Roasting Ruby AI Workflows with Obie Fernandez

Obie Fernandez (00:03)
so my head's the same size as you guys.

Joe Leo (00:06)
Yeah, more or less. All right, V, you with us? ⁓

Valentino Stoll (00:07)
Cool, we are live. No, I'm just kidding. ⁓ Hey everybody, welcome back

to another episode of the Ruby AI podcast. I'm your host today, Valentino Stoll. I'm joined by my cohost, Joe Leo.

Obie Fernandez (00:17)
stuff.

Joe Leo (00:27)
Hey everybody, and since we're not live today, I'm just gonna tell you all that I am. I'm wearing the coolest Ruby programming t-shirt you can imagine. Just picture that, that's what I'm wearing. I've upped my game and I don't wear collared shirts anymore.

Valentino Stoll (00:39)
but you've got like the best Ruby shirts I

think I've ever seen.

That's right.

Joe Leo (00:49)
Yeah, that's an actual good Ruby t-shirt. Yeah.

Valentino Stoll (00:53)
Yeah, so we're joined today by Obie Fernandez. Obie.

Joe Leo (00:54)
Ha ha ha ha

Obie Fernandez (01:00)
yeah, so was I supposed to introduce myself?

Valentino Stoll (01:04)
You know, we're pretty

Joe Leo (01:05)
Yes,

Valentino Stoll (01:05)
like relaxed here. think we all know each other. ⁓

Joe Leo (01:05)
that was enough. It's just the intimidating stand.

Obie Fernandez (01:08)
Yeah, I'm Obi-Bernandez, principal

engineer at Shopify and all around, you know, kind of done Ruby for 20 years and wrote books and stuff.

Valentino Stoll (01:13)
Why don't you give yourself an introduction?

Joe Leo (01:22)
Yeah, it's the all around part that we're gonna, that we want to get to, you know, like, it's kind of hard to know where to start preparing for an interview with Obi. First let me check, you still with us? Yeah, all right, we can just get started. I guess the first thing I want to say is thank you for writing the book, the...

Obie Fernandez (01:38)
I think he's gone.

Joe Leo (01:51)
patterns of architecture, ⁓ patterns of AI architecture in Rubybook. About half my team right now is reading it and ⁓ I'll make the other half read it when we're done. It's very good, we're really enjoying it. ⁓ And I guess we'll start there. I how has the process been for you to write it using LeanPub and using Olympia as your guide?

Obie Fernandez (02:19)
⁓ Well, that book was written over a fairly short period of time, late last year. ⁓ Or was it early this year? I don't know. Everything is going so fast. Makes your head spin. ⁓ Yeah, the folks at Lean Pub are great. Peter and Len, ⁓ known them for years. I was one of the first authors on the platform a long time ago. ⁓

proudly written in Rails. ⁓ You specifically asked how was the process. The process was great. mean, it was very inspired of, as everyone can tell now, mean, we're in the midst of one of the biggest revolutions, I think, ever, probably since machine code to assembler. ⁓

Joe Leo (02:51)
Yeah, absolutely.

Yeah.

Obie Fernandez (03:16)
you know, or a sembler to see or, you know. And it seemed pretty obvious to me, hey, you're back. Maybe.

Valentino Stoll (03:29)
Hey there.

Joe Leo (03:29)
baby.

Valentino Stoll (03:32)
I'm back.

Obie Fernandez (03:34)
So I guess you're gonna edit all this stuff. so we were just talking about, so we're in the midst of this huge revolution, right? And like, I could totally see the writing on the wall, you know, in terms of what was gonna happen. I mean, obviously not exactly what's gonna happen. I'm not saying I was like an oracle or anything like that, but I mean, I could tell it was gonna be big and...

Valentino Stoll (03:37)
I'll edit it.

Joe Leo (03:37)
yeah.

Obie Fernandez (04:02)
The last time that I felt like something was gonna be really big ⁓ was probably serverless stuff, and I even started writing a book in that case as well, which I never finished. I'm so sorry if you bought that book on Lean Pub, that I never finished it. But anyway, but I could tell this was gonna be like orders of magnitude bigger, and I was like, okay, I wanna jump on this wave. It's what I've normally done in my career.

Joe Leo (04:15)
It's okay.

Obie Fernandez (04:27)
The angle that I took with it though, was not so much about agent assistant coded, know, or co-pilot or like all this stuff that helps you write code faster. It was the angle and the reason that it's called patterns of application development using AI is ⁓ because I, there's this concept which was recently just validated, ⁓ having a brain fart right now. But one of the main guys talked about software 3.0.

⁓ at the YC conference, like three days ago, ⁓ his name is escaping me, but, but he's, he's talking about specifically what I cover in the book, which is using essentially black box AI components as discrete components in your code that sit there and replace, you know, dozens or hundreds or even thousands of lines of code by being this smart entity sitting in your architecture. Right.

I mean, that's the crux of the book. The book has a couple other, you know, kind of extra things, peripheral things, and it's got a little bit of like fundamental about LLMs and, you know, prompt engineering and stuff like that. But I think the really revolutionary and hopefully somewhat evergreen part of the book and the reason that it's still a valuable book now, six months later, you know, six decades, six years later in AI time is because it's that's still.

Joe Leo (05:49)
this time later.

Obie Fernandez (05:55)
That's still unproven and it's still kind of bleeding edge. ⁓

Joe Leo (05:59)
Yeah, I think

that's right. mean, we dig into that a little bit more with respect to the black box. mean, as the people that are building the box, what does that functionally mean? Like, what is it replacing?

Obie Fernandez (06:17)
I mean, I call it a multitude of workers. I use the catchphrase, multitude of workers in the book to describe this sense that you have the ability to put near human-like intelligences into your code. Like little workers that are sitting there, little clerks, you know, making decisions and transcribing things or passing on messages. Which is kind of retro, almost steampunky, you know, kind of thing. But in practice,

Joe Leo (06:43)
Yeah.

Obie Fernandez (06:47)
I think can be quite revolutionary, right? And there's all sorts of objections you could come up to it, you know, if it's too slow, it's non-deterministic, stochastic parrot kind of thing, you know, it doesn't actually reason. And, ⁓ you know, no one would ever put this in production and security concerns and...

It's like, gee, this is now the third time in my 30 year career that I'm hearing these kinds of ⁓ objections to a new technology. The first time was in the nineties, the nineties about Java. And Java ended up being the biggest thing ever, which I owe my career to, cause I was in early on it. And then Ruby and Ruby and Rails came along. And as soon as it started to threaten the status quo, it's a toy.

Joe Leo (07:26)
That's fine. Yeah.

Obie Fernandez (07:39)
It doesn't scale. Why would you ever use that instead of the traditional way of doing, you know, secure, you know, yeah. ⁓ but the whole doesn't scale thing. It's like, really? At this point in the game, like we're still saying it's something doesn't scale. Like, do we not realize that human civilization is just an ever cycling, growing, ⁓ you know, process of using more and more computation.

Joe Leo (07:44)
And it's not secure, got that a lot too.

Obie Fernandez (08:08)
Yeah,

Joe Leo (08:09)
Yeah, yeah, least recent civilization for sure.

Obie Fernandez (08:12)
I was just looking at an article this morning about ⁓ cornfields in the Midwest that like Amazon's building these 50 or 100-hectare data centers, which are going to be the biggest data centers in the world for entropic. ⁓ So yeah, doesn't scale is not a valid reason to not do things that are really cool and amazing. ⁓

Sorry if you're an environmentalist and, you know, object to the climate change implications of using all this energy and all this water and stuff like that. Like, I totally admit that there are valid reasons to be concerned about that. But those objections aside, it doesn't scale, has proven to be just a red herring whenever, whenever this comes up, you know, a ruby herring, if you

Joe Leo (09:02)
Yeah, I was gonna say that exactly.

I will. Well, so you did bring this up though that I think is important because I think even people that are bullish about AI and the possibilities of AI when they think about architecting with it or for it, the first thing that comes up I think is that it's non-deterministic. And so I guess for people listening and for people who, you

What do you recommend? How do we handle this?

Obie Fernandez (09:39)
Don't think of it the same as you think about your deterministic processes. That's why I refer to them as little workers. This is...

This kind of coding and this kind of architecture is truly Greenfield in the sense that you can now replace operations of your company, which is less... ⁓

is less appetizing. mean, there's different reasons why this is not necessarily appetizing, but I can tell you that to big enterprises, this is very, very appetizing because their biggest expense is human. The operational expense of having humans ⁓ kind of shuffling information around and making decisions, and that is literally what you can replace. And humans are not deterministic either. In fact, humans are highly non-deterministic. ⁓ And yet,

Joe Leo (10:22)
Yeah.

Yeah, absolutely.

Obie Fernandez (10:40)
things function with humans in the loop and sometimes function quite well. Why? Because in aggregate, systems work and can be reliable and can be predictable even if they're not wholly deterministic. yeah.

Valentino Stoll (11:01)
You know,

I've heard a lot of talk of like the comparison of microservices, like, you know, AI being like kind of like the blow up of microservices everywhere. And even maybe act a little bit the same with the encapsulation of the service and things like that. Like, well, what are your thoughts with that drawn parallel? Like is it.

Much like microservices at this point, it seems more like it's something completely different to me personally.

Obie Fernandez (11:36)
That's a good question. I mean, it certainly can be. I could imagine how they would be. The thing is that right now we seem to be in the part of the cycle that is where microservices are shunned. And it's all about monorepos and monoliths and bringing things together.

because I think a lot of big organizations had a bad time with the complexity and indeterminism of ⁓ large microservice deployments, which I find to be a shame because in the right hands, with the right teams, microservices, architectures, ⁓ evolutionary architectures featuring microservices, I think are super, super powerful. ⁓

Valentino Stoll (12:28)
mean,

Amazon's doing pretty well.

Obie Fernandez (12:31)
Yeah, yeah, I mean, I was there, I've talked about this in other podcasts, but I mean, I was there when, you know, I thought works where some of these things were getting fleshed out and being experimented with at clients. ⁓

notably at some finance, you know, big finance ⁓ operations, know, like hedge fund kind of operations and things like that where microservices were being, you know, kind of written and shipped almost daily. And they lived and died based on performance and things were not upgraded. They were just replaced. You so like you had a thing.

and you have a better way of doing it. And so you put that new thing out there and start giving it some traffic and let the other one die out. And sometimes those old microservices were not even necessarily decommissioned for a long time. It's a completely different way of thinking about your big ⁓ architecture. could we see something like that emerge with... ⁓

with AI, mean, yeah, if you have systems that can be decomposed into agent behaviors and ⁓ where you can have components of your system acting as microservices and I mean, it's really, really easy to anyone who's got a grounding in like original OO, small talk style OO and message passing and that sort of thing.

It's like, wow, now the messages are literally messages. They're literally language messages being sent. you know, and in the talk that I give to go along with the book, on the slide where I talk about collaborating ⁓ AI discrete components, I put little speech bubbles to represent like, you know, user ID, blah, blah, blah, has not logged in in a while.

Joe Leo (14:10)
Right.

Valentino Stoll (14:11)
Yes. ⁓

Obie Fernandez (14:34)
you know, as a message that's going out and then you have a, guess you could say like Valentino mentioned that microservice, you know, which is an LLM based service that goes something along the lines. I should do something about that, you know, and then looks up, you know, their activity and, you know, tries to make a determination about how best to reengage, you know, that user. ⁓ And then maybe does it itself or sends out some messages like,

Joe Leo (14:59)
Yeah, it's fascinating.

Obie Fernandez (15:04)
Like, Hey, I believe user, you know, one, two, three, four should get a, should get an email, you know, and it's, you got your email clerk, you know, who's got a queue. It was like, Whoa, okay. Need to send them an email. But now instead of being hard coded, ⁓ you know, like a deterministic, you know, procedural kind of thing. ⁓ you can write a pro you can write a system. That's a hell of a lot more.

Valentino Stoll (15:17)
you

Joe Leo (15:25)
Okay.

Obie Fernandez (15:32)
resilience, certainly a lot more innovative in the sense that that mailer who's responsible for engagement emails can be prompted to say, here are tools that give you ⁓ access to the full breadth of information that we have on this user. And here are some guidelines about how to communicate with them. Do your thing. Right. And then it can go and look at, you know, purchase history. can go look at.

Joe Leo (15:57)
you

Obie Fernandez (16:03)
You know login records and you know anything that needs to look at to create a very very personalized message, right?

Joe Leo (16:12)
Yeah, and would trust it, know, promptly about as much as you would trust a, you know, a junior to mid-level employee, you know, to do the same thing. No. Okay.

Obie Fernandez (16:20)
It's actually not true though. I'm going to disagree with you. You

would trust it to the extent that you trust an expert ⁓ marketer.

who has perfect recall, perfect accuracy, and perfect command of the domain of writing engagement emails. Properly done, obviously, for that caveat, but.

Joe Leo (16:45)
Okay. Okay.

Obie Fernandez (16:48)
But AI systems are not nearly as non-deterministic as people think. Right? That's a common misconception. I call that out on some of my talks as well. And I've seen this firsthand, you know, for the last couple of years. Well-prompted, well-prompt engineered, proper context, low temperature, you get very repeatable results well within the margins that you

Joe Leo (17:02)
Well, if you...

Well, that's an interesting thing because although you mentioned that it's fundamental, at the beginning of your book, I actually found it very interesting, even as somebody who does a fair amount of prompt engineering, because you do take a look, you do dive into the details of like, okay, what should the temperature be? What are the parameters that I might want to edit? ⁓

adjusting my prompts or why I'm trying to narrow the path as you as you talk about in your book which I Found to be really interesting

Obie Fernandez (17:48)
This still doesn't seem to be common knowledge. mean, I run into cases. So in my job, I help teams across the board, across the entire engineering organization with their internal AI, you know, jobs at Shopify and still find cases where people have the temperature turned way up on things where the temperature absolutely needs to be at zero or 0.1 or...

Joe Leo (18:11)
Mm-hmm.

Obie Fernandez (18:17)
whatever because you're just trying to get some sort of consistent results.

Joe Leo (18:21)
Yeah, yeah, absolutely.

Obie Fernandez (18:23)
And

a lot of the frameworks and tools don't do you a favor on that.

Joe Leo (18:27)
No, think default out of the box for chat GPT is at 0.7, I think, right? Or at least it was last I checked. Yeah.

Obie Fernandez (18:32)
Yeah, which is great for creative writing and for

having fun with the chat agent. Not so great if you've loaded a bunch of MCP's and are trying to run some workflows.

Joe Leo (18:44)
Mm-hmm.

Valentino Stoll (18:46)
It seems like we're trying to make micro assistance in a way. ⁓ I mean, so what they're kind of two trains of thought here that I see emerging once like, ⁓ you know, these agentic encapsulations where the agents are sort of kind of co-communicating and possibly with an ⁓ overarching agent managing all of their workloads. ⁓ And then there's another.

Obie Fernandez (18:53)
Yeah.

Valentino Stoll (19:15)
kind of train of thought where, okay, it's almost better to have like, ⁓ you know, your orchestration planner and then delegating the tasks to the things without any overarching communication and that then, you know, come to an agreement at the end and collaborate. I know there are different use cases for each.

But there does seem to be a driving industry force of trying to pick one over the other because of the system implications of managing it all. Where do you see the orchestration layer evolving? where are the strengths of each as you start to build things out from Shopify where you have these discrete units ⁓ that are co-communicating in ways?

Obie Fernandez (20:12)
⁓

The approach where you have an orchestrator that's using, that's leading a team with different experts is very, very powerful. And I see, I believe that it's going to get used more and more as the costs come down because the costs do get exorbitant. Especially using Opus 4, which is very expensive, but is in my opinion, the most competent model.

However, what you can accomplish with that approach is extremely powerful, extremely, extremely powerful. ⁓ And I've seen that up close as well. And the reason is fundamentally because you're narrowing the path more effectively than you can with just using one agent. ⁓ So I've been really into using Claude Swarm by my colleague, Paulo Arruda. super underrated and slept on at the moment.

Joe Leo (21:12)
Yeah, we were just talking about that before the show.

Obie Fernandez (21:18)
⁓

Maybe because you really need a max plan or a deep pocketed corporate expense account to use it.

Valentino Stoll (21:31)
Yeah, I've got an impending pull request coming for a llama support. Thanks to ⁓ the, the Claude swarm, which I, this is pretty wild. ⁓ I saw, I saw this Claude swarm, dot YAML file in the repo for roast. I was just like, what is this thing? and I saw you had mentioned Claude swarm, ⁓ somewhere else. ⁓ and yeah, it was just a list of kind of like these members that were going to be collaborating on.

the of the ROS repository looked like, with like a testing person. ⁓ And yeah, I just, I have this little tool where I create ADRs for something that I want to do. ⁓ And so I had it create an ADR for a llama support and ⁓ it just ripped through it all in like 45 minutes without any intervention from me at all. It was pretty incredible.

Obie Fernandez (22:03)
Right.

It just works better. Yeah, it works better ⁓ because of you get ⁓ this collapse when the context gets too big and you sidestep that issue quite a bit with these agent swarms because the main...

Valentino Stoll (22:30)
just ⁓

Obie Fernandez (22:53)
context for the coordinator doesn't get as bloated as it would if it was doing all the work. yeah, if you just have a single agent, then everything that needs to be done is in that main conversational context. And every time that you run into limits, it has to be compacted and then you lose detail. ⁓

Joe Leo (23:00)
Just a single agent, you know.

Obie Fernandez (23:21)
So you're getting screwed on both sides of that extreme. There's a happy middle where it works well, but there are these sad ends where you either have too much context and then the agent gets confused and you lose intelligence. And there's recent, you know, white papers about that as well that we could find and put in the show notes. ⁓ But on the other side, post-compaction. ⁓

you lose a whole bunch of detail and some of that detail might be relevant, right? So by splitting out and delegating to individual experts, you can find this sweet spot where none of the individual expert ⁓ contexts get too big and they're able to retain the detail that they need. And this is on top and above and beyond.

Joe Leo (23:53)
Alright.

Obie Fernandez (24:16)
The other benefits that you get from splitting, you know, kind of like the mixer of experts kind of model in this respect, which is that your team can be composed of agents that have a specialty in a much narrower, again, narrowing the path, ⁓ domain, right? So at one point I made a swarm for, for Roast and I had a Ray expert and the Ray expert was constrained to just having access to the Ray directory.

that's checked out on my machine, but it has full reign of the Ray source code. So it can answer any question really well about Ray, but it doesn't have to do anything else. And truly the process of going through and reasoning and figuring out how to answer the question about Ray is irrelevant to the bigger architectural thing, you know, to the bigger process of trying to get a feature.

Joe Leo (24:46)
interesting. ⁓

Yeah, yeah.

Obie Fernandez (25:15)
So why have

it in the context in the first place?

Valentino Stoll (25:19)
You know what? It's funny. I feel like this introduces the same problems that humans have, right? Like if you give like full domain scope to one person, like then that domain becomes wrapped up in that person, right? And it becomes difficult to tease it out and share the knowledge across the team in the same way I feel like. You know, how do you like kind of try to counterbalance

Obie Fernandez (25:28)
Yes.

Valentino Stoll (25:48)
those aspects of locking in that behavior.

Obie Fernandez (25:54)
Sorry, how do I counterbalance?

Valentino Stoll (25:57)
Yeah, like do you do to try and prevent these agents from consuming the context? So like as an example, some other agent may not have the context that it was reasoning about for how to implement Ray and it needs to use Ray in a certain way. And then it has to go and read about Ray and they all end up consuming more and more tokens on their own, all kind of reading the same material.

Obie Fernandez (26:28)
so, you know, this is all, this is all pretty bleeding edge. I'm not going to claim that I'm an expert in it, but it's, it's the way you are. It's the way you configure your swarm. Right. And I don't think you want it too deeply nested, right? Like you pretty much want to coordinator and this there's gotta be a sweet spot, you know, like maybe there's like a LLM equivalent to a Dunbar number or something, ⁓ you know, where you want it most, like maybe five or six.

Valentino Stoll (26:38)
Okay.

Obie Fernandez (26:57)
collaborating agents or they start stepping on each other or the communications overhead becomes greater than the benefit that you get from breaking it up in the first place.

Joe Leo (27:06)
So this is interesting to me because this is something that we've done. ⁓ For Phoenix, we've built out separate agents that all have discrete responsibilities. But what Swarm does is it abstracts the construction of all of that ⁓ so that we can just write a single YAML file and have it do its thing from there.

Obie Fernandez (27:35)
Yeah, and it used to be more difficult because you had to write the YAML file yourself, but now you can do Cloud Swarm Generate and it'll, you you can discuss the configuring team and it'll generate the swarm config for you. So the other thing I did this with, I don't mind saying this, you know, so I wrote a swarm, I configured a swarm to write a book about roast. ⁓

Joe Leo (27:44)
Yeah.

Valentino Stoll (27:58)
Nice.

Obie Fernandez (28:00)
So, and I probably should do a blog post about that and post the swarm config for it, but I wanted to see what would happen if I set it loose to write a book on roast. I gave it the, so I made a team and it included the lead who was prompted to write like me. And it included, ⁓

like a technical writer, kind of things you would expect, technical writer, an experienced Ruby developer, ⁓ a quality assurance person of some sort, I can't remember exactly what it said, ⁓ and an editor. So I configured it, ⁓ I sim-linked Roast into the manuscript ⁓ directory, and I said, here you go, here's the Roast thing, and go, and set it loose.

Joe Leo (28:38)
Yeah.

Obie Fernandez (28:54)
to write a technical book. And ⁓ almost $50 and an hour and a half later, I had a 500 page draft of the roast book, including everything from the table contents to the preface ⁓ to figures and example source code and well...

Valentino Stoll (29:04)
Wow.

Obie Fernandez (29:22)
thought out and a glossary and like literally the only ⁓ so so i went through it and it was really interesting because it generated ⁓ since part of my profile is like hey ⁓ really approachable writing ⁓ anecdotes war stories you know things like that so obviously it made those up

Joe Leo (29:25)
Well, who did it shout out in the acknowledgments? I'm curious.

Valentino Stoll (29:30)
Yeah.

Joe Leo (29:53)
you

Obie Fernandez (29:53)
So

it was 2 AM and we were debugging production logs, know, a lot of shit like that, which is easily remedied because you can say, okay, go through, find all the war stories and replace them with TK, you know, so that I know to, to, but the other thing, which was profoundly revolutionary and almost newsworthy and I want to blog about it.

Joe Leo (29:56)
Yeah.

Yeah.

Obie Fernandez (30:20)
Is that it took it upon itself to decide that roast should integrate with rails. Because roast is a of configure convention over configuration applied to this particular problem of orchestrating workflows. ⁓ so given that it took that approach. It wrote about that, which is not something that I have really.

Valentino Stoll (30:31)
Interesting.

Obie Fernandez (30:49)
thought about too hard or introduced. And it also did a lot of work on the Rails integration for like how you would write Roast workflows that integrate in Rails using a Ruby API, which again doesn't exist. That's because Roast is supposed to be cross platform and you write your workflows in YAML. And as I was looking at that, I was getting goosebumps because I was like,

Valentino Stoll (31:12)
It was taking on your persona.

Joe Leo (31:15)
Yeah.

Obie Fernandez (31:19)
this actually makes a lot of sense. This is the way that I would do this and this is the way that it should work. This would be really, really good. ⁓ at that point, I said, ⁓ hey, this real stuff doesn't exist, but it's pretty cool. Meaning I chatted with the thing. ⁓ Go ahead and look at the existing issues. ⁓

Valentino Stoll (31:27)
a little creepy.

Obie Fernandez (31:48)
on the repo and create a series of epics that represent what you wrote about in the book that doesn't exist in the code base. So if you look at the latest issues that are filed in the Roast GitHub, ⁓ it should contain a bunch of things related to Rails integration.

Valentino Stoll (32:09)
That's awesome.

Obie Fernandez (32:09)
And then after, and then after that was done, you know, at some point I said, okay. Um, so I generated a PDF of that cause it's all in lean pub. then I put it on a, on a branch, which I intend, know, like a future branch. And then I said, okay, now go through and remove everything that doesn't exist. Um, that doesn't actually exist. And, uh, if it's in the roadmap, meaning it's in the issues, then.

Joe Leo (32:10)
Yeah.

you

Obie Fernandez (32:38)
don't remove it wholesale, just mention that it's in the roadmap sort of thing to get me closer to real. And I fully intend to do as little manual authoring of that particular project as I can. I mean, I've already started going through it and it's, I could potentially have a roast book in a matter of way before.

Joe Leo (32:50)
Yeah.

Valentino Stoll (32:59)
That's crazy.

I think we should put into perspective here, like, Obi, your last book was written in a similar fashion, I would say, ⁓ where you had used AI to help you start to draft the book, ⁓ which I believe took you much longer.

Joe Leo (33:03)
Yeah, you'd write it yourself, yeah.

Obie Fernandez (33:19)
Well, in that case, I was using completion, right? So I wrote the book using Copilot and then Cursor, right? And I was putting chapters into Opus, Opus 3.5.

last version of Focus. And it did cost me quite a bit personally, you know, just I don't, I can't remember what the figures were. But I mean, it costs enough where I was like, Oh, okay, don't want to do this willy nilly thing. like, I literally had the whole markdown of the book in context. And I would say, Okay, this is what I want to cover in this chapter helped me create a draft, and then it would create a draft. But then I have to rework it because it doesn't

know what I know and that model even though was insanely powerful is not as powerful as the new one. ⁓ So it wasn't writing finished.

It wasn't writing finished material. What it would write would not be fully fleshed out.

Valentino Stoll (34:27)
I guess what I'm getting at is I feel like we will know a good sense of the progress of AI by how quickly you can publish a book. ⁓

Obie Fernandez (34:37)
Great.

But ironically, no one will read the book. They'll just have a personal AI digest the book.

Valentino Stoll (34:42)
Right, digest

Joe Leo (34:42)
Read it. Yeah

Valentino Stoll (34:44)
it, get the relevant bits to what they need.

Joe Leo (34:45)
Well, so that actually is that leaves me a question that I've been I've had in my head for a while. I guess what I want to know is ⁓ How much how much code do you write yourself these days? Like where you were writing, you know in Ruby you're writing

Obie Fernandez (34:47)
Yeah.

Very little. Very, very little. Like shockingly little.

Joe Leo (35:18)
Okay.

Yeah, maybe not as shocking as it was even a couple of months ago, but I'm starting to get this sense from lot of very experienced, very forward-looking engineers. So what's your process like today?

Obie Fernandez (35:35)
I think you just have to periodic. So you either have to come up with a, you have to come up with a prompting regime like in your Claude MD. But by the way, I hardly use cursor anymore. So don't use cursor anymore. pretty much use, I use Claude squad and I now have like three, four, five, even six concurrent Claude code instances going, doing different things. Not necessarily at the same time, but they're sitting in different work trees.

Joe Leo (36:04)
Okay.

Obie Fernandez (36:04)
you know,

move at different stages of PRs. ⁓ And it will, so by default, it will cut corners and try to get you to what you're asking for as quickly as possible, even if that means making spaghetti code and God classes with hundreds or thousands of lines of code and things like that. ⁓

And people use that as a, as a reason to say that it sucks or that it's a junior programmer or that doesn't know what it's doing, which is completely and absolutely the wrong way to look at it. The problem is you have it narrowed the path. You know, your, your con your, your prompting sucks essentially, except that it like, we're just learning this now.

You know, there's no shame in your prompting sucked right now because it's literally all of this is being invented right now. ⁓

Joe Leo (37:03)
Yeah.

Obie Fernandez (37:09)
So think of Sandy Metz, think of Uncle Bob or don't, if you prefer not to. But know, clean code, you know, think of Martin Fowler and refactoring. Think of all the knowledge that we've used ⁓ to learn about what it means to produce good code. Okay? So picture that in your mind. You've got bookshelves, you've got YouTube videos, conference talks.

Joe Leo (37:17)
Okay.

Yeah.

Yeah. I am, yeah.

Obie Fernandez (37:38)
Literally anything that has not been communicated to you one-on-one, know, from literally from Ward Cunningham or Sandy or something like that, is somewhere out there in written form or video form and has been digested over and over again by these state-of-the-art models. It knows. In fact, it knows at a level that completely

Joe Leo (37:50)
Uh-huh.

bread.

Obie Fernandez (38:06)
is it knows at a level that is not imaginable to the ordinary person.

Joe Leo (38:11)
Mm-hmm.

Obie Fernandez (38:13)
It knows it arguably better than Sandy, better than any of these individual people. So all you have to do is just check yourself. You know, it's like basically, okay, you're making a mess, realize you're making a mess. And this is why it pays to be senior, because you know you're making a mess. A junior person doesn't realize they're making a mess. But the senior person who's done it before knows they're making a mess. Okay, I admit it.

Joe Leo (38:31)
Yeah, that's helpful.

Obie Fernandez (38:39)
And if you go back and you look at the history of the Roast source code, for instance, it was a mess at a bunch of classes, lot of procedural code, not well tested, not well factored. I took a weekend because we were moving as we were working on a team as a team on it. Several people, I took a weekend of my own time. And I said, fuck it. And didn't do any personal stuff and like took most of a Saturday and like a couple hours on a Sunday to re to refactor the Roast code base using cloud code. Um, by prompting it to say.

Uh, okay. We're looking at that. We're taking a long hard look at this, uh, from the perspective of solid principles, way that Sandy Metz would, you know, critique this code and like, we need to break this up. No more God classes. want everything well factored and tested. want to apply design patterns where appropriate. Let's go. And that involved a lot of hand holding and a lot of monitoring and things like that, but we got through it. And if you look at that, there's a PR somewhere in there in the history where.

Joe Leo (39:23)
Mm-hmm.

Obie Fernandez (39:37)
that refactor is presented and that refactor was also vibe coded. I did not actually write most of that. Maybe I wrote 10 % of it, 20%. I just guided the agent to do things the right way. ⁓

Joe Leo (39:44)
Right.

But that's interesting

because you're using vibe coding in a way that I think most people don't. Like, vibe coding seems to connote something to some people that is negative. But I think the way you're using it is a little bit different ⁓ because you started by saying that you've learned and are learning to narrow the path, prompt correctly, get the results that you want out of the vibe coding session.

Obie Fernandez (40:27)
Yeah, like, vibe coding, if you've got 30 years of experience and worked with Ward Cunningham and Martin Fowler and, you know, people like that, I guess is different than...

Joe Leo (40:35)
Yeah.

Valentino Stoll (40:36)
It's really funny you mentioned the refactoring aspect. ⁓ I've actually started using this as part of my post-generative ⁓ pipeline. So you'll have a session where you're trying to accomplish something and make changes, and it does its usual, like maybe touching on way too many things. ⁓ And then just at the end, you get to somewhere that you want and then just have it say, OK, let's.

let's refactor in place and go through and make the sensible changes and follow these principles like you mentioned. It does an incredible job on that. And I'm wondering like, that the solution to combating the spaghetti code?

Obie Fernandez (41:24)
Yes. ⁓ Yeah. And you can. You can and will, you know, formalize it in the form of slash commands and of, you know, your clot MDs and your cursor rules and things like that. There will be best practices emerging that people use to do this. There's a series of really good. ⁓

just really powerful prompts essentially that you can put together to do red, green refactor, to do review, to break it down, look for technical debt that's sprouting and squash it. And you don't have to type it. You don't have to do the same prompting over and over again. You can codify this into custom commands, slash commands, and cloud code.

Valentino Stoll (42:10)
I was going to say.

Obie Fernandez (42:23)
So.

Valentino Stoll (42:24)
Yeah, and encapsulate

just those specifics of it. Kind of like in its own agent.

Obie Fernandez (42:27)
Yeah, because once you

have some prompts that are working well, you just use them over and over again. Like I said, there's no shame in not doing that right now because like, I mean, literally figured that out. you know, was it yesterday that I was putting them in place? Like, slash commands, you know.

Valentino Stoll (42:31)
Yeah. So.

Joe Leo (42:44)
Yeah. Yeah.

Valentino Stoll (42:46)
That's funny.

Joe Leo (42:49)
Yeah, it is amazing. mean, you said this at the top of the show. It's amazing how fast it moves. ⁓ But I think it also makes sense, you know, the corollary then is that if it moves this fast, then it's, you know, it really is time to jump in, right, for engineers. Right now is the time to make mistakes and figure out ⁓ the kind of prompting and the kind of workflow that works.

Valentino Stoll (43:14)
⁓ So ⁓ which is the best development practice? So we've gone through phases in the industry like test-driven development, behavior-driven development, all these different ways of working through development. Have you found? Yeah, right? ⁓

Obie Fernandez (43:15)
Yeah.

book driven development. Yeah, it's like start a project,

write a book, tell the agent to write a book about it. See what it writes about, go yes, that makes sense, and then tell a swarm of agents to implement what's in the book.

Joe Leo (43:35)
Yeah. Right.

Valentino Stoll (43:36)
Yeah, I you for this.

Joe Leo (43:43)
Amazing.

Valentino Stoll (43:47)
So to that point, have you found that these workflows work better following these specific patterns, or is it just their own thing? Are we creating a new kind of development?

Obie Fernandez (44:04)
I think it's a new kind of development, right? It's gonna...

You know, within sectors, this will not be universal, but within certain sectors of, you know, within parts of the Ruby community and within the community of people who are of, let's say a certain generation, like my generation, your generation, you know, who have the appetite to do things well, right? To write clean code, to, you know, the software craftsmen and craft people out there.

there, I believe there will be an eternal September of.

being able to write clean code like that if you want to. Right? And that may or may not become part of the zeitgeist, it probably won't. You know, the dominant model is younger, faster, more vibey, you know. I had someone...

You know, when I, when I did the big refactor on roast, had someone tell me, why does it even matter? You can regenerate everything instantly if you need to. Like, who cares? Why are you taking, why are you slowing down progress to rewrite things when we can just write it over again? Like, I felt like an old man, but I think I'm right. I was like, no, like we're going to, we're going to do it right.

Joe Leo (45:38)
It is, I do like that as a, like an existential question, right? Like, okay, so it's important, well, you my answer would be, well, it's important for us humans to understand it, to which you would all again say, well, why? is it? What is your brain matter in this? But I still think you're right, but they think that the discussion is probably ongoing.

Obie Fernandez (45:43)
Yeah.

Yeah.

But it won't. It won't matter. It won't matter

anymore. It won't matter any more than it matters whether you understand the intermediary or machine code that's generated by our executables, by compilers when they...

Joe Leo (46:18)
that's interesting to think about. Yeah.

Obie Fernandez (46:21)
It won't,

promise you it won't. And in fact, there are efforts underway, I'm sure at many, many companies, including mine, to ⁓ write tests with AI so that you don't have to write tests anymore. And I'm gonna put a line in the, or my marker or whatever you wanna call it, my line in the sand. I set it here on this podcast first.

Joe Leo (46:47)
All right. yeah. We're friendly.

Valentino Stoll (46:47)
you

Obie Fernandez (46:48)
Okay. And can I curse you guys? Any friendly

Valentino Stoll (46:51)
Go for it.

Obie Fernandez (46:52)
fucking bullshit and it's dead end and it's not going to work for anyone. And actually what's going to happen is that eventually we won't write any implementation code. And the only thing that we will do is essentially write tests, except we won't even be writing tests because it's really just the conversation that we're having that establishes the spec.

Joe Leo (46:54)
Yeah.

Yeah.

Mm-hmm.

Obie Fernandez (47:17)
So come hell or high water, have to communicate the behavior that you want. part, there's no magical brain interface yet that extracts what you want, except maybe there is because I essentially did that with the roast book. So, no.

Joe Leo (47:22)
Right.

No.

Yeah, I know.

Valentino Stoll (47:35)
You see, if we're all logging all of our thoughts

Joe Leo (47:36)
Careful.

Obie Fernandez (47:38)
Yeah.

Valentino Stoll (47:38)
and persona into some machine somewhere, ⁓ you know, it's...

Obie Fernandez (47:42)
I mean,

I'm one of the lucky ones that actually has probably enough of my brain in the base knowledge that I can actually tell it to me. ⁓ Yeah, but okay, for mere mortals, there is no way to just kind of magically get something that you want.

Joe Leo (47:54)
Yeah, I know, I was thinking that. That is helpful.

Obie Fernandez (48:07)
You might get something that's useful, if you literally have business requirements, you have to satisfy them. You have to communicate them somehow. And if you ever want any other humans to be able to review those, ⁓ that has to be captured in some sort of artifact. ⁓ And that artifact is, if you squint, is essentially some form of cucumber.

Joe Leo (48:12)
Right. Yeah.

Yes.

Yeah, it's a test, it's a spec.

Obie Fernandez (48:36)
It's a

spec, yeah. And if you have a spec and you have near limitless resources to implement that spec, why would you let a human try to implement it? There's no reason. There's no reason that you would let a human try to implement it. No more than the ⁓ project requires anyone to know machine code or assembler.

Joe Leo (48:50)
you

Valentino Stoll (49:04)
You know, it's funny, like humans are terrible at optimization, right? Like they're terrible at it. ⁓ You know, we keep creating new ways to try and be more optimal, like in our lives, like in our code, like in so many different ways. ⁓ And we're really just like, fine. It turns out that like, yeah, code, you know, algorithms are better coded, like encoded and then like optimize themselves through math, right? Like machines that we're building like.

They're better optimizers and they're gonna create the most optimal code, the most comfortable. As long as we have these specs like you mentioned, ⁓ I feel like we're just kind of edging out of writing code as we know it. ⁓ So I'm curious what your thoughts are on the evolution of that in the Ruby space seems like we're very well positioned because the code aligns so well with the spec.

writing, right? Because it's so close to natural language. Do you see that kind of parallel thinking? Or ⁓ where is your drop-off point of, I need to hold on to this code for the life of me?

Obie Fernandez (50:18)
to hold onto it, like still wanna maintain it.

Valentino Stoll (50:20)
Like still

want to maintain it, right? Like, cause there will be a point where these, you know, things will be able to maintain code bases of their own, right? Like, and they'll be able to optimize them and make sure that it stays in spec. Yeah. It's already here, right?

Obie Fernandez (50:30)
It's already here. It's already here. It's

already here. So like I have this new project, which is my test bed for what I'm talking about, like not being the implementer, which is this, the dispute board. I have not written a single line of code in that project myself. That is a hundred percent coded by agents. And...

Valentino Stoll (50:46)
Okay.

Obie Fernandez (51:01)
I initially tried to go through and do like full code review of everything that it was doing. And after a while I said, no, this is, it's just too much. I don't have time. It's a side project. You know, it's not part of my core thing that I'm working on. just a fun side thing just to prove a point. Right. And, um,

And then once I gave up, said, okay, well, what I'm going to do is I'm going to write a bin console and then fire up the console and just start playing with it to make sure it works. It does what it says. So I'm basically manually acceptance test, you know, acceptance testing the thing. And Roast has a certain amount of that going on as well, which is why we have so many examples. ⁓ So like pretty much any time that we're

We, the Royal We, writes a feature for rows. Part of that is getting the agent to write an example so that it can run the examples and make sure it works. And then running all the examples functions as a kind of integration suite, acceptance suite. ⁓

But I don't know, I forgot what these original question was, the work, I don't know. I don't know, it depends on the use case, It depends on the situation. And the reason it's all over the map is because in some enterprise corporate situations, by regulation and by threat of losing your job, a human has to review every line of.

code, you know, ⁓ domains that have to do with finances or customer data or PII or, know, things like that. ⁓ And even there, I'm aware, not at Shopify, but at other companies, that there are ⁓ efforts to get rid of that because actually the biggest backlog right now for companies bigger than Shopify, talking like 10x scale. So you can imagine what I'm talking about. ⁓ The biggest problem they have right now,

Joe Leo (52:47)
Mm-hmm.

Mm-hmm.

Obie Fernandez (53:18)
which is actively being discussed amongst us that work on these things, is that the number of PRs is exploding. Because everyone's using coding agents and I've even heard of 5X increase in PR volume, which means you can't human review. It's like the whole system is breaking down. The whole human PR GitHub.

Joe Leo (53:46)
Yeah.

Obie Fernandez (53:47)
look

at the thing, look at the chainset, is breaking down. That part doesn't scale. And you have humans complaining, like, all I fucking do all day is review PRs.

Joe Leo (53:57)
Yeah, which to be fair, they were already complaining about that about their peers and now there's more. ⁓

Valentino Stoll (54:01)
Right.

Obie Fernandez (54:03)
Now they're complaining more, right? So now it's like, okay, well, what's the minimum amount of human in the loop that we can get away with? ⁓ Like I'll give you an example, at Facebook, they have like eight million issues. Like eight million open issues, right? Like, I don't care how big their engineering staff is, they're not gonna get through that. They know they're not gonna get through that, right? So, you know.

Joe Leo (54:05)
Yeah.

Valentino Stoll (54:05)
you

Joe Leo (54:20)
Jeez. Yeah.

Obie Fernandez (54:31)
you can imagine how they're planning to get through that.

Joe Leo (54:34)
You know, I like that as a concept too though, right? Like, cause every single code base, ours included, have more issues than we know we'll ever be able to get to because we know that they are, you know, they're ever growing, right? We can do 80 % of the issues in our backlog right now. ⁓ So some amount are just never gonna get done until we finally do get to them by that time. There's, you know, twice as many issues.

Valentino Stoll (55:02)
You know, I had this thought the other day where it seems like where we're heading is like the point where, okay, you just host a repository and somebody opening up an issue shows a price, like a cost associated with addressing that issue. And suddenly GitHub just becomes an open marketplace for where sponsorships go. Right? Like.

Obie Fernandez (55:03)
Yeah, yeah, probably.

Valentino Stoll (55:31)
It seems like we're getting to the point where just you opening up an issue and somebody tagging it the right way gets it integrated into the project. We're not too far away from that.

Obie Fernandez (55:43)
No, yeah, we're

not too far away from that at all. I there's people working on that right now. I mean, I don't know about the market and the price, but the...

Valentino Stoll (55:49)
Great. And so like what's that?

Right, yeah, not the marketing part, but

ultimately, there's a cost associated with all of it. And so it does count.

Obie Fernandez (56:00)
What

do think roast is being used for? That's the kind of thing it's being used for, People want, and it makes total sense, I'm not saying it's something they shouldn't want, what you want is that an issue, whatever form it may take, pops up and as quickly as possible you have someone on it, quote unquote. That someone is some agents that are figuring out what to do.

Valentino Stoll (56:03)
Yeah, right. Yeah.

Joe Leo (56:08)
Yeah.

Mm-hmm.

Obie Fernandez (56:30)
⁓ But at that point, you know, eventually we just become the supervisors.

Valentino Stoll (56:38)
Now, what I would

love to see is like people's personas, like being the creative engines that also like, ⁓ you know, create all these things as well. So let's say we have Obie Fernandez, the agent, ⁓ you know, scanning GitHub for ideas ⁓ and maybe Twitter, maybe, maybe just like, you know, I don't know what your thinking process, you know, probably having a DJ session, ⁓ you know. ⁓

where your creative juices flow. Maybe it's just like something that, it's an agent that does all of these activities to kind of gather up ideas and then opens up a repository and starts a new project based on all these different ideas. ⁓ then other people come in and they open up issues and maybe collaborate on the issue for, yeah, this would be a great idea, right? Like ⁓ it would be really interesting to see.

how that pans out. Like what actually gets created from all this stuff? And how accurately maybe would it reflect reality? I feel like it would be closer than we might think.

Obie Fernandez (57:47)
I mean, there's a timeline where very soon I become very wealthy because I'm, I can clone, I, I believe that I can lay claim to my, you know, literally intellectual property in the sense of, you know, Hey, you know, if you want me to, if you're, if you're a fan of mine and you want me to, you know, to work on thing, then you can license my personal version of Codec.

Valentino Stoll (57:51)
Right?

Great.

I mean, that

would be great. Hey, I would like pay for OB to come and implement some of my issues.

Joe Leo (58:21)
You

Obie Fernandez (58:22)
Yeah,

and hey, it scales. Yeah.

Valentino Stoll (58:25)
haha

Joe Leo (58:26)
Yeah, this

is like an onlydevs.com kind of situation.

Obie Fernandez (58:32)
I'm sorry.

Valentino Stoll (58:36)
Ha ha ha!

Obie Fernandez (58:41)
I just registered it.

Joe Leo (58:43)
Excellent.

Valentino Stoll (58:44)
You should.

We're not live, so there's no race. ⁓ my God. ⁓

Joe Leo (58:47)
Hahaha

Obie Fernandez (58:51)
Well, in that case, I am gonna...

Joe Leo (58:53)
Yeah.

Obie Fernandez (58:56)
Born here,

Joe Leo (58:57)
That's right.

Valentino Stoll (58:58)
Yeah.

Obie Fernandez (59:00)
It'd be funny if that was available.

Joe Leo (59:07)
V, we should probably let Obi go. We're off and running long in our sessions. I know, it could be.

Valentino Stoll (59:10)
Yeah, I know we've missed on so many topics I feel like we were meant to talk about.

Obie Fernandez (59:16)
No, you don't

have to let me go actually. I cleared my schedule. Do you limit your length or do you let it run like Lex Friedman?

Joe Leo (59:20)
Yeah, yeah.

Valentino Stoll (59:20)
that's great. So we've got another hour. We could split

this up into two episodes, you know.

Joe Leo (59:26)
You

just let it run.

Valentino Stoll (59:32)
We both

have things to do, I'm sure. I don't want to speak for Joe.

Obie Fernandez (59:34)
⁓ okay.

I'm having fun so we can keep it going if you want.

Joe Leo (59:46)
Yeah, I gotta think about it little while. ⁓ I'm ⁓ curious, I brought up the music and I really like music, I had you on... ⁓

stuff on earlier. I was just listening to it while I was preparing for the show. And I am always curious because I look at, I've always looked at software development at its core as a deeply creative process for creative people. And you know, it's a low and behold you find so many people that are.

artists. So you know I'm curious to know like about that artist-developer crossover for you. First I mean how do you see yourself? you see yourself as an artist first, as an engineer first, and what kind of things inform your process when it comes to the art?

Obie Fernandez (1:00:47)
⁓ Yeah, thanks for the... That's a cool question. I do see myself as an artist and I try to bring that to my day to day. I don't like saying that I'm an artist in professional setting because it feels like I like smelling my own farts kind of thing, you know? I don't know, it's a little bit something weird about it. But... ⁓

Joe Leo (1:01:09)
you

Obie Fernandez (1:01:15)
But yeah, I mean, think that's part of what makes it special, right? And like one of the reasons that I stay deeply, deeply committed to continuing to make music, know, no matter what I'm doing or whatever is that it keeps me in that zone.

So, you know, to produce music on a regular basis, you have to be in touch with your muse and without getting too, you know, too hippie about it or whatever. If you want to produce open source, if you want to stay relevant and, you always have a muse, like whenever you're doing these creative things and there's parallels between the music and actually even with the AI stuff.

I've been using Suna really, really heavily to produce. So I love writing lyrics. Like I've written poetry on a personal basis for years and I have, you and I've written tons more stuff than will ever get published. know, my kids are listening to this or their AIs, you know, like go searching, you'll find it and maybe see my slick or whatever, you know, there's good shit in there.

But anyway, but what I was getting to is that with the music world is changing as well, even way more to the extent that people realize. Like I personally am friends with a DJ in my music niche that was at one time a top five DJ in the world and who is still very actively producing and they're using Suna every day for hours a day.

Joe Leo (1:02:52)
interesting. Yeah.

Valentino Stoll (1:02:54)
Why would you not? Just like us, why would you not use cursor or like clog?

Obie Fernandez (1:02:59)
I mean, you know, I would never, I would never dox them or like, would never divulge who I'm talking about. And luckily there's, you know, number of people I could be talking about, but I think it's scary to people. And especially the artists way more than it's scary to your average developer. ⁓ because you are, you know, you, you are lifting, you are using an unauthorized way.

intellectual property that is not yours. It's just that it's a thing that now exists in the world. So it's kind of like it's not it's literally not going to go away. Like this is the timeline we're on. This is the way that it's going. And hopefully this doesn't speak too badly of me, but it's like it's a tool that exists. So. One. You know, I don't feel I don't feel like it's putting

people out of business per se. the world is evolving and jobs come and go and that's the nature of jobs, you know, over historical time anyway. So this is just the latest way that technology is disrupting things. And you know, if you lean conservative, not in a political sense, but if you lean in like, let's keep things the way that they were, you're not going to like that. And if you're lean towards like, fuck it, you know, let's see what comes next, you know, then you're going to, you're going to like it.

And I definitely go that way, right? Like I want to see what comes next. like personally as a, I've been producing a lot more vocal music because I like writing lyrics and I like hearing my lyrics performed. And now I have performers at my fingertips whenever I want it. So, so people are like, how do you do all this stuff? It's like, I don't watch TV and I try to limit my social media consumption and, you know, as much as possible, which means a lot of my spare time I spend on things like Suno and making music.

Joe Leo (1:04:38)
Yeah.

Obie Fernandez (1:04:58)
You know, but when I'm on Suno, I'm basically learning how to prompt engineer Suno and how to produce, you know, things that like give you inspiration. And then in some cases I've published music that is just really lightly post produced, you know, like mixed and mastered, but it's like wholesale to Suno. So I like, have a whole project called Saindra, which is that. ⁓ but for my own music, which I take

a lot more care of, you know, in many different senses, there you have the ability to break, you know, to separate the vocals. And now you have really, really good stemming capabilities. You can break something that you generate with AI into like five, six, even 12 different quote unquote stems, where the individual components of the track, which means that you can integrate them and use them in your normal traditional music making process when you're producing music. And that's what I'm doing. ⁓

Joe Leo (1:05:43)
Yeah.

Obie Fernandez (1:05:57)
And the results are mind blowing. Unfortunately, the long-term effect, I think, is just to further entrench this log distribution, whereas if you were famous before, you're already famous. You're going to continue to be famous. And everyone else will see the value of what they produced even further devalued. But that's literally happening across the board, not just with music. Right?

Joe Leo (1:06:12)
Yeah.

No, I think it's been

happening ever since.

I read about this a few years ago in relation to Spotify. It democratized the ability for people to get music out there. As a result, me as a music listener, I've hit a bonanza. I get to hear a million different ⁓ artists in a million different ⁓ genres. It's amazing. But for the individual artist, your job got 10 times, 100 times harder to stand out because now everybody has access to the same.

Do you see, is this a similar, does this rhyme with that?

Obie Fernandez (1:07:01)
It's a similar thing,

but it makes the situation even worse because the people who are already established, it's even easier to create the output than it was before. So the parts of their job that were actually hard and that actually took talent are getting easier. But again, that's not unique to music or art or anything like that. That's literally happening everywhere.

And you could look at me and you could say that's a case study, right? Because I have every intention in the coming months and years to start producing a ridiculous amount of product. ⁓ have a personal vendetta, know, a quixotic vendetta to prove against Python. Like I want to find every significant Python project that should be in Ruby and have a hoard of agents rewritten in Ruby.

Valentino Stoll (1:07:49)
You

Joe Leo (1:07:49)
You

Obie Fernandez (1:07:57)
and put it out there so that no Rubyist has to suffer the indignity of having to write Python again ever. Yeah, I'm personally offended, but I will be able to do that in part because I already have some, I'm no DHH or whatever, but I already have a measure of popularity to build on as a base, which means that it's exponentially easier for me to get coverage and...

Joe Leo (1:08:03)
Now we're talking. This is why we have to keep this interview going.

Valentino Stoll (1:08:05)
That is so great.

Joe Leo (1:08:20)
Sure.

Obie Fernandez (1:08:27)
but visibility into whatever I do and get multiplying exponential returns on that. It's an unfair advantage, even at my modest levels, because I don't think I'm a super-mis-person.

Joe Leo (1:08:30)
Yes, that's true.

Valentino Stoll (1:08:36)
It's really...

Joe Leo (1:08:41)
But you, nonetheless, you want to press it. I understand.

Valentino Stoll (1:08:44)
You're

also sharing the wealth with all your open source too. In a way, it does have a trickle down effect. In your case, anyway. ⁓

Obie Fernandez (1:08:48)
Yeah, yeah.

Yeah,

I don't want sound like too conceited or whatever, but I mean, I'm trying to stay across the board, you know, any domain that you're in, that's kind of happening, right? It's like happening with influencers, happening with, and it's going to get harder and harder to break through, you know, because of the...

Valentino Stoll (1:09:10)
Yeah, I could see that. You know, it's funny you mentioned

like the breadth of like production value that you personally get like with ⁓ Suno as an example. I remember ⁓ this guy that had created the Echo Nest that Spotify built that now is like the recommendation engine. ⁓ It's just like this brilliant guy. He had like found mathematically how to like split apart different aspects of the audio.

so that you can then take like John Bono's drum pattern and apply it to any song as an example. ⁓ So I wonder like if like there's a parallel to this style of like creation that will evolve out of this where because there's so much ⁓ ease to generate and blend that we'll kind of see like a new like hip hop style evolution of all of this where.

It'll just be so readily available to mix and extract and fuse together things ⁓ that will see a lot more, not only uniqueness, but like, you know, the creativity, like explosion. And, you know, maybe yes, people like the value of it may go down because the quantities increase, like it's a supply and demand kind of thing. ⁓ I don't know. ⁓ Maybe these platforms are the ones that lead the charge.

Obie Fernandez (1:10:20)
Hmm.

Valentino Stoll (1:10:35)
⁓ like what, where do you see that creativity flow? Like turn it, like, are we really looking at a future where creativity no longer like creates monetary value?

Obie Fernandez (1:10:50)
⁓

No, I wouldn't say that at all. ⁓ In fact, creativity, true creativity, is gonna become the only valuable thing in this world.

because until the nature of AI changes, is really only, ⁓ they're only statistical models, right? So they're only going to give you an average, or an extrapolation of what already exists.

Like fundamentally, they can only give you an extrapolation of what exists. ⁓ Because, you know, and if you get philosophical about it, you might say, well, everything already exists. But I don't think that's true. The latent space is unimaginably large, right? And very, very sparse. There's definitely lots of things that don't exist. ⁓ And if I... ⁓

You know, like if I combine a kumbaya with, ⁓ you know, Japanese folk music and, know, give it a hip hop, lo-fi hip hop beat or something like that. Some might say that's creative, but at the same time, that's still just putting together things that already exist. Right. You can do that right now, but that's not super. You know, the value of that is questionable. So that human curation, the creativity of like figuring out.

what is actually appealing about putting that knowledge together is the only remaining valuable commodity in this world, I think, is where we're going. Everything else that is ordinary in any way will get automated.

You know, and that I think will have such a profound impact that no one is even really talking about it in any sort of mainstream way. It is still far outside the overtime window. Like it will completely and utterly revolutionize everything about everyone who's working in any way with information. Because literally anything that you need and that goes on this screen.

can be instantly made. If the cost drops significantly enough for the kind of technology that we're working with now, and I went into this in Las Vegas in the talk I gave that led to the book. ⁓

I said, software is going to be completely different. you need, if you, if for God knows what reason you actually need some new software, you're going to talk to an agent and you're going to say, Hey, I think I need some software to manage my ledger of blah, blah. And it's going to go, Oh, um, that sounds like this. It's got A and B, you know, it's like idiocracy style. Oh, it looks like a, yeah.

Joe Leo (1:13:59)
Mm-hmm.

Right.

Obie Fernandez (1:14:06)
You click

on A, and it goes, cool. You've selected A. After A, you would normally get C or D. You go, I think I want D. And it's not writing specs. It's literally implementing your little choose your own adventure thing.

Joe Leo (1:14:11)
Yeah.

Valentino Stoll (1:14:12)
Ha

Joe Leo (1:14:17)
Yep.

Hahaha ⁓

Obie Fernandez (1:14:31)
Because as long as it already kind of exists, if there's a model for what exists, for what you need, it's already there. It already exists in the AI, you know, the knowledge of what it is that you need. The only thing in that sense, the only thing that becomes valuable are the things that no one has thought of yet that they really need. And for that, I think fundamentally based on the architecture of the LLMs and the way that human thought works, only humans can do that right now.

Joe Leo (1:14:41)
Right.

Obie Fernandez (1:14:59)
You know, and I happen to think that we're probably still a ways off from like true AGI, know, ASI sort of thing where they replace us in that way. ⁓ Which is cool because if you're used to being an ideas person, if you're used to being creative and tapping into the muse and however metaphysical, the more metaphysical you get about it, the better.

Joe Leo (1:15:21)
Yeah, yeah. Yeah, I do.

Obie Fernandez (1:15:22)
You know what I mean? Like

if you're used to dropping three tabs of acid and going totally metaphysical and figuring out what the next thing is and then going and doing it, you are way ahead of the game, my friend. You know what I mean? Because you are already fundamentally operating at the level that will be the only thing of value left in the coming decade.

Joe Leo (1:15:29)
Hahaha

Yeah.

Obie Fernandez (1:15:48)
Literally everyone else that works in knowledge is going to be fucked.

And I don't know what the consequences of that are, but I know where I want to be on the side of that divide.

Valentino Stoll (1:15:56)
So how do we?

Joe Leo (1:15:56)
I'm just taking all this in.

Yeah, I know, Me

too.

Obie Fernandez (1:16:07)
Yeah, we're.

Valentino Stoll (1:16:08)
So, mean,

so what stops people from just cloning the right personas and amplifying them?

Obie Fernandez (1:16:19)
⁓ What do mean?

Valentino Stoll (1:16:21)
Like, let's say if we're talking like, okay, you know, being these most creative, innovative, you know, ⁓ in tune with like your surroundings and the things that are getting created around you, like at what point do you just clone somebody like that and have it corner the markets for you? Like, yeah, I guess what I'm saying is like, it seems like we're

either like heading toward like this crazy monopolistic future or one where like everything is decentralized and crumbled apart because nothing really matters anymore.

Obie Fernandez (1:17:06)
I mean, I think it's gonna be the latter because your average person doesn't care about being creative whether they want to or not. They got bills to pay and they got crushing jobs and they have anxiety and they have kids and you have, you know, all sorts of things that keep them from tapping into the muse. And it does get metaphysical. You literally have to get philosophical and metaphysical to think properly about the implications of what this means.

I don't think there's any other way. That's my opinion anyway. you know, it's like the... ⁓

You know, I don't know. Think of a normie. You know, think of your average person who just works a job and doesn't necessarily give a shit about what it is. You know, they're just paying the bills. How much creativity do they have? You know, how much do they have to really offer in the way that we've been talking about? I hate to say it, but, know, usually not very much, which is really sad. you know, so, you know, we could work.

Joe Leo (1:18:00)
Right.

Obie Fernandez (1:18:07)
We could work towards this utopian vision of the future where the fact that, you know, there's less toil due to these technologies means that the average person can get more metaphysical, ⁓ can get more in tune with their creative side, can ⁓ produce cool shit, you know. I assume most of it's going to be artistic in nature, you know, but, but hey, inventions.

I'm 50, right? I'm Generation X. Like, when I was a kid, it was super cool to want to be an inventor. You remember that? When's the last time you heard a kid say they want to be an inventor?

Joe Leo (1:18:44)
Yeah.

Valentino Stoll (1:18:49)
about my son just the other day. That's true.

Obie Fernandez (1:18:51)
that's good, but it's your son. Most kids

Joe Leo (1:18:51)
there you go. Yeah. ⁓

Obie Fernandez (1:18:56)
have their face too so far buried in an iPad that the last thing on their mind is to be an inventor. Maybe people can want to be inventors again. When we were kids, we wanted to have patents. I know I was a nerd, but it was a little bit more of a widespread phenomenon.

I feel like if you ask an average teenager nowadays if they would ever want to have a patent, they'd be like, what the fuck is a patent?

Joe Leo (1:19:23)
Well, you know, in fairness, we spent a generation or so telling people, you have to learn how to code. ⁓ And that doesn't suit most people either. But learning to code is, I think, as we've just kind of described, ⁓ is not the same as learning to be creative and to get metaphysical. Although there are those of us that do that, it's not, one does not go hand in hand with the other.

Obie Fernandez (1:19:50)
Right. One does not merely become a software crafts person. That's true. Yeah, no, that's very, very true. And it probably goes hand in hand with the dumbing down of the American educational system and the death of liberal education and the liberal arts and things like that. Because without that building block, you're not even aware.

Joe Leo (1:19:55)
Yeah, right.

Right, of what's possible, the kind of thinking that's possible.

Obie Fernandez (1:20:19)
that what's possible you're like what's a muse

Yeah.

Valentino Stoll (1:20:27)
You know, on that note, feel

like we're definitely missing the, you know, the creative educational spirit of like, why the Lucky Stiff era of people, right? Like, ⁓ how can we like start circling back to that in preparation for this new giant change that's about to happen, right? Like, how do we start encouraging our, you know, our kids and the young people like graduating now?

that are getting into this stuff, like, what can we start building now to prepare it for better situation?

Obie Fernandez (1:21:02)
I

reject the basis of the question. We're old. By definition, we don't really get to decide what's cool. The best we can do, I think, to the extent possible, kind of push for...

Valentino Stoll (1:21:06)
We're old.

You

Joe Leo (1:21:15)
Yeah. ⁓

Obie Fernandez (1:21:25)
these foundational building blocks we've been talking about, know, unfortunately, I think it's unfairly kind of cast into the political divide, and at least in the United States, you know, like one side represents liberal values and the other one doesn't. And it's a shame because it becomes a tribalist sort of thing where in reality, it doesn't need to be, you know, it's really about human ingenuity and things that are equally applicable and, and, ⁓

attractive to a Jewish New York lawyer as they are to ⁓ a farmer in Kansas, you know, working on their tractor or, ⁓ you know, a mechanic working on a chopper in Orange County. I mean, you know, like the fundamental basis of creativity of tapping into the muse. Literally, the only reason that I'm not more utopian, you know, utopian and

and idealistic about this beautiful future coming forward is because psychedelics are illegal. It would seem to be that the part of the people don't want the masses to get so creative because they might get ideas about throwing over the current, the way that the world currently works.

Joe Leo (1:22:41)
Right, yeah, getting creative about

how we live and how we relate to one another is also on the table.

Obie Fernandez (1:22:45)
Yeah, exactly. But as

far as giving the kids tools, we already are, right? Like the fact, you know, for better or worse, you know, we all have diminishing attention spans and especially the youngest generations. I saw this when I was trying to teach my son to code. Look, he's a software engineer now. My son Liam, ⁓ you know, he works in React and TypeScript, you know, he's doing well and...

I tried to teach him to code. I had no idea he was going to be a coder because when I first tried to sit him down at a computer, was probably like, I don't know, nine or 10. And like I was working on a command line thing. There was some Ruby runner or something where you're a little man, you made him run back and forth. And like we were writing scripts to do that. We had absolutely no interest in it because it was boring. It was literally boring to write, you know, 50 lines of code to make a man run a little bus, you know.

Joe Leo (1:23:36)
Alright.

Yeah,

left or right, yeah.

Obie Fernandez (1:23:43)
run back

and left and right. ⁓ But now, with Bobcoding, I have seen videos of kids getting super excited about writing video games because you can actually write something that's, you can materialize a video game that is at a high enough fidelity that it actually matches their expectation of what a video game is. The reason that my son, 15 years ago, was not interested in a command line man is because he already had a PS4.

Joe Leo (1:24:02)
Mm-hmm.

Yeah, that's good reason.

Yeah.

Obie Fernandez (1:24:14)
You know what mean? I think so taste your cash would be and to be a little bit more optimistic about it. We already are building those tools so that a future why could come along, right? And, and work on it. So, but, but I mean, there's nothing stopping those of us that are older other than time, right? So like kind of get crazy with it. I mean, I, I, I, I'm, I think that there's a lot of green, there's so much greenfield opportunity right now.

for just completely re-envisioning the way that you go about creating green field software, like creating software from scratch. Why are we doing Rails new? Like, why aren't we doing something like, okay, what do you need? You know, and then just materializing it, but materializing it in that way that we talk, you know, bring it full circle to what we talking about earlier, where we're telling it what we want and it's kind of deciding what, you know, what to make, right? Like that's where you open up.

Joe Leo (1:24:57)
Yeah.

Mm-hmm.

Obie Fernandez (1:25:14)
the opportunities for more democratized software. Unfortunately, the implementation details start to matter less and less to the user. And the implementation details, at least for this stuff that we write is in Ruby, which means that Ruby starts mattering less and less. But also Python and also Go and everything else starts mattering less and less in this new era. So it seems that eventually that's all going to go, you know, be forgotten, you know.

I mean, I remember, I remember when me and some of my friends that were nerds, you know, like took pride in the assembler we were writing, you know, and like compared notes on, you know, how elegant it was or whatever. But, you know, that's just the artifact of its time. You know, it's in the, it's in the waste bin. And I think that the beauty of Ruby and Matt's is nice. So we're nice and all that stuff is, you know, eventually slash potentially very soon going to become irrelevant, right. Which, makes me sad, but

Joe Leo (1:25:55)
Mm-hmm. Yeah.

Obie Fernandez (1:26:15)
Also, what comes next?

Joe Leo (1:26:18)
Well, I think you're

a good example because you're kind of showing both sides, right? With music and with programming, ⁓ less about is it, do I want it to be the case or am I afraid of it being the case and more about, well, it is the case and we're handling it, right, in reality today.

Obie Fernandez (1:26:38)
Yeah, ultimately, ultimately the thing that you shouldn't be is afraid. Because fear is paralyzing, you know, and, and, and fear is the anti-muse, you know, it's like, it's literally the worst thing that you could do right now is to sit there and be like, man, I'm so cooked, you know, like I'm so screwed. This is all over. I'm going to have to change everything and that sucks.

Joe Leo (1:26:58)
Yeah, yeah.

Obie Fernandez (1:27:07)
That's the worst possible approach that you could take because you're completely overlooking how much opportunity there is.

Valentino Stoll (1:27:16)
Yeah, you know, I

feel like what we're missing out on is the early days of, ⁓ you know, the Pixar folks or IDEO, right? Like Tim Brown from IDEO. He used to, I don't know if he still does, he used to hold these sessions, creativity sessions. We're trying to like reset everybody to their childhood ⁓ for inducing the creativity. And I feel like that's kind of what we need is like...

a way to induce creativity in people so that they will experiment in ways that aren't currently being experimented in. And just continue to do that.

Obie Fernandez (1:27:51)
I mean, there's really good technology for that, guys.

There's super good technology. It just happens to be illegal in most countries.

Joe Leo (1:27:58)
Hahaha

Valentino Stoll (1:28:01)
You

Obie Fernandez (1:28:02)
My first week at Shopify, a colleague sends me a Slack and they're like, hey, Obi Fernandez, so amazing to meet you. And I thought they were gonna say something like, ⁓ you know, I read your book or, you know, fill in the blank. And they're like, I noticed that you go to Burning Man. Do you do acid? I'm like, not on Slack. What the heck?

Joe Leo (1:28:28)
I know, right? Yeah.

Valentino Stoll (1:28:29)
Ha ha ha ha ha ha ha ha ha ha

ha ha ha ha

Joe Leo (1:28:32)
It's a permanent record. ⁓

Obie Fernandez (1:28:33)
I'm to discuss this topic. ⁓

Joe Leo (1:28:39)
Shit.

Obie Fernandez (1:28:41)
But yeah, no, there are ways to induce, yeah, there's certainly ways to induce creative states, know, through breath work and meditation, you know, and other things. And yeah, that's part of the success as well.

Valentino Stoll (1:28:59)
Yeah, get out there and be creative, folks.

Obie Fernandez (1:29:01)
Yeah, 100%.

Joe Leo (1:29:01)
Yeah.

Obie Fernandez (1:29:03)
One of reasons I like Ruby is to the extent that there can be an alignment between being kind of this free spirit, artistic, creative kind of thing, we have a good overlap with the Ruby community there. Even in really innocuous ways like Aaron Patterson's and all that stuff. I don't know of other programming communities that...

Joe Leo (1:29:24)
Yeah, I totally agree with that, yeah.

Obie Fernandez (1:29:30)
actively encourage and celebrate that sort of thing. think it's part of our legacy with why, you know, so, so, so yeah, you heard it here first. I want to like double and triple down on that myself because I think that's what we need, right? Like it's literally the only thing that's going to matter eventually. Yeah.

Joe Leo (1:29:34)
Yeah, characters. Yeah.

Yeah.

Valentino Stoll (1:29:52)
Yeah, and I mean, for sure, none of, you know, it doesn't matter what you make, you know, don't go, at this point, like it doesn't even matter if it's like productive or fruitful in any way, like just the experimentation process and like trying something out can eventually become fruitful, right? And it can eventually produce something, right? And I feel like a lot of people try and focus too much on.

all right, well, I want to create this specific thing in the hopes that like, it makes my dreams like come true, or maybe it like makes a big business and really like that's not where you should start. Right? Like, if people should just do. Yeah.

Obie Fernandez (1:30:32)
No, you just, you just kind of put

stuff out there over and over and over again. And that's why, you know, that's one of the reasons that it's good to do things like music at the same time. You know, and the reason that it carries over is because with music, you realize that 90 % of what you put out is either crap or no one's going to listen to it. And you do it anyway, and you just keep putting it out and eventually you do get better.

Joe Leo (1:30:54)
Yeah.

Obie Fernandez (1:31:00)
and things start to hit and you start to get breaks and things like that because you get better. But if real artists ship, that will never get old, Like that will never get old. So yeah, let's get our Ruby people fired up with the...

Joe Leo (1:31:10)
Yeah.

Yeah.

Valentino Stoll (1:31:21)
Yeah,

we've got to reinvigorate the strange Ruby crowd to get them out. Get them active again. That was such a

Obie Fernandez (1:31:29)
Yeah,

maybe we should have a strange Ruby conference, know, that's kind of like built for weirdos.

Valentino Stoll (1:31:36)
Yeah.

Totally. I have a very heavy AI track and just see what comes out of it or into it, I guess. Yeah, right. Awesome. It's happening. All right. think we've covered it all here, Obi. We've gone from both ends of the spectrum here, ⁓ all over the place. I think we've covered pretty much every base you can imagine in AI.

Obie Fernandez (1:31:48)
Calm down. Calm down. Let's do it.

Joe Leo (1:31:49)
Yeah.

Obie Fernandez (1:31:55)
All right.

Valentino Stoll (1:32:09)
where it's heading, where it's going, ⁓ what's getting created. I mean, we did miss out on a bunch of your incredible projects. ⁓ You just released Ray 1.0, which has some incredible features in it. I would recommend people checking that out. And also the Desiru, am I pronouncing it right? Yep, the dspy Ruby implementation. ⁓ Soon people won't even need to write prompts, ⁓ just specs.

Obie Fernandez (1:32:29)
Yeah.

Yeah.

Valentino Stoll (1:32:41)
So I have definitely been playing with it and it's been working well. So I look forward to messing around more with that. Yeah. ⁓

Obie Fernandez (1:32:45)
Okay.

That's cool. And roast, roast is the big thing that I'm pushing down. ⁓ and the best thing about roast is they let's see. I, so I'm seeing a pattern emerge with roast. need to blog about it. So a lot of the production roast usage that I'm feels like it's like 80 % not AI tasks in the workflow. So, so it's, it's a little bit more workflow orchestrate like typical workflow orchestration and scripting than you would think.

Valentino Stoll (1:32:51)
and Rose. ⁓

Joe Leo (1:33:09)
Ha!

Valentino Stoll (1:33:10)
This thing.

Obie Fernandez (1:33:16)
except you can easily sprinkle little bits of prompts and AI smarts into it, which makes all the difference in the world. You know, so, so that is new to me. I didn't realize that was going to be the case. Like we, we, we kind of, like, realized that we needed to put inline bash prompts, like, you know, basically a command. So, so we added that and then we started using it and then we started using it more and more and more. And now I'm realizing that actually what

Joe Leo (1:33:29)
Yeah.

Obie Fernandez (1:33:45)
what's really valuable is being able to like easily do a script and put prompts in to insert prompts in that script in a methodical way. Is that interleaving of deterministic and non-deterministic tasks? Because a lot, when you're talking about repeatable workflows that you need to run at scale or that you need to get the same results, kind of results over and over again, ⁓ you have this process and there's a quote from my colleague, Schmidt that, ⁓

Joe Leo (1:33:50)
You

Obie Fernandez (1:34:14)
Um, I don't have in front of me, but it's basically, goes, you know, you can start with a big prompt and then you start to decompose it and you start to pull the parts out of it that don't have to be AI. And that's the power that roast gives you, right? It's like, can easily pick apart a bigger prompt that, and you don't even have to start in roast. Like you can just create your workflow, like, and give it to cloud code or create as a command as your way of prototyping it.

Valentino Stoll (1:34:26)
Yeah.

Joe Leo (1:34:27)
okay.

Obie Fernandez (1:34:42)
Or you put it into roast itself as a big prompt and then you run it and you kind of get the top of it running. And then you start decomposing, you you start picking it apart and almost everything that I've seen includes a step where you go run this command. Anytime that you're saying run this command, you probably don't have to waste tokens on that. You know, you can probably, you can probably tell the previous step to prepare the parameter.

Joe Leo (1:35:11)
I see.

Yeah.

Obie Fernandez (1:35:11)
that go into the command

and then you can literally ding, ding, run the command like in a script way. You don't need the AI to generate the command that you're going to run. You just let the AI provide the input parameters, the command loop parameters, whatever. A lot of these things involve GitHub. So you're quite often just using GH command line. Because the problem is if you let the agent generate the GitHub,

Joe Leo (1:35:16)
Yeah.

Mm-hmm.

Obie Fernandez (1:35:42)
It's going to work while you're developing it. It may work, you know, in low volume, but you start running at tens of thousands of times a day or an hour, it's going to fail. It is definitely going to fail. It's not going to work. So there you go. That's in a nutshell, you know, why Roast exists and why I think it will be successful. Cause that's very, very needed right now.

Joe Leo (1:35:50)
Yeah. Yeah.

Valentino Stoll (1:35:52)
Ha ha.

Yeah,

I'm looking forward to messing around with the Claude code aspects of it as well. ⁓ That it's like integrated into the workflow mechanisms. That's going to be super valuable. I'm already thinking about a ton of stuff.

Obie Fernandez (1:36:19)
Of course, you know, it's all not going to matter in the long run because of the bitter lesson, probably. You know the bitter lesson.

Joe Leo (1:36:29)
What's that?

Obie Fernandez (1:36:30)
it's like basically that every time the AI takes a leap forward, that kind of like obsolete all these work around. ⁓

Joe Leo (1:36:37)
Yeah, yeah,

we talked about that on an episode not so long ago, yeah.

Obie Fernandez (1:36:42)
Yeah, because

essentially everything I just described for last 10 minutes or whatever regarding roast is essentially a workaround for the fact that you can't just give a workflow definition to the agent expected to do it every time. So sorry, I take it all back. Fuck it. ⁓

Joe Leo (1:36:55)
Yeah. ⁓

Hahaha!

Valentino Stoll (1:37:03)
All right, Obi,

we're gonna have to have you back on to dig into some more, ⁓ you know, ephemeral topics and metaphysical ones. You know, I feel like we can talk to you about all different kinds of things all day long here. ⁓ Yep. ⁓ Is there any imparting wisdom you wanna give or projects that you're keeping an eye on that you want folks to know about?

Obie Fernandez (1:37:10)
Yeah, anytime.

It's always a pleasure.

Cloud Squad is awesome for letting you quickly run Cloud Code and different work trees. I barely even knew work trees existed before a month ago. Turns out they're a thing. There's this, I think it's open source called Cloudception. It's kind of like a storm sort of thing, but it hits a registry of don't know how many thousands of agents that.

Joe Leo (1:37:40)
Hmm.

Yeah.

Obie Fernandez (1:37:56)
this guy, like link in the show notes probably. ⁓ Yeah, other than that, I'm not so sure. I mean, I'm going to be giving a lot of talks in the near future. ⁓ I'm speaking at Yoruko, I'm speaking at Ruby Africa and Nairobi. It's mostly gonna be on these same topics that we discussed on Roast, et cetera. ⁓ Not speaking at Railsworld, but I'll probably give a lightning talk, cause I'll be there.

won't be in philadelphia for the last rails conf because they didn't ask me to keynote to fuck them

Joe Leo (1:38:35)
Yeah, I'm

Valentino Stoll (1:38:35)
You

Joe Leo (1:38:37)
thinking right now of somebody you could have replaced.

Obie Fernandez (1:38:40)
Yeah. But well.

Joe Leo (1:38:43)
But, alright.

Obie Fernandez (1:38:47)
It's all good. Thanks for having me on the show. Yeah, I really appreciate it.

Valentino Stoll (1:38:49)
Yeah, I mean, you got it. Yeah, I mean, for everyone

Joe Leo (1:38:50)
Well, you'll be missed. Yeah, I appreciate it.

Valentino Stoll (1:38:55)
else listening to, Joe and I will be in RailsConf in Philly. come say hi.

Joe Leo (1:38:59)
We'll be there, it's cold

comfort for you OB fans out there, but we'll be there. Yeah. ⁓

Valentino Stoll (1:39:03)
Yeah.

Obie Fernandez (1:39:06)
Right on.

Valentino Stoll (1:39:09)
But yeah, thanks for coming out, Obi. We appreciate it.

Joe Leo (1:39:11)
Alright, thanks Obi. ⁓

Obie Fernandez (1:39:13)
Yeah, you're welcome.

Appreciate it.