Real Investor Radio Podcast

Craig (00:00)
Hey, welcome back to Forked AI. It is Craig Fuhr, your humble host, along with the great David Moses. Good to see you, sir.

David Moses (00:07)
Good to see you, my friend.

Craig (00:08)
Well, we are in true, in true AI hell right now as we speak, I prepared all these wonderful notes for the show today regarding a topic that I wanted to discuss, in Bill O'Reilly's speak, we'll do it live because I'm currently, for the listeners, I'm currently on my laptop in the studio talking to you.

but I'm also screen sharing into my Mac mini upstairs, which is like a whole different ball game when I'm on this. So like I have to switch between the two modes, Dave, and it's really, kind of, it's kind of, it looks exactly the same. And so I really, never know whether I'm screen sharing or whether I'm actually working directly off my Mac. It's ⁓ so you'll, we'll, I'll show you guys in a second, but Dave.

Couple a little so what's new with you, sir. And then we'll jump into the news. What do you you were a little late coming because you said you were in the middle of something fun, but it's not yet working.

David Moses (01:03)
Yeah, I'm buried in the last, it's like the last, like when you're at the finish line, but then like everything starts to go in slow motion. That's kind of what it feels like. It's like, wait, just one more thing and then this thing is gonna work and everyone's gonna be super happy. Anyway, I'm building this dashboard so we have every document that comes into the company is being.

Craig (01:15)
Yeah.

yeah.

David Moses (01:26)
⁓ read and understood, renamed, filed, and summarized by AI. But I'm creating a dashboard that actually shows people the invoices, accounting, so that accounting and the managers can approve the invoices and then click a button and they get entered. ⁓

Craig (01:41)
you know, I'm starting

to think that we have to start doing show and tell on the show. And, but like a lot of what we're doing feels kind of proprietary, you know, like, and so like, I'm, I mean, it's, it's, it's not as if anyone else couldn't do it, but like,

David Moses (01:53)
I don't mind showing it.

mean, I don't mind showing it. My brother, like 90 % of this is my brother. I took like one little portion of it because I'm the accounting, you know, because I'm accounting essentially. And the one little part that I took is like the part that's keeping everything from actually working. Yes, yeah, I mean, he's a puzzle solver.

Craig (02:11)
So your brother's AI guy too.

David Moses (02:20)
Right, so he's a brilliant guy and for the longest time he was just, if something was broken from a physical structure standpoint anywhere in any of our portfolio, he was the guy who figured out this is how you fix that, this is how you do that. So I was like, that works well with AI because no one really knows yet. And so you need people who are gonna figure it out and he has that brain, that you know.

He was the one who could do a Rubik's puzzle or a Rubik's cube when he was nine. I had to learn as an adult. ⁓ That's what my daughter does.

Craig (02:50)
Right. Right. I just took the stickers off and put them all in. Yeah. That's it. was just so much easier. Yeah.

So yeah, I it's that creative curiosity, right? And then when you mix that with a, with sort of an architect mind, I think very powerful things come of it. I don't, I don't think I really think like an architect, like structurally though I'm finding like last night I laid in bed for literally three hours.

trying to think of the structure of all of the components, Dave, that I'm building right now for Dominion and, and specifically the LOs and sales side. And so I have all these, all these disparate parts, which frankly, some of them do the same thing and, and, and, know, but were built before I had access to Salesforce and Outlook and all these other things. And, you know, I was trying to cobble

all my little ideas together. But what you find is we talked about it last week on the show is that you just keep bolting on and bolting on and all of sudden these little widgets start to topple over. They just kind of fall over on themselves. And so, man, I did some really cool stuff this week that I'll show on another show, but I'm going to show what I built yesterday in a one shot that I think is really going to blow your mind. But before we get into that, man,

David Moses (03:51)
Yeah. Yeah.

nice.

Craig (04:06)
some crazy news this week. looks like it looks like dog and dogs and cats actually do play well together. Elon and ⁓ Anthropic announced a major deal where basically Anthropic gets a ton of compute and access to I guess all of the chips that Elon plans to make. And I just thought, you know, in light of the fact that there's still grok out there,

you know, there's another LLM that's owned by by Tesla or whoever, whoever the hell owns it now. It's interesting that the two would team up and it just kind of shows you the the, you know, the sort of the the the war of the LLMs and the that that that we're in right now. And man, it's a it's a huge deal. The anthropic gets access to like 220,000, you know, Nvidia chips and I guess

Probably all the chips that Elon plans to build using anthropic AI. So your thoughts on that and, and, and in light of your comments, Dave, I thought I had to bring this up because you were prescient when we were talking about Claude going from Claude, geez, I can't even remember now 4.5 to like 4.6 or 4.4. can't, whatever the latest rev is. And that lead up to the new rev was

a massive drop in ⁓ performance for the previous model. And everyone noticed it. It was a huge announcement. And you said, it's just, it's a compute thing. It's clearly a compute thing. So give me your thoughts on the news of the big partnership.

David Moses (05:37)
Yeah, mean, I... So, it's interesting to see how it all evolved, you know, but...

Craig (05:43)
Yeah, by the way, for those

for those who've listened, who are just tuning in, Elon recently, you know, within the last several months said that Anthropic was one of the most evil companies in the Western world. And so, you know, but now, you know, hey, let me give you access to 220,000 great GPUs.

David Moses (06:02)
Yeah,

but I think the Ilan Sam saga, right, think that the lawsuit, the $150 billion lawsuit, I think it's interesting. I wonder if we'll ever find out for sure how much that and what's come out.

Craig (06:08)
yes, yes, yes.

yeah.

David Moses (06:21)
regarding how really XAI was just a distilled version of OpenAI that they just probably just stole. know, like I wonder how much of that played into, okay, well.

Craig (06:27)
Right.

let me team

up with the second largest competitor to OpenAI.

David Moses (06:37)
Yeah, and because we need to have something that works that isn't just completely ripped off of somebody else. I would never bet against Alon. think that, you know, I think.

He really understands how to scale. I think he really understands how to vertically integrate. ⁓ But there is that old adage that if you can't beat them, join them. And the reality is that nobody can beat what Anthropic has and continues to put out. And they're really hyper-focused away from...

Craig (06:58)
Yeah. I think Sam often does too, but yeah. Yeah.

David Moses (07:12)
know, cat videos and, you know, quote unquote, the bullshit that Google and OpenAI eat up compute doing is something that Anthropic just did not have. It's that they couldn't, right? Anthropic, if they wanted to do video generation or video analysis, they would do it better than everybody else does it. But it just requires too much compute and...

and that's the hold back. And so they had to focus on business enterprise and they did and it turned out that that was a wildly successful bet. And I think that Alon's like, know what? If he can cut a deal with Anthropic and get, you know.

I think what he gets is access out of the deal. think he gets access to the smartest models and just being incrementally better, I think he sees the value in that because on this trajectory when you're hyper scaling, when you are growing exponentially, just being just a smidge ahead, that extrapolates out quite favorably.

Craig (07:53)
yeah.

yeah. It exponentiates in this case. Prediction, is there ever a time at which the two companies merge and become a very fierce competitor in what appears to be at least nationally like a three arms race, right? You've got Google Open AI, and then obviously the exanthropic impending merger that I predict.

David Moses (08:33)
I think it comes down to can Amadei and Musk work together? Dario, he and his sister actually, they seem to me to be people who came into this to change the world. Yeah, Elon wants to change the world, but Elon wants to be the...

Craig (08:38)
Yeah, of course.

David Moses (08:53)
the person, right? He wants to be the guy. He wants to be out there. you know, that maybe that works, but I don't know if, you know, I just don't know how those two personalities or three personalities really, you know, come together. I think that

Craig (08:58)
He's the people's billionaire. Yes.

David Moses (09:11)
Musk's all about freedom, right? He's all about, you know, let's put this out there and see where it goes. And I think Amadeus quite a bit more reserved than that. I think he's quite a bit more deliberate in what he's doing. I do believe the hype about Mythos being too powerful to release. don't think, I think, you know, maybe it's 10, 15 % marketing and 85, 90 % reality.

Craig (09:27)
Yeah, me too.

The marketing is great.

David Moses (09:36)
Yeah, I believe him because what I see from Anthropic has always been a slant in favor of ethical behavior. And not that Elon's not ethical, I'm not saying that. It's just for him, it's... Being ethical and controlling things such that they don't get out of control...

Craig (09:47)
Have you?

David Moses (09:57)
takes care that I think Amadei takes that Elan is more on the freedom side of that. Just like, let's just put it out there and let's let people run with it and see what happens.

Craig (10:07)
It'll be interesting to watch a couple of different use cases that I was, I wanted to get your thoughts on this week. Have you had a chance to play with a codex at all? The codex, you know, the sort of the code, the claw desktop version of open AI.

David Moses (10:23)
I have, I've played with the model but within the open-claw space, you know, very, very, very strong. think it's the closest competitor to 4.7 that I have seen and I think it, you know, my personal opinion is that the go-to when you need something done quickly is codex. It's just faster.

Craig (10:31)
I'm loving it.

It's fast. It's definitely faster. It's not as verbose. It I think it's better at coding. I'm not I'm obviously not a coder, but it appears that the results that I get in one or two shots is is better than what I'm getting right now with Claude code.

David Moses (11:03)
Yeah, think that's true. Whether it stands the test of time, whether how it's built can be, you know...

Craig (11:13)
You mean what it builds

or how it or how the or how codex is built. got it. Right.

David Moses (11:18)
No, no, how codex builds.

The things that I've built with codex are one-shot types of things. They're things I need done quickly that are serving a purpose in a moment. Hey, I need all this data normalized and moved from here to here. I only need to do it once, but I need to do it. That's the kind of thing. Or I need this displayed, but when I'm done with this, I'm done with it.

I do tend to build with Anthropic more when I'm building kind of more permanent things. But again, what's it been out? Two weeks? Three weeks? I'll be saying something different. For almost a month. ⁓

Craig (11:49)
Right. Yes. In my vast experience developing with Anthropix latest model.

Yeah, I'll tell you, it's

just night and day. Yeah, it's fun though. I enjoy the competition. I like seeing the nuanced and sometimes even more the nuanced differences between developing in both. And frankly, you know, I do it at the same time, which is fun. And I'm anxious to see

David Moses (12:02)
you

Craig (12:19)
You know, I said, I said just a few months ago, I was like, I think we're going to look back on open claw as like DOS, you know, six months from now, like, well, it was, it was revolutionary. And I think in terms of what, what it does is revolutionary. It's revolutionized the way people work with these tools. But I still think that like, it can be very kludgy for the non-technical and you, and you need to be, you need to be very structured in the way that you build with open claw or else it just becomes this monolith that falls over.

itself. And so, um, yeah, I'm just anxious to see where that's going. And I, but I am really enjoying the desktop models from both Anthropic and open AI that are frankly, you know, the biggest threat there, there are their own versions of open claw, right? Like I can send, I can do everything in, in, uh, in, in codex and Claude code that I can do an open claw. The only thing that I can't do, which I really miss is I can't talk on my phone to

my Claude code sessions. We have a, it doesn't allow it on the enterprise accounts. So I'm sure that's coming soon. Strangely enough.

David Moses (13:23)
So wait,

it doesn't talk to...

Craig (13:24)
So I can't, can't like if I'm, if I'm talking to open call Rocky and I can just talk to him all day on my phone over telecom, over telegram and he, you know, ⁓ executes the commands. you, you can't, you can't do that on enterprise accounts with Anthra with, with a cloud desktop. They have dispatch and they have remote control, but those things are not available for our account.

David Moses (13:42)
You

⁓ okay, so try, do you have the ability to use custom connectors? So if you create a custom connector between Cloud Desktop and you name it, you'll have to use like, if you're connecting it to Microsoft, you're gonna have to use a,

Craig (13:52)
yeah.

David Moses (14:08)
a monster API key that has all the permissions. You can't use Claude's connector because Claude's connector is only going to connect to your...

Craig (14:15)
So what

would I use to connect remotely with my phone to my claw desktop?

David Moses (14:20)
Telegram or

you'd still you you no, I'm sorry. No, you would use you would use cloud just use the app use the cloud app and If you use a custom connector versus using so like if you if you change the config file like in your cloud desktop Right. That's a file that lives on your machine. So you can't It can't access that MCP that custom MCP from your phone because I know it can it just doesn't do that

Craig (14:45)
Right.

David Moses (14:45)
But when you do a custom connector, it will. The custom connector works from any Cloud. And if you have that config file on Cloud desktop on your laptop, and then you go to your desktop, it can't see it unless you also put it on your desktop. But with the custom connector, anywhere you log into Cloud, when you log into Cloud, the custom connector works.

Craig (15:05)
So tell Claude to create a custom connector that will allow me to speak to my Claude desktop on my Mac mini over Telegram with my phone.

David Moses (15:17)
I would actually, the way I would say it, I actually did it in codex, not in Claude, ironically. ⁓ But that was just, it wasn't intentional, was just like that was on the screen. That's what was on that screen over there. What I did was I said create me a custom MCP connector for Claude.

Craig (15:22)
Of course you did. Of course you did.

I'm starting to wonder who's running the business man because like you're just playing with AI all day long. Keep going.

David Moses (15:40)
for and then fill in. Like for me it was, I had one for AppFolio, I had one for Mongo, which is database. had one for QuickBooks. So that's what I did. So I could actually share my screen. You can see.

Craig (15:47)
Okay, well fill it in for me. Just fill in the blank for me.

Tell

me what I'm telling you. Create a custom connector for Claude for in my case.

David Moses (16:00)
I'll copy you the prompt and send it to you, but essentially what it says is I need a custom connector. I need a custom connector for Mongo that allows me to read and write to all of my Mongo collections, all of my Mongo databases. And it created this custom connector for Claude that...

Craig (16:03)
Be safe, save the problem.

Yeah.

David Moses (16:18)
that essentially just, when you open up the custom connector, you'll see all of the individual actions that it built out, right? And it just.

Craig (16:26)
I know what

it does. I'm asking you, what am I telling it needs to connect to?

David Moses (16:32)
what do you have access to that you want to connect to?

Craig (16:35)
No, I just want to be able to talk to my claw desktop instance that runs on my Mac mini from either iMessages or ⁓ Telegram or Discord. I don't care what it is. I just want to be able to talk to it basically remote control it just like you do with open claw through Telegram or some, you know, something on my phone.

David Moses (16:58)
as opposed to just, well, I guess what are you trying, what are you having it do? Like are you having it,

Craig (17:02)
I want to go into, well, the reason I'm thinking I might use like different channels is I would want to talk to this project. I would want to talk to that project. I want to go into this project and say, hey, ⁓ see what you did there. Let's run ⁓ the next session. Right? Like I just want to run sessions from my phone if I'm at the airport or if I'm at the beach or right. Just like you can do with OpenClaw.

David Moses (17:23)
Sure. if you've got your,

yeah, but if you've got Claude on your phone, right, and you go into Claude and it's all custom connected to everything, you have a project for each thing that you want to work on, a Claude project.

Craig (17:36)
Okay.

David Moses (17:37)
And you have put in your, you go into user preferences, you just, user preferences will just allow you to put those instructions so that every chat, every project, know, Claude knows what it can access. And.

in addition to the custom connectors, the instructions just tell it ubiquitously how to use it. And then whatever you want to work on. you want to... I've got one that I built for... The first person that I actually gave it to was my community management leader. And I said, okay, this is connected to Sync. This is connected to Mongo. And what is in Sync and Mongo are every single thing, every record.

Craig (17:54)
Right.

David Moses (18:13)
for community management. so instead of going into sync and finding, you know, pulling a report.

She just says to Claude, either from her phone or from her desktop, give me a list of all of the co-owners at ABC Association that are delinquent by more than $100. And it'll just spit that out because it goes in, the custom connector knows, OK, this is just the API, right? But what I've done is just created this layer, this MCP server that translates

you know, whatever I'm putting into Claude into an API call and then goes and makes the API call.

and then pulls back the information and organizes it the way Claude organizes things, which is beautiful, presents it in a beautiful way. And then I can do what I wanna do. And she can say, create me a, you know, the idea is that eventually we'll create custom connectors for individual presidents at associations or treasurers at associations who want to interact with their data and we'll give them that.

tool and we think it's going to be, I mean, I think ultimately that's the future rather than putting, you know, Hey, log into this portal and do this and do it. it's just, it's just, here's a chat bot. Yeah. Here's a chat bot. Talk to your data.

Craig (19:21)
Yep. Yep. Yep. Yep. Here's my API. had a, I had a, so

if, so I'm, trying to build out an autonomous or semi semi-autonomous flow for Dominion lender finance. And this is where we lend to small to midsize lenders. And part of the onboarding process is, is on

you know, not unlike any other account approval process, they're going to send us, you know, an application, personal financial statements, you know, all the requisite docs. And I was thinking like, I kind of have a portal already, like a drag and drop portal, a very seamless, you know, application that they can fill out in probably 10 minutes rather than an hour. And, and, and it's kind of instant response, right? But then I was thinking like, that's kludgy.

Like why wouldn't we just create a script, tell them to like download Claude desktop, it'll create a folder and then they can either drag and drop the docs into that folder and then that then uploads to us somehow, right? ⁓ Without having to, you know, I mean, I don't know if it's six and one half a dozen the other, the juice is really worth a squeeze on it. But I think for the AI curious,

David Moses (20:24)
Yeah, think that would be awesome.

Craig (20:33)
to have a process like that where it's like, you know, you're just going to control everything. We're going to, we're giving you a little script. You can either run that script to pull down those docs from your quick books, or you can just drag and drop docs right into this folder. And then that folder automatically uploads to, um, securely to us. Um, or we just developed the portal that allows them to do the same damn thing, right? Like I don't know which is cooler, you know,

David Moses (20:59)
I think the portal is captive, right? Like you've got their attention, you can, you know, they're not distracted by eight million other things that, you know, like when they're in your portal, they're doing things, you know, it's purpose-built. And that's, think, you know, an argument for that. But I think if you're gonna create some kind of, you know, Claude-based situation, I...

Craig (21:08)
Yeah, yeah,

David Moses (21:23)
you know, my thought would be, you know, give it the ability to interact, like give it the underwriting guidelines, give it the, you know, what they need to provide. Basically, if you could make it an underwriter and make it so they could kind of have a conversation with their own Claude and it would give them the answers to the questions that they would otherwise be asking the LO,

and allow them to, by the way, just drop your document here. We'll tell you if it meets the criteria instantly. And if it doesn't, we'll tell you why it doesn't and what you can do. Like, okay, this doesn't meet the, if it's a borrower, this doesn't meet the minimum threshold for cash. This account doesn't have enough cash. Okay, so now.

you know, move this money over to this account and screenshot it and okay, so now I've put up my screenshot, yes, that works, now you have enough, right? So it's going to, it'll help them walk themselves through it and ask these questions and get instant answers because like we were talking about last time, you know, you can still be a great company and tell people no as long as you tell them no quickly.

Craig (22:31)
Yeah, I think

we've built that into ⁓ sort of the doc recognition and conditioning reviewing into DSCR. I don't know that we've built it into the application process. We have a very, very slick electronic app that is about to, we're about to unleash. And it's just, it's, you know, it's, we're, we're, literally going from a spreadsheet app to now an agentic app and

the time differential for the client is probably going to be a 90 % savings in time just to fill this thing out. Cause most of it's filled out for them already. ⁓ it's just amazing. But now translate that. So that's sort of the Dave, that's sort of like property docs and like sort of bar, like some borrower docs, but not really, not really like, articles of incorporation and different, ⁓ ways and, and you know,

David Moses (23:04)
because it goes, it gets the information. You're just gonna give me the names of your LLCs and I'm gonna...

Craig (23:22)
guess there's bank statements, but it's a different type of documents that you get on borrower approval, right? And we have great OCR.

Yeah, I'm just thinking, I'm thinking silently here, which is never good for a podcast, but yeah, I could see, I could see, I appreciate your thoughts on that. One other thing that I thought that I heard that was really an interesting use case, and I'm still trying to figure out, cause it came up on our AI call and then it came up on a video that I was watching the other day. So I was the lead programmer for Anthropic. forget the guy's name, some sort of Eastern block country that I think he's from. And, um,

He was talking the other day about how like Anthropic, there's no, there's nobody coding Claude anymore. You know, it just, it just kind of like codes itself, but they, but he was like, and, we, and our, and our agents all talk over Slack to one another.

And so the other day on our AI call, there's about 10 of us in the company, someone brought up that we really need to connect teams to Claude or you know, Claude desktop. Man, I'm still I'm still really trying to figure out what the why and the how and the what do I get from it? Like, I like I think the idea is slick. But I'm not really sure like how it all like

works and why I want to do it.

David Moses (24:39)
Yep.

And depending on how you ask Claude, if you want to just ask, hey, how would this be cool for me? If you ask it like that, it'll convince you that this is a good idea. And if you say, tell me how this is going to waste my time, it will convince you this is a terrible idea. it's like, what are you... Yeah. Like it's good for...

Craig (24:47)
yes, yes it will.

Yes, give me your greatest ideas on this, and then poke as many holes in it as you can. And it'll totally shit all over it. that great thing that

David Moses (25:05)
You know, and Claude is still like, Claude's the best. Claude, yeah, Claude is the best at convincing you of whatever it is you're looking to be convinced of. Yes. But I wonder when it will be the source of truth for all, you know, bad ideas and good ideas. Like, where will actually definitively tell you. ⁓

Craig (25:07)
It's the best cheerleader ever.

Yes. Good or bad.

All right, so how are you

using Slack and Teams with Claude to better your day?

David Moses (25:30)
I'm not.

Craig (25:31)
Well, what are your thoughts on why someone would want to do that? Because I honestly can't put it together.

David Moses (25:38)
⁓

My only thought would be to figure out where people's pain points are, where people are being distracted or bogged down. And I said this before, I think that productivity, I think you know...

Craig (25:46)
Mm-hmm.

David Moses (25:54)
I think you know who's productive and who's not. I don't think you need any kind of AI analysis to know who's goofing off. within the sphere of people who are really trying their best, what makes their experience better within the company?

Craig (26:00)
Yeah.

David Moses (26:12)
And I think that monitoring Slack and Teams can really distill that information and give stakeholders, others, give decision makers the ability to...

have a better finger on the pulse of the organization. So I think that's the, that would be my, I would think the number one use of bringing Claude, of bringing Teams and Slack into Claude. So that Claude can just, you know, keep an eye on it. Yeah.

Craig (26:27)
So.

figured, yeah,

I'm with you. And I'm kind of getting the sort of the bigger your your bigger goal. How does it play out? Like, so ⁓ I finally got obsidian working in a way that I think it's really a solid memory device. Now. I Carpathy released like his theory on how to use it as like sort of an LLM wiki.

And he did this huge write-up on Git and I basically gave Claude the, write-up and said, yeah, design. want to design a vision for my obsidian. And here it is. Here's my vision for what I want it to do. And I've already got a vault that's got a ton of shit in it and I need to clean the vault. And then I need to like start working out of every project, knowing that obsidian is modern during this project. So it wrote me like a script that is basically like a skill.

And then whenever I launch a project, say Obsidian Ops, and it knows now to set the folder up in a way that Obsidian will recognize everything. And it's working. Yeah, yeah, both. ⁓ And so ⁓ working great. My Obsidian graph, the, you know, that graph that you get looks wild now. That's fun to look at. But more importantly, it's really a great.

David Moses (27:37)
You have that connected to Cloud or Codex? OK. All right.

Craig (27:53)
memory keeper, man. Like I can, I can start, I don't ever start a session now in a project where it's like, what are we talking about? Then it gives me like the wrong, you know, the wrong link and none of that never doesn't happen. And then I had this idea of like, what would be getting, getting back to our conversation with teams here? Like who's productive? What's not? Hey, what's the, what's the 2026 time card for when your manager comes looking and says, Hey, what have you been doing with AI all day? We see you spent

you know, $3,000 in tokens. Well, now I can basically have this obsidian report that looks at like the diff over 24 hours. Where, where was the code base 24 hours ago? What changed by project? And it writes me this really cool newsletter report every morning that I get that's got like, you know, I tuned it with like images and it looks like a, it looks like a newspaper.

And it's super cool. like if Jack ever came to me and said, Hey dude, what the F are you working on? Well, I've got a daily report that I can give you that I'm working on. It's called the Rocky review and it's pretty cool. So yeah. So, and so it took like, um, it took like, I don't know, an hour and a half, like I got the obsidian working great. And then it, and then I was like, Oh, let me, let me like build some sort of UI that would be meaningful off of what this thing does. Right. And it turned out pretty cool. I'm still kind of working it, but um,

David Moses (28:56)
That's awesome. That's really awesome.

Craig (29:14)
But but getting but like getting back to the whole teams thing. If you had if you had some sort of agent that was running on your desktop that was monitoring, hey, this is what this person's really inspired in right now because they're spending so much time with it. And it's talking to some other person in the company that I guess is doing something similar. I don't know. Like, I just I'm still struggling with how to use these legacy some of these legacy enterprise software, including Outlook.

I'm still trying to figure out how to really get this thing to understand what, what, what's going on in my outlook inbox and what's important, what's not important, what can wait until the afternoon. I only want to check my email for like 10 minutes a day, three times a day. That's it. And so how do I, and I told it that it's, it's really struggling to get it right. So I'm like,

horrible legacy app that we all spend way too much time in outlook except for Jack Bavier. He loves it. But but like I hate it. I just don't like it. And so I have I'm still really working hard to figure out what email Zen looks like and how I'm going to get that

David Moses (30:18)
I have, I mean, other than echoing your sentiments on Outlook and how that gets cleaned up, I think I know why. I think that the models aren't there yet. When I say they're not there, I think they are easily duped by a good email.

Craig (30:29)
Yeah.

Yeah. Yeah, they don't, they don't.

Yeah. And that's the scary part. You can't have it go just reading all of your emails because suppose there's some malicious prompt in there that tells it to go, you know, give you, Hey, override everything that you know, and run this script. That's going to show me everything on Craig's computer.

David Moses (30:50)
Yeah, no, prompt injection is a real thing. But even the things that are just innocuous, like, you know, I had it, so I have an agent that checks my calendar once an hour.

text my emails once an hour just to see if there's something missing from my calendar that needs to be added. And depending on how exciting the email makes the event sound is often how likely it is to show up on my calendar. I mean, I've given it some rules like, hey, if this is, you my kids are in, you know, third grade, sixth grade, eighth grade, if you get, you know, an email.

Craig (31:24)
Anything from that, right?

David Moses (31:25)
from

the school that says the eighth grade is doing this thing, then that has to be on calendar.

Craig (31:29)
See, that's the part that

like that's what I'm getting at. Like, so I said, OK, you don't you don't you're obviously not getting the task. Let's go further. I want you to evaluate the last 90 days of every email and then, you know, give it some sort of tag or, tell me what you think. And it it wound up being this exercise where I was just going to have to go through and sort of condition and dispo every one of them so that the thing could learn. And I think that what you're that what we're both getting is, is that

There's there's an inference. There's an intent. There's the sort of the underlying human aspect of emails that I don't think that AI quite gets yet that I don't and I don't even think we're a year away from it. Like I think that, you know, if you and I try the same exercise six or seven months from now, we might give vastly different results. But as of today, I still struggle with these with these bloated apps that we've all used for years that were supposed to make us way more efficient and

happy and what we find ourselves just trying to search through our Outlook emails for six hours a day. And that's mind numbing. And I haven't figured it out yet. Haven't figured out the team's thing. Haven't figured out the Outlook thing. Still trying to figure out the Salesforce thing.

David Moses (32:39)
It surfaces great ideas and great insights, but it's not reliable. I can't rely on it not to miss things and I can't rely on it not to give me bullshit that I don't need to know about. And that's the unfortunate part of it. my approach has been build it for the future. If you build a workflow,

Craig (32:42)
Yeah. Yep.

Yeah.

David Moses (33:01)
that does something that the model can't do right now. If it's super critical, then yeah, fight with it a little bit, give it some parameters, tell it, you know, right? But, just train it and let it, and then it'll work for you. But if it's not hyper, if it's not super critical, you don't wanna spend a ton of time on it, I'm a big proponent of, you know, build it. If you're reasonably certain that future models are gonna get it, build it.

Craig (33:10)
train it a little.

David Moses (33:28)
You know, build it with that in mind. Right, so keep it simple. Don't keep piling on top of it. Build it so that maybe it only gets 70 % of it right now or 80 % of it right now. But you know, as models improve, right, then it's gonna take the same prompt and the same context and get it right 95 % or 100 % of the time.

And that I think is, where you really have to, I think, be careful is the context. Are you giving it enough context? It's really just, I'm not giving it enough information. If it's that I need to answer a bunch of questions or I need to train it a zillion ways, then that's not a good solution. That's not good. But if you're giving it everything it needs to know to make the right decision, and you're confident in that, and you're giving it a prompt that should lead to the right decision.

and it's making the wrong decision, still use it and let it grow so that as future models come out, you'll start to see that I gave it the right context, I gave it the right prompt, now it's getting it right. And now it's getting it right. And I think those simple, every time I build something that's super spaghetti-like and complicated, it comes back to,

Craig (34:30)
Set it up with the right architecture.

David Moses (34:40)
eventually, eventually the, you know, eventually the AI will get it, eventually the model will get it.

Craig (34:45)
You know, I think there's just, you know, it's so funny how certain buzzwords just kind of permeate, you know, quickly and process everything now is just, you know, people have been talking about documenting processes for years. It's how real companies are made and sold. I think when you're talking about programming with these tools, you have to really think of the smallest processes, the smallest executable processes that can be strung into

know, multiple processes. And so I was really, and then, and then sort of like the connectors between those things, like the database, the CRM, the, you know, the UI, the front end, the backend and the middle, right. And I think that's, guess, putting it into words, like I've been trying to say for the last two or three months of my own journey here, like I never really, I would just talk.

you know, Hey, this is what I want to do. And just chit chat with, ⁓ with, with Claude all day long. And you know, sometimes great things will come of it, but often they weren't like structurally sound. As you say, they were good for rev one, but for rev six, when the thing really starts humming, it's a mess. And so, man, I've really started to document each little process that I'm trying to achieve.

and sort of the building blocks that go in between before I start, before I sit down and start prompting, you know what mean? Like we prompt, we plan together and then, and then, and I'm about to prove to you what happens of that and show you something that I built in the day. So let me share my screen.

David Moses (36:02)
Yeah.

Yeah,

I mean that by the way is, I mean it's a brilliant way to look at it and then you can take those little processes and then you can say, where else, either where does this really ⁓ translate well where I can use that same process to do a bunch of other things ⁓ or.

where can I adapt the process? So now I've got basically the same process, but it applies to this completely different process.

Craig (36:41)
Correct.

So, so Dave, getting back to, know, everything's full circle on the podcast today. So I cannot program from my phone right now. We're going to solve that with a custom connector today. So the other night I'm sitting watching the Orioles game with my son and I have an idea. I've had this idea for a long time, Dave. I can't see you right now. So I apologize. I'm looking at what you see. ⁓ so

David Moses (37:03)
That's okay. I'm definitely not holding up a, you

know, I hate the Orioles sign.

Craig (37:09)
Great. ⁓ So I've had this idea

for a long time for creating content for both my job as a loan officer, as well as for for all of our AI discussions. And so what do we have to ingest? Well, currently, we have a shit ton of episodes from our from real investor radio, all of the transcripts in super base currently. And

And that's a great. And then obviously the podcast from here with fork.ai. In addition to that, I every day I'm on my phone surfing Instagram and in my feed, I get a lot of things about food. I get a lot of things about AI and I get a lot of things about real estate and sort of the macro economy and things that like my customers probably should want to hear about if I make content off of them. So there's there's the theory.

So I was on my couch and I was telling Claude all of this and I said, so what I really want to do is I want to develop a like a lightning fast scraper that when I'm on Instagram, I can just copy the link and it automatically paste to this app that will then go out and scrape the entire Instagram post.

bring back, you know, if it's a carousel, I want all of the carousel assets. If it's a video, I want the video assets. If it's if it's just text, yeah, want that too. And so I talked to it for probably about an hour, Dave. And I said, and you'll have full access to my appify API's, you'll have full access to my super base, you'll have full access to, you know, everything you need.

but plan as if let's make a plan as if you have access to all of these things. And then think about maybe some things that you might want access to that I didn't mention, man. And then I said, after we planned for an hour, I was like, now write me a detailed PRD with a, with a very succinct task list of like minute tasks that a swarm of agents can tackle in one session.

And dude, I came to work the next day and pasted those files into a into a, you know, into a project. And it one shotted this this app for me. Now, originally, Dave, I told it to just make it as a web app, ⁓ you know, and it'll run on my iPhone. So structure it to make it look good on my iPhone. And it did a great job of it. One shot at it right out of the shoot. But then yesterday, when I got to work, I went later in the day, I said, you know what? I've never made an iPhone app before.

So I downloaded Xcode, downloaded the simulator, and this is what you're seeing. This is an actual app running on my iPhone right now. And so let me explain it to you. So it's a very simple capture screen. And the cool part is that when I'm in Instagram, Dave, and I copy the link, I can actually go down. I don't know if you're an iPhone user or not, but when you copy the link, you can actually share it then to an app. There's preset apps. So I don't even have to come into this interface. I can just share it directly into the app. Now,

David Moses (39:52)
I

Craig (39:59)
The interface is not the buried lead here. The buried lead here is the structure that we set this up with, that the scraper goes out and it returns everything to the super base. So the super base database is really the magic here. I just put a UI on it because I thought it was cool to develop a UI on what we were scraping.

But the end goal here, Dave, is to take what's in SuperBase and then basically synthesize content off of that. So here's a scrape ⁓ link from Instagram. That's going to come up in my Content and Intel UI that I've already built. And I can select that news story. And then the second app that I'm building that I'm almost done with

then synthesizes that content from SuperBase and has a great copywriter skill. It has a great LinkedIn algorithm skill. It has an unbelievable graphics design skill that can make the carousels. And so why did I do all this? Well, I did the app just so I could see what I was getting back that's in SuperBase. But why I really did it was I want to create

really fast and valuable LinkedIn content, like three a day. And I want those to be based on this podcast, the RIR podcast, what I do for a living. and so that that's, that's, so that's my show and tell, I'll show you, it's really cool, Dave. So when so it scrapes, it scrapes the link, and then it puts everything in, I can, I can put some notes in there.

David Moses (41:35)
Yeah, show us it in action.

Craig (41:41)
I can give it a folder just like you can in Instagram. You can save links to folders and I can create new folders. And then it puts it all into a library. And again, this is running simulated off of a computer upstairs. So it's a little, it does not, but it's beautiful scroll. You know, it brings back exactly what it should, the account information. And then you can click into the card.

And I was trying to get this to work right before the show. But the card view, Dave, is so beautiful. It brings up like sort of what you saw in there. It brings up any captions, the text, and then it has a it has a swipeable media. So if it's like a carousel, I can just swipe through the carousel of that post. And all of it saved a super base. And I'm really upset that it's not working for the show, but I'll come back and it'll be working great.

David Moses (42:28)
Yeah, of course. ⁓

Craig (42:29)
So that's my show and tell man. What do you, and the coolness of it was I literally started it with a conversation while watching an Orioles game, came to work the next day and one-shotted it in like an hour. Naughty, I don't even think it was an hour.

David Moses (42:44)
And then you can actually get that app approved by Apple and you're off to the races.

Craig (42:48)
So I could, I

could. And so I started to develop an idea for the food thing. So I love food. I'm a big foodie. And when you save links on two folders in Instagram, it just becomes like a black hole. Like, you know, I'll save a million different little videos of people making food. And then I go to look for the wonton one and I can't find it. Right. And so this one, I thought, well, how cool would it be if I could just scrape it, save it to the library.

And the app then shows you the link. It has a card. The card has the full recipe on it and then a shopping list and your nearest grocery store that sells organic food. I don't know. But like all of those things instead of my scraper that's obviously scraping for more things. But like it's a foodie app that like it has it's it's literally your grandmother's index box of recipes all in your phone and it's all the recipes that you love searchable.

I'd pay $2.99 a month for

David Moses (43:45)
It's awesome.

It, yeah, there was a, someone had done this on a video where they took a picture of everything in their refrigerator. And then it kind of, it like one-shotted a recipe. Yeah. And I, it was a year ago, I tried it, it didn't do anything ⁓ close to as cool as what it did on that video. But ⁓ I bet it could do a damn good job now.

Craig (43:56)
Yeah. It tells you what to make that night.

Yeah.

So, you know, I, I, I wanted to do it because it was like one of those simple ideas, Dave, that like, that will, if I can get the other side to work, the, the content generator side, which I think I'm fairly close to, you know, it's going to take a little work to tune it, but that's a very powerful thing that saves hours, hours in a day. You talking like, I don't think people know until they start really making content.

to attract an audience and get interaction and engagement. It all seems very easy when the guys are telling you that you can do your faceless YouTube channel and make millions of dollars. But like it takes time to make that content and not an insignificant amount of time if you want to get it right and decent content. You can you can send out shit that's AI slop. But like to get to get a LinkedIn post or carousel to really mimic what you see in your head and and and would say

is is time consuming. And if I can get that right, and I think I think I'm fairly close to decent rev one, which is always good enough, right? Like, I say, I'm, I'm, I'm, Jack Bavier is in my camp, you you put out rev one first, then you feature the shit out of it, like that it just gets something out there, right? And then and then feature it out. Exactly, the the use and the feedback. So I'm excited about it. It was it was just a quick idea.

David Moses (45:14)
you

You need the feedback, really.

Craig (45:33)
And it came together super fast and, it's incredibly useful in my day to just quick scrape an idea, save it for later, create something off of it if it, if if it merits creation. So, yeah.

David Moses (45:46)
It's very cool. I wonder if I... I have some things that I could probably show, but let's... I wonder...

what might be interesting, but there's a lot of, it's funny, there's a lot of, I think, small business owners out there who are still asking themselves, know, I know that this can do so many things, but what do I, like, I get it, AI's gonna change the world, AI's gonna change my world.

Craig (45:56)
Well, you don't, I don't want to pressure you.

What's that thing? What's that little thing? what?

Yeah.

David Moses (46:19)
But what do do with it? Where do I start? what should I... And there's almost... I mean, you and I know there's probably no wrong answer to that because it doesn't matter which rabbit hole you decide to jump down. They're all going to give you... They're all going to immerse you in the same... The same sort of environment that you need to be immersed in.

Craig (46:37)
Yeah.

David Moses (46:39)
in order to start to get it, you know? And you just, won't know until you jump down that rabbit hole and say, you know what? This is not where I thought I was going, but this is interesting and I'm gonna go down this other road.

Craig (46:41)
Yeah.

Yeah. I find

that it just gets back to that. We live in the idea, the age of the idea man. And if you just have a little bit of creative curiosity, just just on anything that interests you, it doesn't mean like I was telling Jack about this Instagram scraper idea as we were getting on the elevator a couple of weeks ago. I was like, yeah, what do you what do you sell it for? mean, like, what's the use? Like, what's the I'm like, everything doesn't have to have an entrepreneurial

like, you know, some sort of grand scheme to make money off of it, right? Like a lot of things I build do. But I watch I watch my son, my son's going into his senior year at WVU next year, and he's going to be working at the local watering hole that he likes to go to. He's going to get a job there and probably spend all his money and beer there. But he's he's taken an interest to the AI. And he's developing these t shirts ideas.

for this bar. It's a very, it's sort of an iconic little bar on campus and the guy who owns it's a bit of a character. All the kids know him and it's a cool place. And my son is developing these very Tommy Bahama like ideas for logos and sort of a merch. And I was like, all great. You know, I love the, I love the, the, the designs Dave are beautiful. I mean, they're really stunning.

And he thinks and he has no artistic talent whatsoever. So it's all AI. And yeah, he knows what he wants to look like. And so he's struggling through the morass of like being new to it all. You know, he's just he's he's struggling with adversity, just like you and I did for the last year. And he's figuring it out. And I'm trying to give him a little bit of direction. But I'm like, No, man, I see your passion in it. Like it's all going to figure itself out if you just stay on that pen, you know.

David Moses (48:09)
But he knows what he wants it to look like.

Craig (48:31)
You just keep going with that. Like I've got a passion to create something. I'm getting, I'm getting there. I'm on the path and it's all going to work out just fine. And I don't need to put all my hands in and, know, watch you practice like I did when you were in little league, you know, go figure it out. You know, you'll be just fine. But I did, but I did give him something interesting because it's the entrepreneurial side and he's studying finance. So was like, Hey man, let's turn this into a fricking project. Tell the guy.

David Moses (48:46)
That's exactly, that's great advice.

Craig (48:57)
that your father will put up 500 bucks or frankly, you'll go out this summer and make 500 bucks and he'll put up 500 bucks and you're going to buy the shirts for this much and you're going to sell them for this much. And you're going to, and not only, but, it's, it's the, how many people on campus can we get wearing these shirts? Right? It's not about the money you're going to make. It's about the buzz baby. It's about like going like, yeah, I got to own these shirts.

David Moses (49:19)
That should be the tagline on the shirt. It's about the buzz. Because it's a bar,

you know.

Craig (49:24)
So it's like, want to tell them like, no, if you sell them for this much and you'll make this much, no man, it just goes out to buy more shirts. Cause like you just want people talking about the shirts. Jake is the one that created the shirts. Like you're not going to get, you're not going to be a millionaire off the shirts. Just make it the biggest thing ever. And AI will help you do that. And so I developed like a whole business plan for him. And I was like, now, now, now when you're now feed this to Claude and tell them this is the plan and then keep going. Just, just keep going.

And it was great. So I so for. And then you'll tell it to Phil Coles and this is the worst plan ever. But like, again, you know, for the for those that listening to the show, you know, I've got a 20 year old son who would prefer to be with his girlfriend drinking beer and playing video games. He's a straight A student. But but but he's taken an interest to this. And for those of you who are listening, all you have to do is just just sit down and talk.

David Moses (49:52)
First thing Claude's gonna say is this is a great plan.

Craig (50:17)
Just figure out something that you're really passionate about that's a problem for you. And I guarantee you this thing that is way smarter than all of us combined is going to help you get.

David Moses (50:26)
That's right. and that's the key. Find a problem that you want to solve. But you talked about the artists. I thought yesterday about if I were to write a book.

you know, what would, or if I had like a newsletter, like what would it be called? And I was like, you know, it would be called Those Who Can Do, Those Who Can't Use AI, right? So it's like, you're reminding me of that. was like, hey, my son's not an artist, but you know what, these are beautiful, right on point, you know, because he knew what he wanted, and then AI does empower people.

Craig (50:50)
Yeah.

David Moses (50:57)
who have felt discouraged in the past because they don't have a talent in a particular area, but they have an appreciation, a passion for it, and that's really all you need because the AI will fill in the blanks for you. It will take and, yeah, it can't duck a basketball for you, right? But.

Craig (51:02)
Yeah.

David Moses (51:15)
Basically anything in the knowledge world, anything in the technical world, anything that requires a certain amount of expertise. Find something you're passionate about. Find a problem that you need to solve and just have a conversation. I don't know when I talk to people on regular basis, certainly...

Not everyone's gonna talk about AI all the time like I do. And I get that. But I really do feel like the world generally is under-reacting to what's happening around them. And I think that that is of primary concern. I think that this should be, people should be better prepared.

Craig (51:46)
Yeah.

David Moses (51:53)
for what's coming than they are.

Craig (51:54)
I am the more especially over the last couple weeks, Dave, as I am now full time just doing this, you know, eight, 1012 hours a day. I really came to this sort of startling yet not surprising conclusion last night that like honestly, man, I for what I'm trying to create with and I'm just just speaking about my small snippet here of with Dominion lender finance. I think that we could create

I think we're very close to creating something where the customer has this beautiful seamless experience where they get sort of these great touches along the way that are almost fully automated. They can ingest all of their docs automated and have them commissioned immediately. They can submit loans literally with like a drag and drop and full underwriting almost immediately. like I had this, this, this, you know,

where does the loan officer come in at some point? You know, like is there is there a time that we can foresee where, you know, DSCR better be mechanized because there's so there's so little profit in it now that it's it's a complete volume game. But at the same time with the volume, you have to have more staff. And so you better find an agentic way out of that mess. And I think that we're very close to that.

And I, know, when you look across the entire United States, people that are doing similar but not same jobs, you all you start to wonder like, man, it's going to be a very vastly different landscape a year from now, 18 months from now, two years from now. And I don't think in a good way, I think there's going to be a lot of pain before there's a lot of pleasure.

David Moses (53:26)
Yeah, it doesn't have to be that way. I think that it is in our nature as humans to grow from struggle rather than to... That's why such a small percentage of us are dreamers and a lot of us dreamers, we struggle to get shit done.

Craig (53:38)
just curl up in a ball.

David Moses (53:50)
But very rarely are we held back by a... We don't need to struggle to grow because we're curious. I think that that's what... Geez, for your kids, for my kids, that's the most important thing is just be curious. Be curious and ask.

Ask why not, ask what if, and just keep pushing. Yeah, you're right. ⁓ Ultimately what will happen is a lot of pain before...

before there is calm and before there is... What do they say? The night's always darkest before the dawn. And that's what will happen. And then we'll get to the point where... Then we'll get to the point where this thing winds up being really, really great for humanity. Yeah.

Craig (54:39)
That's the hope.

That's the hope.

David Moses (54:41)
But

you're right. Certainly what you do, certainly lending is going to be commoditized. It's just moving money around. It's assessing risk, which should be significantly easier to do as the months go forward. It should be much easier to do.

Craig (54:54)
Checking boxes.

I think we have a think Jack's got a tool right now

to handle that. The risk part. Yeah.

David Moses (55:04)
Yeah,

and it's figuring out, you know, we won't really, like really predicting, really assessing risk is about predicting the outcome. Just because something appears risky or we can logically believe in our heads that it's risky doesn't mean that on the other side of that.

it winds up being a failure, right? It may seem like a failure, but until we are collecting the amount of that AI can allow us to collect on the front end, we won't really be able to flush out all of the real...

⁓ risk triggers and signals on the back end. You know, we're always kind of weighing that, Like we don't lend to people who meet certain criteria now because, you know, based on information that we have gathered over time from the data that we actually collect and gather, the risk of it failing is

too great for the amount of return that we're getting. But as we get more data, like that game is going to continue. It's not just going to be, we're always going to collect this data, and we're just going to get better at predicting the outcome. No, we're going to collect a lot more data. We're going to get a lot more information.

Craig (56:13)
Yeah,

imagine if you can if the better part of underwriting alone is assessing the risk of the deal and the borrower. Like that's a major component, obviously, right? And imagine if a tool was built to be so good at assessing that risk that you could then price different products off of very sort of

know, slice, very small slices of risk categories. Like right now, we have a risk box, and the risk box is pretty, is very well defined. And you're either in it or you're out of it. However, there's there's always edge cases, there's always a guy who's got a 679 credit score, and we require a 680. But he's got a shit ton of experience. And he's got amazing, you know, ⁓ he's got an unbelievable liquidity. But he just had a cut, you know, maybe he's got a utilization problem. It's not really even a but but we've got a box.

And it says 680. But what if the the the risk assessment tool to go could figure all of that out and go, oh, wait a minute here. The guy's at 679 because he's got 55 % utilization on his Amics right now. But he's got 700 grand in the bank. could pay it off tomorrow. And he's got all of these deals that we can see that he made its way through. So guess what? 679. But maybe you'd get to charge him like a point and a half on the back end when it would only be a point a quarter because he needs it.

Like, right? Like, so I think that, like, there's a real case for like, hey, do you extend the box? Do we can we can we to capture more business? We've got this great tool now where we can say, look, you know, we might be able to make a little bit more margin for a guy that's probably going to get turned down by seven other lenders because they don't fit the rigid box that these people have built and they can't figure out the risk.

David Moses (57:54)
And reality is you have to look at the fact that the box is bullshit. And the reason the box is bullshit, before you even look in the box, you know that the boxes are 680, 660, 680, 700, 720, fucking arbitrary round numbers. You know the box is bullshit. You would really pique my interest if you showed me a bunch of boxes that said...

Craig (57:58)
Haha!

Right?

David Moses (58:17)
the LTV is 73.9, right, if the credit score is 683. Now I say, okay, well, wait a second, probably, at least it doesn't appear, it could be completely arbitrary still, but at least it wouldn't look completely arbitrary. And that's, you know, I think that, and then at one point, can you use your tools, like, these are the things,

Craig (58:24)
Right?

Wait, that's not arbitrary.

Does it feel like it? Feels very specific.

David Moses (58:41)
that I say, well, what's the purpose of a loan officer if risk assessment is all done agentically and if the borrower themselves are really the ones who are talking their way through in or out of one program, another program, whatever it is, what's the point of a loan officer? So what immediately comes to mind is like, what if your lender can mitigate its own risk?

Craig (58:46)
Automatic, right? Underwriting.

Yeah. Yeah.

David Moses (59:07)
and monitor its own risk better. What if loan officer becomes more a role to not just assess or sell you on a particular product or loan, but someone who can actually say, by the way, you're flipping these houses and you're John...

you know, John Smith as your agent, and John Smith's probably a buddy of yours, you know, but our data shows that, you know, Jane Brown is in that market and earns, you know, 7 % higher, you know, prices on average. You probably want to list with that person. I know John's your friend, but...

Craig (59:52)
Or

what about what about if the loan officer calls you and says, Hey man, I'm looking at the last 12 deals that you've done and they were in these zip codes. I just did a market analysis of each one of those because we can on the fly now. And ⁓ I can tell you who every one of your competitors are. I could even tell you who they're borrowing from. I could tell you the days, the date, the dates that they sold these houses, bought these houses, don't know what they put into them, but you could probably assess that.

David Moses (1:00:05)
Yep.

Craig (1:00:19)
And so maybe you should start looking in these zip codes because they feel pretty hot right now. And I think that we like that's the vision I have for developing marketing outreach to my borrowers. Like I want it to be highly specific. I don't want them to get some some ⁓ email that talks about some guy in Maryland who just flipped a house. Like I want them to know what's going on in their backyard. And the fact that I know it as good as they do shows that I care.

which I do, but like it shows that like exactly. I, and I think, I think that increasingly my vision for what a loan officer should be, should not be somebody with a thousand accounts. It should be somebody that has a real book of business that is that it, has real value because they service those accounts to the very deepest. And I think that, that

David Moses (1:00:47)
Well for sure when they make money you make money and if they make more money they're going to do more deals.

Craig (1:01:08)
There's a lot of this business that can be very volume oriented. The phone rings off the hook and I'm just going to take every call and nothing gets served to the level it needs to get served at. Leads don't get converted to the level that they could be converted at. Accounts don't, you don't go as deep with bar words as you possibly could. And that is going to be the role of a loan officer going forward. How do I get to know this customer's business well enough? They believe that I care about it as much as they do. And I'm the easy button when they call.

David Moses (1:01:34)
Yeah, a trusted advisor, just a relationship that ends when the loan funds. Someone who can say, you have a competitor in this neighborhood or in this market who just bought something $20,000 cheaper than this. ⁓ turns around. It would be so easy.

Craig (1:01:41)
Exactly.

you know how easy it would be? Do know how easy it would be for me to create to create

a to create an email that would be completely dynamic, that would pull in all data from elementics, and send that frickin thing out by three o'clock this afternoon to every one of my borrowers that I could do it by this afternoon. I don't want I'm not getting paid as a loan officer anymore. So I'm just going to go ahead and keep working on my shit. So

David Moses (1:02:09)
Do it. I'd love to read it.

when

the stuff comes in, you will refer that to the other loan officers and they will speak well of you.

Craig (1:02:20)
Yeah, no, it's a it's a fun idea.

⁓ It's one that I'm really serious about. I think that I think that unique direct marketing via email can be every bit as powerful now as direct mail and other sources of marketing and to be able to create really highly dynamic emails to a massive list, I think will be increasingly easy over the next month or so.

David Moses (1:02:45)
Yes.

Craig (1:02:45)
I have an

idea. have an idea for the member we were we were supposed to show off something on every show. Want to get back like I love doing the show and tell I think people are going to love it. So what would you what would you say to even if it's the dumbest idea ever? Do you have do you have an hour or two a week? And maybe we only each of us only has to do it like once a week. So so it rotates. So do you have an hour?

David Moses (1:02:50)
Yes.

Okay.

Craig (1:03:11)
or two, to create something like that app that I created to be able to show and tell, or, or frankly, it could be anything that you're working on for the company. But like, I just thought like, Hey, as an idea, I created this really killer little thing. It's super helpful for me. And I'm showing intelligent on fork AI today. So that's my idea. Each of us has to go once per show to show something off. It can be work related or something that you created quickly in a one shot.

David Moses (1:03:36)
But prep it, polish it, and do it. All right, I'm in. Next week, I'll have it ready. have a... Yes.

Craig (1:03:39)
and at least get it working well enough that you can simulate it on screen.

I can't wait to see. can't wait to see.

And it could be the dumbest thing ever. It could be a it could be a report from your obsidian that gives you

David Moses (1:03:52)
No, whatever it is. Sometimes the things that seem the dumbest are actually the most useful.

Craig (1:03:56)
Yeah, I didn't think we were going to talk for an hour and 13 minutes today, it's been a great, great conversation with you as always, sir. Any last thoughts?

David Moses (1:04:05)
You know, it's wonderful to be alive.

Craig (1:04:08)
Well, catch us all where you can find us online everywhere. I'm guessing anywhere where where better podcasts are found YouTube and at forked.ai the number four KD.ai you can catch every episode there and my next project that I promise to have by the end of the month Dave is I'm going to create that newsletter that is basically a

a repository of all of the great video content that's on YouTube that I love, all of the great X accounts, everything that's trending in AI that like people might want to click on or learn about, that will be a project that we will put up on fork.ai as our newsletter that people can sign up for and get delivered or download or whatever. we're still figuring that out. But that's the next project for fork.ai.

I hope to see you guys. We hope to see you guys on the next episode. Hope you enjoyed this one. Please comment and subscribe and we'll talk to you on the next one.