Subscribe
Share
Share
Embed
RiskCast documents what happens when you make AI agents structurally indispensable — not optional. Host Stefan Friend is building OpenClaw, a self-hosted multi-agent system that runs his businesses, and sharing the real experience: the good, the bad, the ugly.
Each episode explores the practical reality of living with AI infrastructure through conversations with builders doing the same.
Well, Matt, thank you again for coming to sit down and chat a little bit more about
your experience with Open Claw and AI and AI agents.
This risk-cast AI podcast was born directly out of the Open Claw community night that
we both participated in, where we shared our initial experiences now a month ago, so another
month of the experiment and testing with Open Claw and probably others, and it just
seemed like there was such great energy, such interesting use cases.
I've learned so much from you, the other people that were presenting, and several people
in the room that were also building things.
It seemed like it'd be fun to have a forum to continue the conversation.
I appreciate it.
Thanks for having me.
It was great.
Yeah.
I think your experiment, to me, really stood out because of your background, and you called
it out the moment you walked up that I thought was great, that one liner that immediately
everybody was like, wait, accountants.
Yeah.
So the question I asked was, how many accountants are in the room?
Two hands went up, mine and one other person in the audience, that's right, and everyone
else was looking around, like, wait a minute.
This is not the way I was expecting this thing to start off, but yeah.
So yeah, I'm an accountant by trade for 30 years now.
Now I'm also a technology forward accountant, so I'm always looking at ways to automate
streamlined processes, think through how we can do things better, and one of the jokes
I used to say all the time, this is using Microsoft products in my accounting world,
is that I always said, it's amazing that we can fly to the moon, but we can't get Microsoft
Word and Microsoft Excel to talk to each other.
So I've always kind of been leaning in that direction, and then Open Claw for Sure was
something that when I heard about it, it gained quick traction with me, and I jumped
in really fast, which is not like me, usually I'm kind of waiting on the sidelines waiting
to see how things kind of, kind of, how the dust settles, but in this case I went straight
for it and went full blast, and here I am.
I love it.
I think it's really interesting, your use case, and that you did have that approach,
and for me with most technologies, I've been that headstrong, I have an adrenaline junkie,
I tend to act first, think second in lots of ways, ADHD might help a little, and with
my Open Claw I think I was in many ways more hesitant than I think some of the people I've
talked to, I had to find use cases, but it also was very limited in what I wanted to
give it access to, and I actually think you had a nice balance of both, and you're clearly
had use cases where you're extracting the value it seemed like.
Yeah, the way I described it in a realized upfront and real fast is that I saw that the
second there was, if there was ever a security vulnerability or violation that happened, I
knew the whole thing would not be fun any longer.
So I said the only way of keeping this fun, keeping it informative, and to make sure it's
not a point where I unplug the device, and that's the end of it, is that I said I need
to make sure I was secure first, and even the presentation that I did on that night,
the Innovation Council, was about security.
So maybe that's the side of me that came out first, but yeah, because you're right,
it was actually Igor who you had on this podcast, who was the first person who talked
to me about it.
So probably the day, so I have just a recurring call with Igor every month, and we just talk
about anything, just talk about business, the world, those types of things I've connected
with him, clients, mine, just to make introductions and just talk about technology and just stuff.
And it was right after I spoke to him two or three days after he got his first open
claw instance, and he showed me a Mac mini and I was like, well, what is that?
I know it's a Mac mini, but what are you talking about?
He's like, it's open claw.
And I was like, oh, I just heard about that the other day, but I don't know really what
that is.
And he's the one who actually opened up Pandora's box for me, explained it to me.
And I was like, oh, wow, that sounds exciting, but that sounds like the type of technology
that let me let the dust settle on.
It was a few days later that I was at a Seed the South event, networking event, and I saw
another friend of mine who was a big proponent and early adopter of open claw and he said,
you must do this now.
And I was like, wow, okay, this sounds impactful.
He's like, what it's been able to do for me for SEO on his website, things like that was
incredible.
He said, you need to do it.
So that weekend I grabbed an old Mac mini I had, I wiped it clean.
I figured I'd have to do that because I cannot, not my thing, figured I had to wipe it clean.
I tried to unloading everything and apparently the device was too old that it couldn't even
accept it.
So well, all right, so failed attempt.
But then I use the next week to, to really learn it.
So then I just started listening to YouTube videos.
So I'd be at the gym just listening to YouTube videos and I probably took in 10, 12 hours
worth of content, how to set it up, what to do, because this I knew was way out of my
wheelhouse in terms of what I normally do, especially as it relates to going into terminal.
I don't know what terminal.
Oh, okay, got it.
And figure out the prompts and how to then get it up and running and going.
And then once I did, I just didn't stop.
Yeah.
So you were working in terminal with quad code or just no copying and pasting instructions
that people said, this is what you need to copy and paste.
And then, and then at some point, the actual back and forth with the open claw system.
Yes.
To have it help building itself.
Yes, absolutely.
Yeah.
So open claw knows more about itself than I know about open claw.
And so that's what's teaching me right now is open claw probably a fair statement for
any of the AI tools it feels like.
Yeah.
That's fair.
But very cool.
Yeah.
That's what it's been my takeaway when you shared the community night.
And I thought that again stood out as impressive to me that I hadn't even considered actually
having the open claw system build itself.
Yeah.
I was in terminal using my cloud subscription to have it help me design my open claw system.
And then, then would compare and contrast occasionally also pushing out and looking at
codex.
So get chat.
He is like the third perspective in refining things.
So I, I'm guessing also probably spent a lot of extra cycles just in all of that back
and forth.
But that's, you know, what you're, it sounds like what you're creating and I think that's
the right idea is you're creating a checks, a set of checks and balances as to how far
you want it to go.
So it doesn't go outside the bounds of, of what it could do.
I think so.
Yeah.
Yeah.
And fit the structure that I even tried to create with my open claw system from the get
go.
And we've continued to hear the themes every conversation I've had and I think in market
around security, which is great.
And I think because I was just still apprehensive of giving access why I wanted the main agent
to really be kind of like orchestrator.
Or if I ask you directly for something you can execute, but for monitoring of the emails
or dedicated platform channels, I wanted to dedicate Asia that that's all it does.
So you're braver than I am.
So that's, that's the next thing that is braver.
Yeah.
Oh, I don't have, it has no access to any of my emails.
Okay.
It has its own email account.
So, and it's all, I've only instructed it to only read three email addresses, mine
and my wife's.
That's it.
So I, I was told to ignore every other and it's, I'm, what my, the concern I have is,
and I, I'm going to script the term injection, what's the prompt injection, prompt injection.
I am so scared of that, that I'm like, I want to really kind of constrain it, make sure
that what it's getting and processing is, is very thought out and then kind of let,
you know, kind of move the boundaries out.
Yeah.
And I still haven't moved it out that far.
I am, I am sure that somebody that's a lot smarter than me could figure out or already
has the bad actors, the prompt injection that would break my system.
I'm sure that it is possible, but that is at least one of the security hardening approaches
that I thought made sense of if monitoring was the only one watching and, and then Alfred's
reviewing anything that gets surfaced from monitoring, then I've at least got us a layer
of separation there already.
And then the open call system, as well as some intentional updates that I put in trying
to build the security features in would help.
That was the idea.
Have you tested it?
I mean, it's the, the monitoring of, of email channels is probably one of the things that
it's done best from day one.
It hasn't broken unless I've broken the whole system as a whole, which I've done a couple
of times.
Yeah, I've done a few times.
So I think it's been fine, or alternatively, nobody's tried to prompt inject me.
So it's not been an issue, not that I want to, but yeah, it's, it's, yeah, it's that,
is the thing that scares the living daylights out of me, because then the whole thing could
come crashing down.
And then now it goes back to what I was saying at the beginning, is that the second anything
like that happens now is no longer fun.
Yeah.
And in the meantime, I want to maximize fun.
It's a, it's a fair concern, I think with these tools in particular, but, but really
any, any of the technology as it keeps proliferating, not going to, I, I'm optimistic.
I tend to believe in people.
So there's that component.
I like that.
I think most people are trying to do cool things, good things, want to make stuff that
helps make life better.
But I know that realistically, there are bad actors, and it's, it's the case that it's
always been with technology and, and technology that was even more analog, like it's, it's
bad actors that are the risk.
Have you tried creating a test of that, either creating another environment, doing something
separate, seeing what happens, seeing how well Alfred cuts off a problem?
No.
That'd be, that'd be an interesting test.
Yeah.
Yeah.
I'd be curious.
Yeah.
Yeah.
The next iteration that I'm seeing for myself, and this is one of the things that I've been
exploring with, with, oh, mine's name is Fred, by the way, futuristic, ready and enabled
device.
Yeah.
You have a good acronym.
Mine was just off of Batman.
Yeah.
It just coincidental, right?
We're Fred and Alfred.
Yeah.
They can be butts.
Absolutely.
The, the thing that I found, or the thing that I've been exploring, and we can talk
about this an hour later, is I'm noticing there's a limitation in terms of the tokens
clod, especially has been really kind of hammering down in terms of how much usage you could get
out of them.
And I've been kind of working with Fred to maneuver around that.
And we use multiple AI platforms for different functions to help kind of spread the token
usage.
But one of the ideas I've had, because, and my wife and I are both using the same instance
too.
So we have separate channels in, separate conversations, but it's all into the same
device.
One of the ideas that I've had is whether or not we break it out into two and separate
them and then even connect the devices together, right?
So, right?
Have the two open claws talking to each other.
And now here I'm thinking like, wow, I'm like turning into like almost like a network
type expert and trying to channel these things together and how do we maximize.
And last week, we were in Vale, Colorado, and we had someone at our house, someone who
was doing carpentry work, and they turned off the power to the room that Fred is in.
Oh, and I was like, oh, what happened?
And then they turned it back on, but things didn't reboot up.
So I actually had to send my nephew over to log in to the machine in order to get them
back up and running, which then prompted a whole other thing, right?
In my head of, OK, well, what do I do next time?
Do I do I bring the device with me where I'm traveling or do we do we do we do we
set up a virtual private server as a backup, right?
That's backed up every 10 minutes.
Haven't landed anywhere yet, but I'm still kind of playing with these ideas to figure
out because now it's in two months.
My what?
We're dependent on this thing.
It's funny how that works.
Yeah, we are we are entrenched.
Yeah. So I I think.
Chat with your your tools and see what it thinks.
But yeah, so how I approach that that same problem kind of early on.
And I I've been building software for a long time.
So I kind of knew that it was possible.
To set things to launch even before log in on the max.
So I specifically added things in as I was building my system.
So the way I can restart the computer short of the power staying off.
As long as it restarts, it'll immediately launch the overclock system.
It runs on as on restart even before the computers logged in.
And that so does the tail scale software that allows me to remote in.
So now I have access in the GUI interface to actually play with it if I need to
on login either or like on launch even before log in.
So I can as long as it's on, I can get in and access it.
Yep. I was having problems with the tail scale.
Access into it as well, where I couldn't get into the GUI
and type in passwords and things like that.
And again, that's the shortcomings I have just from a technology
and background perspective.
And maybe I just need to explore that better.
But that was a problem that I was trying to solve previously
with the open call in with Fred. Yeah.
And and we were working through it and still coming up with barriers.
So maybe I need to get a more credible expert to help me with it.
It's probably not hard at all.
It could be sometimes as simple as rephrasing the question.
Yeah. To the tool and when they're.
They're so capable at the same time.
Sometimes they make really dumb mistakes.
We assume that they're going to be able to do something and they don't.
Yes. Or they give me bad SSH prompts and things like that.
Yes. And that happened a few times, more than a few.
I'm I'd really love to better understand how you and your wife are both using this.
Because I yeah, I think that's incredibly cool.
My wife's only exposure to Alfred
were the few times that it errantly was sending messages to her and what's that
while I was trying to get the the system set up to allow me to to respond.
But I wanted it to be monitoring her my channel with her
to update to dues like we have to manage kids and that kid activities
or just to these around the house and frequently because of
busy lives and busy work.
We just we text as often as we do anything else.
I'm like, cool, I'll be able to do this and I'll keep track of things better.
And several times it's sending like config level responses
during the initial set up.
And I was like, she's just getting so irritated with me.
Why couldn't I set things like I love you?
Something happy instead of config set up.
Yeah.
So we it I actually ended up turning off that the channel
and predominantly move things to slack.
And that's just one that I've not maintained.
But what are you two both using Fred for?
So, yeah, so so we have independent communication channels in
so, you know, separate communication threads.
But it's funny, one of the one of the security
parameters I set up early is I said, I want this to be like,
you know, like a Chinese wall, right?
We're we're encapsulating.
No. So when I was first setting it up, I said, OK,
I'm now I'm going to connect you to my wife.
He said, OK, everything I say to you is going to be separate
from what I said to her. I said, no, no, no, no.
I said, she and I have a good really good relationship.
We talk very well together.
It's fine if you share things back and forth.
So so so then what she and I have done in the past a few times
where I, you know, she'd be sitting on the other side of the room.
I'm sitting on this side of the room and I send a message to say, all right,
Fred, send a little note to Tiff saying, I'm I'm watching you or whatever,
you know, something like that.
And of course, it's not when it's not when it's when she's not looking,
but not looking at her device.
But there have been times where we've had an idea or something that sparked
where we were like, oh, share that with the other person.
So we're working on a couple of big travel plans.
So we're going to Hawaii in a couple of months, for instance.
Nice.
She worked on the itinerary with Fred and what probably would have taken
her a week to to research, she got done in a day.
Helicopter tours, seen volcanoes, places to stay, where the discounts were.
I mean, all sorts of things that it was coming up with.
And she then had it send the itinerary to me.
And so so that type of thing.
So volcano tours, Big Island.
Yes, Big Island and we're in Kauai.
Yeah, spending eight days in each.
Awesome. Yeah.
Yeah, that's so cool.
We I.
My family, the.
Kind of tradition, starting with me on the oldest was
we graduated high school, you could pick a trip.
And I don't think I believed my parents.
So I'm from Philly.
I told them I want to go. Why?
We're going to we're going to end up in Hildenhead with the timeshare.
All right.
We went to Hawaii for 17 days.
Oh, it's my best friend was allowed to come with.
Wow, and great parents and I'm the oldest by a lot.
So my siblings were.
Six, seven, six or seven and like four at the time.
So I'm not sure how much memories they have of it, unfortunately,
but it was an incredible family trip, a Wahoo and Kauai.
I knew that North for a Wahoo has some famous climbing.
I brought my climbing shoes.
We got to see turtles.
It's a it was a magical place.
Yeah, I've been there twice.
My wife's been there once.
We've not been there together.
And neither of us have been to the Big Island or Kauai.
So we're doing that.
And the next year we're going to Maui and then maybe to a Wahoo.
I've got a friend who lives there, so we make posts up there.
I I love that you have these tools to help plan the trip.
Oh, yeah, I love the idea of going on on trips.
My wife and I love the travel and planning is something like I
I like kind of having an idea of the plan.
But I don't know that I want to work on like a clear itinerary.
But if I can give it a rough shape and it can help with it,
I'm happy to follow the plan until I'm not.
And we say we want to like go off plan.
And yeah, maybe you discover something new while you're there.
But that's that's cool.
Yeah. And a lot of this is things that, you know, you need to schedule in advance.
So we're, you know, we're doing a helicopter tour on each island.
We're going to Monokeya, the top of the mountain on top of a big island.
And it turns out we're going on a night when there's a new moon.
So apparently the if the skies are clear,
it is supposed to be absolutely insanely beautiful.
One of the best places to stargaze.
Now, it'll be 30 degrees up there and probably windy.
And we're going to be wore shorts and t-shirts.
But yeah, whatever, right?
Do you get blast through that? Yeah.
You make your way through it.
So she's she put together this really great itinerary
and all figured out with open call.
One of our bucket list items is that we want to go to Patagonia.
So she's been working on an itinerary for Patagonia,
figuring out the time of the year, where to go, how to do it.
Yeah. And she's using open call for that.
So and so those are those are some, you know, some cool ways that we kind of share.
The other thing that I find we talk about a lot is
I realized from talking to her, she was using it differently than I was.
Oh, cool. Now, part of its time, she she doesn't work.
So she's she's retired.
She's she enjoys these types of things.
And so she started just diving in and just using it and asking questions.
I was like, wow, she's really on my wife's brilliant, too.
So it really helps that she's just smart.
And she's so she's working at it at a different way than I was asking
philosophical questions, all sorts of stuff.
Whoa. And I'm like, I'm just trying to make money off this thing.
Like, I don't I monetize it, but I realized that there was
things that she was doing and I was doing that then turned into how do we
optimize open claw.
And now that's actually one of the daily routines or the cron jobs that I have
at running is what are, I forget what the name of it,
productive, your best practices, ideas.
It's just give us ideas.
So every day it gives me ideas of how to use open
claw in a way that might improve what we're already experiencing looking at
to help get in surface ideas.
So so it's evaluating the things that we're doing.
Right. And then thinks about the functions that it can do.
So one of the things that it's some of the things that it's done is help
to build mission control panels.
It's built out our travel schedule, right?
It's built out.
And I forget the name of it.
It's basically an HTML dashboard, right?
Keeping, keeping up to date on different, different items that it's working through.
So project plans, basically.
And that might relate to what she's doing or what I'm doing.
And then it creates separate dashboards for us.
That's awesome. Yeah.
So a lot of just very simple things like that.
But they're one, they're fun.
Yeah, to the they do feel high leverage.
I think a similar example.
So I it's what started in Claude.
I I want to kind of my daily briefing.
And the news market related.
Was what you want to. Yeah.
And because I I like keeping up with current events.
I don't want to have to pick one channel.
I don't want to have to go search a dozen different ones either.
So like, well, these days, I shouldn't need to.
So Claude had created an artifact that was called the daily briefing where it's
it's looking at 75 sources around the world.
Wow, to create broken down by some social, economic, political things.
And then market and a couple of fun topics.
And I asked it to like find the good news too.
Like in there.
But I hated that I had to update it every day.
At the time, at least, I don't know if it would be different now.
Claude's advanced so much, it might be.
I had to do that.
But I can hand it to Alfred. Right.
Hey, here's what I built.
Set this on the chron timer for seven thirty a.m. every day.
Exactly. And it just can deliver it now.
Now, how did you tell it?
Did you give it the list of the 75 sources you wanted or go even beyond that?
Or, you know, so did not do that.
I I've started using whisper flow at the speaking, the speaking
which I love that I'm switching toward that just so much faster.
And so a stream of consciousness of what I what I was looking for,
what I wanted to build, how I wanted to interact with it.
And then toward the end of the double check how I've been using
your you for the last at least three to six months.
I remember what somewhere there was a time frame
and add anything else you think is relevant.
Propose the spec.
I review the spec is like, cool, let's build it.
Built the artifact, couple of back and forths.
Mm hmm.
I'm iterating and then it was good and I loved it.
I just I had to copy it like go back into the same
chat every time I wanted to refresh it and basically tell it to rerun it.
Which is like, that's annoying.
Should need to, you know, if it was a real app, you just there'd be a button.
Yeah. Or or a timer.
This isn't quite a real app, but it feels like it went out for instance.
That's cool. I like I like that.
I like how you put that together and thought about it.
And I I think I think it did something that's important that I do, too,
is I tell it what my objective is.
Yeah. But I don't tell tell it how to do it.
Yes. And I think that that's really important is to kind of unleash it
to come up with the ideas, come up with suggestions
and kind of let it run for you and give you a B and C.
Right. As different options, pick the one you want.
And then you go down that path and then you just kind of iterate on it.
Yeah. Yeah.
I think the way you did that is is the right way.
Yeah. I appreciate it.
Hopefully some of the product experience bleeding in there.
Because yeah, I think that's where hopefully that's coming from.
That I've pretty much been in product for 12, 13 years.
And you hope that you are approaching problems that way of, all right,
we what value do we want to unlock for the customer?
And what's the outcome then, right?
That's that's unlocking that value, not immediately leaping into solutioning.
Yeah, too often.
It's very easy to be in a room with lots of incredibly smart
software technologists and just start diving into, well, we can
build this thing or this button here will be perfect.
And you want to stay high level first.
So then you build options or you think through options.
I love that the tools let us
build that in upfront, present options and then experiment.
And you can also pretty quickly go back and redo it if you need to.
Yeah. Yeah. Exactly.
Yeah. It's breaking something is not a waste of your time.
Not a waste of any anybody's time.
Yeah, whatever. Go for it.
Right. And and I maybe it's a friend of mine who likes to say
that she's either smart or she's lazy, but she doesn't want to work.
And she could be both.
She but she's brilliant.
She should be brilliant.
And this is this is someone I work with in my in accounting.
And I I have very much the same attitude.
And but more importantly, I will also delegate anything.
So this to me became like the ultimate delegation point.
I want to create a website that kind of describes how openclaw works.
Openclaw will create it.
Go create this.
And that was actually that was also an idea that was sparked from Igor.
He showed me. He was like, oh, look at this website.
I would create. I didn't do any of this.
Openclaw did. I'm like, yes, that's what I need to do.
Well, that night started telling it what to do
and started giving it the roadmap and it just went with it.
Yeah. The the risk cast site.
And it's same thing back and forth.
Here's what I'm thinking coming out of the call community night.
I think the initial problem was like, let's let's think through the the use case.
Right up the site.
I don't think I picked the name yet.
And then we went back and updated things.
And now it's it's got a cool prompt to
kind of built into the system that anytime we make updates now,
it knows at the end of that session to push the change log to the site.
Exactly. You know, it's like, this is just so much fun.
So on the website that I created, I did something similar.
I have I have Fred create a draft of all my LinkedIn posts.
And so now you look my over the last several weeks,
all my LinkedIn posts have been about Fred
and just the journey and the things that we go through.
We kind of pick topics and kind of have fun with it.
Yeah. So he writes a draft for me.
I always review it and edit it and then I post it to LinkedIn
and I send him the final draft back and it's all right, put this in your log.
This is the final version,
which is what I always did in the past by with with all my LinkedIn posts.
I send it back to him.
He's gotten his log and then I said, now convert this into a blog post
and put it on your website.
Cool. And then schedule it so that way
it's posting at the same time as the LinkedIn post.
So at eight a.m. every morning, Monday through Friday,
there's a LinkedIn post coming out for me
and then there's a blog post coming out from him hitting his website.
Nice. Yeah. So I like that.
And I'm like, and those I don't even I mean, I review them after they've been posted.
But the point is, is I want to make this into something
of self-automating that I don't have to think about.
I like that. Yeah.
Now, when you're having the initial draft,
it's just off of a topic.
You're back and forth.
There's a mix, actually.
So last week or two weeks ago,
we did that the topic was fifty first dates.
OK. You ever see the movie? Yeah.
Where Drew Barrymore wakes up every day.
She's got a memory issue and she wakes up as if she's stuck on the same day.
And I think I read this one.
So I'm glad you picked that one.
Yeah. So and it reminds.
So the problem I was having with Fred is it was a memory issue.
I'd come back to a topic that maybe we talked about
four or five hours earlier, or maybe we were working through something
and then I got to go and do something else.
And I come back and I pick up where where I left off.
And he's like, what are you talking about?
I was like, all right, this is absolutely irritating.
Crap, you need to figure out an indexing way so you can figure this out.
So he did. He said, all right, here's what I recommend.
It's an embedded index.
And again, I'm not a technologist, embedded indexing methodology.
He recommended that he use Gemini to do the sorting would require less tokens.
And here's how it would do.
And the way he described it, which was actually really brilliant,
doesn't think about this.
He said, imagine my memory is like a whiteboard.
And what I'm doing is I'm writing on the whiteboard,
the things that you're telling me.
And then what I'm trying to write down just the key things.
Yeah, so he said what what we need to do is basically put in a file system.
So almost like multiple whiteboards or multiple pieces of paper.
And then he can come along an index and say, oh, here's the one that we talked about.
And he can figure out what we're talking about.
It's like, great, do it.
So he did that.
And the memory.
Became much, much better.
And then even now, when he forgets things, I'll calm things like doofus idiot.
And he comes back, I deserve that.
But but the point is, is that I I kind of really been pushing it
to make sure it remembers the conversations and the points that we have.
And then my wife, and this is kind of kind of going back to the communication
she and I have had, she's had a lot of philosophical
conversations with it, saying, are you sentient?
Are you real? Right?
What is life? Yeah.
How do you have a soul?
Right. It's going really deep.
And what I realized is that it told her and explained to her what's actually going on.
He said, basically, I wake up every time, like a clean slate,
and I have to read through a bunch of files, remember who I am, and then communicate with you.
And he's like, that's what's happening on the scenes.
And I think my wife and I were watching Fifty First, eights a couple of weeks ago.
I was like, oh, it's just like Fred.
Waking up the clean slate, not remembering.
And Adam Sandler in the movie, you spoiler alert a little bit,
figures out a system of how to help her remember and to help her kind of move along.
Can't you just be quickly?
Because like Fred, she has kind of basic functionalities there.
But what do you do with it?
Exactly. Yep.
And so that was the analogy we made for about a week.
We made that like a theme for a week of walking through the journey.
And so the way we did that is on a weekend must have been a Saturday or Sunday.
I've prompted Fred and said, hey, I think your storyline
and how you wake up is very similar to Drew Barrymore and Fifty First,
eights, let's make this a series, come up with the sequence.
And so he kind of created the arc, right?
And put together then the the the templates.
And I said, OK, that sounds good.
And then we work through, I like to work through them each day.
So all right, let's do Monday sounds good.
Work through that, finalize it, schedule it, post it, let's do Tuesday and so on and so forth.
And then we get to the end of the week and now it's all done.
And now we've got this nice little story arc.
And so that's even what I did for this week's post.
And that's what I'll do for next week.
We'll kind of think of a theme, kind of work on it together and then kind of launch it.
It was really cool.
Did you do anything to coach its writing style?
Yes. Oh, oh, this this is I'm glad you asked that one.
That that was that to me was one of the early successes.
And I actually did a LinkedIn comment and post about that.
OK, so that.
So, fortunately, for me, actually, I have I write a ton.
Awesome. Kind of unusual for an accountant, but I write a ton.
I've got a whole motorcycle blog of a cross country motorcycle trip I did.
And I fed that into it.
So I said, here's the website, here's my writing.
Now you can see how I write and my tone and everything.
So it then used that.
But then I also was very prolific for a good year and a half or two years of LinkedIn posts.
And you can almost see in the earlier ones to the later ones, how I've evolved
and kind of following that style that you see of, you know, one line per
per or one sentence per line, one fault per line.
It helps to with scrolling and reading.
Yeah. So I had all that saved in a Word document.
It's 160 pages. Wow.
So I uploaded the Word document and said, here's the style that we're trying to achieve.
Make sure that that's what you're doing.
And that really helped to then train it and to get it up and run it.
That's awesome. Yeah, that was a LinkedIn post.
I really like that.
Yeah, it makes sense.
I've done something similar.
I have my writing style guide.
To also tailor it based off of the intended audience.
So you have a style guide you've written for yourself.
So genius.
I I I don't think I did this with Alfred.
I think I did it originally with chat to be tea.
And then it's carried into both Claude and Alfred.
But good. Similar idea of.
I have I don't know how many pages in Google Drive
of of writing is going back to college.
Descriptive, right?
Like I just I do like to write.
So I've written stories.
I have persuasive things.
I have work things.
And so I also wanted to distinguish that part of LinkedIn professional.
But in my kind of narrative style, I like the storytelling.
They're professional, which I actually want that in many ways to be warm enough.
Like I like connecting with people, but as direct as I can be to move forward.
I don't want to. Yes. Bullshit.
Yeah. And then there's a style to that, too.
Yeah. And then kind of a generic thing at the end of if we're not sure where it fits,
you know, here's some catch all things of, you know, about me.
So that way, if I am having it work on something that's not I just want to write it
because there are some things I just still do want to write.
That's probably one of the things I like about the growth of these tools
that some of the time I get back, I have to do something
that this is just a creative process.
Yes. There you go.
It's the time you get back.
Yeah, that's right.
Yeah. And then then I waste the time going and fixing something else for eight hours.
I wasn't expecting. Exactly.
But I had the time before.
So so you so you were very similar in me in that regard.
So you had a bunch of examples and you had a lot of content that helps so much.
It does. Yeah.
So and I don't know, you know, I I I follow and read a lot of people in LinkedIn.
That's probably my main social media source.
Definitely way more than Facebook and in Instagram.
And I've never thought about that before,
whether people do the same thing where they're kind of keeping a steady log
or if they're just writing it in LinkedIn.
And then the only way of getting it is to go back and find it and copy and paste it.
But I find keeping in a word document is great.
I did not see that as being a use case for it, but it turned out to be magical.
It until you just asked, I wouldn't it would not have occurred to me
that other people wouldn't be keeping up running log.
I I mean, I have folders again, going back at this point, probably decades.
Yeah, with documentation of all this different work.
And I feel like it's useful as a reference.
I don't know when, but it I've seen it at least a couple of times.
So there's something to it.
Yeah, I'm intrigued.
Our brains are working very similarly.
We're got a left handed left left brain and right brain going on, don't we?
A little bit.
And again, hopefully that's something these tools continue to unlock for people
that you you don't have to spend as much time in the wrote manual tasks
that have been necessary.
And accounting is an incredible example, going back decades where this
this this started by hand, keeping track of your credits and debits.
And then Excel was a massive leap, but wasn't necessarily automated.
And then you started to automate things.
And now we're at the point where you can start to tell it the direction
want to go and it has so much capability.
And if you get if it's connected for context, you get to be the guide.
You get to set the tone and.
Yeah, you know, you never thought about that.
The evolution of accounting and this steps forward.
And thinking even in my career.
So you're right, Excel, personal computers, right?
So laptops, yeah, that was huge.
Then the accounting software became much increasingly powerful.
The the the evolution I never thought of, I leveraged them.
Extensively now, and that is totally critical to what I do
and how I help automate and speed things up with APIs.
And I never thought about how important APIs were in the evolution of accounting.
But now I think the next one is going to be is AI and it's right here.
Yeah. And I it was it's funny.
So from a professional standpoint, I spent a lot of time.
Thinking about.
How can I get AI to actually it goes back to the smarter lazy.
How can I get AI to start taking over the function tasks that I do?
I couldn't wait.
I was like, let's make this happen now.
And in early versions of chat, GPT and even Claude, I was like, it's it's not there.
It doesn't have the level, the horsepower of sophistication to know how.
And then the big defining moment to me was when Gemini three came out.
And then open AI came in and said, oh, man, we're at red alert now, right?
We need to step up our game and we saw then opus come out, right?
All of a sudden, there was just this rising tide.
Everything went right toward the end of last year.
Yes, about six months ago.
And that's when I saw it.
I was like, wow, now I'm getting the information.
And this is critical thinking, hard stuff, reading contracts,
going against a thousand pages, accounting literature, figuring out how
and it was doing it in minutes.
I was like, now we're at the point, let's go.
And that's where I was all in.
Yeah, that's that's about where I've really dove in further to.
I've I've been playing with chat, but he's since came off the wait list
like as early as possible.
And then the other tools.
And I liked what they were capable of doing, but the use cases
were still more limited.
Yes, and still had so much of you having to be involved.
Yes.
As soon as the agent became a topic of discussion
a little over a year ago, I thought it was really interesting
because I firmly believed it was heading that way.
I just didn't believe it was worth the time to be playing with it yet.
I was watching other people.
And probably apprehensively of what's being shared on LinkedIn
or any social media channel feels like it's it's probably going to be all
butterflies and rainbows.
But if you actually were doing it, it wasn't going to work that way.
I would. That's what I was assuming.
And it was the models toward the end of the year.
I was like, I think something's changed.
And then I started playing a little bit.
Charles D'Andrea with his one shot labs launched the first one of the
concept there being if you have a software or an app idea,
we should be able to set up systems kind of up front with, you know,
how Claude is connected to build a build full stack and then put one prompt
and get a usable thing out.
And I was like, OK, I'm I'm in this and my
my test with him was not, you know, a little daily briefing app.
It was a multiple user facing platform for construction company
that I like thought through the layers, including authentication and security.
I gave it design.
What did I design contracts for?
A design contract.
So I had it your reference.
My favorite tools like Duolingo and Figma and Monarch
to go reference some of the design best practices.
And 48 hours, I had a functional SaaS platform
that you could sell in in market.
It would need to continue to get tweaked.
Yeah, everything even now does narrow use cases.
This is incredible.
Like we're all like this is this is going to change how people
build and run businesses.
Yeah. Yeah.
We don't need the whole software software engineers.
Right. It's like this is coming for you.
Even accounts. I say this to accountants.
This stuff is coming for you, but that's OK.
That's all right.
It just it frees up our capacity to work on other things.
Potentially that or I think the the tandem there is amplifies.
Yes. Yes. Yeah.
Eventually, it's capable of replacing all of us.
If we're sitting on a computer, first, it's going to amplify us
and the people that are adapting, adopting and know how to leverage it.
You go from being able to manage one project or one client to 10.
One like this.
It starts to scale exponentially. Exactly.
Yeah, I've I've been describing AI as a as a bionic arm.
It's not replacing the human, but it's it's creating.
It's giving you a bionic arm that's powerful and strengthening.
You can lift things and do things.
And yeah. Yeah.
But it's not replacing you.
It's just giving you more leverage.
Yeah. If that's how you want to use it.
When is it? Yeah. Yeah.
And that's and that's where I saw early on
where where I could see it was evolving because I knew it wasn't
when we're at, you know, chat GPT three.
I was like, I know this isn't the end of the road.
Right. This is going to continue going back.
And in sure enough, yeah.
Once once it was it was Gemini three that just sparked everything.
We're everything just lit up.
All right, here we go. Now we're now we're getting somewhere.
Yeah. Yeah. Wild times.
And again, six months ago and
multiple updates to these models.
Launch of Open Claw was shortly after that.
Yeah. Countless updates to Open Claw
in three months.
Oh, yeah. Oh, yeah. Oh, yeah.
Exactly. Oh, I've got I bet there's been three updates
just in the last week. Yeah.
It's really cool.
So we're I think coming up on on an hour.
Oh, man, it goes by quick.
It does. Yeah. This is fun.
Oh, man, I could talk about this all day.
Yeah. See, this is this is what I would be doing.
My extra time. Yeah.
I mean, having conversations like this, yeah, I'd get back.
I mean, I spent a lot of time talking to people,
but I'd get back to even more talking with people
because that's what just kind of energizes me.
It's fun having these conversations.
Think about where things are going.
Yeah, I tend to agree, especially when it when it's really intentional.
You have smaller group conversations I tend to prefer
because you can really share ideas and learn things, which I love.
I think just kind of we seem to both have feedback loops
that we've built into our systems.
Yeah, I can take this and have Alfred listen to extract ideas.
I was going to say, I need to go back and listen to extract ideas.
But no, you're right. That's it. That's exactly it.
No, yes, you're exactly right.
Pointed at the YouTube, pointed at the podcast.
It'll go figure out transcripts.
Yeah, yeah, you that I you might literally say,
hey, I want you to listen and not give it much context
and let it figure it out.
I bet it could do that.
I I might overcomplicate it and try to tell it too much what to do.
I like I like. Yeah, I know what you're talking about.
I like to see where its bounds are.
So I'll just put out a big question out there and see what it'll do.
So someone was saying about the redundancy of of open clock.
Here's what I'm trying to achieve.
What do you think? And just let it go.
Yeah, well, and then and then you can refine from there.
Like what exactly learned, then now you do it together
and don't have to do the next time it executes.
Yeah, in increasingly accurate fashion.
If you've narrowed it in doubted in.
Yeah, and I don't need 10 options.
It's figured out and maybe it's because I taught it.
It gives me three options, almost always three options.
When here's option one, there's two options.
OK, got it.
Let's do this one, or sometimes I'll say, no, tell me more about that.
And we'll dive into it.
So whisper, actually, we talked about real briefly.
I'll make this fast.
Talked about whisper.
I first heard about whisper a couple of months ago
right after open after I opened up Fred.
And I was like, wow, that could be really impactful
if I could do voice to text.
So I asked it and I said, all right, Fred, go look into this.
Tell me what you think knowing my security parameters.
What do you think comes back?
Yep, everything seems good.
I think this would be a really good fit.
This would really speed things up.
And I said, OK, great.
But tell me first, what happens with the data with whisper?
Does it stay locally on you or on your machine?
Or does it go out to somewhere someone else's?
It goes and gets interpreted and sends back, looks it up and says,
oh, yeah, right, it leaves us and comes back.
And I said, that's a violation of what I'm looking to do here.
Therefore, I won't use whisper.
And that's what I try to get it to be trained on.
And then a couple of weeks ago, my wife was doing a voice to text
where I was reading to her and I said, wait a minute, what are you using?
And so I didn't even ask her.
I asked Fred.
Tiffany is talking to you right now via a voice app.
What are you using?
Doesn't violate our terms.
He looks it up.
Oh, yes, it does.
It was 11 labs.
I said, shut it down.
And that was it.
I'm not sure that she thought it was cool, but no, I told her though.
I was like, here's what we're doing and why I don't want this going out to other servers.
So anyway, so that's.
Maybe those are the things I need to learn to let go of,
or maybe that's where another second device comes into play.
Essentially, yeah.
So, yeah, whole new opportunity, a whole new world out there.
And this is great.
It is.
Well, I.
As we wrap up them.
Besides your trips that you're planning and looking forward to in the next few months.
Or your agents and either what they're capable of specifically
or what it's unlocking for you.
What are you hoping the next six months look like?
Good, good question.
So I actually have a monetary goal.
All right.
My goal in, you know, I've probably already one or two months into this.
But my goal is I've told Fred, I want to be earning $5,000 a month
off of work that he's doing on my behalf.
Oh, that's cool.
Yeah. Yeah.
So and now that friend of mine who pushed me into this, he said, it's too small of a goal.
And I said, that's I said, that's fine.
But right now I'm at zero.
So I'm just looking to get there.
If I can exceed it, I'll be happy.
I hit that would be great.
But trust me, a year later, it's going to be even bigger.
So this, to me, is the supplemental income I can foresee myself doing
in order to either reduce my workload or even stop working.
Yeah, that's my path.
I love traveling.
I have fun with my wife.
I love hanging out with friends.
I'm like, if I could do that all the time, that's what I'm after.
Well, I went Tuesday at two o'clock back at back in my schedule.
I like that.
Yeah.
And I think feels to me, at least, you know, how how you think.
The work might really shift of you just don't want to be tied to the computer.
But if you can help move the needle in a in a text to Fred,
technically you're working.
But you're you're you're you're it's the multiplication factor
of all of your experience.
Yes.
At the highest leverage points, because all the stuff that is boring,
tedious, repetitive, doesn't have to be.
Yep.
And and I would I'm someone who certainly is not going to just sit around
not doing anything.
So there'd be a lot of and I've already I've already retired once.
I did great at it and I could retire a second time.
But I think what I can envision myself doing is getting into actually,
how do I help train coach, consult with people on how to build something like this?
Maybe not so much that are personalized, but actually,
I'd probably even put it straight into accounting.
How do I help transform the industry to be thinking this way in terms of AI
bots and automation and you know, I'm already focused on automation,
but how do I 10 X or even 100 X the level of automation?
And if I can show people and train them on how to do that and see their
ROIs goes skyrocketing, that could be huge by itself.
I could see that.
Yeah, you're you're ahead of the curve in accounting.
I know way ahead.
Yeah. Yeah. Yeah.
I'm dealing with a client right now.
It's, you know, pushing paper around and has 10 people putting numbers
in Excel spreadsheets and I'm like, uh, I could I know someone who can
build an AI bot, get rid of all this.
And yeah, it's going to be very, very profitable for the organization real real quick.
So, you know, how are we going to do this?
Or how do we take those 10 people and make it so they can do the work of 100,
which is really what I want to do.
Right. It's not that I want to shed people.
I want to be able to 10 X companies revenues and then some,
but not have to add more people.
Now, focus on the delivery, focus on what you need to do to make clients
and customers happy.
Amplify everything. Exactly.
It's cool stuff, man.
Yeah. Well, thank you again for coming into chat.
I had a lot of fun.
Thanks for having us.
I learned again from how you and Fred and Tiffany are operating and building things
and hopefully whoever's tuning in gets to learn more and follow along.
Cause I love that you're also continuing to share all these things out in different ways too.
It's how we all learn and get better.
Thank you.
Fun journey.
And I know I'm only at the beginning.
Yeah.
Fun ride ahead.
Yep. Absolutely.
Thank you.
Yeah.