navarre.aiClarisTalk AI · the FileMaker AI podcast
Are Developers Becoming AI Managers?
Cris Ippolite is in San Francisco for AI Dev Day 2026, joined by Chris Moyer, Vince Menanno, Marcus Swift, and Kate Waldhauser for a live conference debrief. The conversation captures the mood coming out of the event: software development is changing fast, but the group is not buying a simple “developers are dead” story. Instead, they talk through a more nuanced shift where developers increasingly become managers, editors, architects, and reviewers of AI-generated work.“Developer as manager” – is this a destination or just an awkward transitional phase? The panel compares today’s AI tooling to
Apple PodcastsSpotifyYouTubeAll episodes
Transcript
Transcribed by machine (whisper, large model), 5,672 words. Names and product terms are sometimes misheard; the audio is the record.
Hey everybody, welcome to a very special edition of the podcast. Yes, we are in studio. Yes, we're live. And oh my God, we got a ton of guests here. This is going to be so much fun. But the most important thing is we're in San Francisco, California, here reporting from AI Dev Day 2026. So first I want to introduce who we've got on the panel. We've got a couple of different rounds. We're going to start off with a little conference debrief. But I'd like to introduce our guests for today. You're going to recognize all these faces. First of all, Chris Moyer, you know him from AI News fame. Glad to have you join us here. And Vince Manano, thank you for joining us as well, too, all the way out from Florida.
Marcus Swift, also, you're joining us here from? Nanaimo, British Columbia, Canada. That's what I was wondering. There we go. The international podcast here, actually, just so that we're clear. And then Kate Walthauser, who we, of course, did our last podcast together in the studio in Austin. So welcome to you from Austin. And so the reason that we're all together here today is because we all attended AI Dev, not day, because it's actually two days now. And this is the third conference that Deep Learning AI has put on by Andrew Ng. And I happen to think it's one of my favorite AI conferences. They're generally pretty hit or miss. This one I enjoyed quite a bit.
The first time they went with two days, they opened it up to a much bigger group down at Pier 48 in San Francisco. I think it's about three times bigger than the first one that was here a year ago. And even the one in New York. So what we're going to do today is kind of share everybody's perspectives. On the sessions, we only went through one day of it so far. But I think there's some really enlightening things. We want to bring it to the community and kind of tie in what maybe some of the relevance might be. So I think what we should start off with was the keynote that, frankly, I missed. So I will be the one learning here. So why don't we start off with the keynote?
So the format of the keynote was like four or five. It was just like a general session with multiple speakers. Yeah, yeah. And they all, you know, had their topic. They flowed into each other. But right in the middle, they had a panel of. Of four speakers. And so I'll back up. I'm a developer, but I also run a number of businesses. And I work a lot with the business community. And we've been hearing a lot about the fear of what AI is bringing to the table and what's going on. Yeah, we got to talk about that today. Yeah. One of my big thoughts has always been, you know, whenever we've seen major tech evolutions, the floor definitely rises. You know, and jobs get lost and things change.
But we don't talk about right now, which is where I've been spending a lot of. My time thinking is the ceiling also rises. And so where is the ceiling rising? Where is this whole industry shifting to? So as business owners, we can start to prepare for that. We can start to find the opportunities and the problems we're going to need to solve at that at that ceiling space. And so as I was listening to this panel, one of the things that really hit me was that doing is going to get automated. So, like, we're talking about everything from coding, fingers on the keyboard, coding. That's already starting, right? That's done. Not doing anymore. Right, exactly. And then but like actual work, like knowledge, work or office work or whatever.
I think that's that's that's also going OK. And and then as robotics hit, even physical labor jobs are going to also start to be affected. So I really start to see over the next number of years, we're going to see doing getting automated. So then the role of the of the human being in that space was what really hit me out of this panel was that were the entry level job is going to be managing. Is going to be how do I I'm going to be given a problem set or a new worker is going to come in and say, this is your problem to manage, not to get it done. Your job isn't to do the doing. Your job is to manage either the bots that are going to be doing that piece and you make sure that they work, that they're coordinated, that they're getting the job done well, that you're defining the problem.
Because one of the things we're learning right now is you can't outsource thinking, but you can outsource. But what I'm starting to conclude is we can outsource. We can outsource doing. And I think we're going to be able to do more and more of that. And so then as a business owner, it's it's how do we start to work with and prepare for our entry level people and training them and getting them ready to be more managers? Because I think the entry level is going to be managed. So there's a two tiers here. We've got managers. So so doers become managers. Correct. And then entry level. What's the opportunity for entry levels become the managers managing is the entry level.
So my question is, what happens if you don't want to be a manager? Well, my only response to that is what happens if you don't want to drive a car and you want to stay with the horse? There's still space for that. That becomes very, very specialized, doesn't it? Right. Well, yeah. But see, what if we're not? Here's my here's my theory is what if we're not at the car yet? What if we're or what if we are? But what if we're at the car that they developed that didn't have windshield wipers, didn't have, you know, didn't have, you know, all these things, you know, brakes, brakes are important. There was no there were no horn was just or what if or what if we're not like what if we're just somewhere in between?
Like I still like my philosophy is that we're still really, really early. So being a manager might be a stepping point to where we're headed. And it's it's where we are now. But is that what the thesis was in the keynote? Is that what they were saying? That no, this is certainly piece. This is that like shoes were dropping for me. I don't think that was their thrust, but it was like I was coming with this lens thinking about already. Where's that ceiling going? And I want to agree. I don't think we're there yet. I think we are in some spaces. I think as developers and coders, we are more and more being pressed around. Should I be doing the coding anymore? Is that the right space for me?
And more and more, it's going to be the answer is going to be no. Well, the interesting thing about the timing of it, like so Chris and Vince, like the role that we all play, is we're intermediaries between the technology and the customer, right? So are you guys already hearing from your customers? Is the tech there already? Like, is it should we jump in? I mean, first of all, are you getting inquiries? Actually, I'm very curious, Chris, how many of your customers are coming to you proactively asking about AI for starters? There's definitely interest. Heidi and I had an interesting situation a few months ago. Actually, another FileMaker developer in Michigan came to us because they were getting asked about AI by their customer.
I think they sell hot tubs. They're something like that. They had five stores. And they had this whole laundry list of stuff. Like, I wanted to do our 401k contributions and calculate payroll and stuff like that. I was like, that's all real deterministic stuff. Bad choice for this. That's important, by the way. Let's put a pin in that, too. One of the use cases, though, was reviewing sales calls. You know, that's a lengthy process. There's hours of calls. And their provider actually had a thing where it could summarize the sales call and say, here's the points they hit or do the analysis for them. If they say, make sure you mention the promotion or whatever it is that they wanted to do coaching on.
And so there was one great use case. So we thought, well, our work here is done. We've told them, you have deterministic stuff. Hire your FileMaker developer to add some workflow logic in. And then just last week, we talked to the same developer again and said they got open clause and stuck Mac minis in all the stores and just YOLOed the whole thing. Didn't listen to a word of advice. Well, at least hopefully you're not culpable. I don't know what to take away from that. You didn't sign anything, did you? I don't know what to make of that. It's like open clause is probably not the best fit for someone who's not technical and doesn't know what they're doing.
And they just did not care. As a bit of a, first of all, great framing for is AI there yet? It might be too there in some of those cases. And great framing for this episode. We're going to finish off the last segment with some hot open claw talk. And so for anybody that's not familiar, we'll get into defining what that is. But I think for the purpose of this conversation, that's the most aggressive. AI there is basically. Right. So when we talk about is AI there yet? Well, the combination of maybe not totally and it being completely in YOLO mode is probably quite dangerous. But Chris, this ties in as well to, you know, the owner's job was to manage the AI.
They didn't do a good management job there. Right. They didn't use the right tool to get the job done. And that's, I think, coming back to more and more of the skill set that we need to be learning and we need to be providing solutions for. Is that, you know, you're not supposed to use a sledgehammer when you need a hammer. You're not supposed to use a hammer when you need a sledgehammer. And that becomes management tasks. But I think that's the interesting role that all of us here play is we're the ones that know the difference between the hammer and the sledgehammer. Right. Like Vince, are you having these conversations with your customers, too? Are they coming to you proactively?
And are you do you find yourself responsible for determining, like, what AI can actually do? I mean, to some level, I'm dabbling in it and I'm learning it as everybody is. And it's growing very quickly. But there's a lot of other people on our teams that are just doing a lot more in that space. And so that's that's happening. There's there's definitely demand. There's definitely interest, et cetera. I think going back to the to the keynote or the or the panel discussion, it was interesting to see that it was kind of like, you know, very positive oriented. It was eight, nine and ten. Like when they were gauging. What was the question? The whole team. The whole idea of the thing was what's the future of software engineering is like, is it dead or what do you think?
Oh, and so for humans to be software engineers, it was sort of a doom index, like on a scale of, you know, one to ten, with the ten being the future is bright and one being we're all doomed. It was very high, from like seven to ten. Oh, really? It was the gamut of four panelists. Was it? Well, and Replit CEO was ten out of ten. Yeah, yeah. The one who was. Who would. So Replit is like automated coding and he was like, absolutely. AI engineers are software engineers are. Well, he was speaking to his book, though, in fairness. Right. I mean, he needs that to be true. And I'm not saying it's not, you know, Kate, you and I had a conversation the last time we did an episode together where we were talking about like the doom scale, like, you know, where you kind of put yourself on there.
Let's switch it up a little bit. Where do you see software coding on as far as like done being ten? Is that what the framing was? During the panel? Software is just done? No, the higher was the brightness of the future. Zero being the future's doom. Interesting. Ten being there's it's a really bright future. Oh, wow. For software engineers. For people. For human engineers. For human software engineers. Ooh, let's talk about that. Yeah. Yeah. Kate, where are you at on that? I mean, I think I think it's like how we're we're changing how we define a software engineer. Yes. Basically what they're meaning and to the point of the panelist. And I did this. So, so the way I went through the conference today was I use Claude and I, you know, took some notes uploaded, you know, screenshots of the slides.
I was seeing and then had Claude research each speaker, you know, where which company the work for what could their angle possibly be because you realize anybody that's going to go speak at a conference. They're going to have an angle, right? So, of course, thinking back that oh, yeah, all these people would rate it a high rating. Yeah. I mean that that how could if they said one, right? Well, there we go. There's an opposition research. So I think that there could be something that they're saying, yes, it's highly, you know, optimistic for software engineers, but I think software engineer is going to mean something completely different, you know, even if a few years than what we think of a software engineer today, we're just, you know, or maybe it's going to take on a new term.
We do we have a context engineer prompt engineers, right? But but I don't know, you know, because software is now like AI is software like that's that's it, you know, in our current state, you know, we are. Marcus, so this conference was put on by Andrew Ng from Deep Learning AI. He's like basically a celebrity. Anytime you can't get through the exhibit area, it's because he's there and all the fans are taking pictures of him. He got plenty of applause when he came out. Oh, for sure he did. But he also has a very aggressive stance on, you know, when people ask him whether software, you should still learn software. Like young people should still learn software.
He says yes, emphatically yes. So what's your take on that? Somebody new right out of the, you know, right out of the box, should learning software be a thing, a priority? Well, I'm going to go around a little bit and then come back to what you're asking. What somebody said on the panel that a junior developer is going to get more and more complex tasks than a junior would have got five years ago. Because they have the knowledge right at their fingertips, you can give them a more complex task. You don't have to worry about. You're talking about because they're, because they. Are embellished by the tool sets. They've got code or whatever, whatever harness they're using.
And so because of that, you can give them more complexity. And this has been my personal journey. I mean, you and I met a year and a bit ago and I was like a neophyte in this, in this space. Right. And in a year and a half, I feel like I've come a long way. And, and this is what they're talking about, that the, the journey from junior to senior gets compressed because you can start to interact with these models that teach you so much. So do I. I'm a better at languages now than I started a year and a half ago because of how much I'm looking at the code that, that, that the AI is generating. So you take that away. I couldn't write. I'd be really, really struggled to write a basic Python.
You've gone through immersion, like immersion learning, like you would learn a language. You just go to that, that country and. This concept reminds me of when I first got started in the computer industry, I was really interested in HyperCard and I didn't know how to write HyperTalk or all that stuff. And I just. Started taking apart example, HyperCard stacks that already existed. And I learned by deconstructing existing stuff. And I think there's a real opportunity for new software developers to do the same thing, have the LLM build something for you and then pick it apart and say, now, why does this work? And so for those, uh, people who are curious about, you know, what, why does the clock tick kind of a thing?
I think there's a real opportunity for them to learn by osmosis as it were. What's been foundational. For me though, is the knowledge I brought from FileMaker because FileMaker teaches you how to, how to think about data, how to structure data. You, you, as a FileMaker developer know the full stack and you can see that layer and how it interacts with the front end and how the front end, or you don't call it that in FileMaker, but, but that's really what you're seeing. And so that ability to visualize the data has really translated well to now me writing in JavaScript and Python. Right. Um, there's no doubt. Um, I mean, I can share with you my. Observation from like the first couple of years of doing services is that we were just thinking, yeah, plug your data into the language model and amazing things will happen.
And then we realize all it does is amplify bad data. And one of the things, like, you know, the, the FileMaker databases are sort of like the HTML code of databases. Like they're very forgiving. Like you can have a lot of garbage in there and a lot of columns you never needed, or just kind of sort of forget about things that stack on top of it. It'll just still work, but it's not so forgiving when you get down and when you're connecting language models to that. So. I, I think on some level. Like as a FileMaker developer for many years, I think one thing that's really fascinating to me when, especially when we got started with 42 you and you were part of that journey, right?
You meet these people and they come and they get grab, they gravitate towards it because it's like, wow, it's multidisciplinary. It's like, it, I get to do, I just, I get to not just do code, but I get to interact with people. And it's like, so it's not for everybody. And so for some people it's like, Hey, just give me. The inputs I'll give you the outputs. That's that's all I want. You know, like just slide the pizza under the door and, and I don't have to worry about it to interact with people. But I think for some people there's, there's a, there's a beauty and, you know, coming into work and you're working for a Joe's garage one day and you're working for a financial services Institute the next day, you get to be like an actor.
And that's so rewarding because it's like it's whole dynamic, it's different. It's changing. It's it's. It's, it's something that a lot of the people that I've come across who were talented, that were talented in that multidisciplinary thing. And I think that has a big impact on our future because being not just like a coder and being more broad gives us the ability to, you know, have empathy with the customer, have understanding in terms of how you solve those, their problems, et cetera. I think that's, that's going to have a lot more longevity for, for, you know, our background. So there's no question if we're wondering if you have all those skills, right?
Let's just say that's your harness, like your career harness and you have all those skills. Well, then you're thinking about which ones peel away in AI era, which are, which are going to serve me well. And you're talking about human skills. I'm talking about humans. The thing is to Kate's point, like not everybody wants to talk to people and, you know, interact with the customer and gather requirements and, you know, do the people things. There are some slide the pizza under the door folks. And like, what, what future does that hold? So I think that's really kind of where. Yeah. Yeah. Yeah. Where the, why they, I mean, first of all, that was the first topic out of the gate at the conference, right?
So obviously this is on the top of minds of a lot of folks, but there's easily 3000 people there who are building and involved in the industry itself. And also with that in mind, what were some of the other topics that you guys heard bopping around? I would say that last year at this conference, I really started to hear context engineering for the first time. And that was like almost de facto in every one of the sessions. Any other things that popped out at you guys? Well, to answer your first question, sort of another kind of word that was thrown around a lot during the keynote was taste. Oh, yeah. Taste. Yeah. Yeah. We bring the taste. So we're still important.
Right. We get a cookie, you know, that kind of thing. But as far as the theme goes, though, much has been made of the cloud's nice and all in the cloud is useful, but the cloud can go down sometimes. So the best approach is a hybrid approach. And the models are getting small enough and capable enough that we really need to take seriously the idea of having models running at the edge. Hybrid, by the way, edge, local and putting it in the cloud. No question. Several companies, Redis is talking about, let's have our storage being close to where the data is being generated, not up in the cloud. Is there talk about, like, does local equal trust in those? Was anybody talking about that?
It was really about latency. This was about speed. It's like, you know, if you have a data. Lake, some of that data is old and you can mine it for trends and things like that. But if you're trying to find out, is my equipment about to fail, you need real time data and to have it, you know, trucking, you know, terabytes of data up into the cloud and bringing it back down is expensive. So saying, OK, I have this big gov of data here locally on prem, get my AI infrastructure wrapped around it locally. So the latency is really, really tight. And I can get real time determinations of. Whatever I'm trying to figure out. I'm shocked that people aren't hammering the trust thing more, honestly, because that was literally the only reason anybody ever used an open source model in 2021 is because you didn't trust open AI, basically.
Right. And there's definitely advantage to that. Vince, where were some of the. I mean, I attended the last session that was talking about her agents as herding cats. And it was banned. I think it was. And this fellow from from Greece was giving that talk. It just fascinated me because. At one point he had a slide that he was talking about the evolution of agents and how agents evolved for from being like something you you directly interact with to all the way to to the point where they have this kind of network. I don't know if anybody else participated in that talk where they they kind of like a broker of of agents that you would talk to and it would find the right agent that you could talk to.
And then, you know, you would have that communication with all those different. So, yeah, that was that was fascinating to me just to see that whole evolution and where that's gotten and to the point where it's like I could see, you know, maybe maybe where it's heading in terms of, you know, if you're a company and you want to build a product or a service and you would put your agents into the pool or mix and then it brings the question of, like, how good is each agent? You know, it's like almost the same as other people. Right. You rely on this person. Is that person going to do the right thing, et cetera? Where those specialties come from and whatnot.
And again, we're going to have another segment here coming up in just a few minutes where we're going to dive into the claw discussions and we'll kind of explore some of those things. Marcus, what was you mentioned some of the stuff you talked about outside of the opening keynote stuff or some of the keywords? I think another key theme was was memory, memory, memory context, right? Or memory engineering. So we certainly context engineering was a buzzword. Jericho right now they're talking memory engineering was huge. And then we're just starting to talk about harness engineering. I've heard a lot of that, but I think that's the word the phrase to watch for in the future.
But a lot of presenters and booths on memory management find that really an intriguing idea. I will say that my second observation from the services world is like data matters. But then also, are you kidding me? Everyone's like blown away by memory, like statefulness. Like we like we've been living with in state trying to figure out what to do with state for 20, 30 years. Right. In our in our careers. And here this community is like, wait, you mean I can store things and I can retrieve them in another conversation. Kate, your observations from some of the other sessions. Well, you know, so I again, like I was using Claude. So so as I did this, so I did a little.
I know. I know. So Claude and I, we came up with some themes. Right. Of what. But again, this is my personal experience of the conference because we didn't go to all. The same sessions or I was taking notes on certain things. So again, like this is this is the AI of one. Right. But this is my personal, you know, but but again, like I I'm I'm coming from the responsible AI perspective. And to me, this was a conference of responsible AI, but nobody called it responsible. Because the topics were observability, reliability. I mean, there were several sessions security. But but what was interesting to me was going to multiple sessions about, you know, you know, using rag, using embeddings, using.
You know, graph. Yeah. And different ways that are all, you know, either can work, you know, in conjunction or apart to achieve the same goal is like having less hallucinations. Right. Having more trustworthy AI, which, again, is like. And so I've always thought like responsible AI, the word responsible is going to go away. Like this is AI. This is what it is. Table stakes, though, right? Yeah. I mean, from your perspective, is this good that responsible AI has become AI? It's just it's just yeah, it's just this is what we talk about. And it is. It's bringing back in all the determinism that we're used to because you can't have just an LLM put it out there.
It's like you need all the other infrastructure around it. And I think we're still at the you know, there's sessions about, oh, you do this. You could do that. You could do that. And I don't think there's one right or wrong way. It depends on the situation. And then, you know, what tools you have at hand. And then, of course, what budget? Nobody was talking about budget here. Nobody talked about cost. And I was thinking, well, how much is it going to cost me? So I think there's a lot of factors that that are still go into. And we as consultants, right? This is what we have to do is know all the resources we need to know and have all the tools at hand. I think it's a great observation.
And I know that's that's an area that you spend a lot of time in, too. So it's particularly curious having that be. Well, actually, what Claude told you. Yeah. So as we wrap up the segment here, we're just going to do just a rapid round here. What for all the other FileMaker folks, all our friends in the community, what's your top line thought, Chris? What would you tell them about what you're seeing? Out here and at an AI conference, what what is something, you know, some advice or a takeaway that you would give to FileMaker people in particular? One aspect of FileMaker is that it's kind of a joyful thing to work in. You can move so quickly in some areas and there's a lot of fun to it.
The original vibe coding, the original vibe coding. And there's a lot. I mean, Google just ran through all their stuff and they made like a Hindi techno song about organic chemistry and just, you know, just in a world. And so there is a lot of fun to be had just playing with. You know, get back to play and play with some of these tools because there's there is joy to be had there. I love that. Vince. Yeah. Explore curiosity. Those are words that kind of ring in my head. But I thought at some point, just like Chris was saying, we were doing this with FileMaker so many years ago. Like iteration was a thing back then, like layout mode, browse mode, layout mode, browse mode.
And you have like the beauty for the customer is like, wow, you just did that. Oh, that's so, so exciting. But, you know, we. We've dabbled there and AI is giving that to us in a very different level. You know, so I love that. Marcus, your thoughts. Marcus Kirsch: This one stumped me. I'm glad they went first. But it's come together. I mean, when you, when you approach AI in a playful way with your years of knowledge, because you can't outsource knowledge, right? When you're working with AI, you're still in charge of telling it what to do of managing it. When you take all of your experience as your FM developer and approach AI. Playfully, magic seems to happen.
And that's what got me hooked. That's what got me hooked in FileMaker. That's what got me so hooked with AI was it just, it's so fun. But now I come to it with knowledge that, that a vibe coder doesn't have because of my years in FileMaker, which is a superpower. Kate? Kate DiCamillo: I guess, um, oh yeah, this is tough. Um, I don't know. I think the word trust is coming up for me and I think that's maybe trust your gut. Trust your intuition, trust your human intelligence, but also like learn, you know, and I think, I think AI, yeah, it's, it's a whole new world. Right. And, and, um, I guess I think, you know, we all have to move at our own pace with it.
And I think it's, but it's also important to listen to people that have been like, like us, like we're out there, we're listening, we're gaining information and, and there's a lot of experts and there's a lot of, there's a lot of movement. There's a lot of movement to just make it better and better. And like, there's a saying like today's AI is the worst AI you will ever use. And it's like, that's true. You know, that, that, that we're just, we're in a process with it. So Chris, what are your thoughts? Well, I would say, um, you know, if you don't trust AI, maybe you can trust some of your, your fellow constituents here from the AI community who've ventured out and, you know, dove into this space.
The come on in the water's warm, uh, is really what I would let everybody know. So you guys are really appreciate you. First of all, it's been a lot of fun being at the conference with you guys look forward to day two. We've got a whole bunch more observations, uh, and, uh, and then next we're going to put together a segment on, um, uh, open some hot, open-claw talk. So thanks a ton guys. Uh, look forward to another day of the conference with you guys, appreciate this.