Judy Warner (00:00.994) Hi, Colin. Thanks so much for joining us again on the ecosystem. Before we get started today, why don't you take a moment and introduce yourself and tell us a bit about your background and then we'll dive a bit into what Xeno does. Collin (00:15.783) Absolutely. Well thanks for having me on, Judy. It's always a pleasure. My name is Colin Stoner. I have an electrical engineering background. I've been a a double E consultant for about 15 years, designed around 250 circuit boards in the last seven or eight, and I run an EMS shop locally. I also co-founded Zenode, which is an electronic components intelligence AI. Judy Warner (00:39.16) So tell us a little bit more about what Xeno does specific to, you know, our audience who are engineers, like what problem you were you trying to solve as an engineer and what specifically does Xeno do for engineers? Collin (00:57.223) Yeah, so doing a lot of engineering is always a lot of research. and there's a lot of work that always felt like it could be automated, but never really could. And then when AI became, you know, when it came out in 2023 in kind of its most recent iteration, it kind of became clear to me that perhaps you can automate some of this work now. I I spent a huge amount of my time finding data in data sheets or searching for that little footnote while I'm working on a board. you know, why why is the power not coming up on this rail when I expect it to? There's a tiny little footnote on page 372 that says, hey, by By the way, if this thing is turned on, this happens and so, you know, that's my problem. Finding that thing could be two or three days, you know, of poking around at a board, but if we could surface that right away, I could save myself all of that time. So that that was kind of the initial reason that we started Zenode. And we found so many issues like that. they're not the big sexy problem, you know, let's generate a circuit board from scratch. But they're the ones that I spent most of my time on. so that's what we started Zenode for. Judy Warner (02:11.69) Yeah. Judy Warner (02:15.245) So you mentioned early that, you know, you you thought these things should be able to be automated. And I think that's true of a lot of younger engineers that grew up with software. It's like, why is this so hard? and so manual and it's and I think it's young engineers that are ultimately gonna help solve this problem. But it wasn't gonna be chat GPT in twenty twenty three. And so I think there still lives a lot of skepticism around that and push back. I certainly get it here on the podcast. So I want to be very frank and honest with our audience. So between say 2023 and those, you know, basically using something like ChatGPT as a chat bot, what's happened in these last three years and when you talk about the problems you were talking about, like a power rail issue, how's that changed and how has it affected the way you and your team have developed Xeno? Collin (03:22.642) Yeah, yeah, it's it's changed so much, even in the last year. it's it's honestly completely different today than it was a year ago today and the difference between twenty twenty three and now is staggering. So if you tried ChatGPT in twenty twenty three, you know it was a cool demo for a lot of people, but not earth shattering or certainly life changing. what I saw in it was a a glimmer of a possibility. and in the last three years that possibility has become some kind of reality. So in 2023, it really was a chatbot. You put some text in and it gave you an answer from its own memory in one shot and that's what you got. And so it was prone to misremembering things or hallucinating, just giving you wrong information. And so for most engineering workflows it was interesting but not useful. what's changed in the last couple years is that it's not a single shot anymore, and it's not just regurgitating something that it was trained on. It's trained on how to go research, how to find the data that you need, how to surface that data, how to self-correct. And it doesn't just output text to you. it knows how to use tools, it knows how to write a Python program. So if you need to transfer one spreadsheet to another and change the Judy Warner (04:39.16) Mm. Collin (04:55.604) form, it's not going to go read the one and then manually write the next one. It's gonna write a program. Judy Warner (05:01.006) Mm. Collin (05:02.728) To do it so that way it's perfect. It's not, oops, I hallucinated this one value in 300 of them. and so because they're setting up systems, they go back and they have feedback, they look at their work, read it, and confirm it. you know, where did I get this information from? Is that actually what it says? Judy Warner (05:03.863) Right. Collin (05:26.116) That's the difference between before and now and they've been trained to be good at that sort of thing. and so the it's a world of difference, even in the last couple you know months, being able to maybe go read a data sheet at the end of last year to we can reliably go and pull pretty much anything out of a data sheet, and it's not an issue anymore. And so now it's not can we do it, it's now that we have it, what do we do with it? Judy Warner (05:42.766) Mm-hmm. Collin (05:55.953) That's where all the exciting stuff happens. Judy Warner (05:58.776) So the pushback I get all the time, and I understand it, and by the way, to add icing on the cake you just set up about the twenty twenty three, twenty twenty four and it being a chat bot, and then there was annoying marketing people hyping it and saying, the world's gonna change, you wait. Next year it's gonna design all your hardware. And so I think engineers are naturally and we're so glad you guys are. cynical, but man, I get a lot of of rightfully earned pushback when I interview AI founders like you. And lately I feel like these conversations are more substantive. And I feel like that's sort of what you're hinting at too. So let's talk about the big elephant in the room, which is Where does the engineer need to own their design work and maybe pieces of their research, whatever that looks like, Colin? And that they should really never give up that that very human, unique thing that engineers do. And what parts are safe and reliable that we can hand off to surface important. pieces and not spend three days digging through data sheets. Like where's that line for you now sort of three years in? Collin (07:27.964) Yeah, yeah. I'm I'm glad we're talking about this 'cause we're We're seeing opinions on this, like even from our customers crystallize. and I think the hype machine that and there's so much of it does a real disservice to the technology. it doesn't cure everything. and in my opinion, it shouldn't either. So the line that I draw is the AI is good at searching for things, finding information, and doing it quickly. you can spin up a hundred AIs and have them search for. Judy Warner (07:41.58) Yeah. Collin (08:00.448) stuff and get results back in minutes, what would have taken days. They're great at doing that. What they're not great and I would argue shouldn't be used for, even if they were really good at it, is making the engineering judgment about what it is that you're trying to do. So I I see a lot of criticism about well, it can't design me a board or it it can't tell me, you know, this is the right chip. And my pushback to that Is you're you're using it not in the right way. You're asking it to make an engineering choice for you. What you should be doing is saying, I need to do X, Y, or Z. Help me find parts that do something for that. Or maybe more specifically, you know, you make the choice about what kind of parts you need. Have it go find some that meet your specs. And those specs don't have to be, you know, let me change the dials on DigiKey to get whatever results. Go into the data sheet, read and find if it has this feature, this feature, and this feature. Or if it can do some kind of thing, maybe something that doesn't surface in the parametrics. or, you know, a lot of data sheets have marketing fluff in there. So, you know, their headline spec, it can do this, you know, this many amps. But then you go read the fine print and it's like we can do that many amps at 25C. but then you get, you know, into any realistic operating temperature and it falls off a cliff. you know, and how far it falls off a cliff is probably more what you care about than what the headline number is. So all of the digikey specs or all of the top level parametrics that get published are from the marketing team and they're at 25 C. But there's a chart in there that shows what happens at 85 C. Judy Warner (09:37.516) Right. Collin (09:55.697) Have the AI go and get that for every part that you want to look at, even if it's a hundred of them, it's good at doing that search. And you can do all hundred at the same time. And then you can make the real choice. what part do I want to use given this information? What kind of you know, microprocessor should I use given all the data I did with the AI or the the data that the AI surfaced from a search about, you know, I need this, I need that, you know, help me find one that has. You this kind of interface, whatever it is, use the AI to do the search. Use it to find information. Use it to do the stuff that's routine. If you could write out three or four instructions to an intern and have them do it, have the AI do that. Use your engineering judgment to make the choice. Don't give that up to the AI. Judy Warner (10:42.517) Mm. Judy Warner (10:48.959) I like that picture of an engineering intern, right? Go do the grind part so I can focus on the intelligent and the architecture and engineering judgment that could take you years to learn. And there's so many trade-offs that you learn over time. You talked about how AI is different and how it's better, it's searching, it's more intelligent. Is that because AI in general is getting better? Or is that because companies like Xenode are training AI how to look at this data or a combination of both? Collin (11:28.816) It's it's a combination of both. So we we do some work to make it better and present good data to it so it knows how to use it. But for pretty much any application, whether it's electrical engineering or writing code or finding information in a financial statement or whatever it is, it has to understand how to navigate a document. It has to understand, you know, this was said somewhere and this was said somewhere, so I should be looking here instead of there. Making those choices, that's what AI is trained on now. It's it's less about how much of the internet can we shove into its memory and more about given this situation, which choice do you make? where do you go? Do you go to the table of contents? Do you go to the index? what do you search for? If that didn't work, what else should you search for? That kind of stuff. and so with that improvement comes huge leaps in capability for even small amounts of improvement there. And that's what we're seeing lately. Judy Warner (12:29.087) Okay. So let's talk about where the rubber meets the road. So you started off by saying you were an engineer, you felt these pain points, it felt like it should be automated. And so you decided with you and your co founder that you'd take on the problem and see what you could do. So do you have any examples? Again, I'm thinking about my cynical audience here or just healthily cynical engineers. Like can you give some use cases that can illustrate sort of we did it this way and it spent this much time or it gave this degree of reliability or whatever the use case is. Can you share, if you're allowed to, any use cases from your customers where you're like, that's it? they can really illustrate in a real way and not a hype conversational believe it just 'cause we said so sort of way. Collin (13:28.604) Yeah, yeah. So because AI can do all this research and it can do it in parallel and relatively quickly, research is cheap. which means that you spend less of your time doing the research and more of your time doing the engineering and being creative about how do I want to choose some parts. So I used to make power supplies for welding equipment early on in my career. so big switch mode power supplies. this was about 15 years ago. So switch mode topologies were, you know, firmly established, but a lot of technology was changing. It still is. so you know, there's a whole bunch of different kinds. Kind of fets that you can choose, and between the FET and the magnetics, and you know, a few other details, those kind of set what the design can do. so choosing a good FET is pretty key to making a cost-effective or capable, high-performance, whatever it is, power supply that you need. so choosing that FET becomes a really big deal. so does choosing the magnetics, and whatever you choose sets the rest of the design. So typically you You want something that's on some kind of Pareto frontier. It might be, I need the best performance. Bar none, that's that's what I need. Or maybe I need the best performance per cost if it's a lower cost design. whatever it is. So typically for a power supply, you're trading conduction loss versus switching loss. conduction loss is the one that everyone looks at kind of right away. you know, I need a FET and I need RDS on to be as low as possible. and if you're switching slowly, that's kind of all you need. so you can just pull up digital. Collin (15:15.636) key search you know sort by rds on and away you go but when you get a really low on resistance you're trading against the size of the FET and by size of the FET I mean like internally how many how how big is the the channel. Usually what they do when they design these things is they make one FET, you know, on the die and then they array it a zillion times. So you have lots of little fets and because fets can share current, it flows through them evenly. So if you get a big FET, quote unquote, one with really low RDS on, what that means is the die is big and you've got lots of copies. So your gate charge goes up. And when your gate charge goes up, that means every time you switch the FET every Judy Warner (15:57.421) Mm-hmm. Collin (16:04.128) every time turn it on and turn it off, you have to charge or discharge that capacitance. So for a switch mode power supply, you to make a small one and usually a more efficient one, you want to switch fast, as fast as possible, usually. And so the faster you switch, the more often you're discharging and charging the gate capacitance, the more power you're dissipating in doing that. And so kind of a general rule of thumb is that you want the conduction loss to be equal to the on-resistance loss, the conduction, or you want the conduction loss to be equal to the switching loss, roughly. And I'm sure there's some power supply designers in the audience who are saying, well, but But general rule of thumb, you know, you're trading these two things against one another. You get better RDS on, a lower number, you're gonna get a higher gate charge and a higher total charge on the Fed. if you want a low total charge, you're gonna have a higher RDS on. So you gotta trade these things. But there's a frontier for that. And That Frontier is not the same for every part. Some designs are just better than others. some materials are just better than others. So in twenty twelve, when I when I was doing this in corporate engineering, silicon carbide was a the brand new kit on the block. And it was just better. and so if you looked at the that Pareto frontier, you could see that. But to do that, you'd have to go into the data sheet and you'd have to get all the numbers and then that would give you one data point. and so you know, you might go look at five or ten data sheets and kind of get a feel for where things were. but there's you go on DigiKey or Mousser, there's like 10,000 vets in one small category. Collin (17:55.461) I'm not gonna go through all of those. I'm not even gonna go through a hundred. If I get through fifteen or twenty-five, like that's a lot. so you kind of just didn't do that typically. you'd look at a couple, you know, from a couple different series and be like, okay, well this one is roughly better than this other one. And so of the couple that I've looked at, I'm gonna choose this one and that's where we're gonna go with. I'm simplifying a little, but not that much. so now, because research is cheap, And you can make custom analyses for pretty much everything you do and creating those analyses is also cheap because of the AI. Have the AI out go out. Get the top hundred, top five hundred results based on a couple different searches for roughly what you need. I need an N channel FET and a D square pack, and I need to do at least a hundred volts. And you know, we're looking at an application switching around 50 amps, whatever it is. give me the top 500 results for that. Good. Now with those 500 results, I want you to go get RDS on QG total, Q, you know, the total gate charge for everything, the total charge across the FET. I want you to get the safe operating area. All of the specs that I could possibly need. I want you to go get all that data for all of these parts. And it'll go and dispatch agents to each one, all in parallel. It'll go pull all that information. For a couple hundred, it might take 20 minutes, maybe 30 minutes, possibly an hour, depending on how much you're asking for. Collin (19:32.433) You wouldn't have been able to get that information in days before. Now that you've got it, let's plot it. You know, plot RDS on versus QG. And you'll see that Pareto frontier come right out. And a lot of parts will be on it or close to it, and other parts will be below it. So you can immediately throw away those bad ones and you can see the top couple. so if you're doing some kind of you know high performance design, you can pick from that top corner. But maybe you've got some other constraints. Maybe price is an objective for you in a commercial product. It usually is. So I want the best fed. But it needs to be within my realm of price. So I'm willing to trade some of those specs for price. Price goes up pretty rapidly for the best possible stuff. So, you know, show me the best FET, but with you know the breath best performance per dollar. and you know, maybe show me the difference between silicon carbide GAN and silicon. You know, let me see what the differences are. Judy Warner (20:19.104) Right. Collin (20:40.55) And then your creativity kind of becomes the limit. What is what do I need for this application? And this is where you apply your engineering judgment. You know, so hand the AI the circuit and say, This is what we're doing. I need you to extract all the details from the circuit and let's make a spice simulation so I can see how GAN works here, how silicon carbide works and how silicon works. And that way I can help choose, you know, is it worth it to go to silicon carbide to get no reverse recovery? Maybe. Let's find out what that buys us. and so you're making the judgment. I want to test this thing. Let's go test the thing. It only takes 20 minutes now. I don't have to sit here and spend a day or two setting up a simulation and getting all the data. I can just do it in 20 minutes. And while that one's running, I can set up another one. So I'm still making the call, but I'm making that with way more data to do it and way more experiments and tests to see, you know, does this actually perform better, or is it just marketing hype? Or if it does perform better, what do I need to do that? And what am I paying to get that? and so now you've got a lot more information. You can make a much better judgment, but you're still making the judgment. The AI is not doing that for you. And I think that's the key. Judy Warner (21:36.79) Right. Judy Warner (21:53.899) Right. Judy Warner (22:04.296) I think that's what you mean is research is cheap now. Because before before if you had to do that manually, which sounds like you did, how much time would that take you? Hey, let me back up a minute. You said something about you can plot it. Is that something engineers do, or have you set up sort of inside Xenode like a scatter plot or some kind of plot that surfaces it and That kind of user experience. Collin (22:35.196) Yeah, on Xenode we do set something up so you can you can plot anything you want in kind of any way that you want. You know, for this it would be a scatter plot. you know, I've got the spreadsheet of all the data from the research. So you could download that and just pull it into Excel and make the plot yourself. it would take a minute or two. the research is kind of the big deal there. But if you have the AI do it, it takes two Judy Warner (22:42.827) Okay. Collin (23:04.146) 15 seconds. most of that's time for the AI to think about what it needs to do and write the program to make the plot. and then if you want to change it, it's it's fast, but by no means do you have to use the AI for it. but you can do anything. You can make bar charts of stuff, you can make a radar plot given, you know, some kind of thing, you can make a pie chart to see of the options, you know, show me which percent fall under this and under that, however you want to split it, and and that's where the creativity comes in. What kind of analysis do you want to do? How do you want to see this? What will best inform your judgment? What do you need to know to do? that. Have the AI run the work part of it for you. The rote, you know, just this is the workflow, I need you to do it, and I need you to do it over all these parts. Meanwhile, you can think about so what does that mean? And given this information, what kind of other analysis do I need to know to make the best choice? Judy Warner (23:47.051) Mm-hmm. Collin (24:04.676) Is this even the best way to do this? Should I be looking at something else? Spin off another search. You know, we're we're looking at you know silicon carbide right now because it has no reverse recovery. But you know, we need to do something, something different. Is GAN maybe a better choice? What does that involve? Spin off an AI, let it go find some GAN parts, do the same plots, and you can just tell it, hey, do the same analysis for GAN. Just go do it. And it'll take what you already did, do the search for GAN, set up all of the the research agents, get the data, plot the stuff, and it'll come back to you with here's what we found. you know, GAN is X percent better in this, here's the plot, and here's all the relevant data. you know, or you know, maybe it's something more complicated and this is This is why I can't really tell you exactly, because it's your application, it's custom for each one. So you get to exercise that creativity and a lot more of your time is spent doing that, which is way more value add than asking an AI to design the circuit. It you don't need to do that. It's Judy Warner (24:55.093) Right. Judy Warner (25:09.822) Yeah. Well, and you're asking it to surface massives amount of research data that you can actually check yourself, right? As well if you need to. But that what's so messy about engineering is every specific application has all these trade offs that your engineering judgment is playing against. And I think that's why most engineers just call a big BS on AI. It's because they're like, I'll never do it. But what I hear you saying is we're not asking you to do it. We're just surfacing the data so you can look at a much more narrow data set to get to your and spend more time doing the engineering judgment instead of the stuff that an an intern or somebody, you know, pouring through data sheets just as a Time and brain suck, it's there's nothing. It doesn't take skill and engineering judgment to do that. It takes engineering judgment to know what these things mean and if it's marketing hype, but so I mean, I can really see that you've built this around the parts of your job that you hate it as an engineer, which is why I think, you know, I've been interested early on from you and your co founder Brandon, because you guys were engineers and I think that's a pretty compelling case that you want to build it in the way that you want it to work and your fellow engineers want to work. so that's very cool. So like I said early on, I get a lot of AI skeptics and I'm skeptical skeptical as well. And I think you and I share this in common, Colin is I feel like AI is here, whether we like it or not. Like we're going we're there. We're already in it. And I think just to push back without just without a lot of thought, that's gonna hurt your career as an engineer. Like it's coming. And I think the bigger decision is how we're gonna use it and how do we protect the human creative. Judy Warner (27:37.471) really unique things that engineer does that you cannot replicate in AI today or in the foreseeable future. So how do you speak back to so where I say we have this in common, I try to look for things that give a positive use case for how to use it and but yet how it'll protect the humanity and the creativity of engineers. I don't hear anyone saying that engineers should let go of ownership of their design, not one person. Nor do I think AI can do that now, nor do I hear anyone saying that's gonna happen anytime soon. So what do you say to skeptics? I mean you're talking to customer and people, I'm sure, every single day. So for those who, you know, I get people put me on blast on YouTube or social media. And I understand it and it doesn't offend me at all. But how do we how do we manage this in a more productive way where engineers are still staying on track? Like we're in AI whether you want to or not. So how do we use it productively? Are there rails we need to put on it? Or how do you think about it? Collin (28:54.748) Yeah, so I'm not a huge fan of putting guardrails on it because I think that it's just gonna freeze it in place where it is now. And You know, the advancements are so incredibly useful if you use them in the right way, that I think that does a disservice to everybody. instead I think we need to start thinking about how it actually should be used versus the hype machines that are unfortunately so common. so like I was saying earlier, don't let it replace your judgment. You spent decades building that and even if it could have judgment, should it have judgment. I don't I don't think you want to give up the product design to it because then you can't control what you're actually building. and that's that's not the point. So when you're using it, use it as a tool to help you, not as something to do your job for you. You know, when when I'm talking about doing that analysis, I'm still the one figuring out what I need to do. It's just the the heavy duty lifting that anyone could do. I don't I don't need to do that anymore. So not needing to do that frees me to do all of the other things and exercise that judgment where it has value and actually have the time to make that judgment. So it it's a force multiplier. And so I think that's what we need to be thinking about. How can it make your job faster? What parts of your job aren't really value add? And how can you use the tools that we have now to accelerate those pieces so you can spend more of your time doing the things that are value add? Collin (30:46.344) So those I think are the big ones. And unfortunately the the hype out there is probably still going to be, you know, AI is going to replace you and take your job and it's gonna be a dystopia because apparently that that sells, which is unfortunate because I think it does a huge disservice to what it actually can do for you. You know, just as a aside for everyone, I use AI in my normal workflow for everything that I do now. you know, when we're developing Xenode on my other projects, and the speed at which I can move now is incomparable to what I could do even a year ago. honestly, even six months ago. and if you'd asked me that question six months ago, I would have said the same thing. So the the curve is going up. I'm doing a lot more engineering. Judy Warner (31:13.803) Mm-hmm. Judy Warner (31:25.557) Right. Collin (31:36.377) Engineering now. Because I have the AI tools to accelerate all the things that you'd still have to do are still very necessary, but they're just time consuming and they take away from actually doing the engineering. So I can spin off a tool to go, you know, do some research for a board design. I can also hook up an AI to the existing board that I just got for another customer that. firmware's having an issue and there's it's gonna take me days to figure out what it is. And you know, while that does involve some debugging skill, most of it is just sitting there banging on it with a hammer trying to figure out what's broken. AI go do that. It'll solve it in 20 minutes. And then we can move on to actually bringing the board up and and testing the rest of it. so and I can be doing three or four of those projects all at one time. Judy Warner (32:16.832) Right. Judy Warner (32:23.04) Right. Collin (32:30.736) So the the increase in skill, if you use it as an enhancement and an augmentation for your skill, not to replace your skill, that's that's kind of the huge lever that you get if you use it properly. But if you try and say, Hey, AI, go, you know, do my job so I can sit back and sip some coffee, you're gonna get bad results because you're relying on it making the judgment calls instead of you. Judy Warner (32:37.95) Yes. Yep. Judy Warner (32:52.534) Yeah. Collin (33:00.474) And that's I think that's the big difference. Judy Warner (33:03.444) Older engineers are worried that if we do that we're just gonna make bad young engineers. Like they're cheating. And I I I think that's true. And so I think it's really dangerous to to outsource that because it's gonna build badly engineers and they're not gonna know why. Some things Collin (33:10.096) I think so, yeah. Collin (33:25.318) Yeah, and I I I wanna touch on it a little bit, especially for the younger folks in the crowd. you know, 'cause I know there's a lot of concern among People still in college, or if you've just graduated, you know, that AI is going to replace your job. Because even we were a few minutes ago talking about well, if you'd have an intern do it, and you know, if you were just that intern, or if you're looking for an internship now, that's a scary thing. the AI does what I was gonna do. so you know, we do need to be cognizant of how we're using it in bringing up the new engineers. So, what I would say to people who are in that boat, or if you're an older Judy Warner (33:51.66) Mm. Collin (34:05.472) engineer concerned about that, you know, bringing up the next generation, using the AI as a tool to help you find and understand what you're working on, that is also something you can do. So using it as a learning tool, that's I I use it for that all the time. So say you're working on that same power supply and you're looking for the parts you you could have it help you guide, you know, what what is the thing that I need to care about? And it'll give you information about where you should look look next. So, you know, it it can tell you you're probably concerned about these things. you should look at at these different items. Further reading, you can look here, you can look here, you can read this book, whatever it is, because a lot of what I learned early on Judy Warner (34:58.219) Mm. Collin (35:02.84) started with some kind of material I read from somewhere. I remember one of my coworkers handed me a power supply design book that I read voraciously for the first month and a half of that that first corporate job I had designing power supplies. and so that opened my eyes to the things that I need to care about, then go chase those down. So in the same way that now I work as an engineer, the stuff I'm chasing down is all the research data. If you are a student or a new engineer or learning in general, even if you're not a new engineer, you can use AI to help you find the things that you need to learn and then go find those things once you've learned enough to know what you need to know. Have it go find those things, then go read those. Use it to further your understanding, not do the understanding for you. So it kind of Judy Warner (35:52.362) Right. I think that's a good nuance. I just got a message today on LinkedIn by a guy who's a senior engineer. he's an embedded engineer out of Michigan, and he sent me a really nice note and he was saying he was trying to s help a colleague solve an EMC like a compliance issue and he was talking about it. And this younger engineer said, You sound like that guy on Signal Integrity Journal. He goes, You mean Dr. Eric Bogatin? He goes, Yeah, that's the guy. And he said to me, Judy, those free master classes you have on your website, and you mentioned Don Telling's book and Eric's books, and I read And I bombed a job six months ago, and now I'm the go-to on SIPI. So in this guy's case, I think he he went and read and did the research and it made him better. But it didn't sound to me like he was against automation or anything like that. He was saying, thank God there's these tools out here. And I think AI is just one of those tools, right? But there is good, there's books, there's websites out there that can also help you go deeper. so you do retain it. And I think that's a really good point. Well, I have taken so much of your time and but before I let you go, Collin (36:58.055) It is. Yep. Judy Warner (37:15.167) Where can people go to learn more about Xenode or your product or give it a test drive? Or what would you recommend for people who want to learn more about Xenode? Collin (37:26.14) Yeah, so our our website is Zenode Z E N O D E. You can sign up for an account and try out some of these workflows that we talked about today. If you have your own workflow or you want to understand how maybe you could use AI, even if it's not Zenode, in your own workflow, I'd be happy to discuss it. You can email me directly and we can go through your application and see if Zenode can help or how you might be able to automate it with some other tool. Judy Warner (37:57.151) Very nice of you, Colin. Well, again, thank you so much. It's been really fun to be on this journey with you and with Brandon. I remember meeting you what what has it been two or three years now? Collin (38:08.782) It's yeah, I think three years ago at PCB West already. Judy Warner (38:11.601) Unbelievable. And you guys were just a tabletop and a early tool. And it's been really a joy to watch your journey and watch you guys really build this out and build a customer base around it. So congratulations and thank you so much for sharing all your knowledge and your insight with our listeners so they can keep learning. Collin (38:25.917) I do. Collin (38:32.542) Thank you, it's always a pleasure. Judy Warner (38:34.731) For our audience, I'll put the links that that Colin mentioned down in the show notes. but the easy part is to go to their website. Say that website again, Colin. Collin (38:46.622) It's zenode.ai, z E-N-O-D-E. Judy Warner (38:52.039) Or also I'd recommend you follow them. They have a company page on LinkedIn and then connect with call in. so you have a resource to dig more into this topic. I hope and trust this has been helpful to you. We'll see you next week. Until then, remember to always stay connected to the ecosystem.