Anthropic vs. The Pentagon + AI Agents Are Rewriting Software | TSG Ep. 1029

Hey, everyone. Happy Friday. Wow.

Think, let me say that again. Happy Friday. It's good to be Friday.

I, I thought my weekend was getting off to an early start, but it's not, but I'm happy to be here on the gang, especially with this great gang of folks we've got here. Let me introduce you to them. Some of 'em I haven't had the pleasure of speaking with in a few weeks.

Our, our friend Wiki Wang Wiki. How are you? How are you?

Good. Yeah, good. Also, I haven't, I haven't seen her in a very long time, my friend Tracy Reagan.

Hey, Tracy. Good to reconnect and have you on today with me on at the same time, because we always have fun. Absolutely.

And then these two, I feel like we've been on, I don't know, three or four times this week with you, Jack Poller and, uh, John Swartz. Yeah. Nice.

Hi. Alright, let's jump into things on this fine Friday. You know, you gotta give Dario a Modi credit, right?

The, the big bed wolf huffed and puffed and threatened to blow his house down and, and the, uh, they took a stand and he stuck by his morals and ethics and positions. I wrote about this earlier in the week before a modi's, uh, latest stand, you know, outlining it. But John, what, what do you think?

Well, you know, in the classic refrain, this is a developing story, and there's a new, and there's a new little, um, angle here, and that's Sam Altman actually, and we'll mention that in a couple seconds. But first, you're right, Alan, we talked about this yesterday. There's this big confrontation, or face off, it's almost like a, a high stakes scheme of chicken between Anthropic and the Defense Department over the use of Claude and this $200 million Pentagon deal.

Uh, hegseth, the defense secretary wanted, uh, some guardrails, lessons, maybe softens and anthropic acquiesced somewhat in a policy statement that they issued. There's a deadline of Friday today, actually, in a few hours on this deal. And, and what was pushing Anthropic was they, they, uh, they're in very big danger of losing this deal to either open AI or Google or, or someone else.

And, um, they need the revenue. And so anyway, there was this head-on collision, and it looked as philanthropic had backed away. It looks like they didn't seize, as you had written, Alan, they didn't seize their apple moments going back to 2016, and we can talk about that later.

Since then, there has been a staff meeting that Sam Altman held at OpenAI last night, and he says that that company OpenAI is working on a deal to resolve the impasse between Enro Bank and Pentagon, according to the Wall Street Journal. So we have this classic kind of conundrum where you wanna maintain an philanthropics case, the AI safety first, uh, reputation they have, which they have really, really ridden hard ala Apple versus Go government contract that if they not only lose it, they are in danger of being quote unquote blacklisted and losing co contact and contracts with, uh, suppliers. So it's a, it's a classic confrontation and anthropic to its nature, and we'll see if it, if it stands behind what it says, if it doesn't change things again, is, is putting a stand and, and standing up for what it's, uh, thinks is a, uh, responsible use of ai.

Jack, you usually help the, you know, uh, not help, but you usually support the governments and, and the government. Oh, no, no, no, no, no. I just What Do you think?

Well, I I, I have two major thoughts on this, but, uh, first is that this is really a, uh, what's the polite way to put it? This is a p*****g contest between two very big people of a very big, um, personalities trying to exert who's the boss here. Uh, in general, I understand the both red lines that Anthropic is taking, and I agree with them.

I don't think what the government should be involved in mass surveillance of American citizens. And most especially, we've had, you know, I don't know this, the, the, the concept of what happens when AI goes wrong in autonomous weapons and stuff as something sci-fi writers have been exploring for a better part of a century now. And it never turns out good.

The ais are way too early. So I do believe and agree with the red line that we should be liming ourselves to having human in the loop on anything and everything that AI does in terms of far as target selection and web weapons control and who it attacks and when and what and where. Um, I don't think the Pentagon, the DOD wants to use these things that way, but I also believe they don't want have their hands tied.

And so, like I said, this really is, uh, uh, uh, who's, who is you? You don't, you don't think the Pentagon wants to use that? No, I, not in, I do not.

Pentagon and so forth. Wait, well, there is a difference. I'll be, it's a subtle difference here, but there is a difference between deciding what targets should be attacked and getting information about them and autonomous decision making without a human.

And Lily, and I think that this is the argument here, and I believe that, you know, o Modi's Red Line, and I'm sorry if I'm mispronouncing that, but I believe his red line really is about making sure there's a human in the loop. And you cannot have an autonomous, uh, actor in there without a human. And, and, you know, look, AI hallucinate all the time, time, and I love Philanthropics products, and I think they're very, very, very good.

But they do hallucinate, and it would be really bad if they hallucinated and said, Hey, based on all of the, oh, I don't know, let's say all of the, uh, um, LLMs being, uh, uh, uh, attacked that they're going to attack Pete Hegseth in the Department of 10 defense because they've been triggered to believe that he's evil. Right? And do that autonomously.

And so I think that right, that, that that whole concept of making, you know, guiding decision making is different than making the decisions without the human and then acting upon those decisions and pulling the trigger. And I think that's a red line that the Pentagon should be willing to live with, and I don't understand, other than it's a personality clash, why they're not willing to live with these. Yeah.

You know, what, we talked, I'll be really quick. We talked about this yesterday, a group of us on the show, and our kind of conclusion was that there was gonna be a lot of posturing and perhaps some kind of meeting somewhere in between. There might be some sort of negotiation or talking through the elements of this, where this was salvageable.

But now it seems as if I think open AI is just gonna try to swoop in and and acquiesce to whatever is asked of them. Yeah. I I you Have a question here.

Like, if the AI Foundation model become kind of like a public utility kind of business, who should be the, like, party to, uh, guide and build up the standards? This one, I'm not quite sure. Maybe we can discuss a little bit more.

I, I not, I would not trying be, but he's dead. I'm sorry, what? Chay and Tropics trying to be that right?

They're Well for their own, for their own right. Yeah. They have the clawed constitution.

So here, here's my take on this for whatever it's worth. I wasn't on yesterday's show, John, I was out. But, um, I did write the story.

I think I picked it. Yes. Um, you could agree or disagree with Emoti and Anthropics position, but at the end of the day, it's their company and their right to decide how their tools are used.

And if the Department of Defense doesn't want to use it gi with those rules, they don't have to. But here's where it gets heavy handed. And this is typical of this administration, like a petulant 5-year-old who needs to get sent to the bedroom without dinner.

They say, ODN anthropic, if you don't do this, not only are we not going to use you, not pay you, but no, we're gonna basically declare you persona non grata, right? We're gonna declare you a supply chain threat that like by statute cuts you out of a lot more work than just the DOD stuff. They're looking to be punitively pub, publicly punitively punishing this company.

'cause they won't roll over and play dead to boxcar Jesus or whatever this guy calls himself when he is not drinking. Okay? Amen.

Amen. And, and that's really, that's heavy handed. That's totalitarianism.

That's wrong. We should not even be having this conversation. We as a group should not, should not even be in the news.

This should be gentleman making and gentle women making decisions in a, in a, in a board room sitting down and talking things out and figuring this stuff out. Nobody wants nobody wants a Rob. I'm Actually, I'm actually glad they're talking about this publicly.

'cause I think anthropic No, the point is John gets to a backbone is and stand up for something. I mean, all these other guys are pushovers and I, I really do like this. And if philanthropic wants to be considered the apple of its of its era or of its industry, this is its moment to show that.

Yeah, Yeah, but this is not needed. Pushovers, They want Money. This is not needed.

Yeah. They don't need to blacklist a company because the company says our product's not ready. That's, That's what this, that's The problem I'm having.

Yeah, that's the problem. And it, and with this administration, everything is like that, which means that they're not gonna get the best product in their hands because they're not working with companies to say, this is what we wanna do in the future. What do we need to do?

So we don't have a Robocop moment. Seriously. I, I don't have a problem with the Department of Defense saying, any tool we use, we are gonna have the right to use it as we see fit.

And if a company doesn't wanna acquiesce to that, that's okay. You shake hands. It's a big market.

There's a big market outside of the DOD. Right? And you go on, you don't go out to punish the company because they don't tow the line.

You know, when you're asking them to salute The, the iron, that's problem. The irony here is the defense department is taking to, uh, completely, uh, 180 degree orthogonal positions. Right?

They're playing to Tric. Yeah. Right.

They, right, they're, they're declaring that, uh, anthropic is a critical infrastructure thing that therefore they can use the defense product Act to, uh, force them to, to acquiesce, which is sort of A wartime thing. But that's typical of, that's what We used, we've used it Elsewhere, agency in a war. Right.

We've, we've used that elsewhere. And other, you know, the Biden administration used that power too, to force people to do things. But you, you, it, it seems counterintuitive to declare that anthropic is super critical to the Defense Department and a supply chain risk at the same time that just, that just doesn't make sense.

Right. It's either supply chain risk or it's not. Yeah.

It's outta Every, every, so, you know, budget, They can Yeah. And like I said, I, I think this is, this is not going to end well for the defense department regardless whether, whether they get what they want out of anthropic or not, they've already lost in the bottle of public opinion. Yeah, exactly.

That's, anthropic has already won because, and at this point, anthropic, if they acquiesce to the Pentagon, they're gonna lose a whole lot in the public opinion, I think. And they'll actually lose smartness then they win. You, you're right, you're right, Jack.

They realize that really quickly. I mean, somewhat quickly. And you know what, I think in the end, this is gonna be actually do wonders for them, not beyond Yeah, no.

They're gonna make their bones around this. Right? Absolutely.

I I lean heavily Up something though. I I just, before I jumped in here, I, I was working on an article, I probably won't go out till Monday, but it's based upon some research by the folks that I didn't even know how you pronounce it. Ipsos, I-P-S-O-S, Ipsos, Ipsos.

Very interesting thing here. China, 83% of people surveyed in China, whether you believe it or not, we could argue, or another day, 83% of people in China very excited by the potential of ai rest, 17% or whatever hesitant, you know, about AI and, and its potential to impact themselves and humanity. And you know, the downside, if you will, in the US only 39%, a clear minority are excited by ai.

A majority of folks are frankly afraid of AI and, and are hesitant to embrace it. And for something that's gonna be so disruptive and change the world and everything else, that's not a good number. And, and by the way, the US isn't alone.

If you look at the us, the uk and I forgot what other, you know, western style country, the numbers are a little better. The US is like kind of down at the bottom of, of excitement around ai. But, you know, is this one of the reasons why we, we see this stuff playing out.

We, we, you know, we hear so much about all of these things. I don't think we've ever had a technology so politicized, So polarized And polarized, right? But it's, it's being used as a political, uh, tool now.

Um, I've, you know, we, we had a couple of, uh, of these talks back. We talked about will will AI be used and the political, um, you know, discussions rolling into November saying it's gonna take our jobs, kind of like immigrants taking our jobs now. It's gonna be ai, we're making ai the enemy at the worst possible time that we could and discussion.

We also Yeah. We're making, we're making it the enemy too, about affordability in terms of like energy costs and Well, Yeah, the whole data center thing, the internet. Yeah.

And, and to me, a again, we should not even be having a discussion about what Anthropic and the Pentagon's talking about, because it should be a working partnership between all of these companies to, to come up with some agreements and how the stuff is gonna work. It is just a pol. Everything's po, everything is polarized.

Everything's political. Political. And I am so done with it.

I am so sick. Welcome To the last decade. Yeah.

Oh my Goodness. You know, that's down rabbit hole. That's the world we live in.

But it's essential for us to embrace AI and to get it right. It's essential for our economy, it's essential for our citizens, it's essential for our government. It's essential for our health, our medicine.

And if we continue down this road, we're gonna get it wrong. And we don't have the option of getting it wrong. We have to get AI right, that is for everybody.

That is for everybody on the entire planet. This is an important shift in our technology and how it's moving us forward, and we must get it right. And kudos to anthro, uh, you know, through anthropic for saying, Hey, we don't have it right yet.

We are not ready for that. And they shouldn't be blacklisted for saying it. Well, we gotta move on, guys.

But it is a, it's a heated, emotionally charged question. And I think, you know, the bigger issue of how do we view AI friend or foe is, is something we're gonna have to come to terms with. But, um, let's move over to the rise of the autonomous speaking of ai, right?

The rise of the autonomous workforce, right. And Tracy, you mentioned, are we afraid of our jobs? Well, maybe Jack, what's this one about?

Well, I think, you know, I was gonna bring that up if you didn't. I think part of the problem is, um, the language we use to describe these things. So ServiceNow made an announcement about their latest feature and incorporating ai, and they're talking about it as an autonomous digital, a AI workforce.

And I really think that that's a poor framing of it. But what they have done is they have basically said, let's use AI as a natural language interface and agen ai, excuse me, on the backend to help with what would be considered tier one, level one, um, uh, AI help desk type tasks, things like password resets, account lockouts, uh, all of the very routine stuff that, uh, it people handle on a daily basis that takes up an inordinate amount of time is very easy to solve with, uh, automated workflows. And so what and what, uh, ServiceNow has done is basically added some automation and some AI agents, uh, around it.

What the real story here though, is that the value is in the governance and the control plane and the platforms. I think Wiki mentioned this earlier, talking about platforms. There is a drive from ServiceNow and Workday and others to build an entire platform that helps people automate their, their basic business operations and their, their, all of these things that are very repetitive, very simple tasks that can be helped by ai, which can gather additional, more context, can work faster as long as you put in the right guardrails.

And I think, uh, as I said, the real value here isn't in the workforce, but it's in the control plane that controls all of the AI agents and being able to put the right guardrails, the right security around it. Yeah. Uh, I, I personally am worried, uh, I feel less science, very positive actually.

You can see the step by step what, uh, ServiceNow doing, right? Like last year I did a podcast for ServiceNow and uh, Salesforce regarding how they try to be a Gentech platform. And you can also see they acquire some company like identity security or something, try to make the, uh, AI agents become more governance and also make all those workflow works.

Even though like, people is kind of like scared, okay, AI agent will replace their work, but think about those works like ServiceNow try to provide, right? It's like you can't do like a 24 hours monitoring in person, like at your sleeping time, like take turns or something, which is really hard for human. But now, like yeah.

Agent can help you. This is part of the helping part. So I, I like this part which make me don't understand for the current market is even the, sorry, it's not this so much, right?

The stock stock still done that much for me. I don't understand the logic at all. I dunno, how do you guys look at this?

Like, If we understood the logic, we'd all be making a ton of money. I'm not here. Well, Lemme tell you some of the logic from a company, you know, someone who's running a small company, small company, this is really important.

I want, I want this to be successful for them. Yeah. Because I want it to become common.

I don't need to personally for a small company, I don't wanna have to pay, uh, first level support people. I would love all of that to be automated and at two in the morning, right? Because then we have a global reach.

This is part of what I was talking about I think last week, is the democratization that AI does bring across companies. Because we can act and look more like a big company if we have AI doing the same things as big companies are doing. So I'm, I was like, yay, I want this in my company.

Yeah. That was using this, That was something I was gonna ask. Uh, all of you.

I mean, Tracy, Jack Wiki this idea, I keep getting these announcements, you know, and I get a little bit jaded by, a little bit jaded by ServiceNow and Salesforce inundating me and Mike every week with something. And it always seems as if it's kind of an incremental or a recasting of what they're trying to do because they're trying to reach an audience. And we refers to the, the stock price.

I mean, and there's a perception about AI agents not really catching on or gaining traction as fast as we expected this to happen. And I'm wondering if, I mean, I'm hearing two points of view on, say, the ServiceNow announcement. I just wonder are they kind of like looking for a venue or an avenue to reach more customers?

Are they trying to work this out? Or is this, is this a logical progression of what they're doing? So I see, I'm glad you brought it up, John.

I, there's two pieces to this. One is these agents themselves, AI specialists, right? Autonomous agents themselves.

To me, they are scared to death of open call and open source dime, you know, free dime store agents that you can make, do anything, right? And so they need to show the value of why their specialized agents are somehow much more valuable and quite frankly, something you could deploy for free as a, and this is classic open source disruption, right, chase? Yes.

That's what I'm thinking. When are we gonna use those for my company? Open source.

The the other angle though is as Wiki mentioned, and John, I think you touched on, you know, you gotta give ServiceNow and Salesforce credit. 'cause from day one on this, they recognized the threat and they clearly staked out their high ground, right? Their defensible ground as being the control plane, being the manager of agents.

And, and the question is, do they only manage their own agents? Will they manage open claw? Will they manage each other's, right?

How, how wide is that extend in terms of this agent control plane agent management? Because no matter how many AI specialists ServiceNow spins up, they'll always be other specialties and other agent uses that they don't have, but that they still don't want to give up the control plane for, for, and I, I think that's what both Salesforce and, and ServiceNow are angling towards, right? And on one hand, the agents themselves may wind up becoming Linux, right?

Or Kubernetes or something where they're open source and, and you know, people are going to use whatever they want without regard to, uh, that but on. And then on the other hand, they realize their real opportunity is the control plane. There's, there's another part to this though, Alan, which is that, you know, I think, John, you were asking, you know, alluding to this in the questions, and so is Wiki in terms of the stock market, which is how do we position and market this?

And it's really all about putting an AI tag on top of what they do. And, and I'm not saying that they're not using ai, they are using ai. But if you look at the business model for ServiceNow, Salesforce, Workday, and these other companies, they are all about automating very boring routine tasks in the business that are all necessary tasks.

But I can tell you that no HR person wants to have the 44th conversation today about what benefits that a, an employee gets when the employee can find out that information or an agent can provide that information for 'em. You know, you are two, you know, you're a 40-year-old employee with two, you know, a wife and two kids. Here's your our partner and two kids.

Here's what benefits you get. You know, I don't need to go over that with you. And new employees don't need to hear the pitch over and over and over again, or sign up for the 401k or whatever.

It's, right. So they are in the business of automating, they're now going to put automation, they're gonna put AI tags on all of this as sales and marketing efforts. And I think that, I don't necessarily think that's a bad idea.

And they're not AI washing, they actually are using the ai. But this is a lot of, how do we get out there and get more customers or inform our customers of what's, what's available. Um, as, as Tracy said, as you know, I'm also a small business owner.

Do I wanna deal with that stuff with my employees? Absolutely not. I would much rather be able to have some automation do this for me.

And there's a lot of value in that. People also don't wanna be the first level of tech support, doesn't want to answer the phone call and say, oh, you're locked out. Okay.

Or, or my computer doesn work. Have you tried rebooting it? Nobody likes being on the other end of that call either, right?

And so we go, I mean, Tracy be in agreement, right? Nobody likes that job. It's not a job.

People, nobody just says, I want my life to be spent day in and day out saying, reboot the Computer. Jack. I I, I agree with you, but let me ask a question.

All of us here, all of us have interacted with AI support VE website by this point, right? Whether it's an airline or software or, or whatever. Have you been happy?

Uber's a big one for me. Uber and Uber Eats. Have you been happy with the service that the AI support is rendered to you, Right?

Yeah. I do have, I do have one example. This is like X, right?

I try to change my password, but, uh, for some reason I haven't set up my like, uh, code generator. So I'm kind of like a block there. And I call their customer service.

Um, they only have two options. One is reset up your password. The second one is talking about the two factor authentication.

My question is in the middle. So like, no, no, no, no response for that. And there's no human in the loop.

Like, guide me to the other route. So I just feel like maybe if, if you want to use ai, right? I, I guess maybe need to bring human in the loop so the customer service side, I can get like better the other experience or something.

Wiki. This is a really good conversation to be having around this point in particular, even around marketing, because we haven't gotten to a point where we're using agents truly like robots. Like, uh, since first reason, second act third.

So what you're saying is when you, when you have two options, that's more of a menu driven, uh, rules-based AI system as opposed as a system that's listening and sensing reasoning what you're, you're asking for and then acting upon it. And this is what we're gonna, I think this is where we're gonna start seeing real, uh, change in ag agentic ai where that reasoning, sensing and acting process or the sensing reasoning and acting process is gonna get better and better. And it's gonna make many things easier for small companies like ours, marketing support, any outreach that we do.

But right now, I don't think, I don't think it's perfect. Well, and I, I'm gonna jump in here 'cause I can see Alan sort of shaking his head and I've got, you know, I'll tell you that I've spent, uh, about over the past day and a half, I've spent about four hours on the phone with Samsung and T-Mobile trying to upgrade my to the latest Samsung phone. And I have talked to the ais, I have talked to the humans, I know what the bug is in their system, but I can't get anybody to listen to me, right?

And the humans go through the exact same checklist and steps as the AI does. And right now, the automated AI are automating the existing workflow and steps the humans do without adding, as you said, Tracy, the sense and the reasoning part first. And so at this point, I'm about to say, screw it, I'm gonna do something else because I can't get satisfaction from either a human or the ai.

But, but you know, Jack, to be fair, you ask the people at Samsung, they may tell you, well, that's good training, right? We, we, we, we trained the humans and, and we took that right to then into the agent. They both do the same thing.

It worked, right? We, we, that was what we were shooting for, right? It may be a bad result, but, but that's what we were, that's how we were designing this, that to have the consistency Yes.

Between them. Yeah. But You mentioned, Jack, you mentioned, uh, really, this is maybe off topic, but you mentioned AI washing.

We're gonna be hearing a lot more of that in the next few days. I, I, I suspect after what happened at Block. But anyway, just, yeah, I'm gonna write about it, but oh my God.

Well, that's a, no, that's a whole nother topic, but Yes, it is. Let's go to a whole nother topic though. Um, you know, certainly coding has been at the tip of the spear of ai, uh, practicality of AI usability, but now we're seeing it extend further down that software supply, pipeline, supply chain, and, uh, Tracy Autonomous app dev with the ability not only to write code, but to test it.

What's this? It sounds, I don't know, too good to be true. I hope it's true.

I hope it's true. I really do hope it's true. Crossing your fingers, huh?

I am crossing my fingers. I, you know, um, I think this covers many topics. One, are we gonna lose developers?

Are developers gonna lose their jobs because these tools are out there? Um, how is this going to impact the DevOps pipeline? Hopefully it, it puts a grenade in the middle of the DevOps pipeline.

I am so waiting for a disruption in this process. I really think that we're way behind. This is our stuff.

What we do now is long in the tooth. Um, but there are bad sides of it. I mean, what, what Cursor wants to do is, you know, I, I think what's important in this story around Cursor is how they're managing VMs, right?

So what they're saying is, you, you, when you're sitting at your computer and you're working, you're doing one thing at a time, and you're max, you could be maxing out with one LLM your tokens, but you, you, you're gonna max out your environment. We're gonna instead allow you to have many, many, um, you know, sessions running in this LLM world so that you can do more things at once. And our agents are gonna go do more than one thing, because we can do that.

That is really a great idea, right? That's an amazing idea that they're, they're, they're starting to spread the workload and make things, uh, more efficient. I think that's the real story in, in, in, in what they're talking about.

And it's just applied to software development. But at the same time, how expensive is it gonna be? I I, I mean, I, you know, I hear all the time our team talking about we ran outta tokens, we ran outta tokens, we ran outta tokens, and we're, you know, we spend enough on, on, on AI at this point.

So Too bad someone didn't ask that about cloud 10 years ago, Right? Right. How long did we fi You can't control the cost, right?

That's where it becomes a problem. And you're still gonna have to have technical people manage this stuff. So you're gonna have technical people potentially in the beginning using more expensive tools, but they may be able to get things out faster.

And in general, we're still not seeing the whole picture. We're talking about building one component at a time. And when those components get built one at a time, and they have their own pipeline and their own push to a production setting, it doesn't show us a big picture.

So that still has to get fixed. But in terms of using these VMs to spread the work and spread the love, I think it's a very interesting di direction. I do like, oh, sorry, do you wanna go first?

Go ahead. Yeah, I, I do like this direction actually. I can see the big change about like how we arrange the work and, uh, the, the, the whole Slack, right?

You can see like in the past we, we did do development and then we consider the security and then we go smooth on the marketing or other stuff. But now they try to restructure the way people do the programming or even think about architecture. Like for like, uh, uh, I try some like, uh, um, ASRA code as well.

Like recently they invited the, the, the, the security scanning during the programming side part, right? So, uh, it's kind of like change the way, like the, the, the, the, the step by step. But they embed some, some portion to the left side or early side, which change the structure.

And way, I, I like this way, but like, again, I agree with you, like some developer may, may worry about it. 'cause currently not only developer can do the developer work, non-developer, like at the business side, if they have idea they can do that themselves. Well now we're gonna just pile the cost of cloud on the top, on top of the cost of ai.

And I, I feel like that there's gonna be a problem with expense to really get the adoption of this, even though I, it's cool. It's just has a way to go before it's gonna be affordable. But I'm, but I'm glad you brought up the expense though, because, you know, I'll give a personal story.

I was doing an analysis of some survey data. So I have a giant Excel file with a whole bunch of stuff, and I'm using Python to manipulate it. And friend told me, just use cloud code.

So I started trying cloud code and, and I was like, wow, this is awesome. I'm doing so much more than I could possibly ever have done before. And I ran through my $20 of tokens in, you know, a day and a half or whatever it was.

And clots basically said, well, would you like more? I said, and I'm like, well, what does that mean? Is that a hundred dollars, 200?

He was like, $5 and you get, you know, you get another whatever. And I did, and I was, it took me all of a half a second to say that $5 was well worth the spend because, you know, from my productivity perspective, it was a 10 x improvement on my productivity. And I think right now, the challenges, and we, we had this very early on in, in, in cloud, and we had now have, what is it?

A whole new industry called finops or whatever to, to manage the cost. But we have to weigh the productivity improvements versus the investment. What's the cost of a soft, fully loaded software engineer, a hundred, $200,000.

The amount of money you're spending on Quad to improve the productivity of that engineer is huge. And it's, it, it, it can be a huge ROI now do you have to manage it? Absolutely.

You wouldn't give your your programmer permission to go out there and spend $5,000 on a and on an editor, right? Or something that wouldn't have value. But you do have to, we do have to start thinking about what the ROI is right now for me that ROI, and for a lot of people, I think is huge and it's well worth the money at some point, we're gonna spend so much that it may not be.

So, Jack, how many more nickels have you bought? That's between me and my credit card statement. Because, you know, I, The school night I came from, the first one was always free.

You Know what I'm saying? Yes, Absolutely. But you there is that, yeah.

If you think about this, right? Uh, if we try to do a protocol or like a personal choice, right? It, it's very easy to go this way, but, um, what if the, the things you did by yourself, you want to make it scalable, right?

We, we, we have to consider the price, how much we spend. Like, it, it will to be totally different model. If you do it by yourself locally or you put it on the cloud, let other people use it at the same time.

So the, one of the, the, uh, examples they use, which I thought is, is interesting, um, example was taking the, um, the doc, the cursor doc, and have it go through and do testing to make sure the documentation and the, and the, the gooey front end basically was the same, and it took 45 minutes. And I'm thinking to myself, Hmm, I think a hiring a person to go through and read the doc and and test it is probably cheaper than 45 minutes worth of, of AI time. That's a long time.

45 minutes of running expensive. That's an expensive proposition. Agreed, agreed.

Guys, we're about outta time today. What a great, what a great way to end this week. You know, sometimes I look back over the week, I, I need to start looking back over two or three weeks because John, you're on, Jack, you're on a lot of these with us.

The, the, the pace of these developing stories is accelerating so much. I mean, this, These cycle is in pay is in parallel with the technology. And it's, you know, I was gonna mention something.

I, I've talked to this guy, a couple of executives in the last couple days. One from Vos data, one from Nvidia, and all they talked about, they say, you know, we know this subject well, but the pace, we're we learning ev something new every day? And I feel Alan, like, we're writing and chasing things that we never imagined we'd be, we Be doing.

No, no. Is it, you know, I think I mentioned this. I was talking to David Brynn, the sci-fi author, who's eing for us at our, our RSA seminar on that Monday of RSA week about AI native dev securing AI native dev.

And I said, David, I'd like you to get maybe some of your other sci-fi author friends, I'd love to do a podcast with you on. Mm-hmm. Did you ever think you would be living to see the stuff you were writing about as sci-fi becoming real?

I mean, we're, we're living in many ways a bit of a sci-fi future, whether it's Philip Dick and, and the, you know, the Electric Dreams and Blade Runner, or Asimov's Laws of Robotics or, or you know, what have you, or, or Frank Herbert, uh, the Machine Wars, the Thinking Machine Wars, right? And Dune, uh, it's like living in a sci-fi reality. It's crazy.

Anyway, The robocop conference scene, Yes, the Robo, we Don't want that to happen. I'm afraid that, well, Mr. Not a philanthropic old have something to say about it.

Anyway, Wiki, Tracy, Jack, John, thank you for joining. Thank you for joining. We've got our tech strong TV replay immediately following.

Have a great weekend, everyone. On behalf of everyone here at Techstrong, we're out.