Last year we needed to have a new fence built. Ours was falling apart and we had
a new puppy who would easily get out. So we had a few options.
Do nothing. That's always an option. Even if it's not a good one, hire somebody to
do it for us at great costs, like $10,000
or do it ourselves for $2,000 plus one weekend.
Option three also required skill expertise and the right tools, which I don't have,
but luckily my father-in-law and brother-in-law do.
So we opted to build it ourselves.
Now we faced a similar choice a couple of years prior when we had to replace our
hvac.
Except here we couldn't do nothing
and we didn't have the skills, expertise,
tools or access to the raw material.
We had to hire somebody to replace it for us.
Now I just got back from EEC Cam's creator camp where one of the biggest topics of
discussion was the very idea
around
DIY versus paying for a tool but for software.
And that's what I want to talk about today. Hey everybody and welcome to another
episode of the Streamlined Solopreneur. My name's Joe Casabona and today I want to
talk about
when to vibe code, your own tool
and when to actually pay for the software.
So what is the major question that came up at ecam M's Creator camp and really what's
been coming up a lot lately,
it's this
now that AI can write reliable software for us,
should we replace our software subscriptions With tools we've vibe coded,
it's easy to think. The answer is almost always yes.
The temptation of replacing our tools
or our monthly subscriptions is really good.
I would love to not have to pay
thousands of dollars a year for software.
And to be honest, I have been vibe coding
a lot more of my own tools lately.
Here are a couple of examples.
I coded a replacement stream deck plugin for Apple Music
because El Gado removed theirs.
I even made some improvements to it like adding playlist button support and dial
support for the stream deck. Plus
I also deleted El Dodos camera hub app
and created my own teleprompter app. I am using that right now. It's so much better.
Elga O'S Camera hub app was supposed to have this thing called voice sync where the
teleprompter would move based on what you were saying and it did not work even remotely
reliably.
And the tool I am using does work reliably. In fact, I am
ad-libbing off the script right now and
the app is telling me, are you ad-libbing?
So
I coded this with Claude
in a half hour.
The biggest projects I've replaced are web-based.
I replaced Seja with my own testimonials collector
and I built my own beta reader software.
I did this because honestly with Seja I wasn't collecting enough testimonials to
really justify
the $30 a month
and nobody was using the Submit a video testimonial.
I know that some companies like Ecamm use it for customer research and for a lot
of other things or a lot of people get a lot of video testimonials. And if I were
in that boat,
maybe I would keep paying for Senda.
But $30
a month is a little much for the one to two text-based testimonials I get
and the beta reader software, like I don't even know if people are gonna beta read
my book.
And so it didn't make a lot of sense for me to
start paying for this software.
And it could be the case that once I get users into this Beta Rita software, once
I start writing my book,
I might want to pay for the software.
But I didn't at this moment.
But like I said, the temptation to replace every subscription with software we need
to support
shouldn't blind us to the fact
that there are some things we shouldn't replace.
It's really nice to think there's a world where, oh, I will just pay a hundred bucks
a month for Claude and five bucks a month for lovable
and I will replace thousands of dollars of software.
But that's not
tenable.
I'm not gonna say it's not possible and I'll talk about why in a minute. But it's
not tenable. For example,
I will likely never replace
Kit which sends my emails.
I don't want to deal with deliverability.
I don't want to deal with checking automations or managing payments for subscriptions
or spam
Ecamm,
which is my recording and live streaming software.
I don't want to have to deal with concurrent streams and massive overlays and managing
resources and reliably recording files.
Do I trust myself to vibe code something with AI that will reliably record and save
isolated versions of all of my files? I don't.
And the guys at Ecamm
know this stuff really well. They've been doing it for a long time.
And finally, my podcast hosts rss.com and transistor,
I will not replace them because again, I'm serving up audio files or video files.
Now
I am doing the analytics and trying to account for spam. These are just things that
the
three companies that are
the four companies that I just named have a lot of expertise in this.
And so now we get to the crux of the matter.
How do you decide if you should replace software you pay for
with something you have vibe coded?
And it comes down to those things I mentioned earlier, skill,
expertise and the right tools
as well as your dependency
on the software.
So with defense,
and not at all with the hvac but with the fence, we had the skill
being able to build anything, use tools properly like, like
carpentry tools.
We had the expertise knowing how to build a reliable fence. This was something my
brother-in-law apparently has a lot of experience in.
And we had the tools, namely in Auger, which allowed us to sink the posts without
having to manually dig three foot holes.
But in this case, we could also do nothing.
That is, we didn't have complete dependency on the fence.
We could take the dog out and put him on a leash.
When it comes to our hvac, like I said, we didn't have the skills, the expertise
or the tools,
but we also have dependency
on our hvac,
especially
in the bitterly cold winters we have here in the Northeast.
So doing nothing was not an option for us.
We didn't have any of the skills, expertise, or tools to replace the HVAC ourself.
And so we
decided, well, we didn't.
It was decided for us, right? We had to pay somebody to replace our hvac.
But let's move away from the home examples.
What does this look like in software?
Let's break it down one by one. Number one skill.
Do you have this skill
to do this? You don't need to write code anymore.
This is true, but you still need to understand software security and maintenance.
You can't just launch a tool into the ether, possibly one that,
has usernames and passwords or your own data
and just trust it's going to work.
You still need to at least ask the right questions or know the audits
to do the security. And you need to know what to do if things break,
right? If you make an iOS app or an iPad app
and then the next version of iOS comes out and your app breaks,
you need to be confident
that you can fix it
or the AI can fix it
or that you can just stop using it. So that's the skill part.
Number two is expertise. You need to know enough about the problem and how to solve
it to build a solution for it.
I think we've seen time and time again, right?
Vague prompts
lead to
incomplete or inaccurate responses.
The first thing I do when I create new software
is create the spec.
And I use that using one of the quote unquote smarter models.
And then I will feed that to Claude Code after reviewing everything and going through
and make sure it touches
everything I needed to touch.
But if you're just like, Hey, I need a way to send emails,
right? I wanna replace Kit, I want a way to send emails.
There's gonna be a lot of stuff that the large language model will not be able to
help you with. In fact, when I replaced Seja,
it's straight up said,
Hey, do you want the video part? 'cause that's gonna be super hard.
So you need the expertise. And I had the expertise to know
I'm not gonna want the videos.
All right, number three tools.
The tools to actually deploy the software are much easier to use now,
but do you know how to use those tools? Do you know how to use Netlify or Versal
and Superb Base or
CloudFlare, there's a lot of free options now. And knowing which one to use is important
because
for example, I just replaced my personal website, casabona.org. It was a WordPress
site, now it's a flat file site.
I'm hosting it on Netlify,
which means that every time I write a blog post I have to deploy and it uses 15 credits.
If I was writing multiple blog posts a day,
I would've made an incredible mistake
and
I probably
should have kept paying
for my WordPress host or move to a different thing like Ghost if I was sick of WordPress.
So that's something you need to understand.
And then number four, perhaps the most important question, right?
Because the first three, like maybe they're not insurmountable. Maybe AI can help
you quote unquote, or maybe you have friends who have expertise in this area, but
dependency.
If the software stopped working and you didn't know how to fix it,
how would that impact your business?
I was quote,
coding an iOS app
back in March and I really wanted this feature and I thought it would be neat.
And
that version of Open AI's Codex
couldn't get it right.
it's fixed. It wasn't fixed, it's fixed. It wasn't fixed.
And then it told me how to fix it. And I had no idea. I have never programmed
with Swift. I had never written an iOS app before
and luckily I had the skill and expertise
to make feature branches
so that
if something went awry, it didn't kill the whole app,
but it also wasn't a required feature.
So that's the dependency thing. So let's look at my examples again.
First, let's do the stream deck plugin and the teleprompter app.
I have the skill
expertise and tools
and I have zero dependency on these.
If they disappeared tomorrow, I'd be mildly inconvenienced.
So I was very comfortable vibe coding them,
if my
stream deck plugin stopped working, I could go pay $8 for one.
It maybe wouldn't do everything I wanted to do.
But also I have the media keys on my keyboard, like that's not the only way I can
control Apple music. Same thing with the teleprompter. There's a lot of teleprompter
apps. I vibe coded this one 'cause there are some very specific features I wanted
and I have them now and it's working great, like unbelievably great,
but I have zero dependency.
Okay, number two, send your replacement. And the beta reader software,
beta reader software man,
that really does a number on my accent. I try to keep my New York accent on a low
burn, but beta reader software really kills me.
Okay? I have the skill, I have some expertise
and I have most of the tools.
So I felt very comfortable doing this. And again, virtually no dependency on these.
Would I be inconvenienced if
my Senda replacement went away? Yes, I'd have to update
a bunch of embeds across my websites
and that would be annoying,
but I would spin up a Google form
to keep collecting testimonials
and the beta reader software. Forget about it. Like
I'm not even using that right now.
Okay, so those are the things
I've coded. Now let's look at Kit
EAM and podcast hosting.
I do not have this skill for these types of software,
nor do I have the expertise.
Things like deliverability
or concurrent streaming,
file storage, spam management,
server management.
There's a lot of things that can go wrong with all of these, which is why I trust
Kit
ecamm rss.com and transistor.
I have some of the tools.
I know how software works
and so I know enough to know that I don't have all of the skills and tools and expertise.
But that leads me to number four. I have deep
dependency on all of these things. I am using Ecamm right now to record. I sent an
email with Kit this morning. Both of my podcasts go out later this week and I know
they will reliably do so.
If they went away tomorrow,
I'd be screwed.
Yes, I could find a kit replacement.
Yes, I could find a podcast host replacement. I honestly don't believe I would find
a good enough Ecamm replacement. Like yes, OBS exists, but
I would be really left in the lurch,
which is why I happily pay for all of those things.
I guess disclosure, here I am,
a content creator and evangelist for rss.com. And so I do get that account for free,
but I host actual podcasts over there.
Now. I
suspect
I'm at a slight advantage here
because I have software development expertise,
I have a master's degree in software engineering.
Maybe some would say because of that I have the burden of knowledge
and maybe just go out there and try stuff and see what happens.
These decisions are going to be different for you.
Maybe you don't have the dependency I have on kit until you could try creating your
own ESP, though I really wouldn't recommend that.
It's up to you to evaluate if you want to spend time
coding something, testing it, deploying it, and most importantly
supporting it.
That's
the big one.
Because yeah, you might launch something and it works great today,
but what happens
a month from now, two months from now, six months from now
where something goes wrong and you don't know how to fix it.
Again, I know the temptation is great,
having the discernment to know when you should
code a replacement
or keep paying for something
that is well supported and reliable
is important.
if you need help with this or if you've been wondering
how you should approach using AI to write or replace software for you,
I am cooking up something special. It's a cohort and course where I will walk you
through how to figure out what to replace and how to use AI to replace it
and
how to make sure it turns into a long-term solution for you.
We'll have an example project to walk through before you start building your own.
And that's really my goal For the end of this cohort. You will have a tool that you
have fully replaced with your own version. Thanks to Vibe Coding with ai, this is
something I have saved literally thousands of dollars on. I've replaced websites
where I was paying a ton for hosting. I've replaced tools with high monthly subscriptions
what I've done for myself is to replace the tools
that
I know I won't need in six months if they go away, or the tools that I don't have
heavy dependency on. 'cause I don't want you to switch
and
then end up having to switch back. So I will be with you every step of the way to
help you.
If you're interested, you can head over to streamlined fm slash waitlist, scan the
QR code to join the wait list
and you'll be the first to be notified when presale opens. There will be a special
presale price. Again, that is streamlined
fm slash waitlist.
So that's it for this episode of the streamlined solopreneur. In it, we went over
the basically four criteria to figure out
if you should look into vibe, coding, a replacement, or just pay for the tool. I
am a big proponent of paying for software that's worth paying for. I pay for convenience
all the time
and I pay for expertise.
But
if there's a tool that you can reliably replace to save yourself some extra money
or simplify your tech stack,
I think that's really worth looking into.
So let me know over@streamlinedfeedback.com. Have you vibe coded your own tool? Are
you worried about it? You can send me a voice note. I'll play it on the show. But
But that's it for this episode of the Streamlined solopreneur. Thanks so much for
listening and until next time, I hope you find some space in your week.
