AI Economics Hangover

Welcome back to another episode of the Knot Brothers podcast where today we're going to talk about developing situation in marketplace. And that's the potential implosion of some of the AI tokens and AI usage. Recently we've seen companies like Microsoft canceling their cloud code licenses in favor, in part in favor of using a cowork, which

God bless anybody who's forced to do that. But also we're seeing other companies coming out saying, hey, we're not really seeing the benefits from AI that we once were or that we thought or that we expected. even have Sam Altman with OpenAI being like, you know, that stuff we said we're going to be able to do, I don't actually know that we're going to be able to do that now. And that's on top of things that we've talked about before, like the redefinition of AGI and everybody is trying their damnedest.

to redefine what AGI means because I think they've realized that they very much have over-promised what the capabilities of the current technology is. So we're going to rant about this for a little bit and see what we wind up. Mark, where do want to start? What's most interesting to you?

I from a business perspective, one of the most interesting things to me is the over-promising that we have heard from the tech companies and maybe the most interesting and disappointing at the same time thing has been...

the CEOs and executives of especially big companies across the board, kind of being early adopters and early proponents and just kind of taking the tech bros word for it of what this technology can do and what it's going to become. Right? Like there, there have been massive, massive, massive moves made across big companies, not just in big tech, but just big Co's across the, across the U S and all under the guise of

you know, whether they say it or not, all under the guise of AI is going to deliver and it's gonna, it's gonna deliver massive shareholder value and it's gonna fundamentally change businesses. And while some of that is true, I think the extent at which it can do those things is being challenged right now. And that's what we're feeling in the marketplace right now. And some of the some of the things you started the podcast saying, which, so you know,

I'm interested to understand what are these executives gonna say next about like, oops.

mean, this is the challenge, right? And there's lots of like, I mean, there are plenty of accounts that I've followed that have poked fun at this the whole time. And like,

I think this is always the way that it is, right? Because you're presenting something that seems like it solves every problem to somebody who doesn't have.

the knowledge to understand what they're being told and they don't have the knowledge to sniff test it, but they're used to being really, really smart and making really good decisions, right? So feeding the idea that AI is going to solve everything and it's going to 10X your profitability and it's gonna, you all of these things to an executive. I mean, it's like taking candy from a baby. They're just gonna lap all that shit up.

And it's interesting because you see people talk about this like, you know, now you have companies spending thousands of dollars a month on tokens and even encouraging and requiring people to spend thousands of dollars a month on tokens. When if we rewind the clock only a few years and you were like, Hey man, can I get a new laptop? And like, I really liked the one that has 32 gigs of RAM. They're like, well, that's going to require VP level approval. And we're talking about like 500 bucks here.

We're like, nope, you get the three year old HP piece of shit model that's been passed down three different times. That's what you get to work with. And now we're like, you want to spend half a million dollars in tokens a year? Sure. Sounds great. Go for it. It's a, you know, and I think that like that total that drunkenness, right? Like the we've talked before about like how AI models get drunk on tokens. And I think the people have gotten drunk on tokens as well.

Yeah.

And I think there's maybe sort of like that a little bit of that sobering up that's happening currently where we're realizing like, hey, we're burning a lot of money and I'm looking at the yield from this. It's not really making sense. And that's when you have companies start to think, look at things and go, you know what, it was actually cheaper when we had people do this than it is trying to make all of this stuff do this or.

And I don't think this is universal. think that's the problem. That's, think, most probably what most people look at immediately. And they're like, OK, AI is just going to go away. I don't think that's the case. AI is incredible tool. If AI never got any better or any smarter than it is right now today, it is still an incredible tool that shifted an entire industry forward by an order of magnitude. And that's fantastic. It enables a lot of really great things. I think the challenge is when we start to see it as like,

This is the pathway to universal high income in the fucking Wally world, like the world that is in Wally, where we all just float around in our fat asses on these carts and nobody wants for anything. And I don't think we're quite there yet. I don't think the technology is going to provide that for us. But can still do some really fantastic things. Like you want to parse through 10,000 pages of shit and pull out some information. Great at it.

Yeah, I think the so there's a question of our companies making this sort of, you know, three steps forward, four steps backwards moment in AI because of fiscal years coming up. There's an argument to be made that Microsoft in particular is doing that because they haven't had super strong earnings and a an easy way to prop up their last quarter profits is to cancel a bunch of cloud licenses and promise that, you know, for the for the next next fiscal, it's going to look different because blah, blah, or

and then just buy a little back. Well, and in their case, also, Microsoft's an interesting one because they're canceling their Claude licenses and in part, they have their own. So they're like, well, you don't need Claude, you can just use Microsoft Cowork. Yeah, it's great. So like, or Copilot, I said Cowork, not Cowork, Copilot.

Yeah, when in vinyl all that or

Which is just as good. No, it's not.

Go pilot.

So they're an interesting case, but I think that's not exclusive to them. And I do think to an extent you're probably right. We saw Amazon do this at the end of the year last year where they laid off a whole bunch of people and just happened to also buy a whole bunch of GPUs around the same time period. And it helped make their books look better, propped up the stock a little bit, and freed up operating capital to go purchase a bunch of GPUs. And they've probably hired back a bunch of those people.

but it's just kind of the games that we play for Wall Street.

I think that's definitely true. But I think, I think the underlying tone though, regardless of the gamesmanship that you play in, in Wall Street, I think the underlying tone of how powerful AI can be for organizations is hitting that sobering moment and probably the best way possible in my mind, right? Cause what we found over the last five months of 2026 is kind of where the limitations are with things like, you know, not just the AI technology, but also things

like OpenClaw or Claude's version of OpenClaw or any of the other massive, know, Frontier Models version of OpenClaw that have kind of emerged. And what we found are limitations. It's incredibly powerful, but there are definitely limitations with what it can do and what it can't do. And so what it does incredibly well is anything repetitive and anything you can teach it to do over and over and over again. What it doesn't do so well is make anything brand new.

And so those limitations are fundamental differences as to what you would need humans to continue to do and invest more time and energy into and what you wouldn't. And honestly, that's not really that much different than any other evolution that we've been through in terms of human capital and workforce, right? Think about manufacturing jobs and automation of manufacturing jobs over time, or the automation of customer service jobs over time. Like those are all things that can be repetitive in nature. You can build a process around, you can teach either a lower level

individual to do that work, or you can teach machines or robots to do that work. And this is kind of the next evolution of the exact same thing, where some of those lower level activities or lower level thinking things are things that can be done over and over again, or to your point, massive data structure things, I need to go pull out this thing and instead of a human spending a week researching it, I can get an answer with AI in a day. Like those are things that are wonderful use cases for

AI and at least where we are right now with model development and where we see the models plateauing probably for the next maybe forever, who knows. But I think you're reaching this point where I think most that are close to what this technology is capable of are starting to say this is probably about what we should expect. And so what can we do with this versus what we thought this was going to be?

What I think that we, some of it's probably fatigue as well, right? Or just like realization of reality to an extent, I guess. Where, you know, we've been told for...

four years now, I think, that, you know, we're two months away from AI writing all of the code and replacing all of these sort of jobs and functions that people perform on a daily basis. And that goalpost has just continually moved time and time again.

And the speed at which we're seeing developments has slowed to like a snail's pace, right? At one point, it was like we were getting new models all the time. We still are getting new models. But you were getting new models that were quantum leaps from the one before. Now we get new models. But in Anthropix case, it's just the old model released under a new number.

And then like they are better to an extent, but they're not leaps and bounds better. The biggest improvements that we're actually seeing right now, in my opinion, are on the open weight models and specifically like some of those smaller scale models where we're seeing more like highly tuned and just capable models on that side. The frontier models aren't doing a whole lot, but I think that becomes the challenge of like.

Once you've consumed the whole of the internet and you've run it through your autocomplete algorithm, what's left to do? And eventually, maybe that's as good as it gets. And you're fighting for the last little bits of optimization that you could possibly get. But at the end of the day, it's a pattern matching machine, and there's only so many patterns to match.

I kind of equate this to the evolution of something like the iPhone, right? The original iPhone came out and it was light years different. It wasn't the first phone. It wasn't even the first smartphone, but it was light years different from what was already in the marketplace. Much like AI originally came out and it was light years different than what was available in the marketplace. We have machine learning for a very long time, which is essentially early days AI.

But this was incredibly different. was able to reason, least the appearance of reason, in non-complex chat form. It could be well understood. Very different. that's not gotten just incrementally improved. It's gotten light years different. So compare the original iPhone with now what would be on iPhone 17. And if you compare those two, they are fundamentally different products, yet have a similar form factor.

But the difference between an iPhone 14, 15, 16, and 17 is not that much different. It's a couple of hardware specs here and there. if it wasn't artificially degraded from software, it wouldn't really even be known to be that much different in the user experience case. And so I think models are approaching that kind of evolution where, we'll have new models that are incrementally better over time. But it's the difference between an iPhone 16 and an iPhone 17.

And what we're really looking for, I think, to your point with some of these other models are like keeping on the Apple ecosystem are brand new products that are maybe part of a similar ecosystem, things that sit alongside, right? So an Apple watch, an iPad, stuff that supports that ecosystem, but it's not necessarily the model itself that's driving such improvement. So I think about that as like layers. So you have some of the layers of software that sit around all these things. So essentially the

the open call equivalence, things that are opening up and being connected to these things in a way that makes them more powerful. Or it could be smaller models that are adapting similar model structures, but are very, very individual use case-esque. Think like medical practices or even code or anything else that has an individual application. It's not this massive model that I can do at all. It's much smaller model that's highly capable and highly tuned within a specific data set.

So that's where I see the value coming in the future.

Yeah, I definitely agree. think that's where I think that's where we're going to go next. I think we're going to see these the like large scale generalist models that promise to take over the world. They're just going to kind of be a little more quiet about that. That message, right, because I think they've kind of figured out they can't keep peddling that message anymore. It's why they keep trying to move like, you know, the CEO of Nvidia claimed

that open claw was proved that we've achieved a GI. I'm like, you know, it's not. That's not what was promised. That's not what a GI is. It's not what was promised. It's not what we've been talking about. You goalpost moving fuck. And like, it's, it's not that open claw and like the application of these technologies and things are are bad by any stretch of the imagination. They're great.

They're really, really cool applications of what we have now. And I think that even, you know, the guy who created OpenClaw went and, you know, he got snatched up by OpenAI and is working with them. I think since then we've seen, like I've seen, I've been a pretty big like codex hater for, since kind of the beginning, right? And like codex five five or not codex, GPT five five came out sort of after he went there.

And there was a lot of tuning that went on with the open claw harness. And I have to imagine working with just the engineers and whoever are, are, are tuning the actual problem because it used to be the case that GPT would just like, I mean, it was like pulling teeth to get it to do anything. It would just, you'd say, Hey, go do this thing. And it's like, all right, I'm thinking, I'm thinking about making it blue. Would you like me to update the colors? You're like, for the, for fuck's sake. Yes. I told you to build the whole thing.

And, and like GPT-55 is the exact opposite. It'll just churn. And like I've had it turn on stuff for like an hour where I just throw it at something that's really difficult. I'm like, Hey, go do this. Then have this agent review it, fix all that stuff. I'll be back later and I'll check in on it. And it's just running in like a TMUX session. And I'll just sit there and churn and churn and churn and produce. Like the output is really, really, really good. Good enough that it makes Opus four seven look really stupid.

trying to tackle the same task right now, which is surprising.

Well, I think the economics have shifted how the frontier models are trying to get those models to run to, right? It's better user experience for sure in most cases for the models just to churn. But it's also better tokenization for those models, which are relying on those model usages, right? So Claude kind of pioneered that with, you prompts along the way, but it was largely pretty good at being kind of self

sufficient as it as you gave it tasks to go do, it would figure out what to go do and how to go do it. Whereas GPT, it would kind of sort of go try to do that. But to your point, it would give you a lot of false positives and think it did things that it actually didn't do. And there's a lot of rework and it wanted to think its way through the process over and over and over again, as opposed to like finishing something and then moving on to the next thing. And I think in large part, that is a model progression. Yeah.

Well, if you're time to time claim that it's done. That's my favorite one is when it's like, all right, I'm done. Like, I don't think you are. Did you just lie to me and tell me you made this thing? He's like, yeah, he caught me.

Yeah, those are the best moments. like, did you?

Did you just make all the tests say return true?

because you couldn't figure it out. I'm like, yeah.

my favorite was I was doing market research for a product that we're getting ready to launch. And I'm like, okay, make sure you actually do the market, do the research and actually go find this stuff and cite your sources. And it comes back with like this list of things. I'm like, did you actually go do the research? Or did you just return like, you just make stuff up? Because I you caught me, I just made it up. It's like, you serious? I literally told you not to do that. You did it anyway.

But this maybe articulates some of the challenge and maybe part of what people, maybe what's leading to this sort of like shifting of winds, right? When it comes to the confidence that companies and executives have to putting time and energy and money into these systems is, you know, when you use it, like we've talked about before, they're fucking dopamine machines. Like that's.

That's what they're tuned for, right? They're tuned for that dopamine hit where you throw something in it and you say, hey, go do this thing, or you ask it a question and it gives you an answer and it's confident and it makes you feel like you're just the smartest person in the world for even having asked that question.

And then later, you start to find the cracks. You start to realize that the market research it did, it didn't actually do market research. The answers it gave you, well, they were all hallucinated. So you find that, I can't really rely on this data, and I need to go back behind it anyways.

And like there are ways, like there's certain tasks you can identify that it's good at. There's ways to kind of prompt your way into it. can pair other agents to review it. And like there's all sorts of strategies and things to do. eventually, especially when you're talking about people who are not very technically savvy and not very technically capable and not very used to that cycle of working, their solution is like, well, I don't trust it anymore. So don't use that. It's kind of like when I was in school.

the I remember one of the things that's like laughable now is like Wikipedia is not a trusted source. I'm like

I mean, it's pretty reliable. There certainly could be some things on it that are not quite correct, especially like more niche things, but there's a lot of good information with actual citations on there. And now, you know, it is probably accepted at most schools to cite Wikipedia as a source, right? And what I did is I just went to Wikipedia anyways, and then would just cite the source that was on Wikipedia.

Ha

made perfect sense to me.

Wikipedia and Reddit it's probably the two most like primary sources that have trained AI anyway.

Yeah, I mean, Reddit is an interesting one because like Reddit is always, you know, it's like one of the most searched terms is, know, something, something, Reddit. Because, you know, that's the only place to get like truly authentic. You're going to know how people feel about something if it's on Reddit. You can watch all the review videos.

Instantly.

And it's a real challenge, man. Like I have, like I watch, you know, things all the time and see like, you know, what do people think of this product? And like, I have products that I know are actual dog shit that I watched somebody just like make a review video and just gush about how great it is. Like, this is, this is an awful product. And I don't trust you anymore to make any product recommendations because you're not willing to call the baby ugly. And like, that's, that's a challenge. But on Reddit, I never have to worry about that.

if they're gonna call the baby ugly, because they will.

That's true. Well, so getting back to our discussion on AI and economics, tokenization, those sorts of things. One of the other areas that have happened recently is Sam Altman, the biggest proponent of AI is going to take everybody's jobs. You should definitely be ready for that. And economically, we don't know how we're going to handle that yet. Just we'll figure it out together. It's pretty much been his stance, which was a theory.

It was probably even a relatively good theory for some period of time. And even he's coming out at this point and saying, you know what, that was a theory. And I don't think that theory is going to hold true any longer. And it's largely because of what we're talking about. have just limitations with what the models are going to be capable of. And that's okay. I think the general acknowledgement just needs to be, all right, well, if these are the tools in our toolbox, how do we use them?

for the benefit of our business, of our employees, of our people, of our customers, and how do you get the most out of it? That's, think, the trend that we're going to be on for the rest of this year. don't think we're going to be in this continued wave of AI hallucination. Yeah, well, it is going to change the world, it has, but I don't think it's going to have the negative economic implications that...

Yeah, it's gonna change the world.

have been pontificated for the better part of probably a year, year and a half.

I mean the question is like when's like do we think

There's like multiple questions, right? To stack up. One of them is like, it a bubble and it's just gonna pop? Not necessarily think it is. don't think, I think this is, I mean, there was the, we saw what happened with the .com bubble and like that was a different animal driven by a lot of things. There are similarities for sure. But the one thing that we can look back at the .com bubble and say is like, I mean, all that happened, it just was a little bit too early.

Like if all of that happened today, it may have never burst. Technology took some time to catch up. That's the advantage that think AI has had this go around as technology is already here. Distribution layers are already here. You can get access to people to use your AI products on their cell phone while they're doing anything anywhere.

Yep.

And that just wasn't possible back in the dot com bubble era.

Mm-hmm.

So I think that makes it fundamentally mentally different. I do think that the extrapolations of what's going to happen and the issues with power and AI data centers popping up everywhere, I think those are starting to erode a little bit. And hopefully that means some benefits. Hopefully that means that we can buy some fucking memory for less than $1,000 in the not so distant future.

you

But you have like real ripple effects, right? the between the promises that OpenAI made and the promises that Anthropic made, you had the largest manufacturer of consumer memory close and say, hey, we're not making consumer memory anymore. We're only making a memory for AI. And that has wrecked, absolutely wrecked NVMe and RAM prices.

So I would love to see some of this stuff come like normalized just to bring some of that back to reality. Open AI is already backed out of their stuff. They've said, you know, all of that, all of those data centers we promised like, yeah, it's like, we don't actually need that much compute. We can just rent it instead. Anthropic did the deal with X, which was I think was really

interesting, right? It kind of raises the question of like, well, what the hell are they doing with, with Grok, right? Like, are they just kind of giving up on Grok? Or is it just gonna always be like a third tier sidecar? And they're not really vying for frontier capability anymore, because they've leased out the entirety of one of their data centers for Anthropic. And I'm assuming we'll probably look at others.

Yeah, there's no doubt that the winds are shifting and it's not a matter of we're to have these massive players building data centers everywhere and trying to figure out how to power them. I think it's going to slow down.

I think, I think largely those deals have already been struck, which is why we're not hearing as much about it, right? I think, I think they are still, I think either they've been struck or they're in near finalization. Like there's this massive one that's supposed to be going out in Utah that I think they're still trying to figure it out, but it's like, it's huge. It's like 50 square miles of a data center. Like that's a, that's a real thing, a real commitment. That would be the largest one, at least in North America. I don't know about the world, but it'll be gigantic.

So I think that's some things that's slowing down. The part that I don't think is vaporware and I don't think is going to slow down is the consumption side of AI utilization. think we've been talking about like frontier models and over-promising and all those sorts of things. But I think we are still at the early stages of the implementation layer of...

what we can use AI in, how often, how frequently. We're going to very quickly get very used to in every application and every place that we interact with technology, just being able to ask a simple question and getting a complex answer back, regardless of where you are and what you're using. It's the same way that we've sort of been trained over some period of time by Amazon to expect an Amazon experience when shopping for anything anywhere.

Yeah, if it only takes more than two days to get here, it's something that's gone wrong.

And you better have everything available. Like, my gosh, you don't stock that? Why don't you stock that? I don't understand. And so it's this rewiring of the user experience. And I think that's going to happen everywhere in the implementation layer. And that's going to drive demand substantially across model utilization. Now, is that frontier model utilization? Is it those specific models that we've been talking about? I think it's probably on more fine-tuned models over time.

It'll start at the frontier layer and then it'll, it'll, it'll drill down into those more fine tune model utilization. But you have, you can have groups like Google, which Google started in a very behind position in the AI race. And I think you can make an argument there. They're, they're, they may be number one now with some of the, the, the promises that they started making last week and some of the changes they started making last week, at least in the consumer layer, because what they've been, what they've now done.

Mm-hmm.

Is they've introduced a whole bunch of new capabilities set to all their product pieces that are at the consumer layer, right? Consumer or business use layer that weren't there before. And so they're extremely well positioned to become like become well adopted as compared to like an open AI, which kind of sits as a sidecar on top of other things. This is now built in the, as a fundamental piece of all the Google products.

Yeah.

And it uses Gemini, which is their model, obviously, as the they kind of own both pieces.

depend on the and we'll probably see like shifts there have always been like these like what each model is good for even within the frontier class models right like Gemini has always been a little bit better at like creative things look and feel design like that has always been like Gemini's bag

I think, you know, because as you as I think about it, like if you say like Gemini has, you know, maybe the best model, I kind of like scoff. I'm like, no, they don't. But I'm also thinking about it from my vantage point and my use cases and my use cases are mostly writing code and technology like technical, right? From a technical problem solving perspective and writing code like Gemini. Ass.

And I would not choose like codex or GBT-5.5 specifically is like, that's the God tier right now. At least for a moment, that'll change next week. We are right about the integration. think the thing that Google has done is they've been able to integrate that into their product. So when we think about, like we've been in the Google workspace ecosystem forever. I think that's what they're calling it now, workspaces.

They change the name like every other Tuesday, so I'm not sure.

Yeah, Google Apps back in the day. That's what I still call it most of the time, but I think it's Google Workspace now. So they're able to bring their model capabilities directly to the surface, directly to the user inside of those platforms. you know, I've noticed they have sort of that copilot approach is really what they're doing.

and making it where you can ask Gemini to help you find files or to organize something or do something in Google Drive, which can be beneficial. And it knows how to read the file types, especially the Google pages and Google Sheets and that sort of stuff, which are at atypical file type and other things to struggle with. Because they can't access them.

So from that perspective, I think it is good. I think the other thing is they're building more AI in a search. made this, you know, I have a prediction pinned in Slack from two, three years ago where I kind of outlined like what my prediction of like what Google is gonna do. And they're on like step four or five of what I outlined because now the AI option is on the homepage and

In many cases, it's the default option that you get.

Well, they just extended that even last week. I don't know if it's released yet, but what the next version of some of this stuff is coming is basically the experience that you get with GPT or with Claude, where you can give it extended, extended instructions, extended, whatever. And it will by default return the AI results unless the search results are better.

And so you're essentially building in that same experience within what we've always known as to be a Google search is now going to be primarily powered by AI, even more so than it already was to the extent that even image creation will be done just like it is in Claude and GPT or other things create like the creation of things, not just the finding of things, which is what we've historically known Google to be.

Yeah, and they have the data. They've been doing this stuff for a long time. They were using machine learning algorithms before anyways to prioritize their search results. This just changes how they present the results and also how they rank those and figure out relevancy and those sorts of things, which has completely destroyed any understanding of what SEO is in the modern era. It's almost like you have to produce actual content that people actually want to get to now.

in order to show up when somebody searches, which I think is fantastic, right? Like get rid of all the bullshit, get rid of people producing things just for the robots to rank you highly, like produce something somebody wants to read. It should kind of be the rule. But I think that the applications are definitely where it's gonna go. We'll see more applications of all the frontier models. And I think we'll probably see some other models creep up.

even ones that are built on top of other ones. like you have, even within the development world, you have like cursor with their custom models and a couple others that are out there with custom models that are really just like, they're building on top of an open weight model or something and tuning it. But it's for a specific task and.

When you do it right, it can perform better than generalist frontier models or untuned open weight models. So I think that's where we go next. And we start to see the real benefits of these. Not so much just promises of utopia, but real applications. And I think this is the same thing that we saw. We've seen this time and time before with.

with every revolutionary sort of event or product that happens, right? You have like the industrialization that we saw back in the day. We didn't see, I wasn't alive yet. But like the predictions of where that would go and where it actually went are I think a little bit different. But they're kind of reaches this point where you sort of reach the plateau where

we can now just start keep incrementally applying this where it makes sense. And introducing other people to the technology because I think that we're also, we eat, and breathe this every single day and we surround ourselves with people who also do. So for us, this is like a big echo chamber of everybody doing the same stuff. But like if we walk down the street right now and we start talking about the same thing, people will look at us like we have a third fucking eye.

because most people aren't interacting at this level even, or have an awareness of, and some people just haven't even used AI yet. Or it's just like, they've just started chatting with ChatGBT. So I think there's still a lot of marketplace to build and figure out before we can sort of it the end or say like, hey, we know exactly where it's going, because I think there's still a lot left.

I think what you just said is, the reason that everyone has been so bullish on the growth of GPT, of the growth of, of Anthropic, the growth of things like perplexity. It's that's the reason is because there's so much untapped user base. That you could, and if you extrapolate that even to like a low level usage across these things, especially by way of, you know, application layers, can, it can.

all of a sudden become something that is extraordinarily different than it is now. And I think that's what everyone's banking on. And I think that's whether the demand is going to be right and whether we've gotten the compute and data systems and all the centers and all these different things right or wrong is anybody's guess at this point. But I do think we're still in early days of overall demand generation.

And it's not going to look like what most people have been using it today as, which is you have your own individual GPT account or you have your own individual cloud account or anything like that. That's not where the power is. The power is going to be in the application.

of people will use them inside of an application of some sort, something they're already using or intending to use, and it's something that augments and makes the application or the interaction better, more so than being a direct interaction, which is largely the case with most tools and technology today anyway.

Yeah, it's not much different. It's essentially, you know, the old promise of a chat bot and any application you've ever tried to go into, but one that's actually good instead of one that's just hot garbage that you're interacting with.

mean, hell, that would be an improvement. If the chat bots that you're gonna force people to use could actually be decent and answer questions, like that's a win, I'd take that.

Yeah, I'd take that too. Product idea.

Alright well, I think we beat this one down today

Think so too. Until next time.

time.