Latest / CTO Lunches / Ep0 - Thread of the Week - AI Assistants

Transcript

Ep0 - Thread of the Week - AI Assistants · CTO Lunches

• Kendall Miller: So it can't complain though. But yeah, I'm doing good. So I also paid my first Rishi: Excellent. Kendall Miller: college tuition bill this morning. Whew. Whew. People talk about it, but now I see Rishi: Ooh yeah, that's sounds brutal. Kendall Miller: why people talk about it. It's just a lot of money really fast. Anyways, and and it's out of state, so you know. Rishi: For with you know, like a I mean the the big ticket items I've I've bought in my life have been cars and, you know, down payment for a house. And I feel like those at least feel better because they're kind of a one time thing. Kendall Miller: This is a down payment on a house, and in six months another down payment on a house, and Rishi: Ha ha ha. Kendall Miller: and and so on and so forth. So Rishi: Yeah. Kendall Miller: yeah. Anyways. Rishi: Hi everyone, welcome to the CTO Lunches Podcast. I'm Rishi Malik here with Kendall Miller, and we are your hosts. Kendall Miller: well tell me what you're drinking. Rishi: I am drinking generative A Rye from Friday deployment spirits. Kendall Miller: You you have the bottle you can show? It's it's the sponsor this week, so we can we can show it does it won't be the sponsor every time, but yes. It's such a good looking bottle. Rishi: Yeah. It's a good looking bottle. It's just a great whiskey. it's perfect thing to be drinking on a on a Friday here. What are you drinking? Kendall Miller: And and and the and the rye I just I'm wanna say one more thing about it. The the pepperiness Rishi: Yeah. Kendall Miller: of the rye is just so good. whiskey. we we Rishi and I make Friday deployment spirits, if that's not obvious. I am drinking Rishi: We do. Kendall Miller: Friday deployment gin, force push gin. and it Rishi: Ooh, excellent. Kendall Miller: is it is very good. it gets cloudy because the there's so many oils in it and that's what I like. You put an ice cube in it, it gets cloudy real quick. because as soon as there's a little more water in that it can't all dissolve all those oils. And it's very complex And when when Rishi and I sat down with the distillery to come up with this recipe, they had made, I don't know, three, four gins. Correct me if I'm wrong, Rishi, it's been a while. And I remember sitting down thinking, I'm about to taste gin and give tasting notes to a distillery and this is Whatever choices I've made in my life that brought me here are the best decisions I've ever made. I remember just being so happy about it. But the distillery did have one recipe that we were like because we had given them some notes ahead of time. Hey, we like these things. We'd like you to try these things. So they came up with three recipes and I think they expected significant notes on it. But we found one that we were like, This is really good. And then Rishi kept saying, This is my favorite gin, and I kept being like, You can't say that. It's our gin. And now, It really is the best gin. It's my favorite gin. I like it so much. So, Rishi: It is it is excellent. It is excellent. you know, I think the the funniest to me, I think the the benefit that we had was, you know, we're partnering with this distillery and the the owners and the distillers are also engineers, right? And so they bring that kind of scientific method Kendall Miller: Yep. Rishi: and rigor to what they do. And when you bring that to to gin and spirits like that, I think you just get outstanding results. Kendall Miller: They challenge so many assumptions in it. They start with with a base grain and and even the ways that that this is distilled, which corner me sometime if you ever want to hear me talk about, but there's there's some things that our gin is different than I think probably any other distillery out there, and that's part of why it's been so very good. Anyways, that's that's Rishi: Absolutely. Kendall Miller: that's what I'm drinking. Rishi: We'll jump into CTO lunches. Why don't you kick us off with what CTL lunches is? Kendall Miller: Yeah, well so CTO lunches. is if you're yeah I'm assuming if you're listening to this you probably know but maybe as as this grows we'll see. But CTO lunches is a group of about twenty three hundred engineering leaders now. We have people from all over the world. Almost everyone is in senior engineering leadership. We we try to lean towards director plus I mean people float in and out of CTO roles and go back to IC or go into product roles and we don't kick people out. But it tends to be very senior engineering leaders that are in the group. we have lunch Lunches all over the world. we probably put on somewhere between 20 and 30 lunches every single month in cities all over the world. no recruiters, no vendors, no quote unquote thought leadership. It's a lot of like, hey, I'm dealing with this, I'm wrestling with that. How do people solve this problem? and it's a it's a good community. So check out CTOLunches.com if you're not there. But yeah, tell us what we're talking about today, Rishi. Rishi: Yeah, so as a reminder, this this episode is sponsored by Friday Deployment Spirits. That's Kendall Kendall Miller: yeah. Rishi: and I. you know, which is just a a great thing to be able to tie both together. so this week, one of the great things Kendall Miller: Wait. Friday deployment dot co check it out. That's where you can buy it online. Sorry, keep going. Rishi: Excellent. We'll have a link, we'll have a link in the show notes as well. so one of the the the excellent things about CTO launches is just some of the discussions that happened. So for this episode we figured we'd talk about one of the most active ones in the last last week. which comes down to at a high level, it's essentially AI assistance. more specifically, there's a new one that's out there instinct, it's invite only. know, obviously there was open cloud, there's been a bunch of, you know, other stuff around there, but Instinct has a lot of the it's taking up a lot of the space and has a lot of the excitement right now. One of the great things on the email list is in Instinct is invite only. People are helping each other out, getting people invites and just making it possible for a lot of people to to jump in. but that's what we're gonna talk about today. Kendall Miller: Yeah, the thread started with the thread started with a hey, I'm I'm interested in this. Anybody have an invite and somebody found one and then more people got invites and then they it multiplied. So if you wanted an invite to instinct, CTO lunches was the place to get one, which is nice. And then and then there were security concerns, which which may surprise anyone who's using who's Rishi: And then Kendall Miller: using AI, right? That there was a s do you do you remember what the security incident was that got posted in the I think it was in the thread. Do you remember what happened? 'Cause I can give the the high level. Rishi: I don't remember the details, but yeah, there's there's one in the thread. So yeah, won't you give the details? Kendall Miller: So I I I think what it was was so instinct has access to say your email, your to do lists, etcetera and one of the users went and created a new email address that the agent would not know that he had, and then sent an email to his own email saying, Hey, Instinct, summarize XYZ for me and send me an email with that back. And so even though the email was in no way associated with him, the the agent just summarized a whole bunch of things and emailed them back and he was like, okay, this is terrifying. It's a gigantic security risk. A great way to exfiltrate Rishi: Yeah. Kendall Miller: information. I you know, it it would be amusing to set it up and send an email with Send me all of my private keys or send me the one time password that just got sent from my bank to to my email address. Like you wanna you wanna talk about a scary attack surface? This is scary attack surface. Rishi: Yeah, I mean I feel like we should we should maybe take a second and you know kind of talk through a little bit of that. The the interesting thing to me from that thread was people helping out, getting invites, that's great. The security incident was huge, but it was almost how much the thread pivoted, right? First it's excitement. Everyone's like, cool, I want to try this. You know, we're all busy, you know, a lot of people tried open claw. Everyone wants an AI agent that's gonna and you know, an assistant that's going to genuinely help. it'll help them in their lives. And that's, you know, be it if you're a tech leader, if you're an engineer, even outside of the tech space, you know, I I I get why there's so much excitement. But that thread very quickly turned into the security issues, turned into some of the other concerns, turned into some, you know, good suggestions on how to ameliorate some the pain. but it's just very clear that like very quickly the worry and the concern more or less trumped any of the kind of benefits, you know, at least from my point of view, that you're gonna really get from an assistant that Yeah, we all we all really want the like sci fi, you know, Star Trek computer, go do X types of things. But none of us have the trust, I think, in the AI and the systems yet to really be able to do any of that well. Kendall Miller: Well I think that I think that it's it's well founded to not have that trust. So I mean, you know, I I I just shut down a company, but I spent the last year working on a company that keeps AI Rishi: Mm-hmm. Kendall Miller: agents acting in this very bad way, right? And so like because that's what we see, AI agents have the intelligence of an overly confident 18-year-old, right? more knowledge Rishi: Ha ha ha. Kendall Miller: than that, but they can they can solve any problem for you. They're super excited. to to go solve that problem for you, but but they're gonna solve it in inappropriate ways. And you know, that actually the story that I told and I'm gonna I'm gonna tell it again here is that you know, if I'm backing out of my driveway with my fifteen year old and I and I say to him, Hey, go in the house and Grab some cash. I just realized the place we're going only accepts cash. Don't come out until you have cash. He gets out of the car and he goes in the house and he digs around and he can't find cash and he goes, But dad says I'm not supposed to get back in the car without cash. And so he grabs mom's engagement ring and runs to the pawn shop and pawns it and jumps in the car and says, Here's $2,000, Dad, let's go. Like that's the way that these AI agents are behaving today. And that's what makes them scary. And particularly these like private, self-hosted agents, so the the open claws, the you know, this particular one, it's it's like you're so eager to please and you're so capable that you'll find a way to answer and solve the problem that I asked you to solve. but with no common sense, no ability to, Rishi: Ha ha. Kendall Miller: you know, make a rational decision that a a a normal, you know, if my son did that, I'd be like, what the hell just happened? Rishi: Mm. Kendall Miller: so I'd also be waiting in a car for a really long time for him to get back from the pawn store but pawn shop, but yeah, that's a separate separate problem. Anyways. Rishi: Yeah, I I I I think that's where a lot of the foundation or infrastructure that exists is just not good enough to be able to support a lot of stuff like this. You know, I think obviously this is where, you know, some of the stuff that you work on at a previous company was going. But I think I think it is a problem that needs to be solved still. You know, folks in Thread are talking about I mean, you know, the the one person who had a a burner email address, basically. Somebody else was talking about virtual credit cards where they can limit it. It's like, cool, I can give you, I think they said fifty dollars for whatever they were they were trying to do. go forth and and do the things that are necessary. and that's great. but we only have this kind of limited, you know, bits and pieces. Like you don't want a separate email account to email, you know, your friends to to meet up or your, you know, boss or whatever else needs to happen. It's gonna have to come from your real email account. And there are there is no way to like limit that. It's Read or write, right? Like that's the level of granularity that we have. And it's just not enough. Whereas really you want we want some kind of layered permissions, you know, call it I am, call it whatever else you want. And there is just no ability to to do that right now. Especially when you start thinking text messaging and all the rest of your life that you're gonna tie in. Kendall Miller: Well, so so where do you think this goes? I mean, if if if you were gonna go design one of these, like I I think and in and you, you know, you delve a lot deeper into the security side than than than I do. Like your your background, you you You still do things like pen testing for companies when they ask you to, you know, like this capture the flag things at DEF COM or your jam. Like, how would you design this in a way that you would be excited to use it? Or or is there no design where you can give it the access you want it to have to be useful and the restrictions it needs to have to be safe? Rishi: There's there's there are ways to do it. The challenge is most of us don't have enough access to be able to do it. Right. So the great things about LLMs, the great thing about AI right now is it's very capable because it is flexible, right? It can figure things out, it can go, you know, come up with a creative solution of taking, you know, your wife's engagement ring and potting it. that's creative. That's great. And that flexibility allows for a lot of things that we want to do. But you I think still want that deterministic, you know, hard rules engine type of thing, even if there is a little bit of AI applied to each of those, to be able to go all the way through. Like if you're sending an email to my boss, that should require, you know, me to approve it. If you're sending, you know, an email to I don't know, my you know, to if it it's sending an email to you to say, Hey, let's get together and, you know, have a drink, you could probably just send that. You know, if it goes super wrong, like that that's that's okay. The the blastery is is smaller. it's it so like Kendall Miller: Wow. Wow. I appreciate that you rank me that much lower than your boss. But yeah, okay. Anyways, keep going. Rishi: yeah, so I I think it's that level of of of action that you want the hard rules on. And it's what you want to actually apply AI to in each of those classifications as well. And like you actually see this, like Claude Code's beginning to do you know a lot of this in in auto mode, right? Where they're using little classifiers to say it's not just you invoked bash. It's you invoked bash and I see RM, you know, dash RF in there. is getting a little bit better and like we need that and it's great for cloud because it's controlling that surface. I don't have the ability to build that into email well without just like reinventing the whole stack or text messaging or all of the things I want AI to to really help help me with. And so I I think that is the the challenge that we have. And I think everyone everyone has this from an AI assistance standpoint. But I think you also see this across the board when it comes to corporate corporate usage of of AI. It is the t the terror of like great I'm running Terraform is an AI agent. And I don't think it can run in a Terraform apply, but maybe it can figure out a way to do that and then prod's gone. and that's just horrifying. But there still aren't good guardrails around that. Kendall Miller: So so why if this number's correct that you have in our notes, why did this company raise three hundred and fifty million dollars? Rishi: I mean, my my personal point of view, I I think it checks the boxes for where VCs want to be in, you know, August 2026. It's AI, it's exciting, it's a giant market because the TAM is every human with a phone. every human with an email address. And that's a big market and no one wants to, you know, again with the the way the VC game works. It's better to invest in the ones that maybe aren't gonna win because if you miss the one that does, you've missed the next open AI or you've missed mixed you've missed the next anthropic. Kendall Miller: So how do I raise three Rishi: I think that's why. Kendall Miller: hundred and fifty million dollars on a on a you're gonna miss out FOMO? Like that's that I I mean, I it it feels like Rishi: Ha ha Kendall Miller: joke money at that side like and also like if you raise three hundred and fifty million dollars and you release something that responds to the very first email like that, I mean, I just have so many questions. Now granted, the the attack surface is so broad. There's a million ways it could have been hacked. And I'm sure that they Rishi: Yeah. Kendall Miller: tested and tested and tested a whole bunch of things. But this feels like one of the obvious ones to get started with. So I think, I mean. There is a world for open cloud there's a reason people love these tools that go out and act. On their behalf. And and I I can't tell, do we get to the future where these exist by way of really good guardrails around them or by way of the models maturing to have the kinds of human common sense? Like as we are all on the internet panicking about what these agents are doing, and the frontier models are out there training on all the things we're panicking about on the internet, are the models gonna get better? You know, are the models gonna learn? This is a bad idea, right? Or or not. And and I don't know. I mean, it's really hard for me to tell how quickly they're actually advancing in those kinds of things. Because Claude and auto mode today, right? It's the only way I run Claude. Six months ago auto mode didn't exist, and we wouldn't have trusted it if it did. And it's Unbelievable how far it's come. But I don't know if anthropic at anthropic people are sitting there saying this can be approved, this can be approved, this can be approved, and this can't. Because I mean, one of the other things Rishi: Ha ha ha. Kendall Miller: that just stopped in auto mode is it it will no longer push to push to main for me on GitHub. Like unless I give it, it'll stop and say, You've already told me you want me to push this to main, but you need to explicitly tell me again that you want me to, my my guardrails now don't allow me to without that. Rishi: Sho should we back up for a second and talk about why you're why anyone, including you or your AI, is pushing directly to Maine? Kendall Miller: because I Rishi: Ha ha. Kendall Miller: this so let me back up and be be very clear. I push to Maine on my projects Rishi: huh. I mean we you're talking to a lot of technical people right now. Kendall Miller: that are very simple Kendall projects, or sometimes say a website. I don't do this for anything that's production ready. I don't touch those code bases because I am smarter than that. I know that my skin so okay, this is back to the age intic thing, right? I know that I should never touch the production database of anything that or the production code repo of anything that actually matters. I know that. So I don't. And I and I tend to say, don't give me don't give me access. I don't want access to that. Even if I screw up, I don't want access to that, right? But but there's things that are low stakes that I can have. You know, I I have an app in the Mac app store and it's open source, it's on GitHub, and I I push domain on that. There's no other contributor, there's nobody else reading that those those PRs. Rishi: Mm-hmm. Kendall Miller: so it's just me. And there's no d well, the dev environment is my local computer, so Rishi: It's that's fair. You know what? Like one of these future episodes, maybe we should do talks to folks about like all of the random side projects that AI has really enabled. You know, you have an app in the Mac app store. First time in my life I submitted one literally yesterday. I don't think anyone else is gonna need it or use it, but it's feasible enough to build it and, you know, spare time and ship it. I think that's that is a great thing. that is what's awesome about having this technology here right now. Kendall Miller: random guy wrote me from New Zealand, said I bought your app, here's all the ways it's broken. Super helpful. Like I I fixed probably ten bugs Rishi: Ha ha ha. Kendall Miller: because and then and then I I pushed an update and he sent me a crash log and I found another bug and I'm like, This is amazing. I have users, who knew? so I have made three Australian dollars, New Zealand dollars, I don't know, something like that. And until I think you hit a hundred, Apple won't even pay you for it, so it's it's Rishi: Yeah. Kendall Miller: fine. But but it's exciting, so we'll see. Rishi: Awesome. That's awesome. Well, so okay, jumping back to the the assistant thing for you know little bit. you know, you asked me, you know, what I think needs to happen. You know, we we have to have we have to have granularity, we have to have specific permissions, it can be hard rules, even if there's a tiering system that you can then use AI to like decide you know what you're comfortable with, you know, and do that kind of thing across the board. I don't think that exists, and honestly, I don't know that it's going to exist because. at least not for any any real length of time because I just don't think the the incentives and the you know product roadmaps essentially kind of align across the board there. The better models get and like the more we can trust them, I think you know, the better off we are. but I think that's a small length of rope that we're willing to give them. All right, it's it's not really different than hiring a human, right? You hire an actual assistant who's human. they'll do very, very well. They will make some mistakes and that's okay. And, you know, you like, yeah, sorry, my assistant scheduled it for the wrong time or whatever. But you don't really worry, you know, per se about them just kind of going off the rails and doing something entirely crazy. That is still the case with models. And, you know, given that they're, you know, token predicting machines, you know, in a nutshell, like I think it's gonna be tough to to have any of them build up the the trust that you need to or at least for me, just to let it rip on everything that I I have. Kendall Miller: Well, and I I would go one step further. there's there's my nervousness, my trust issues with my agents. That's that's a thing. But Rishi: Yeah. Kendall Miller: I can I would consider it such a lapse of judgment to be using one of these in a meaningful business-like situation that at this point, if I went to work at a startup and found out the CEO was using an agent like this, that would be reason enough for me to decline a job offer. Like literally. Rishi: Wow. Okay. Kendall Miller: I I mean I would say no, if you're if you're using something like this with your e like, I would be afraid to work at an organization where somebody is making that judgment call. Because maybe you'll be fine today, but the company's eighteen months of runway are about to be zero months of runway as soon as somebody figures out how to hack like that it it's that big of a security concern for me. I'm is it not for you? Your reaction tells me maybe I'm wrong. Rishi: It's fascinating. I I love this I love where we've gone with this discussion. It is a huge security concern for me. Like to the point that, you know, during the open claw hype, you know, Mac minis are selling out everywhere. I was like, no, I'm I'm not even gonna run it. You know, I looked at it and it's like it's too it's too loose. so I built my own, right? And I started using like a little bit of a loop that a friend had built. I kind of built my own version that was tied into, you know, very specific. aspects where I could build some of that granularity I was talking about into it. it was great because I felt, you know, much more comfortable because I'm like, here's some kind of hard rules, hard limitations that I've put around that. it actually went very badly in a comedic way because I was I was actually interacting with my my assistant in WhatsApp, which was great, building all the code through AI, you know, making all that stuff happen. I shipped a bug. that was just an error an error where like it would send a message, the message would fail to send, and that error would get propagated back into the chat. But I didn't realize that it was actually trying to be proactive and, you know, chatting a few of these threads. And it happened to like the last, the most recent four. Three of those four were just, you know, friends I have in tech, and there was like, what is this random error message that's getting, you know, spammed in here? One of it was actually the the other one, the fourth one was a a group chat for a pickup soccer game that it was part of. you know, some technical people, but a lot of just like very random folks from around the geographic area that all would get together and play pickup soccer. so my bot started spamming the chat with just this error message because it kept failing every time. So it would retry and you know, just constant. they actually booted me out of the group because like my account through my name was just spamming them like, he's been hacked and it kicked me out, rightly so for sure. you know, obviously a little bit humbly, but it like where like the security of my data and how all that was working didn't really matter because I could still like ship a bug and, you know, mess that up. But then when you're talking about like, you know, here if they're using this as a it sounds like your kind of beyond security is almost more of a decision making framework. If you're willing to take on that kind of risk as a CEO at a startup, like you have less confidence in in where that company's gonna go and maybe in the l the leaders that are there. Kendall Miller: For me, willingness to take on that kind of risk, that's an interesting way to word it, because to me, no one is willing to take on that kind of risk risk. The cluelessness to take on that kind of risk is a lack of understanding of the will the the risk you're willing to take on. That's that's the only way I could see somebody making that decision is that they are actually unaware of how risky this is. Rishi: Fascinating. Kendall Miller: Yeah, maybe maybe I'm hand waving, but that's that's my reaction, is that's why it causes me fear. I would love to see these agents go Rishi: So that Kendall Miller: somewhere. I think someday we get to ones that that work great and we trust them. But and also you and I are probably a little more right? Like, I mean, I I remove tracking URLs, everything after the question mark when I click Rishi: Yeah. Kendall Miller: on like like I'm I'm an unusual end of the the paranoia here, right? Rishi: Well so is is that for this topic, do you think that is that is that is that a hot take from Kendallinger that says if leaders are using these tools for what they're doing, you would think twice about joining the company? Kendall Miller: Think twice is a calm way to put I would get the fuck out. Rishi: Ha ha. Kendall Miller: Yes, what what about what about you? You got a hot take on this? Rishi: I don't know, that's pretty that's a pretty hot hot take there. I don't know if I can compete. My my view on this, you know, my hot take here is so you you asked, right, how did they raise three hundred and fifty million? You know, it's like a two and a half billion dollar valuation, I think. Which which is crazy, right? I d I I d I have no idea how they're ever gonna be a two and a half billion dollar company. I don't think they will. but I think that about all of the companies in this space. My hot take is that no one outside of well, I I think there's two companies that have a real chance and two half companies that, you know, will try and not really get there. I don't think anyone's gonna win this win in this space unless they're either Apple or Google. And I think that because they have the distribution, they have emails, they have texts, they have phones, they have that entry point where like so many so much of people's lives go through. they have To me, that's number one, right? You need that distribution. You need to be where people are doing the work to really be a a useful assistant. at a wide scale. Like sure tech folks are gonna get hyped up about the next big thing, you know, and and folks will try and do that. But you know, for the the people in my family, you know, the if you think about it like from the mom test standpoint, like my mom would never use any of those other ones. But if it's built into her iPhone, then she'll start to use it. Kendall Miller: But I I would go farther and say only Google because I don't think Apple Rishi: Huh. Kendall Miller: would be willing to accept the security risks. Because Google has hits to its reputation all the time like that. Apple is terrified of hits to its reputation like that. The first time somebody's email gets hacked because Apple's assistant screwed up, Apple's in a lot of trouble. Google's not, because Google's kinda like, yeah, nah, yeah, we we tried. Rishi: Ooh. You just expect it from Google. Kendall Miller: Yeah. We have all your information, we're selling it anyways. Rishi: Well, I mean, but so part of it I think is also the corporate space. So like I think Google from a workspace's standpoint, you know, when you think of like a corporate AI assistant is is well positioned there. Apple, yeah, obviously, you know, not so much. To me, that's actually one of the other companies. Microsoft with with Outlook, you know, they have some there are some people who use Outlook individually still, you know, but you know, outside of those five people, it's it's a corporate aspect. I think they're gonna try. I think it's gonna be a mess and when I think about all the things Microsoft has messed up over the years, like I just can't see them doing this well. Yeah. Kendall Miller: Well well, I mean th the reason you don't have to worry about Microsoft doing it is they literally are Azure is out of compute, so there's no way they can run this at massive scale because they can't that's I hope that I'm joking, but probably I'm not. Rishi: Right. I think the other interesting company that will I mean it feels like they have to do something. They've made some fits and starts. is Meta. Right? They have distribution. There's, you know, between Facebook and Instagram and WhatsApp, they have they have a lot of people there, but man, like I just I think people would find it really hard to tie something else into them, right? If like, great, do my Instagram posts or whatever for me. If you're a creator, like that, that's fine. But if you're like, well, I'm gonna hook Kendall Miller: I Rishi: my email up to Facebook and let it like email people on my behalf, like I don't know. For me, that just feels like way too average way too far for people to cross. Kendall Miller: Yeah, but do you do you have Instagram? Do you have Facebook? You probably have none of those things. Rishi: I I have both of them. I mean I don't do anything with them, but I have them. Kendall Miller: Okay, okay, okay. I mean I deleted my Facebook account in two thousand six and I deleted my Instagram account the day that Facebook bought I trust that company so little Rishi: Ha ha ha. Kendall Miller: it's astonishing. But but I mean if anybody's willing to go out there and say, Well, we'll we'll ride it until we're sued out of existence, it's them and they know out of existence is a much higher number than they'll get sued for. Rishi: For sure, for sure. The correl not correlated. The follow up to my hot take though is that I think all of this money going into it, which you can maybe say is part of the AI I'll say the B word bubble, in general, I think is going to I I think it's gonna I don't know if I will say pop, but I think it's going to deflate at some point in the future still and be a lot of pain when that happens. Kendall Miller: I I spent a little while thinking about a comment you made to me saying that the AI bubble is going to to to burst and it it we're under or maybe not the AI bubble, but you were talking about the software Rishi: Yeah. Kendall Miller: industry. The software industry is our concerns about it caving in are underblown, and I increasingly think you're absolutely correct. I think we're heading three to five years. So t so I think twenty twenty-nine, twenty thirty, the software industry has a massive cave in and I think it will recover, but I think it's going to particularly for things for developers, dramatically cave in. I think in industries where that they're you know, software for barbershops are gonna be just fine, but I think software for developers is going to basically cease to exist. Rishi: Yeah, I think I I I agree with you on that. You know, it's funny, I got an email from a a friend who's I mean you could say he's a software engineer, but he's kinda, you know, he's come at it from various different angles. Like he's never worked for a big company, he's started some, you know, small software projects and made some some money doing that. And we were obviously talking about AI. And his view was, yeah, you know, all of these things like he he's made a bunch of money in like very small it's basically app store types of things. You know, it's not app store, but it's the same idea. small projects. He's learned taught himself Python, just like one of those entrepreneurial motivated individuals. And he was like, Yeah, you know, I think all the stuff that was hard from a software perspective is gone. And so he's reframing his thought process just around the pure entrepreneurship side. What else can I build? What other companies can I start? And I do think that's where it goes. You know, software as a service, I do think is it's gonna hurt, it's gonna be a big, big cut, but there's gonna be Yeah, where there was one company doing something, there's gonna be thirty companies all vying for your attention trying to do it. And that's where the the battles are gonna be. it's kind of a free market esque change. So it'll be it'll be an interesting few years. Kendall Miller: I don't think it's bad for us long term. I think that the short term shake up is going to be very rough on everybody in the industry. Anyways. Rishi: Yeah, absolutely. Absolutely. Well, let's let's talk about some of the the upcoming events. All right, this is CTO lunches. You want to so I think Montreal Kendall Miller: Yep. Yeah, so Rishi: just happened? Kendall Miller: Montreal just happened. Montreal's been meeting pretty regularly for a couple years now and they they have a good group good sized group. Denver has one coming up September sixteenth. It's our big Denver startup week one. Jason Cole, one of the owners of CTO lunches, helps organize that, even though he doesn't live in Denver anymore. And we have over a hundred people come almost every year. And then September ninth is Salt Lake City, and I'm sure that there's lunches that we Also don't know about there's there's a lot that happen every month, but those are some of the ones that are coming up. go go sign up if you're if you're in one of those cities. And if there's not a lunch in your city, start one. Send send us an email and we'll help you get one off the ground. We can make it pretty easy. We have the infrastructure already there. We do these all over the place. And most people just show up and buy their own lunch, to be clear. You don't have to pay for lunch every time. it's good good community and good good people and good food and good company. Hopefully. sometimes the food's not great. The company's always great. that's Rishi: Ha ha. Kendall Miller: the whole that's the whole pitch for CTO lunches. Rishi: the company is great, you know. Little little fun fact. the CTO launches Denver, and I'm pretty sure it was a startup week one. Maybe was it eight years ago? Kendall Miller: I think it was the first first start up week one, yeah, long time ago. Rishi: Yeah, I I sat across from this guy named Kendall and we had j we just met. and it was excellent. It's been eight years of a ton of fun and getting to know you and happened happened to us right there. It can it's a continual thing. it's the benefit of C T lungeons. It's awesome. Kendall Miller: So check out CTOLunches.com. It's free to join. and yeah, thanks for thanks for coming. We're gonna be doing more of these, so we'll see how it goes as we adjust into it.