Episode 92: Be Your Own Software Vendor

Transcript

I am almost certainly wrong about this. If you want to call this a prediction, if I were a betting man, I would not bet on this.
But I think it's possible, maybe on a small scale, but I find it very interesting to think about what this would be like if it ever did truly reach real scale across society kind of thing.
There's been so much discussion about how this AI revolution is going to remove jobs, particularly when it comes to talking about developers and more technical people.
You know, designers, developers, UX people, that whole like kind of slice of the workforce where either they won't be needed at all or there'll be fewer needed or the expectations will rocket up.
Or, you know, predictions around actually this is just a temporary thing and we're throwing a lot of AI generated slop out into the ecosystem.
And what's going to happen is that down the road, developers will be even more needed because they don't have to come in and clean up all the crap.
Like there's any number of hypotheses and theories surrounding this general thing.
But the overall, the overall sentiment seems to be that, you know, AI will increasingly pick up more and more of this technical sort of work and therefore people won't need to be as technical and there will be fewer technical things to do.
But what I've been thinking about recently, and as I speak through what I just did, I'm not sure that this idea is totally contradictory to that.
But it's almost like a different direction.
A lot of this seems to be all about, well, we're not going to need it anymore.
We're not going to need it anymore.
And what I've been thinking about recently, because I've done a bit of this, I do come from a technical background, right?
So I do understand this, I am a programmer, and what I can say is that in my, in the, just in the last, I'm going to say 16 months, these tools really have made leaps and bounds.
Now, I think it is extremely helpful to still be technical, to kind of watch over what these things are doing.
Because when you end up in weird situations or things that aren't working right, it is helpful to go in there and be able to turn the screw yourself.
But there's been less and less of that.
Like when I first started working with a lot of this stuff, again, maybe 16 months ago, there was still a lot of manual intervention required.
There was still a lot of maybe set things up, hand code some stuff, and then have the LLM fill in the blanks kind of thing.
And more and more over the year, it's become more and more possible where these things really can just run with whatever it is you asked them to do.
Maybe the better way to try to frame what I'm going to talk about here, at a company level, at a corporate level.
You know, think about big companies that produce software and services and stuff.
So, the general sentiment seems to be that AI can replace a lot of their workforce.
And they'll be able to produce even better and even more software that people will want to buy and use.
But here's the thing I've been thinking about and the thing that I've noticed that I've been able to do, at least.
As a person who regularly uses cloud services and cloud software, I have found that I am needing to rely on said cloud software and cloud services less than I used to.
Now, this doesn't count, I'm not talking about the actual LLMs themselves, right?
Like, if I'm, you know, in working on, working on like a project or something and I'm using, say, cursor, that's obviously going out and contacting the AI services.
But I'm talking about like, I do way less having to look around for like utility applications and things that fit small needs in my life.
Because what I've been able to start doing is that when a small niche need comes up,
I can jump into one of these AI powered, you know, code and project editors and have it make the thing for me in a couple of hours.
At least make it for me well enough to use it as a one-off.
And I found this in my work life as well, which I find really interesting.
Because as things and projects and initiatives and whatever come up, and it's like, oh, for this one client, we really need to do this one specific little thing.
We may never need to do it again.
And having come from an organization, you know, those are hard things to sell to leadership.
Where it's like, oh, I need to buy this thing for this one niche little project, but we'll probably never use it again.
And by the way, it costs $10,000.
In this world now, you can kind of just go in and tell one of these AI, you know, development suites, here's what I want this thing to do.
And it'll kind of spit it out.
And if it doesn't work, even after you try to tune it a little bit, you're probably only down a couple hours of time.
And if it does work, you get to use it as that one-off.
And if you never use it again, again, you only spent a couple hours putting the thing together.
It's not like this was a multi-week thing.
So this is almost in contrast to two different workflows, right?
Because again, being a developer coming up for years, there would be so many little things in life where I'd be like, you know, I could fix this with a little app.
Or I could fix this with a little web service.
Or I could fix this with a little application or whatever.
But the barrier to having to sit down at my computer, spin up a project, get all the dependencies together, think through the requirements, blah, blah, blah, blah, blah, and then actually code it,
a lot of times the trade-off wasn't worth it for the small problem I was fixing in my life.
Whereas now, I can sit down and give really some often half-assed requirements, talk through it with the LLM a little bit, and get to a result in very short order.
You know, dumb stuff like, oh, I need to take this CSV file and pull this information out of it and merge it with this thing over here and grab this JSON file and reach out to a service and do a thing.
And then at the end of the day, I want to spit out a web interface that lets me sort and filter things.
Because, you know, five years ago, ten years ago, that might have been a multi-week project.
Get it done in a couple of hours, use it once or twice, never look at it again, and it's fine.
It was still a worthwhile investment of that little bit of time to have this one-off tool.
So, that's one thing it's competing with.
The other thing it's competing with is this thing where you'd then go, maybe you had a use case like this,
and then you would go out and start looking around at software out on the web or whatever,
trying to find a thing that might come close to doing what you wanted to do,
which often ended up being expensive or didn't quite do what you wanted to do or whatever.
I think, and here's what I'm probably wrong about,
I think there is a possibility that we as society could actually become less dependent on big company software and services
and instead get really good at writing some requirements into a box
and for whatever, $20, $50, $100 a month,
have these systems build us these small utilities that remove resistance out of our lives
and make us more productive and help with jobs that otherwise would be extremely tedious.
It's almost like now that no one has to really code so much,
everyone can be a developer.
And while the large companies are out here,
and by the way, I go and I listen to the episodes that I record here,
just to, you know, mostly for sound checking and just see if they come off okay.
And I realize that there has been a trend recently in me just kind of like railing against big corporations.
I don't know.
Maybe I'm just a little extra sensitive to that these days.
But here's another piece, right?
These large companies that produce cloud services, cloud software, social media networks, all this kind of stuff,
they are spending a lot of money right now convincing us as the public
that because of AI, their services are going to be worthwhile and better than they used to be.
And I'm wondering, I am thinking that there is a world, at least at current price points,
with current capabilities of these LLM code editors, that could shift.
Prices could go up.
They could become worse.
Who knows?
But where things stand right now, I think there's a pretty strong argument to be made
that we as people could detach a little bit from this stuff
and make our own stuff that does the things we want it to do as throwaway things, right?
And it's so counterintuitive to how we have viewed software and development, all this stuff for so long,
where it's always been about gathering the requirements for the largest use case
and trying to have as many features as possible because it could then be marketed to the widest array of people
or companies or whatever.
But what if you can just make something uber specific to yourself
without really having to code much of anything and really, in a lot of cases, nothing at all.
Just having a conversation, the thing spits out and it works.
And I'm telling you, these editors now, they can set up your databases,
run your services, install stuff on your computer, spin up the web interface, run the thing, run tests.
It can do all of these things for you.
It's not like it's going to build the thing and then you have to just blindly run some command
in your terminal as a layman.
It can handle that for you.
I ask it to handle that stuff for me all the time.
I'm in my cursor quite often saying, hey, run this for me.
I am perfectly capable, perfectly capable of going to the terminal and running it myself.
However, if I tell it to run it for me and something errors, it can then check the error,
correct it and rerun it, right?
Like, and I think maybe right now, not just anybody, but I'd say, you know, people with
technology and developer backgrounds, maybe notched that down one or two levels in terms
of the general public and that level of people could do this.
And with how massive these companies have gotten and this, how complicated the services
and software have become and how expensive it is beginning to become.
And let's be honest, in a lot of cases, how, how crap even experience it is, because you
go out on the web and you go to something that's supposed to do something for you and
you're just drown in like sponsored sites and ad content and whatever.
Why not just spend a little bit of time and make it yourself?
Like you need something that like converts this kind of image to that kind of image or does
a thing or sorts of spreadsheet or like whatever.
You can probably get that done pretty quickly through like a clod or a cursor for very little
money, use it a couple of times.
And if you never use it again, it's not gonna be that bad.
Now, I say I'm almost certainly wrong about this for two primary reasons.
One, well, let me back up because I, I, I believe that this is a very realistic thing that
people could be and probably are doing right now.
Like this is not pie in the sky.
Oh, if we just have one or two more features, we could do it.
Maybe I'm slightly wrong about the level of technical experience that would be required
to do this sort of thing.
But we're very, you know, the point is you do not need to strictly be a developer to,
to, to, to develop anymore.
You know what I mean?
You don't need to know how to code to develop anymore.
So I think that this is realistic right now today.
And I do it.
Like I'm here to tell you that even as a person who has learned to program for years and years
and years, I do it less and less and less.
And I am now in the habit of when small things come up, that small bits that a utility would
take care of, I just code one up, but I don't have to code it.
I just tell something else that I want it to code it.
And it just happens.
But I say, I'm almost certainly wrong about this for two reasons.
First, this is very reminiscent of so many, this, this line of thought is so reminiscent
of so many other kind of technological advances throughout the course of computing history.
One of the ones that jumps to mind is that there was a tie, if you're not familiar, if
you're not a technical person, there, there's a thing when databases rolled around.
And a querying language was developed on top of them called SQL, SQL, which actually,
as I'm thinking about it, is it standard querying language?
I forget what the S stands for.
I think it's maybe standard, standard querying language.
Anyway, SQL is what people call it, SQL.
There was a thought at some point in time that all people would learn SQL and that that
would be the new way to communicate with computers.
And if you've ever worked in a corporate environment, you know, as well as anybody else, that that
never happened.
There's like three people at your company that do SQL really, really well.
And it's super complicated.
And the scripts are like 5,000 lines long and they're a mess.
And it's always a problem, right?
Like that's where SQL went.
Same thing happened when HTML rolled around, which is how you put web pages on, on the web,
right?
There was real thought that HTML was basic enough that any person was going to be able
to throw together some HTML and get a website up.
Again, if you've ever worked with people who work on websites, that is not the case.
There's like, there's again, like this handful of people who know how to do that.
And it's very complicated and blah, blah, blah.
So this line of thought that I'm going through where I'm like, Hey, if you need something
to do something, pop open, clawed or whatever, and tell it what you want.
And it's just going to spit it out for you.
And all people can probably accomplish this.
Maybe true, but boy, it sounds a lot like that pattern.
I think it's a little different because you get to converse in normal English or whatever
your language of choice is, but still it kind of sounds like very, very reminiscent
of that pattern.
The second reason that I don't think this will come to pass fully again, goes back to
my tinfoil hat about big companies these days, large companies that produce.
software and services on the internet are very vested in making sure you cannot really
do this practically.
And they will spend a lot of marketing dollars to convince you of such.
And they will, they, you know, they have a lot of levers and dials and knobs to pull and
push and twist to make sure that everyone continues to believe that in order to accomplish
productive things with computers, you need their stuff.
And so you're fighting an uphill battle there already to just break off and live in the woods
and make your own software by yourself.
But with all that said, next time you have a little thing where you're like, oh man, I
really wish there was some app or some website or whatever that I could just put this information
in and get this information out or collect a bunch of stuff and, you know, display it
back in some way or do something with it or reach out to, you know, any, anything like
that.
Try it.
Like spin up a Claude or a Cursor and, and you know what, pay for, like pay for like one
month of the service, right?
Cursor is like 20 bucks or something.
And I think Claude's like a hundred, you just pay for it for like a month and just to see
if, if based on that investment in, you can do this because I, I bet you can do a lot
more, a lot easier than you think you can these days and cut those dependencies on things
that are mostly just frustrating and poor experiences by just a little bit here and there.
Yeah.