. And then I was like, that's not realistic at all. Like the actual goal should be 30. She's already booked 25.

11:27.63
Andrew
um it's looking like 100 might be the actual goal.

11:31.60
Sean
Showing you up. I love it.

11:33.84
Andrew
massively which is like exactly what we needed we needed someone who was much better at at marketing than i was um so yeah i feel like you know for years we've seen this pattern of like if you pair a good product person with like a really good like subject matter expert with like industry connections that's like sort of the recipe for success for a bootstrapped business um and so now

11:34.32
Sean
I love it.

11:37.58
Sean
Yeah.

11:42.44
Sean
Hell yeah.

12:02.36
Andrew
with Austin and Jade, we have that. And then I'm just tagging along for the ride.

12:06.60
Sean
ah yeah yes

12:10.28
Sean
Listen, someone's got someone's got to talk to the lawyers and all the others.

12:15.74
Andrew
Hey, I shipped a feature today. i shipped i shipped an update to our WordPress plugin so we can now sync image alt text um to WordPress.

12:18.00
Sean
Oh.

12:22.57
Sean
Whoa.

12:24.02
Andrew
So yeah, I just have this like weird hyper role now where I'm like,

12:25.81
Sean
That's pretty good. that's pretty That's pretty good.

12:30.98
Andrew
Doing like a little bit of sales, a little bit of marketing, a little bit of product, a little bit of ops, like whatever needs doing.

12:38.14
Sean
I mean, i think that's... um I think that's valuable too for what it's worth. I think that's like, um you know, I think like the team needs like a junior engineer. So it's good. to um Do you, are you like, when you, when you vibe code now, um are you like,

13:00.84
Sean
Like, like, what is your, um, like for me, right? When I'm building stuff, I don't look at the shit anymore.

13:04.05
Andrew
No.

13:05.36
Sean
I'm just like only in like, uh, I will look at the, the, the dev instance and then otherwise it looks good. Push that, push that sucker up. Um,

13:15.21
Andrew
Yeah. um Yeah, I test things. Like, I test everything. And ah occasionally, if it's tiny enough, I won't even test it, which is bad.

13:25.81
Andrew
I should test everything. But um yeah, no, I don't look at the code.

13:27.32
Sean
yeah.

13:29.85
Andrew
I will sometimes, like, I will look at the plan and try to make sure the plan makes sense. And I'll sometimes, like, have like a new Claude instance look at the code.

13:40.76
Andrew
um a ah I feel like it works better to have like GPT look at the code, but I don't feel like paying for two max plans right now. So um ah yeah, so I'm just, but no, I do not look at code right now.

13:59.26
Sean
Nice. Do you ever tell Claude that Codex will be reviewing its work?

14:03.64
Andrew
No, but that's hilarious.

14:07.90
Sean
um uh during the like during this period where like ah like opus 4.6 was like shitting itself yeah it absolutely like it got ways like it worked way harder yeah um i don't i don't think it's that i don't think it's um telling it specifically that it's codex that it works harder like i don't think

14:09.48
Andrew
Does that work?

14:23.56
Andrew
That's fascinating.

14:30.100
Andrew
It's just like telling it that anyone's going to be reviewing its work. Mm-hmm.

14:33.82
Sean
Yeah, yeah, like a senior, basically I think like a senior engineer. Like you can see the chain of thought is like, oh, like because my thing will be reviewed, I need to like make sure like I do X, Y, Z things, which is incredibly frustrating because I'm just like, what do you need?

14:47.43
Andrew
Always do those things.

14:48.60
Sean
Why didn't you do this before?

14:50.60
Andrew
It is funny because it it, like, that's kind of the most human thing about it is that it seems to be lazy sometimes.

14:55.17
Sean
Yeah.

14:58.34
Sean
Yeah.

14:59.81
Andrew
um One of my friends, I was catching up with Adam, um who who did some work with Miscrance.

15:06.21
Sean
Oh, oh, that, out yeah, yeah, yeah.

15:07.46
Andrew
Yeah, yeah, yeah. um Adam Baumgardner.

15:09.98
Sean
Yeah.

15:10.02
Andrew
um ah And um yeah, Adam and I were chatting and he was telling me that... ah he finds that like he can get around Claude being like too nice sometimes by telling it that to that it's reviewing someone else's work.

15:25.62
Sean
Hmm.

15:29.86
Andrew
It's like this isn't my work. Like my friend wrote this. Will you review it? And Claude will be more harsh than than when you're just like and be less like you're perfect.

15:34.42
Sean
Um...

15:41.07
Andrew
You're great.

15:42.66
Sean
You're a superstar.

15:44.93
Andrew
Yeah. All of this is kind of funny, by the way, because we're having the opposite problem in MetaMonster right now.

15:48.34
Sean
Yeah.

15:50.81
Andrew
Like, we've been working on our big new content optimization feature. And, like, it's been really hard to get the initial grade above, like, a Yeah.

15:54.98
Sean
Yeah.

16:02.16
Sean
Oh, interesting.

16:03.64
Andrew
Like it is harsh. And like the recommendations it comes up with are always solid, almost always.

16:10.08
Sean
Okay.

16:11.17
Andrew
Occasionally there's some bullshit in there, but like are almost always solid and like make a lot of sense. And so you're like, why'd this get an F? And then you're like, no, well, this all makes sense actually.

16:21.57
Andrew
um But the hard thing is like you can implement stuff, rerun it, and it'll still like, it'll barely improve the grade because it'll just find new stuff. And it's like,

16:30.62
Sean
Right.

16:31.26
Andrew
um So it's it's hard it's been hard to figure out how to get it to strike a balance of like tough but fair.

16:39.18
Sean
Right. Interesting. i ran into this ah I ran into this loop while I've been building, not not like um you know not not for the content optimization, but like telling Cloud to review.

16:51.80
Sean
would have a bunch of findings. It would fix try to fix the thing, asking it review, and it's still lot more findings. So I did run into that loop, whereas I don't feel like I ran into that loop with codecs.

17:02.90
Sean
In fact, what it happened originally.

17:04.10
Andrew
Hmm. Have you been using Codex?

17:06.36
Sean
Yeah, yeah.

17:06.51
Andrew
I've heard a lot of people hyping up like the latest like 5.5 um

17:11.46
Sean
Dude, this is like, this is Codex, ChatGVT is like, our opening eyes, like time. Like, Anthropic is ah is isn' is currently not the ah engineering darling at the moment with 4.7 and like...

17:29.77
Sean
um

17:30.26
Andrew
But it's all so fucking fickle. They also, the Anthropic team claimed that they fixed a lot of the problems with 4.7. I don't know if they actually did, but.

17:38.73
Sean
I use, but I mean, I use both. I think both are, both are good. Like, honestly, um, like I, I didn't actually run into the same, like, uh, I don't think I ran into like the same code quality type issues with, um, uh, that, that other folks are running into, but I also don't think I was building anything like super crazy for what it's worth.

18:02.50
Sean
Um, um,

18:03.75
Andrew
Yeah, I didn't notice a huge decline, but I also like the, when everyone was complaining kind of coincided with my time in Puebla. And so I wasn't doing as much while I was there.

18:15.30
Sean
yeah same same except i guess for me it was more building out things in click up and like project management stuff so um yeah uh that that was my entire weekend is we're switching we're switching project management tools again

18:22.82
Andrew
fun.

18:31.24
Andrew
This is like what the fifth time in the last year?

18:33.99
Sean
Yeah, yeah, yeah, yeah. um I don't, it's not so much that we we're necessarily switching project management tools or that like we're definitely going to ClickUp or something. I think it's more just, um I know enough to...

18:49.22
Sean
like draw, write out all the requirements and exactly how I want things done and like where the issues are. I don't know know to solve for some of them for what it's worth, but like, uh, I can solve for like 80% of them and which would be, which would be such a significant benefit difference.

19:06.62
Sean
Um, and I'm like motivated by desperately trying to pull myself out of the business at the moment. Um, so I can, if I have code,

19:15.75
Andrew
feel like this is just the nature of project management is like, it's kind of like software and that it's never bug free.

19:18.85
Sean
yeah.

19:22.10
Andrew
You're constantly tweaking, fixing, improving, refactoring, et cetera.

19:26.12
Sean
Yeah. Yeah.

19:30.29
Sean
Totally. Totally.

19:32.55
Andrew
Maybe it's just systems that humans use. I don't know.

19:37.48
Sean
Probably. I think it, well, I don't know. I think it's like systems that anyone, anyone uses like, or not even just humans, like systems that agents use as well, especially when the models like upgrade or or downgrade or degrade.

19:52.60
Sean
um

19:56.18
Sean
Cool. um It's good to hear that your content optimization thing is like, Like, I feel like you wouldn't want it to tell you that it's great. Like, off the bat.

20:06.31
Andrew
Yeah, but like if it grades everything as an F, people, that also feels like a little frustrating.

20:07.52
Sean
Right.

20:13.01
Andrew
And like if it's impossible to get it to an A, then that feels like, okay, like, so we, I mean, we've talked about a bunch of things.

20:23.52
Andrew
We've talked about, right now we're still trying to tweak the rubric and like tweak the instructions and and see if we can get it to a sort of system that feels more accurate.

20:34.82
Sean
Right. Mm-hmm.

20:35.83
Andrew
We've also talked about like maybe we don't need grades, but people really like grades. um It's like maybe we could just in include the raw score. Maybe we could just be like good, better, best. Maybe we could just do um ah you know, needs improvement.

20:55.94
Andrew
Go back to like my Montessori grades that weren't grades that like pretended not to be grades. um I was a Montessori kid and we weren't allowed to have letter grades. So instead we have like had like dots, minuses and pluses. And it's like, you guys realize this is just a fucking grade, right? It's just a grade. you just changed it from a letter to a symbol.

21:13.66
Sean
yeah

21:17.66
Sean
i didn't know you were a montessori kid i didn't know that was i didn't know that was a thing like i thought i just huh i thought it was like a relatively new thing oh no that's great i like the i like the concept i mean i didn't know about that but i like the concept of what it's built on of like

21:19.42
Andrew
Oh yeah, explains a lot, doesn't it?

21:28.84
Andrew
No, it's been around since like the 80s, I think. um ah

21:38.79
Andrew
I totally agree. The concept of Montessori is super cool. For anyone listening who doesn't know what Montessori is, Montessori is a sort of alternative learning system um that has a few fundamental differences from from traditional learning.

21:55.72
Andrew
the traditional education system in the US. One of the big differences is that lessons are designed to be more tactile and hands on. um So for instance, you learn numbers with these things called bead chains where you're literally like counting beads and you sort of see how like smaller numbers are broken up into bigger numbers. And We learned division with something called test tube division where you like move beads between test tubes to see how you know if you take four beads and You divide it between four test tubes. You end up with one bead in each tube um And so it's like these very tactile hands-on sort of things But there are some other differences too like and I'm sure it varies from school to school, but at our school we didn't have desks um so

22:43.16
Andrew
We had a few tables, but mostly kids sit on mats on the floor. um It's designed to be much more like child driven and designed to be much more um like kids can learn at their own pace. So it's not like every kid is learning the same thing every day. They can kind of pick up individual lessons and learn sort of at their own pace. um They're encouraged to like keep the classroom clean and like sort of learn some sort of self motivation kind of stuff.

23:10.18
Andrew
You also often have kids of different age groups in the same class. And instead of the teachers teaching everything, you have the older kids teach the younger kids. And the teachers are there to supervise and provide additional support to the people who need it, as well as to provide like discipline and things. um And so, yeah, I think in theory, it's actually a really, really great system.

23:34.22
Sean
Right.

23:34.20
Andrew
um In practice, it can be difficult. um You know, it has its own, its flaws like anything else. um And it can be really difficult for people transitioning from Montessori to traditional education systems.

23:48.20
Andrew
um It's like, you know, pretty jarring to go from this very sort of loosey-goosey kind of like independent world to um often at like a smaller school to like a bigger school where everything's like super structured and rigid.

24:06.20
Andrew
um

24:07.29
Sean
Right. That's like the whole.

24:07.96
Andrew
When I was there, we didn't have homework either. um

24:10.88
Sean
Right.

24:11.88
Andrew
And now I think they've like changed that. They've started like giving kids homework because they're trying to like find that balance of like preparing them for the traditional system while also like letting them

24:19.88
Sean
Nice.

24:23.84
Andrew
you know, taking advantage of the strengths of Montessori.

24:26.68
Sean
Nice.

24:29.03
Sean
Yeah, I wonder if there's, I mean, there probably is like a study of like Montessori kids. when what Actually, another week I think Bezos was like a Montessori kid.

24:39.05
Andrew
Was he really?

24:40.01
Sean
I feel like that's where I heard it about Montessori, but i thought I think I thought that Bezos was funding a Montessori school. But now that I'm thinking about it, I think it was because he was a Montessori kid.

24:50.60
Sean
I don't know. I'm not super sure.

24:52.72
Andrew
That's wild.

24:53.38
Sean
yeah Anyway, I just wonder if there's like a study of like where they end up, you know like like collectively, all that stuff.

24:53.86
Andrew
um

24:58.80
Andrew
I wouldn't be surprised if if it was like, if it sort of enhanced the extremes, um because I feel like I knew a decent number of kids from my,

25:05.35
Sean
Yeah. I think it does.

25:11.06
Andrew
grade and like grades around me who really struggled later in life um because they like, they felt kind of different. Like I said, they weren't like super used to the traditional system. They didn't learn some of like, there's some benefit probably as much as I hate it to like just learning to jump through the hoops.

25:31.30
Sean
Right.

25:31.30
Andrew
um

25:31.42
Sean
Right. Yeah.

25:32.23
Andrew
And like Montessori kind of teaches you the opposite. But then on the other hand, it also teaches you like independence and like how to like seek out intrinsic motivation and learn from yourself for yourself. And so like I kind of wouldn't be surprised if it enhance if it like increased the number of people on either end of the spectrum who are either like really successful or like really struggle.

25:59.59
Sean
don't know, I'll look it up later. um Sweet. um

26:06.18
Andrew
What have you been building with your, with your vibing other than project management systems?

26:10.31
Sean
Oh.

26:14.20
Sean
What have I?

26:14.34
Andrew
Which it sounds like you're still doing the PM stuff pretty much by hand.

26:18.96
Sean
A little bit, a little bit. I'm using Claude and Codex and whatnot for it. and like like ah like it's It's sort of like a

26:27.30
Andrew
Are you using them more as like a thinking partner or more to like actually like move shit around?

26:29.80
Sean
yeah yeah so sparring partner for now but there is a click of mcp um which which makes like the whole like um so my goal is that like we can write sops just in like text form or even like ideate with it and then just like throw it like just like hit the click of mcp and like make tasks um and think there's a lot of like other things around that that i'm like

26:37.86
Andrew
Nice.

26:56.10
Sean
like because ClickUp has an MCP, like I have some thoughts about like time tracking that can actually be like way easier for everyone because like if your calendars sync to ClickUp or or it doesn't matter, right? There's the the Google Calendar API. Like you can, ah like as long as you are,

27:15.88
Sean
like logging the fact that you were doing things in your calendar, even if you're like, hey, like gonna be busy, deep work, whatever, which is how I already use my calendar. It makes it very easy to just pull it down and then push it out to like where whichever sort of platform and run that automatically.

27:32.20
Sean
um ah What else? ah Yeah, but for not right now mostly mostly as a sparring partner, trying to figure out how to like reshape our retainers, trying to figure out, like I really hate the way that we like currently do these capacities.

27:49.28
Sean
I'm trying to figure out how to like not sell time. you know And that's, yeah.

27:51.62
Andrew
Yeah, dude, we've constantly struggled with that. um Yeah, you're doing points right now, right? So essentially credits almost.

28:01.77
Sean
Yeah, yeah, it's just abstractions on top of hours, but so that no one gets punished for like working too fast or too slow.

28:05.32
Andrew
Yep.

28:09.54
Sean
But even then, like, I think it's just too output focused and not outcome focused. But the problem with like, how do you how do you ah

28:22.12
Sean
like how do you do an outcome paced payment if it's not like just on performance and then that's its own like nightmare and i don't want to deal with.

28:27.42
Andrew
Yeah.

28:29.68
Sean
I'm trying to find balance.

28:30.36
Andrew
Yeah, the thing we always struggled with with that is like value based seems great in theory, but it's like figuring out

28:31.90
Sean
Yeah.

28:37.22
Andrew
It it like works really well for certain services, but other services, it's really, really hard to derive a true value metric for um that that you have sufficient control over to like really affect things.

28:45.05
Sean
Yeah.

28:47.38
Sean
Yeah.

28:53.04
Sean
Yeah.

28:53.05
Andrew
Because like you know we like when we were a product design agency, it doesn't matter how good your product is if you have shit marketing.

28:53.16
Sean
Right. Right.

29:03.46
Andrew
um And so like we couldn't rely on product growth metrics or something alone. Though I feel like the only people I've seen do it really well, where it feels like a really, really natural fit.

29:18.96
Andrew
I mean, you can go like project based where you're not like necessarily tying things to a value metric, but you're like quoting an amount that is more tied in the the client's eyes to like the value they're getting out of out of things.

29:32.90
Sean
Yeah.

29:34.85
Andrew
um But then at the end of the day, it's like deliverables still, but it's just not time based.

29:39.16
Sean
Yeah.

29:40.78
Andrew
It's like, so you can go that route or I feel like advertising agencies um that charge just like a percentage of ad spend are like kind of the closest.

29:51.66
Andrew
But even that's like not really a value metric because like the value to the client isn't the ad spend.

29:53.73
Sean
Yeah.

29:58.18
Andrew
The argument there is obviously that like you're spending more only if you're successful. And so but it's it's still imperfect.

30:04.70
Sean
Yeah, yeah.

30:07.40
Sean
Yeah, also, so ah was talking to my new head of, my guess I'm gonna call her my head of finance, technically she's my senior controller, but effectively my head of finance, we were jamming on like different ways to do this yesterday, and it's like, I learned, because because she's had done it at previous agency, like ad agency that does this like commission thing,

30:27.62
Sean
ah it is a finance nightmare to like ad agent, like, cause ad agencies, i think traditionally, like you pay them your ad budget and they hold it in an account and then they spend it for you.

30:39.18
Sean
Cause they're not going front you.

30:40.09
Andrew
Oh, I didn't know that.

30:42.25
Sean
They're not going front the money for you, but they also need to like, they want you to commit that spend and ah you know, like, like, but you know, it doesn't make sense for them to have this.

30:43.85
Andrew
No, God, no.

30:52.29
Sean
Anyway, the point being is that like, that is, like hellish to become the steward of someone else's capital but and like requires a completely different like financial like setup for it um because it's effectively a liability not a not a it's not revenue for you um Anyway, so it's a whole it's a whole thing.

31:16.44
Sean
um But I have been building a lot of like small fun things, as you've seen my Manosphere, ah map the Manosphere

31:26.14
Andrew
Where the fuck did this come from, dude?

31:26.68
Sean
on uh okay so like personally i've always had a so huge interest in it like as as like a pure like anthropo and anthropology study of like like you know because like i i grew up with like this type of influence right and like like um not that i like fell deep into it but like you know outskirts right like like hustle culture is related to it and all this sort of stuff um

31:52.34
Andrew
Mm hmm. What's your like definition of the mana sphere?

31:59.08
Sean
I don't know if I have like a definition to be honest with you. i think that like, I think the thing that is always super like, like interesting to me is just like male centered.

32:16.100
Sean
media and how it like becomes a pipeline to other things that go deeper and deeper, right? Like from sneaker botting all the way deep into like alt-right pipeline things and like the creators and whatever that sort of are, that float around this this area.

32:31.38
Andrew
Mm

32:38.26
Andrew
hmm.

32:38.31
Sean
um I think the interest is one often, ah i've you know, I've had friends who, I've had friends who like ah come out of relationship and like, you know, they they go online and they start learning about all these things and then like, like, have like, right, right.

32:57.32
Andrew
They get red pilled essentially.

32:59.25
Sean
And I have like, definitely have had like conversations before they like went too deep. Like you, like, I need you to block all this stuff right now.

33:06.25
Andrew
like Get the fuck out of here.

33:06.34
Sean
You don't understand. Like, like um Yeah, yeah. ah Like, yeah.

33:11.12
Andrew
Here's the blue pill. Come back.

33:13.32
Sean
um ah Yeah, like like, you know, they start going, right? Because, like like, you know, you grow up as an insecure guy, and then you start looking at, like, confidence things, and you start looking, and then you start learning about, like, like ah the game, and then, like, shit tests, and then, like, all this all this stuff, and then and then and then it gets you deeper and deeper into, like, Andrew Tate-ism.

33:33.31
Sean
Yeah. um

33:35.53
Andrew
I think I've got a hypothesis I want to share in a second, but ah but first, what did you actually build?

33:38.82
Sean
Yeah. ah Okay. So I took everything that I know about the Manosphere, dumped it into Claude, and like gave it like loose instructions about like what things to connect and built like an entire graph of different subspaces that link together.

34:00.64
Andrew
like essentially essentially a topical ontology, is ontology the right word? Like a topical similarity graph. Yeah.

34:07.64
Sean
Yeah, yeah, yeah. it's It's a graph of nodes of like entry points, right, that are just like so like normal like hustle culture-y or whatever things or like into like deeper and deeper, like more so than Manosphere type of stuff.

34:09.44
Andrew
yeah

34:22.40
Sean
um um Yeah, I would say like Manosphere, okay, to like define, I think Manosphere content is a

34:34.56
Andrew
It's like media that's that's aimed at influencing men.

34:41.02
Sean
I would say media that is aimed at influencing men towards the belief that men and women are fundamental have fundamentally different roles in society.

34:51.85
Andrew
Yeah.

34:51.92
Sean
It's probably where ah where where that that lives at.

34:53.31
Andrew
Yeah.

34:57.71
Sean
And then I think everything before that are just like gateways into that. Right. um

35:01.98
Andrew
yeah

35:03.81
Sean
uh yeah so it's been doing that um or i did that i like so uh i made a made a quiz made one of those like like like the bdsm tests or like rice purity or whatever so you can take it yeah yeah literally how red-pilled are you and it gives you like a little cute like astrology thing and it also maps like your interests across all that stuff um it's like a like like i think there's a ah um um

35:14.34
Andrew
Like how, how Manosphere, how red-filled are you or something?

35:25.61
Andrew
Gross.

35:31.46
Andrew
How high, how high do you score if you wanted to build a, ah a supplement um on like a supplement tracking tool for, um for like bio hackers?

35:48.77
Sean
I don't know, but I can, I'll find out, i guess. um I actually don't think I took the quiz for myself or haven't, like, actually.

35:57.44
Andrew
Afraid of the results?

35:59.24
Sean
No, it's just like, if I can map all these things, don't I consume all this content?

36:00.16
Andrew
okay

36:04.80
Sean
Anyway, um it's also a really weird thing in terms of, um, like the quiz doesn't have an icp right like like if you are actually deep into it you don't take this quiz because why would you like want to take the quiz and to tell you that what you're watching is bad and if you don't consume any of this content this quiz doesn't matter to you at all so like why ah but i think it's like a fun like joke thing um so i did that um i oh yeah go

36:30.65
Andrew
Yeah.

36:34.63
Andrew
Okay, so my hypothesis, I feel like Something I have noticed about people who are like deeply plugged into cybersecurity is that they're like not afraid to like look into the darkness and find it fascinating.

36:44.36
Sean
Yeah.

36:51.65
Andrew
like i i When I start seeing Manosphere stuff, I'm not like, this is fascinating, I want to learn more. i'm like, gross, turn this off. I'm going to go over to like a happier corner of the internet.

37:02.08
Sean
Yeah.

37:04.86
Andrew
um But I feel like like Andrew was like this too. Andrew Morris was like this, where it's like he was fascinated by the like really weird dark corners of the internet and wanted to like study them and understand them and like look deeply into them.

37:09.40
Sean
yeah

37:15.36
Sean
Yeah. Yeah. Yeah.

37:19.96
Andrew
And I think that like that is a trait that leads to being successful in the cybersecurity space. um And I think there's like a strong overlap between that, like wanting to understand the like kind of slimy underbelly of the internet and like being interested enough to like pursue cybersecurity at a professional level.

37:41.54
Sean
I think that is very, I think that's very, that's that's a good correlation. I think especially the people who are in like Intel, you know, like you like you have to.

37:51.30
Andrew
Yeah, that makes a lot of sense because there's a big difference between like someone who's working on who like just goes in through like the sort of help desk blue hat or blue team pipeline versus someone who is like in thread Intel research um that side of things.

38:01.96
Sean
Right.

38:07.00
Sean
Right, right. yeah yeah 100 um yeah actually in fact anyone i know who's in like intel like right because you like how else would you know about the telegram channels that like the kids are like you know pulling shit on and like company like hacking companies on or like how would um

38:28.13
Andrew
True, like, I guess it's kind of the job description of Intel to like, just spend your time studying the underbelly of the internet.

38:34.69
Sean
Yeah, yeah. And I think, yeah, I think there's a good portion. It's like, oh yeah, that's kind of what I did as a kid anyway. So.

38:43.00
Andrew
Even people I know, though, who are like super leftist and like very like, ah yeah, like, and a lot of the people I'm talking about are super leftist and and yet are just like fascinated by the like the so like this sort of dark side of of the internet.

38:54.69
Sean
Totally.

39:00.84
Andrew
Yeah.

39:01.46
Sean
yeah it's all weird perversion we have um unfortunately but yeah anyway uh i think i think um uh okay so so what triggered me making this was also because at rsa um ben had watched the manosphere documentary and then we were basically all of RSA was talking about the manuscript because it's not something he's known about ever, right? um um And, like like, I was just, like, dropping knowledge bombs on, like, listening.

39:30.83
Sean
What's talking?

39:30.85
Andrew
we Sean, were you red pilling your you' creative director?

39:35.65
Sean
No, I think I it wasn't telling him that it was good.

39:35.77
Andrew
That's bad. That's not bad, Sean, bad.

39:40.13
Sean
You're right. You're right. You're right. Anyway, we'll be dropping our course on this podcast.

39:49.48
Sean
um uh so around that um more importantly i've been working on a bird app oh

39:58.79
Andrew
The worst way to describe it ever.

40:00.81
Sean
it's a bird app it's um i think if someone goes to sadie.wiki and logs in they will see that it is a bird app

40:02.56
Andrew
is so not a bird.

40:11.33
Andrew
Wait, it's live? You actually shipped something?

40:13.64
Sean
Yeah, yeah, yeah. The homepage sucks, by the way. The homepage is like garbage. um But if you sign in, it actually, you can go through onboarding and stuff. been working on this thing.

40:21.96
Andrew
Holy shit.

40:23.06
Sean
I'm working on this thing called Sadie, S-A-D-I-E dot wiki. um And the idea is, the idea was triggered by... um Kaparthi came out with this thing called LM wiki. It's kind taking everything by storm. People have been implementing it and talking about like memory for agents, all that sort of stuff. Okay, great, whatever. I don't care. um the The things I do care about are ah often we help like founders and and other like technical leaders like write content or come up with like social media ideas. And, you know, their struggles always when I use ChatGPT to like write something, it's garbage.

41:00.13
Sean
um and And then, ah you know, they're like, I tried vibe coding something, it didn't really work. was trying to figure out like why it's so bad at at that, right? Like there's different levels to it. There's like, you have an idea, you give it a chat to you tell the right something, it's garbage. Okay, obviously. So then you like,

41:15.68
Sean
ah you give it more context. You might give it like your voice and tone um or like things you've written before. and like there's there's like different, like you might like create a clawed project for example, and like drop in things you've written, create like sort of build like clawed skills and whatever.

41:34.00
Andrew
Yeah.

41:34.08
Sean
um But then like, it's still kind of garbage and then it gets worse and worse over time.

41:37.35
Andrew
Mm-hmm.

41:38.94
Sean
um ah And I think,

41:41.38
Andrew
Is this, are you building context librarian?

41:45.32
Sean
um

41:45.38
Andrew
Kind of, but like personal?

41:46.21
Sean
Like kind of, but like very specialized. Yeah, yeah. um ah the The thinking here is like, it's because none of these things like truly evolve um continuously with you, along with a couple of other things. So one, the nice thing about the LLM um wiki is the fact that it's an ever evolving context. As long as you input sources and continue to add to it, the idea is that the the wiki self compiles and like creates nodes and whatnot. And that's like a long-term storage brain thing.

42:16.52
Sean
The next part of this is that like,

42:16.52
Andrew
Mm-hmm. Mm-hmm.

42:18.69
Sean
um um ah you know, like basically Sadie's a couple of different like engines. One is that wiki. The other is like, like your soul.md file, but it's a little bit more specific to like, um,

42:34.36
Sean
it's a little bit more specific to ah like your voice and tone. like It creates your voice portrait, it understands like your beliefs, et cetera. It's not like its personality as much as like trying to capture ah portrait of your personality, including like words and and whatnot. There is also like now, it doesn't really work, and like um a more like deterministic policy part of like don't use em dashes, don't use like these types of things, and you see

42:59.40
Andrew
Dude, you should just suck in the the AI humanizer skill that What's-His-Face created.

42:59.81
Sean
see

43:06.46
Andrew
I've been using it and it's been like pretty freaking pretty freaking handy.

43:10.27
Sean
Is it Corey Haynes?

43:10.50
Andrew
prettyrickin' handy No, no.

43:12.71
Sean
Oh, okay.

43:13.29
Andrew
um It's that dude who was doing some like crypto stuff to raise money for like um like research of some rare disease.

43:23.16
Sean
Oh, gotcha.

43:24.31
Andrew
do You know who I'm talking about?

43:25.29
Sean
Oh, ya yeah, yeah, yeah. I forget his name.

43:29.40
Andrew
Blader is his GitHub handle. um

43:31.88
Sean
Cool.

43:32.98
Andrew
Blader slash humanizer is like pretty freaking solid.

43:35.52
Sean
Cool. Cool, I'll use that.

43:38.56
Andrew
Yeah, for like humanization.

43:40.38
Sean
um Cool, last, just to kind of get through this before you have to run. um

43:46.54
Andrew
Yeah.

43:46.66
Sean
The other problem is that ah you doomstroll the internet, you see things and you come up with ideas. So what if we had Sadie, thislip this is is the name of the AI.

43:55.72
Andrew
Yeah.

44:00.90
Sean
good Sorry, let me step back. The point of Sadie is that it proactively thinks of social content ideas and offers them to you. um And I don't think anyone's really solved this problem well yet because it's hard and all.

44:18.66
Andrew
By the way, that is not clear from the header in the home page.

44:18.74
Sean
The home page. Yeah, yeah, yeah, it's it's it's garbage.

44:26.06
Andrew
But I also don't know if that's like, yeah. i don't I'm not entirely convinced that that's that should be the point. But OK, continue.

44:36.13
Sean
All good. It's also not like a thing that I really care to like sell as a SaaS subscription for what it's worth.

44:41.56
Andrew
Mm-hmm.

44:43.40
Sean
it's I'm kind of trying to automate some of the work we do for clients. um Okay. A good idea.

44:50.48
Andrew
Can I skip the waitlist?

44:50.67
Sean
Yeah.

44:53.29
Sean
You can just sign it.

44:54.28
Andrew
Oh, yeah.

44:54.79
Sean
You just sign up with X. There's no wait list. There's no real wait list. There's no like...

44:59.67
Andrew
hilarious

45:00.42
Sean
um

45:01.02
Andrew
Hilarious.

45:01.44
Sean
A good...

45:02.34
Andrew
I've got an idea for you too, when you have a second.

45:03.78
Sean
Yeah, got go, go, go, go. Okay, sure. ah

45:07.52
Andrew
finish Finish your spiel. Yeah.

45:10.56
Sean
um

45:10.88
Andrew
Finish your 10-minute long elevator pitch and then

45:13.19
Sean
um right, thanks, thanks, thanks, thanks. A good idea requires a constant evolving understanding of you, what you know, and ah and and what you see in real time.

45:25.06
Sean
And preferably, if we want an AI to proactively do it, it should do a couple things. It should one, see it before you do, so you can add feeds for it to track, whether it's like newsletters or whatever or X or whatever sort of thing. um it needs to have an evolving wiki of you it needs them an evolving portrait of you and then uh it also needs some like rules about like things that like you would never want to say but the extra layer on top of it is that um what sadie now does is also uh based on like

45:56.44
Sean
Oh, and then the other thing is like it has like some like very lightweight like reinforced learning implementation which is basically every single day when it gives you three ideas you say if it's good or bad and if it's good and you write things with it and publish something in like the studio writer it knows that that's like a really good idea that I gave you versus you telling it's bad so it re-weights itself.

46:06.70
Andrew
Mm-hmm. Mm-hmm.

46:17.54
Sean
um the other last few parts of it is that one sadie also self-evolves as an agent it like um um every once in a while when it's had when it has more and more content it changes its own like internal prompts and weights um and you can like step back by yourself like if you didn't like sadie 4.7 you can go to sadie 4.6 or whatever um

46:41.68
Andrew
Mm-hmm.

46:42.85
Sean
and like lock it if if it's not evolving in the right way. And then finally, thing I just just pushed is that, you know, Sadie needs to know more about a topic that you're interested in more than you do. So what it does is that if it notices that you think about something, it'll just like do the its own research and add it to its own library every single day.

47:07.30
Sean
um And then the very last part is that like depending, like you change as a person over time um and like you have different beliefs. So it it it like knows the decay certain parts of the context um to like not have it as important, so. Anyway, that's the that's the whole thing.

47:24.54
Sean
It's just a personal idea of Bird. Oh, so Sadie.

47:27.42
Andrew
Personal Idea Bird is like a hilarious description.

47:31.08
Sean
Thanks. um oh so Oh, so Sadie. It's called Sadie because if you rearrange ideas, it's it's Sadie.

47:40.61
Andrew
that's very clever.

47:41.73
Sean
Thanks.

47:41.83
Andrew
i like that. I like that a lot.

47:42.76
Sean
Thanks. Thanks.

47:44.10
Andrew
um Okay, so my my idea for you, I just saw on Twitter Spigford talking about some system he's been playing around with with one of his personal agents where he has his personal agent just like ask him a random question every day.

48:01.29
Andrew
um And like he was like, that's been more useful at like, building up context than anything else I've done. um And so like, if there was a way for Sadie to like ping you in Slack or something and, or shoot you a WhatsApp message or whatever, and just be like, like, you know, get to know you, like, not what's your favorite color, but like, tell me something about your life.

48:15.22
Sean
Yeah.

48:26.66
Andrew
um

48:26.84
Sean
Right, right. Now that's smart.

48:28.26
Andrew
Could be interesting.

48:29.86
Sean
I will add that. I am trying to add a feature where you can text or you can text Sadie because i want it. ah The clients that we basically do this for want an experience, not like Sadie, but like with us is they can just send us a voice memo and then get the idea written back.

48:49.27
Sean
And so we do all the research and whatever for them. So that's my like ideal five-star experience. So I think it's in line with, it's good idea.

48:58.08
Andrew
Cool.

48:58.15
Sean
Yeah.

48:59.20
Andrew
My distraction blocker won't let me sign in with Twitter right now, but I'll ah i'll sign up and play around with it later.

49:03.52
Sean
Oh.

49:06.56
Sean
Cool. Let me know what you think of the onboarding. Slash, rather.

49:11.69
Andrew
Okay.

49:12.65
Sean
The onboarding kind of sucks, so let me know if you have ideas on making it better.

49:16.58
Andrew
Yeah, that's the next like big to-do on the MetaMonster. Now that like the content optimization is like mostly working, like while Jade and I tweaked the rubric, Austin is gonna start working on trying to actually create a good onboarding experience.

49:30.52
Sean
Nice.

49:30.69
Andrew
um And then it's probably integration time.

49:33.76
Sean
Nice. Nice. Nice. Good to be back on track-ish. ah

49:40.06
Andrew
Yeah, dude. um Yeah, I'll be in back in the US next week, but should be able to should still be able to record.

49:46.52
Sean
Cool. I'll see you later.

49:48.15
Andrew
All right, peace.

49:48.77
Sean
Bye. Thanks.

49:49.71
Andrew
Bye.