315 AI WILL take our Jobs! with Bryan Finster

This is the Agil Fodds podcast, man, it's fun.

Yeah, I'm Brian Fenster.

I'm going to software developer for three decades or so.

I'm mostly doing enterprise supply chain software,

and then also later on platform.

I've got a personal mission about how do we make software development,

suckless. My focus is how do we improve the supply chain software

developer in the end, because if we can remove the waste

and vein from that, we'll have far fewer sleepless nights.

This is the first episode of a series with Brian Fenster

about how AI will take our jobs.

Welcome, Agil Fodds, if you're listening to this episode

with while nervously glancing at your laptop, wondering,

yes, if chat GPT is about to write your performance review,

well, you're in good company, because today we're diving headfirst

into a topic that's keeping tech workers up at night.

AI, taking your job.

And to help us navigate this digital existential crisis,

I've got the perfect gas.

You want to guess who?

You didn't guess yet. It's Brian Fenster, a DevOps guru

who's spent over two decades in the trenches building mission

critical systems for enterprises so large they have their own zip codes.

Brian has seen more technology revolutions than a carnival tilt world

and he's the ranter and chief that inspired minimumcd.org,

co-author of modern cybersecurity tales from the near-distant future

and currently works at defense unicorns solving software

supply chain problems that would have make a Rubik's Cube weep.

So when it comes to AI disruption,

Brian's got some thoughts and spoiler alert.

He thinks the robots aren't coming for your job.

They're coming for your bad habits.

Let's talk about why that actually might be the best thing that's happened

to software delivery since someone invented the Undo button.

Brian, welcome to the show.

Thanks. I'll say that the one thing that wasn't hyperbole in there

was I actually have worksplaces where they had their own zip code.

Take it back, Brian. None of it was hyperbole.

Come on.

I don't have to go spell check hyperbole now.

I don't know what I put in the show notes, but anyways.

Brian, what is this AI stuff good for?

There's a lot of noise out there.

There's a lot of marketing about how it's good for everything.

You don't even have to know anything.

You just have to just get it prompting.

It'll do everything you need to do.

There's a lot of noise about how it's garbage.

And all it does is generate garbage items,

driving it protected by using it.

No real developer would develop that way.

I mean, no real developer would use anything other than a symbol or either, I guess.

But like anything else, the truth is in the middle.

I've seen all sorts of people say, well, it's good for prototyping.

It's good for Greenfield.

Well, no, it's good for everything.

You just have to know how.

You know, I use it for adding new features to existing code.

I use it for refactoring existing code.

I use it for trying to debug things that have been really bothering me for a minute.

In fact, now I'll just say, hey, why is this defect happening?

Nice.

Because it's just faster than trying to dig through it myself.

I'm getting frustrated and saying, why is this defect happening?

Because it's usually finding reasons.

Yeah.

And so it's good for everything, if, you know, if you know how to interact with it.

And there's one thing that I'm going to take a little bit beyond AI.

I've been working in the developer platform space for the last almost decade.

And one of the things I learned early on when we were building up platform Walmart

was that building on a really good platform that enables capabilities is pointless.

If people aren't trained on how to leverage just capabilities, this is just their platform

and enables new capabilities.

If you're not out there learning how to use this capabilities, the platform never helps anything.

You know, this whole word about artificial intelligence.

I see people trying to get both sides of this.

They start to act like it's more intelligent than people.

And it's flawless, which isn't actually artificial intelligence now.

You're talking about something else.

So if it's actually, it actually contains a lot of flaws.

Like people do.

I mean, but it's not all a chance.

Yeah.

Right.

I mean, this is a thing is that AI is, there's a, I stumbled on a YouTube channel

that had all of these episodes called of the computer chronicles,

which was a PBS TV show based out of San Jose from the started in the mid 80s.

Okay.

And their whole thing was they were, they were talking about this and this new upcoming thing

of the micro computer in PC revolution.

They're like the four, four of the PC revolution.

And they had a box that plugged between your keyboard and your computer.

Oh, it was spell check and they called that AI at the time.

Okay.

It's, I look at it as it has more raw information than I do.

And so the skill is going to be information out of it.

Right.

Yeah, I'm with you on that.

Some people will say it only contains common sense.

But it turns out the common sense of an Audi mechanic.

I don't have.

So I can talk to it and ask it.

What torque I need to have my wrench set for when I torque the nuts on my hand.

I hear the words, you know, it only all the time.

And it only does this, it only does that.

And it's only doing statistics.

You know, it almost we solve problems.

And I am getting, I'm always, I'm focusing on the skill about how to get better

and getting it to help me solve problems.

Right.

This is a skill you have to know where the flight envelope is.

But I'm always exploring the edges of flight envelope.

Right.

Is, you know, where does it not work or does it work?

And when I hit spots where it doesn't work well, then I go and look and say,

was that because of me or because of the limitation of it?

And then I try to figure that out.

Right.

And sometimes it's me and sometimes it's it.

Like right now, I'm not really getting prompting it on how to correct visualization errors on front end development.

But I'm working on it because I know if I'm getting better and it's describing what the problem is

and what I really want it to look like, it'll do it better.

Hmm.

Yes.

And CSS frightens and confuses me.

So I'd rather have the computer do it.

So the whole concept of it, it's only this.

Okay, well, if you are someone who's building these things and you say it's only this, I'll listen to you.

But if you're someone who's playing with them and saying it's only this,

it could be that you're reading bad technical reporting.

Yeah, yeah.

Because if we all remember back in the day when web searching, when Google first came out and there was Altavista,

there were people who were skillful at finding stuff on the internet that other people couldn't do.

And like I remember a friend who just was a little bit further along that space than I was.

I was like, holy cow, I gotta be just like that next time.

And so I had to work at how to say the nouns that I wanted to in order to get the results I wanted.

Yeah.

You know, how to exclude, how to force and clear it.

I mean, all these different.

It was coding, right?

I mean, getting a good Google response, it was like coding.

You just had to understand the language it is.

Yeah, now everybody has that and they've kind of like moved on from that.

Now I feel like I hear similar arguments about AI.

But you know, Brian, there are some things that isn't that I found that surprisingly bad at,

but wasn't that surprising if I thought about it a long time.

It only knows what people posted about on the internet.

Well, okay, I'm talking about coding right now.

So some coding problems I've found.

Of course, it's quite awesome at anything like a kata like or anything that's like a math problem,

or anything that's a coding problem that a lot of people have trained on.

But sometimes, for example, I was on this legacy.net platform,

late 4.8, which is old.

Nobody should be on it, but people are still on it.

And I was asking it how to refactor some code to make it testable

because it had these entities in there that were talking the database.

And it told me very convincingly some steps that I could do to do that.

But the problem was, is that was the newer version of.net,

which those steps wouldn't work at all on the version I was at.

And so I tried to explain it to it.

I spent a lot of time trying to prompt it,

because I really wanted it to solve this problem.

But it ultimately couldn't do it because, and I thought,

oh, heck, let me look for the answer online.

And I couldn't find an online either because nobody had posted a good example of how to do it.

But there was no stack overflow documents.

And so that's why it actually had trouble.

The only contains information.

The information has to exist.

Yeah.

If you wanted to learn something new,

well, you're going to go and find a way to teach it new information.

One of the problems I was having on, remember, early on,

is one of the jobs I have is my, I lead the team,

response for my company's websites.

And they're all written and spelled.

And so that's why I came out.

And we delayed upgrading the spelled five,

just because our primary workflow now involves a lot of AI systems,

just with things out fast.

You know, some refactoring we have to do to go from the four API to the five API once.

Well, it's felt as a project is really,

they drink way too much jolt.

A API was relatively unstable.

I mean, I was having problems for two years with, you know,

the spell changes not being able to, you know,

it's like, well, we just have to hand code this because, you know,

neither cloud nor change, you know,

know anything about it yet.

And then they would and then we could go back to coding and speed.

Right. And so if the information doesn't exist, then, yeah,

of course, you know, it'll either make it up confidently early on.

See, this is a problem.

I was writing a blog post about almost three years ago now.

To explain why it was that we would never be using these tools professionally.

What tools?

AI assistance.

I was writing a blog post about why these were garbage and we would,

they would never do anything to help developers.

Because the experience I had was, it was like,

it was making up dependencies and JavaScript that didn't exist.

It's like, oh, use this in human dependency.

There is no such in-pm dependency anywhere in the world.

What are you talking about?

Not even use it incorrectly.

It doesn't exist at all.

And the code was just garbage, right?

But then things changed rapidly.

And in the middle running that blog post,

I was pulling up information to give coding examples

while you never do that.

And I couldn't get the bad coding examples anymore.

And so then I figured out, well, okay, these tools are improving.

I should probably try to stay current with the tools

instead of just poo-pooing them because they didn't work once.

Yeah.

The tools were out of the beginning.

And that's, I mean, that's kind of where we are now is that

now the tools are really good.

And they're getting better.

And if people aren't digging in,

I think this is a thing that, as somebody who's,

personal mission is to make developers lives better.

If you are telling other developers I should avoid this,

you are now working against my mission.

And I don't like you.

If you are ignoring the tools,

your life will not get better.

If you are, you don't have to be early at this point.

If you are crossing the bridge,

I would love to help you learn how to do it better

while I'm on that same journey.

Yeah.

I think the key point is,

I mean, I agree that what you said,

the key point is the tools are getting better.

And as whatever complaints I have today,

it's going to be better, you know,

for three months from now or two months from now or whatever.

Each each update, they keep getting better.

Yeah.

Like a software developer who foolishly forgets that the software

that they're complaining about is getting better over time,

kind of baffles me because what about their software?

I can't remember when I first started looking at node.

And the node engine was just,

I mean, if you compare the node five to where we are,

well, at 20, 23, I think at this point,

if you compare the two versions,

node was not very capable in node version five.

I mean, you could do stuff with it,

but it was hurt.

And now it's a lot easier to put together

some robust solutions on node.

The tools get better, guys.

I don't know to tell you.

I think a lot of it is just biased.

People have been told that they're going to be put out of jobs

so they don't want to look at it.

Well, if somebody tells you something's going to replace you,

you should probably do the due diligence to have enough information

to explain why they're wrong.

But why they're wrong is that the tools are not capable.

Why they're wrong is because you still have to be a good developer

to make the most of these tools.

Are you new to Agile or Scrum?

Looking for a fun way to pick up the knowledge to become an Agile team?

Go get the novel, Agile Noir.

The dramatic novel about a project manager

who needs to transform his teams to become Agile

because his life depends on it.

This book is available in the US on Amazon,

in India on Pathy.com.

In China, it's available on my WeChat store.

Links are in the show notes.

Check out the show notes

because we will have value there

that enhances your listening experience.

What are show notes?

Show notes show up right in your podcast player

for your easy access.

You can read them as you listen to the show.

If instead you downloaded the show as an MP3 from a website,

go back to the website where you got that MP3

and there you will see the show notes.

Next episode, more Brian Finster.

Brian, in your personal use of AI,

I've seen a clear cut win

where, oh yeah, this definitely saved me a lot of time

and I'm really happy I used this tool.

Oh, sure.