WEBVTT
00:00:00.160 --> 00:00:12.800
I I think the bigger question which is kind of related to this content development cycle is commoditization of the frontier models.
00:00:13.359 --> 00:00:15.839
And which is wild when you think about it.
00:00:15.919 --> 00:00:25.600
And and no, and and look, before so we we came up with this plan for what we're going to talk about today, frontier models turning into a commodity.
00:00:26.320 --> 00:00:28.239
Who's going to win in that?
00:00:28.640 --> 00:00:33.920
And then we've also got the agents coming into play now.
00:00:34.000 --> 00:00:44.159
So you've got Grockbot, you've got Meta, Muse, their agent, obviously the agent layer, which is like kind of trying to make this thing useful.
00:00:44.240 --> 00:00:50.560
You've got dots, openai release dot, and and and so you've got all of that.
00:00:50.640 --> 00:00:54.479
That's probably more of an interesting conversation, I think, for me.
00:00:54.719 --> 00:01:04.319
And then no sooner had we kind of come up with this plan and what to talk about than Google suddenly released Gemini 4 today.
00:01:04.400 --> 00:01:05.280
And guess what?
00:01:05.439 --> 00:01:13.040
It's the now the best benchmarking benchmarks model out there, right?
00:01:13.200 --> 00:01:15.280
And I was for now.
00:01:15.359 --> 00:01:18.799
Yeah, and that and then and do you know, do you know, Ben?
00:01:18.879 --> 00:01:20.159
I'll get your view on this.
00:01:20.640 --> 00:01:23.359
I'll get your view on the kind of agenda as well.
00:01:23.599 --> 00:01:26.400
But I'll there's no loyalty.
00:01:26.879 --> 00:01:29.680
So so so so right.
00:01:30.239 --> 00:01:36.000
I was using Claude, Fable, 5.1, Opus interchangeably.
00:01:36.319 --> 00:01:40.799
If you'd asked me six weeks ago, would you leave this?
00:01:41.040 --> 00:01:42.239
I would have said no.
00:01:43.200 --> 00:01:47.040
And then ChatGPT, something went off with Claude.
00:01:47.120 --> 00:01:50.799
It wasn't even like ChatGP Chat GPT's Astra model.
00:01:51.359 --> 00:01:54.000
I was like, yeah, it's great, but it wasn't that.
00:01:54.159 --> 00:01:57.599
It was something went off on the performance with Claude.
00:01:57.760 --> 00:01:59.040
I don't know what the hell it was.
00:01:59.200 --> 00:02:08.000
Someone else on LinkedIn had a little convo with them because they posted and said, has anyone noticed Claude's kind of got it's got off the boil?
00:02:08.080 --> 00:02:10.240
Basically, the answers aren't so good.
00:02:10.800 --> 00:02:13.599
My loyalty was like zero.
00:02:13.919 --> 00:02:16.479
I switched on a dime.
00:02:16.639 --> 00:02:24.400
It's I've got I've got a subscription kind of problem, but but I that that but for me, that's all part of it.
00:02:24.639 --> 00:02:29.360
That's what because it's a commodity, I don't have any brand loyalty.
00:02:29.599 --> 00:02:31.759
And in fact, I don't particularly want to be with any of them.
00:02:31.919 --> 00:02:34.560
I'd love to be one of the open models, to be honest.
00:02:34.879 --> 00:02:35.919
I I totally agree.
00:02:36.000 --> 00:02:57.039
And one of the things that we have to look at is if you're an enterprise and you know, and you're going through this adoption and you get locked in in six weeks, that multi-million dollar contract that you did with one of the frontier models, you you might be in a or you might be very regretful.
00:02:57.360 --> 00:03:11.599
There's there's no way, and and I think actually this is this actually is hindering AI adoption in general, is because there is this fear about moving forward with any one model just because it is changing so quickly.
00:03:11.919 --> 00:03:14.960
And my answer to that is you need to find a way.
00:03:15.120 --> 00:03:17.680
This is what I do with my day day daily.
00:03:17.919 --> 00:03:27.039
You need to find a way to abstract yourself from a layer from those models so that you can move, you can choose as things move, as things move forward.
00:03:27.120 --> 00:03:38.159
You need to make sure that you're able to take advantage of the latest agents that are coming out, take advantage of the latest models that are coming out, and not have to think too much about it from an enterprise standpoint.
00:03:38.319 --> 00:03:49.680
And I think there's a way, I know there's a way, that's what I do, that to abstract that and make it so that you can control, you have the control, the guardrails, the access.
00:03:49.919 --> 00:04:16.480
So you're the big thing here is it also helps prevent shadow AI, where you know your employees are picking up their phone and they're putting your IP into their phone for their specific, you know, model of choice because you don't provide it by having those models available as they come out with the governance and security and guardrails in place, you end up with a much better experience for your employees.
00:04:16.560 --> 00:04:17.920
They end up being more productive.
00:04:18.079 --> 00:04:25.040
You end up with a more holistic adoption process because you know, program, if you will.
00:04:25.439 --> 00:04:31.439
And you don't spend millions of dollars on something that is obsolete in six weeks, right?
00:04:31.600 --> 00:04:33.439
And and and that's a real fear.
00:04:33.519 --> 00:04:36.800
I I hear about it all the time when I'm talking to customers.
00:04:36.959 --> 00:04:44.399
They are afraid of making a huge commitment to one of these frontier models because next week there might be a new thing, and now they're stuck in this one.
00:04:44.639 --> 00:04:51.360
And from a competitive standpoint, I as a company trying to be more competitive than than my competitor.
00:04:51.680 --> 00:04:57.759
If I can move to the the better model faster, I can get ahead of where they are at.
00:04:57.920 --> 00:05:11.920
Now, if I don't have to think a whole lot about that, because I've set up you know my enterprise platform or go with a vendor that allows for that, that gives you the ability to pivot very quickly without disrupting your business.
00:05:12.079 --> 00:05:14.879
And I think that's going to be really important as I move forward.
00:05:15.279 --> 00:05:36.079
These agents that you you know, going back to the agents, these super agents, if you will, these bots that you know go off and spin up a you know a little computer, whether it's computer Kubernetes in the background or what have you, or some container in the background, and run these processes, install the code that you need to run those things, and then essentially complete projects.
00:05:36.319 --> 00:05:42.079
You know, this is kind of the next wave, and this is something that I'm very passionate about because it is what I work on.
00:05:42.959 --> 00:05:51.839
There, it's gonna be this, and we'll we'll be in this world of this evolving and and going back to your commoditization.
00:05:52.079 --> 00:06:04.000
This will become a commodity where there are going to be racing at who can have the best bot offering where it can go off and do tasks for you and then come back with the result.
00:06:05.680 --> 00:06:11.920
So on that, let me just jump in there because it's important we bring the audience with us on that.
00:06:12.000 --> 00:06:16.720
And and and what we're referring to, so obviously everyone's familiar with the large language model.
00:06:17.040 --> 00:06:19.680
Then if you remember, was it earlier this year?
00:06:19.920 --> 00:06:21.839
Must have been OpenClaw came out.
00:06:22.000 --> 00:06:24.399
Peter, who then moved to OpenAI.
00:06:24.639 --> 00:06:41.199
OpenClaw came out, and it was the first time you could actually just spin up your own agent, have it sat locally, or my mine was on a on a VM, but then just it it you could give it a task, and basically anything you could do on your desktop it could do too.
00:06:41.279 --> 00:06:48.000
And that was it sort of preceded the cowork functionality in the frontier models.
00:06:48.240 --> 00:06:48.480
Yeah.
00:06:48.720 --> 00:06:55.199
And even the cowork functionality in the frontier models, it's not what OpenClaw could do.
00:06:55.439 --> 00:06:58.240
OpenClaw could really, really help you out.
00:06:58.399 --> 00:07:04.319
It's genuinely like a virtual assistant that could do tasks.
00:07:04.959 --> 00:07:21.360
And it's when I really first learned about cron jobs, and and that's that's setting up a job, a specific job that an AI agent will go and do at a certain time, and then you could build these kind of like little workflows for it.
00:07:21.439 --> 00:07:22.480
And it was fantastic.
00:07:22.639 --> 00:07:31.600
And I I've mine is still running and and I love it, and I tell you what, it's much more sticky in terms of user.
00:07:31.920 --> 00:07:43.600
If you've invested some time setting all of that up, that's the sticky part, not the large language model, even with cowork capability, it's not that sticky.
00:07:43.920 --> 00:07:52.959
So then what's what you're referring to, Ben, is we've had a spate of these kind of like AI agents come out.
00:07:53.040 --> 00:07:56.560
We've because they kind of like block and tackle each other, don't they?
00:07:56.720 --> 00:07:57.439
These big companies.
00:07:57.920 --> 00:07:58.160
They do.
00:07:58.480 --> 00:08:14.000
We've had Grogbot, which is when you log into it and download it, you can see it's coming from Cursor, SpaceX bought XAI bought Cursor 60 billion dollars or some loose change they had knocking around.
00:08:14.399 --> 00:08:24.160
And and and so that came out almost immediately, hot on the heels, MetaMuse came out, very different product, and we can talk about that.
00:08:24.319 --> 00:08:32.720
And then the Microsoft, not Microsoft OpenAI dot has come out like literally the last couple of.
00:08:33.039 --> 00:08:35.279
I think I first saw that yesterday on X.
00:08:35.759 --> 00:08:40.639
So Microsoft has also released their own sort of version of this as well.
00:08:40.799 --> 00:08:47.759
So they're all they're all like they're not necessarily on parity yet, but they're going in the same direction.
00:08:47.919 --> 00:08:54.720
And we'll see we're gonna see this evolve for an enterprise for end users, it's a little easier to adopt right away.
00:08:54.799 --> 00:09:07.279
For enterprise, they have to figure out how to put the cardrails around it and as they go forward and make sure, make sure that they don't have things happen like what happened with the agents breaking into Australia Medicare.
00:09:07.440 --> 00:09:09.600
And that's completely possible and doable.
00:09:09.759 --> 00:09:11.840
It's just as they do that.
00:09:12.159 --> 00:09:24.879
My big thought on this is sure, each individual one of these model providers are are providing these bot abilities, but you got that their model.
00:09:25.679 --> 00:09:37.600
You know, if their model goes down, if their model is is under load, I mean, every morning, you know, we I monitor the health of the models, obviously, because you know, we we provide all of them.
00:09:37.679 --> 00:09:51.200
And the every morning you see you often see that, you know, as the West Coast spins up, you know, they you start to see constraints on the that's the ability to use some of these models for some of these model providers.
00:09:51.360 --> 00:09:56.399
Sometimes all of them well well well, Ben, try being in the UK.
00:09:56.559 --> 00:10:06.399
You have this lovely most of the morning, you get great model performance, and then who wakes up East Coast US?
00:10:06.879 --> 00:10:08.080
Yes, that's very true.
00:10:08.559 --> 00:10:24.639
Then then West Coast US, and it's like the model performance, it's just no nose diving, but you know, it's it's it's noticeable every day, so you want to kind of get your big job kind of like things done, like the big tasks done.
00:10:25.440 --> 00:10:30.000
Ben, take a look at your microphone as well, mate, because there was a little bit of crackle on there.
00:10:30.080 --> 00:10:31.600
I don't know if it's jumped again.
00:10:31.759 --> 00:10:33.519
Just make sure you're on the right mic.
00:10:33.759 --> 00:10:36.080
So and then and then okay, cool.
00:10:36.240 --> 00:10:46.159
And then you mentioned co-pilot, autopilot, and oracle's fusion claw, which I saw actually.
00:10:46.240 --> 00:10:47.840
I did I even see autopilot.
00:10:47.919 --> 00:10:49.279
Tell us about those, mate.
00:10:50.000 --> 00:10:53.519
So let me let me get the the details here.
00:10:54.000 --> 00:11:07.360
So if you've had I say let tell us all the if you've even had chance, I'm no, I was you're mentioning things to me I've never even heard of yet, and I I'm on top of this stuff all day, every day, right?
00:11:07.440 --> 00:11:12.559
So it's absolutely fine to go let me sign up the web page and I'll read it out to you.
00:11:12.799 --> 00:11:14.480
Nah, I got it right here in my notes.
00:11:14.639 --> 00:11:19.919
Last week, so Microsoft last week, Oracle yesterday, and then Microsoft's co-pilot, autopilot.
00:11:20.000 --> 00:11:23.360
So it gives each agent its own cloud computer identity.
00:11:23.600 --> 00:11:28.960
So here they are coming along with where other people are, and it moves that agent work to usage-based billing.
00:11:29.120 --> 00:11:31.360
So the seat price no longer caps up bill.
00:11:31.519 --> 00:11:32.399
So that's good.
00:11:32.480 --> 00:11:42.240
But you also want to make sure that you look at you know data usage in these various platforms, you got to be careful with because there's hidden costs, so just be careful there, potentially.
00:11:42.399 --> 00:11:58.240
And then so with that, also Oregon's fusion or Oregon, Oracle's Fusion Cloud is its own agents for finance, HR and supply chain, which that that's actually if you think about that, they're they're going into a niche there.
00:11:58.320 --> 00:12:00.720
So that's I kind of think that's a little smart.
00:12:00.879 --> 00:12:04.000
And so it runs runs it both on Gemini and OpenAI model.
00:12:04.240 --> 00:12:06.639
So there they're thinking about you know sorry, mate.
00:12:06.720 --> 00:12:07.679
Is that the Oracle?
00:12:07.919 --> 00:12:10.320
Is that the Oracle Oracle Fusion claw?
00:12:10.399 --> 00:12:10.639
Yes.
00:12:11.120 --> 00:12:12.720
Fusion core claw.
00:12:12.799 --> 00:12:13.840
Sorry, yeah, go on.
00:12:14.000 --> 00:12:16.000
So that's running running on Oracle.
00:12:16.159 --> 00:12:19.919
Sorry, uh, running on OpenAI and Anthropocet.
00:12:20.720 --> 00:12:22.480
No, OpenAI and Gemini and Gemini.
00:12:23.519 --> 00:12:25.200
Open AI and Gemini models.
00:12:25.440 --> 00:12:34.000
So so you're starting to see that a little bit, and I think we're gonna see that more and more as you know, Oracle, you know, stepping back.
00:12:34.240 --> 00:12:38.720
You see this with some of the big providers as well, like the HVEs and the Dells of the world.
00:12:38.799 --> 00:12:42.559
They've been thinking about this too with their platforms and what they're doing.
00:12:42.639 --> 00:13:00.799
They're they're stepping back, and they have you know the NVIDIA models and things that they offer as well, and they're stepping back and looking at their holistic environment and offerings to their customers, not only with the frontier models, but also with you know, bringing along the models that are that they provide, and then also that the customers are.
00:13:01.600 --> 00:13:17.679
And and speaking of the Dell and HPEs of this world, and IBM and IBM, you've got NVIDIA's Nemo Claw in the mix there as well.
00:13:22.159 --> 00:13:28.559
It's a lot lot lot broader market for enterprise actually when you start going through it.
00:13:28.960 --> 00:13:29.759
And it goes back to the back.
00:13:32.799 --> 00:13:35.759
Let me just ask one more, I'm I'm asking so many questions of you.
00:13:36.240 --> 00:13:39.919
I am assuming Oracle's targeting enterprise.
00:13:41.519 --> 00:13:46.879
Yeah, I I mean obviously finance HR supply chain, that makes a lot of sense to me.
00:13:47.039 --> 00:14:02.000
And they have a they have a very large customer base with lots of you know, people who know how to run their platform and you know so they have they have a very large you know community, if you will, yeah, in the enterprise.
00:14:02.159 --> 00:14:09.840
So you know it makes sense if they would actually focus on the HR supply chain and and those things as they go forward.
00:14:10.639 --> 00:14:18.480
I think we're gonna Can I ask you a question about the cost of these things?
00:14:18.960 --> 00:14:20.159
Oh yes, yeah.
00:14:20.480 --> 00:14:20.720
Right.
00:14:21.039 --> 00:14:23.120
When I set up my open claw, right?
00:14:23.600 --> 00:14:29.440
Anyone who's done this will know, but if you haven't, you have to kind of give it a an AI brain.
00:14:29.919 --> 00:14:39.919
And for my open claw, originally I set it up with an API into anthropic and it started doing like work for me.
00:14:40.159 --> 00:14:40.960
Do you know what I mean?
00:14:41.039 --> 00:14:47.039
It was like doing checking emails, doing daily this, that, and the others, nothing huge.
00:14:47.279 --> 00:14:55.440
And I started getting started getting my API invoices through through Anthropic, and the first one was for twelve dollars.
00:14:55.759 --> 00:14:58.000
I thought, whatever.
00:14:59.440 --> 00:15:01.440
$12, whatever.
00:15:01.919 --> 00:15:07.120
Then in my email, ding, another one for twelve dollars.
00:15:07.519 --> 00:15:10.799
About half an hour later, ding, another one for twelve dollars.
00:15:10.960 --> 00:15:15.360
I was like, switch, switch it off.
00:15:15.919 --> 00:15:28.000
Yes, and at that point, I that was my learning experience of actually using these frontier models, how much it costs.
00:15:28.080 --> 00:15:34.000
And for an enterprise, that is exactly they're probably getting a much better rate than I am, I appreciate.
00:15:34.559 --> 00:15:39.759
But but their usage is is is huge, and that's when, of course, for me it was easy.
00:15:39.840 --> 00:15:42.000
I just switched it to uh Moonshot.
00:15:42.559 --> 00:15:57.919
Yes, uh to the to the Chinese open model, and and and and and it's lived very happily as Moonshot for a long time now, and every now and again I pay them like five bucks, right?
00:15:58.240 --> 00:16:01.679
Like every few months, I oh, it's run out.
00:16:01.759 --> 00:16:04.159
Let me just put another 15 bucks on there.
00:16:04.799 --> 00:16:26.000
And I think that you bring up a really good point, and something that I think about a lot is how can you provide the just good enough great experience for the work that you're doing and not be out there using like Fable 5 all the time and paying tens of dollars for what you usually need to do versus pennies, right?
00:16:26.240 --> 00:16:35.360
So being able to route between those models in the in a not even have to think about you route route between those models is something that is, I think is important.
00:16:36.159 --> 00:16:40.960
As you're doing your work, you shouldn't have to think, oh, I need to change to this model.
00:16:41.279 --> 00:16:49.519
And often when you change to those models, they feed that context that you've had back into the new model, and that costs you more money to do that.
00:16:49.759 --> 00:16:55.279
So having an ability to route between those models seamlessly, I mean, that's what I do.
00:16:55.519 --> 00:17:05.839
But but also as enterprises are building this out, they need to think and think in that sense is okay, I've got this massive growing bill and it's gonna keep growing.
00:17:06.160 --> 00:17:18.160
And and especially when you look at the cost, so that$200 max subscription, it's not it's costing them a whole lot more per month to provide you that compute for that max subscription.
00:17:18.319 --> 00:17:22.799
I've seen numbers between five and fifteen hundred dollars, you know, in compute.
00:17:22.960 --> 00:17:27.039
If there's various things that you hear out there, somebody once has said like 2000.
00:17:27.440 --> 00:17:31.759
I don't know what like what they I've seen what's reported, but it does make sense.
00:17:31.839 --> 00:17:44.240
They are spending trillions of dollars on compute to in this horse race to get to to the front and get the installed customer base and get you know the hooks in.
00:17:44.400 --> 00:17:46.319
So then it's really difficult for you to move.
00:17:46.400 --> 00:17:54.720
And that's why I think it's really important, whatever you use, to think about your company and the money that you're spending in.
00:17:54.880 --> 00:18:14.319
And and also the big thing that I also like to think about is if you look at our trajectory on what we're using as a planet, and you think about the resources and the supply chain, those, the, the, the silicon that goes into it, you know, the GPUs, the the data centers.
00:18:14.559 --> 00:18:19.440
In the US, it's like 1.4, 1.5% of the US greatest all data centers.
00:18:19.519 --> 00:18:27.359
And they're they're saying that very here, very soon in the next, I think past a little bit, past 2030, it's gonna be maybe upwards of more than 10%.
00:18:28.240 --> 00:18:32.400
You think about that, that electricity has got to come some from somewhere.
00:18:32.640 --> 00:18:34.960
All those raw materials have to come from somewhere.
00:18:35.039 --> 00:18:47.119
So, what can we do as companies, as good stewards, is not always use the best model for the mundane work, like the work of moving files or creating emails and stuff like that.
00:18:47.359 --> 00:18:51.200
Should we shouldn't be using those fable fives for that?
00:18:51.359 --> 00:19:13.759
We should be able to deconstruct those prompts and use the pieces that need fable five, great, you know, as we need it, and then all the mundane stuff can be used on a Chinese model or another open source model that even like in the you know the on-prem world that could be an on-prem model that somebody's using with their current infrastructure.
00:19:13.920 --> 00:19:21.519
So yeah, companies are gonna have to figure out what their tolerance is here, but more one more thing on that.
00:19:21.680 --> 00:19:31.599
There's a reason why the stats are out there that say 95% of companies are trying to adopt AI do not have that uplift that they're looking for, and they're not what we call successful.
00:19:31.839 --> 00:19:33.680
Only about five percent are there.
00:19:33.920 --> 00:19:44.799
You see them saying, Oh, our adoption rate is this and this, but if you look at their PL and their feature set for their customers and their increase in you know, increase in sales and revenue, it it doesn't match up.
00:19:44.960 --> 00:19:49.200
There's a very few that out there that actually are getting there.
00:19:49.279 --> 00:19:52.079
I mean, the Walmart, you know, Walmart's a really good example.
00:19:52.240 --> 00:20:00.559
They they pivoted early when they started to see this and they came up with kind of their they don't call it a blueprint, but I have a newsletter on enterprise film notes you guys can go take a look at.
00:20:00.640 --> 00:20:04.400
I also have kind of a you know a little bit about that roof type that you can use.
00:20:04.480 --> 00:20:06.640
That's all free, so go subscribe.
00:20:06.960 --> 00:20:23.440
But that is a really good example of you know a company that would this big ship of a company moving forward with AI, having agents sprawl everywhere, and then be like, okay, wait a second, let's look at what we need to do, let's be proactive about this and then make it fit the business.
00:20:23.519 --> 00:20:31.920
And then they had that like 35% correlational, 35% uplift in in you know, for some of the some of the business.
00:20:32.079 --> 00:20:33.039
That's a big deal.
00:20:33.119 --> 00:20:36.240
That means that what they're doing, their customer experience is improving.
00:20:36.400 --> 00:20:39.599
That that's a good signal for them to continue on that road.
00:20:39.839 --> 00:20:45.359
And also they're controlling their, you know, they're they're really thinking about the holistic and controlling their costs.
00:20:45.519 --> 00:20:54.640
The VS, you know, Uber, who in the first quarter of this year said, we spent all of our budget, we have nothing to really show for it.
00:20:55.680 --> 00:20:57.119
Now we gotta rethink this.
00:20:57.279 --> 00:21:01.359
And and I I can only imagine what that looks like.
00:21:01.599 --> 00:21:03.759
Where how how does it happen?
00:21:03.920 --> 00:21:15.119
That happens because you know, you you think you go after AI because AI, and you don't go after AI because you're trying to achieve certain business results.
00:21:15.599 --> 00:21:17.519
I think I th I think you're you're right.
00:21:19.519 --> 00:21:22.559
It's packed full of information that's really useful there, Ben.
00:21:23.200 --> 00:21:36.640
I think also what kind of a realization was for me last week was so GPT Astra came out, big fanfare, usual thing, it's the best model, blah blah.
00:21:36.960 --> 00:21:52.880
And then I can't remember where I saw it one of the big podcasts, put a slide up just saying a chart, just showing that they had now overtaken the adoption in Enterprise for the first time from anthropic with Claude.
00:21:53.680 --> 00:21:56.319
I was like, but they just brought the model out.
00:21:56.640 --> 00:21:59.599
So wouldn't that take some time to move all?
00:22:00.240 --> 00:23:06.160
of the so if you're running services like customer services or whatever and that's you've got all that based on the anthropics models you're kind of gonna need some time to get those moved across and then I thought well no this is kind of like the tier top tier of all of the models it's not like this one of these like flash models Astra's the top tier and then it sort of dawned on me ah it's the coding isn't it so if I'm a hedge fund or I'm Arcley's bank or whatever I'm not talking where what's driving that growth in enterprise the massive spend where am I never gonna use a model that's slightly behind in the rankings it I might be wrong on this but I'd say it's coding and then it's faster shift to switch right yeah my coders are just sat there and they just go hey I just switched from using this to using Astra gave the feedback internally and then no one's looking at the bills.
00:23:06.400 --> 00:23:32.960
And and you mentioned hedge funds you know traders and things of that nature leveraging these to the nth degree I remember years and years ago when having the fastest network switch did where where milliseconds was not not fast enough and they they would they would sell these switches that were super fast because they wanted to make sure that that next trade was before was before the next guy.
00:23:33.200 --> 00:24:39.599
So absolutely where there's business use I I I agree with you there where there is a business reason to where they need to be out in front you're gonna see that adoption and and cost is not the biggest concern where they where the where that cost impacts companies is I think more so on the services side or on the retail side where they're looking at okay they need to keep costs reasonable and down you can't raise prices significantly because you decide that you're gonna use fable five for your entire retail you know yeah exact exactly exactly so so that's it it's kind of more of a sticky use case within your enterprise environment isn't it and governments too you know where there are different intelligence governments intelligence you saw we saw that China was China Claude reported anthropic reported it that they had got figured out that China was using using their model for allegedly using their model for intelligence and they had gotten rid of part of their intelligence team as a result.
00:24:39.680 --> 00:25:02.000
So you're like okay that's not great and see and that's the reason where they they have access to these great Chinese models and they're still using anthropic because they want to be they want to use the best of the best because that's how you you move the needle and get ahead of your gut you know your global competitor if you will so and they're not the only ones that are doing this.
00:25:02.079 --> 00:25:49.599
I'm not saying I'm picking on China but all the countries are that are have this stake are are more than likely doing these type of things where it matters the most you'll see the early adoption and that will that accelerate the growth where you know money comes into play and you're trying to make you know you don't have limited funds the for the rest of us in in the enterprise world being proactive and smart about it we we want to ensure that our bottom line and our profitability continues and that we don't eat up that profit layer you know that profit margin by spend on using the the top the the absolute top models when we can use much cheaper models to get the same job done.
00:25:49.839 --> 00:26:04.640
And I know that this is on the minds of lots of customers that I've talked to they think about this one of the biggest things that I've heard from even technology service providers is you know what we got to get our model cost down.
00:26:04.720 --> 00:27:28.880
You know we're providing services to our customers and we're providing these platforms to the customers but help us figure out how to get that model cost down because you know it's it's weading into their margins for what they're what they're doing it also slows them down because when the cost is there and the cost is so high you're you're gonna use less right whereas if you can get the cost down and this is the other big point I want to talk about by being smart about how your your AI spend is you get more productivity and more benefit out of the dollars that you're spending so you're you're going to accelerate your business profitability if you do it right right because you're not having as much spend per revenue per dollar of revenue that you had would have otherwise so then you can end up getting more revenue and accelerating those accelerating your business the same thing happens with code you look at code and what we're doing more code is being generated than ever than anytime ever because it's AI enabled the problems come with that from a quality standpoint which we still I don't think we've really figured that out yet like quality how do you handle quality when you're generating so much code and your resilience and reliability because having the human in the loop there becomes harder and harder as that as the number of lines of code that are generated increases.
00:27:29.200 --> 00:27:41.599
So companies are trying to figure that out as well but you know this is a we're on a rocket ship I would say and we just are all figuring out together and that's the whole point of us coming on here is to to share notes.
00:27:41.920 --> 00:27:56.400
Yeah yeah and look let's just thank you thank you Ben for for that insight and it's great for coming from somebody who's actually the layer of access your platform into all those model models for enterprise so that's that's awesome.
00:27:56.559 --> 00:28:29.440
So we've had so we've had meta muse I haven't I have not set up meta muse apparently it's much more aligned to kind of consumer whereas I'd say grokbot which I have set up by the way I've got grokbot kind of I'm just going through the motions kind of giving it little tasks here and there now but you can see it's more I don't know it's just more complex as a product meta I one of the benefits of MetaMuse that I like the sound of is you can communicate via WhatsApp through it.
00:28:29.519 --> 00:28:46.240
So when I set up my OpenClaw and I I communicate via Telegram through through open claw grockbot has its own very good actually very good kind of chat app if you like so I've got it on desktop and I've got it on my iPhone.
00:28:46.400 --> 00:29:34.720
That's been a really nice product actually but Muse I have not used I think we talked about last time so I won't go into it amazon immediately stopped the Muse shopping agent yes going on to its platform because they were like they didn't you they they said Amazon said that the meta people forgot to tell them that they were going to be sending this agent over so that was interesting and we'll do some digging around Microsoft's co-pilot autopilot I need to be co-pilot studio session so whilst I'm doing that I will check out autopilot and then look I'm keeping on time as well I what caught my attention from your newsletter then was Apple.
00:29:35.440 --> 00:30:53.279
I always love it I always love seeing when this is the Klaners and all the rest of it when a company says we're gonna replace the humans with the AI agents I'm I'm almost like grabbing my popcorn sitting back to watch the show so what happened with let's cover Apple and then let's start wrapping up so so Apple Apple was you know they were gonna lay off a bunch of people and and they talked about it earlier on and then they decided not to do that and and they won't actually comment on the why but there it's pretty obvious the you see you know Ford that I wrote about they they brought back their engineers you you you see this across the board where a company is like oh we're going ai okay get rid of all these staff and then they come back and say oh oops that was not the best thing to do because they just it's like they're they're they're using they're using AI to to go after you know to find a problem to solve instead of creating a solution that makes sense for the business.
00:30:54.079 --> 00:30:55.599
So who were these people in Apple?
00:30:55.759 --> 00:31:07.359
What do we know and it's all allegedly because obviously there's no press release on on this but who who was the affected group in Apple I think is the term we use these days.
00:31:07.680 --> 00:31:51.359
Yeah so Bloomberg reported that over the summer Apple drew up a plan to lay off about 5,000 Apple care support workers support and have AI phone and web agents take over some of their work it's now on hold it's on hold indefinitely and yeah that's that's what we know which it makes sense to me I do think that you know as we move forward and understand you know what the business needs are and what the customer needs are sometimes when you throw when you shotgun approach your adoption ai you you you get a rude awakening when you realize how that may impact your overall business.
00:31:51.519 --> 00:32:25.519
And so we just need to take you know a stepwise approach and and step back and and I think you know to to Apple's credit I mean I don't we don't know the internal details of other what's just been reported but to their credit they're probably making the right decision because they are like you know what either we're not ready and or or this isn't a good fit and and being able to step back and make that pivot is important especially for enterprises like Apple taking a misstep there could could have very impactful results.
00:32:25.680 --> 00:33:08.160
We see that you know as big companies stumble and if they stumble it it can have a significant impact in their business and their brand it's always kind of like after sales isn't it's like we've sold it now so Apple care you know let's just let's just take a look at that uh uh automating automating that I I do think the these things are connected as well I think there's a big push out there in these boardrooms to say well Elon Musk who's stood next to Donald Trump saying that in the next couple of years anything that you can is currently done by humans sat in front of a screen is going to be done by AI.
00:33:08.640 --> 00:33:42.240
Now right and so that creates pressure in the boardroom noticeably Dario and Sam Altman have stopped saying that they were saying the same things but they've stopped saying that and and then if you look at so that creates pressure in the boardroom I think that why aren't we being more radical are we going to get outstripped by our competitors here because we're not pushing hard enough on automation and well we said it a bunch on this podcast automation's not automating out the humans is not the goal.
00:33:42.559 --> 00:34:04.319
And Elon Musk and hey I'm an investor in both Tesla and SpaceX so clearly I believe in the in in the leader and innovator and I think the man's a genius but do I think he's genius level in his forecasting ability of when he says things are going to happen.
00:34:04.400 --> 00:34:25.360
No I don't Robo Taxi's taken you go back to when he says they're gonna deliver things no no that's fine and I think that's part of the management style I think he says and and so so that it's commonly known I think in in in the circles that I sort of am in as an investor it's called Elon time.
00:34:26.480 --> 00:34:39.920
So just because he's next to Trump saying this does not well you could argue it makes gonna make it less true and Elon's not like I don't think he's making it up I think he genuinely believes that's what's going to happen.
00:34:40.079 --> 00:36:21.599
But then you have to go yeah but is that going to happen in the next two years or is that going to happen in the next eight years from the management time horizon of most boardrooms which is probably three five years are you actually going to have automated out all the staff that are using a screen and I just think we know we can see the answers no look at the conversation we're having doesn't mean you're it doesn't mean you're not going to use ai to make that same group of people hopefully all of your people be able to be much faster and and and will there be some impact on new hires and all the rest of it but possibly so maybe you need kind of less people but hopefully then you are able to grow your business and do more and and that's a good point because the the good point and and I actually believe this and and and it might be something that's not thought of like I maybe one of the everybody's like oh you know we're adopting AI you know we're gonna have less jobs I don't think that's true if you look go back and look through all the paradigm shifts we've had since industrial revolution you know when electricity came on the telephone etc initially there is a disruption and we're in that disruption right now which means jobs are being lost the thing about it is when we have these generative generativity type platforms and not generative have a generativity type platforms that create more innovation we we end up creating more opportunities and accelerating what's going on so what does that end up in the end result?
00:36:21.920 --> 00:37:18.960
It ends up with not necessarily the same jobs or the same positions but there is more opportunities for jobs because there's more opportunities that are being created and and and let's look at where job so if you look at what what can we look at autonomous driving will that disrupt uber drivers jobs yes yes it will for sure disrupt uber drivers jobs and then the uber drivers are not all of them but a lot of them are driving for Uber not because they love driving for Uber if next time you're in an Uber ask your driver do they love driving for Uber as a company you're not gonna hear them saying loads of great things about Uber I I guarantee that and and so you're going to you're going to get that but that shift is going to happen over a period of time.
00:37:20.239 --> 00:37:58.559
It's not going to be that Waymo Wave and Tesla and there's a few others knocking around right are en masse gonna take all of those jobs overnight but will there be a shift in the workforce absolutely that's what humanity's done when we invented the wheel and the industrial revolution and so and so of course but we've always been like that Ben that's it's always always been that way we've been in technology for bloody 30 years.
00:37:58.880 --> 00:37:59.920
Yeah that's true.
00:38:00.159 --> 00:38:30.480
I and back to your Uber comment it's really funny you say that because it brings back memories of every time I'm at Uber I ask you know I I ask questions because I want to know like where'd you come from what do you do more often than not they you know I met a number of startup guys out there who are driving Uber just to get by while their startup is gets going so the entrepreneurial ship you know those Uber drivers you know they talk to lots of people they learn from lots of people as they're doing what they're doing and I think they'll be okay.
00:38:30.639 --> 00:38:34.960
I mean it it's but that goes back to the shape of things.
00:38:35.119 --> 00:39:00.800
You know we saw you know early on America was very much a farmer oriented country before the industrial revolution where you know we had lots of small farms and that was what our economy was you know really around and then that has evolved as technology has evolved and we're gonna we're seeing that evolution again globally and at an accelerated rate will the shape of what corporate looks like change?
00:39:00.960 --> 00:39:09.679
Yes it is changing I think we'll see a lot more small small businesses like we've had actually in the past which is good because that I think it's a good thing.
00:39:09.840 --> 00:39:10.320
That's good.
00:39:10.400 --> 00:39:48.960
It's the innovation comes out of that so the shape of what the world is from a career you know getting your corporate job and staying there for 30 years and retiring maybe you know you know this is the opportunity where people can you know chart their own way that but that hasn't been the case for I mean I have this conversation with my daughter who's who's gonna go going to go to university with that's part of that whole debate as well of course that the whole oh I go and get a job and stay in it for 30 years not doing anything different that's that's not how it's been for years and years and years.
00:39:49.360 --> 00:40:04.559
About 30 years you know you might obviously you might make a tangentio you don't have to make that decision.
00:40:04.719 --> 00:40:15.679
You have to make if you're a young person you have to make the decision of what's what's it really in of interest to me where can I go and earn some money that's important.
00:40:15.840 --> 00:41:59.119
That's why we all got into technology sales right it wasn't because well certainly wasn't for me that I was getting that I had a PhD in something yeah you go to go go there to earn some money and then you start learning and then you start that process and I totally agree and we'll make it the last point I totally agree with you that so many more young people are going to go to a company kind of get foundations in place and then they're gonna move out and start their own their own gig right their own startup because you can do that you can do that not for free but you don't have to borrow a hundred grand anymore to start a start a business and get a website and make it transactional all of that you don't have to do that and we we saw our you know billion dollar businesses that have that have come along in recent years that are very few people uh you know where where they take an idea and they with a very small staff made a a huge business out of it I think we'll see more of those I mean you see that reported in you know forecasts that we'll see more of those and I I agree I think not everybody's gonna be there but you know the average person that's out there you know trying to make ends meet trying to make a living they will find a way i and we're all very innovative and creative and I think the best thing that you know I think works there and is find what you're passionate about and I think like what you're talking about they start a job they almost to figure out what they like to do and what they're passionate about and then for a lot of people that I've talked to then they go off and build their own thing.
00:41:59.280 --> 00:42:05.360
Yeah but but yeah I I think it's a it's a wild world out there and it's exciting.
00:42:05.679 --> 00:42:54.960
Yeah I I do too we're we're optimists not doomers that is for sure so Ben that was a fascinating conversation I really learned a lot from you there I'm sure everyone else did who's listening show notes links uh sorry links are all in the show notes Ben Pickett go sign up for Ben's newsletter it's fantastic field notes benpicket.com and my daily is swar-ai.com we'll go check us out and let me know what you think yeah enterprise anyone in enterprise if you're thinking that was an interesting conversation and my token costs are getting a bit big these days go go reach out to Ben but Ben I've got a bail so great to see you again my friend I shall see you on the next one thank you bye bye bye bye bye bye
00:00:00.160 --> 00:00:12.800
I I think the bigger question which is kind of related to this content development cycle is commoditization of the frontier models.
00:00:13.359 --> 00:00:15.839
And which is wild when you think about it.
00:00:15.919 --> 00:00:25.600
And and no, and and look, before so we we came up with this plan for what we're going to talk about today, frontier models turning into a commodity.
00:00:26.320 --> 00:00:28.239
Who's going to win in that?
00:00:28.640 --> 00:00:33.920
And then we've also got the agents coming into play now.
00:00:34.000 --> 00:00:44.159
So you've got Grockbot, you've got Meta, Muse, their agent, obviously the agent layer, which is like kind of trying to make this thing useful.
00:00:44.240 --> 00:00:50.560
You've got dots, openai release dot, and and and so you've got all of that.
00:00:50.640 --> 00:00:54.479
That's probably more of an interesting conversation, I think, for me.
00:00:54.719 --> 00:01:04.319
And then no sooner had we kind of come up with this plan and what to talk about than Google suddenly released Gemini 4 today.
00:01:04.400 --> 00:01:05.280
And guess what?
00:01:05.439 --> 00:01:13.040
It's the now the best benchmarking benchmarks model out there, right?
00:01:13.200 --> 00:01:15.280
And I was for now.
00:01:15.359 --> 00:01:18.799
Yeah, and that and then and do you know, do you know, Ben?
00:01:18.879 --> 00:01:20.159
I'll get your view on this.
00:01:20.640 --> 00:01:23.359
I'll get your view on the kind of agenda as well.
00:01:23.599 --> 00:01:26.400
But I'll there's no loyalty.
00:01:26.879 --> 00:01:29.680
So so so so right.
00:01:30.239 --> 00:01:36.000
I was using Claude, Fable, 5.1, Opus interchangeably.
00:01:36.319 --> 00:01:40.799
If you'd asked me six weeks ago, would you leave this?
00:01:41.040 --> 00:01:42.239
I would have said no.
00:01:43.200 --> 00:01:47.040
And then ChatGPT, something went off with Claude.
00:01:47.120 --> 00:01:50.799
It wasn't even like ChatGP Chat GPT's Astra model.
00:01:51.359 --> 00:01:54.000
I was like, yeah, it's great, but it wasn't that.
00:01:54.159 --> 00:01:57.599
It was something went off on the performance with Claude.
00:01:57.760 --> 00:01:59.040
I don't know what the hell it was.
00:01:59.200 --> 00:02:08.000
Someone else on LinkedIn had a little convo with them because they posted and said, has anyone noticed Claude's kind of got it's got off the boil?
00:02:08.080 --> 00:02:10.240
Basically, the answers aren't so good.
00:02:10.800 --> 00:02:13.599
My loyalty was like zero.
00:02:13.919 --> 00:02:16.479
I switched on a dime.
00:02:16.639 --> 00:02:24.400
It's I've got I've got a subscription kind of problem, but but I that that but for me, that's all part of it.
00:02:24.639 --> 00:02:29.360
That's what because it's a commodity, I don't have any brand loyalty.
00:02:29.599 --> 00:02:31.759
And in fact, I don't particularly want to be with any of them.
00:02:31.919 --> 00:02:34.560
I'd love to be one of the open models, to be honest.
00:02:34.879 --> 00:02:35.919
I I totally agree.
00:02:36.000 --> 00:02:57.039
And one of the things that we have to look at is if you're an enterprise and you know, and you're going through this adoption and you get locked in in six weeks, that multi-million dollar contract that you did with one of the frontier models, you you might be in a or you might be very regretful.
00:02:57.360 --> 00:03:11.599
There's there's no way, and and I think actually this is this actually is hindering AI adoption in general, is because there is this fear about moving forward with any one model just because it is changing so quickly.
00:03:11.919 --> 00:03:14.960
And my answer to that is you need to find a way.
00:03:15.120 --> 00:03:17.680
This is what I do with my day day daily.
00:03:17.919 --> 00:03:27.039
You need to find a way to abstract yourself from a layer from those models so that you can move, you can choose as things move, as things move forward.
00:03:27.120 --> 00:03:38.159
You need to make sure that you're able to take advantage of the latest agents that are coming out, take advantage of the latest models that are coming out, and not have to think too much about it from an enterprise standpoint.
00:03:38.319 --> 00:03:49.680
And I think there's a way, I know there's a way, that's what I do, that to abstract that and make it so that you can control, you have the control, the guardrails, the access.
00:03:49.919 --> 00:04:16.480
So you're the big thing here is it also helps prevent shadow AI, where you know your employees are picking up their phone and they're putting your IP into their phone for their specific, you know, model of choice because you don't provide it by having those models available as they come out with the governance and security and guardrails in place, you end up with a much better experience for your employees.
00:04:16.560 --> 00:04:17.920
They end up being more productive.
00:04:18.079 --> 00:04:25.040
You end up with a more holistic adoption process because you know, program, if you will.
00:04:25.439 --> 00:04:31.439
And you don't spend millions of dollars on something that is obsolete in six weeks, right?
00:04:31.600 --> 00:04:33.439
And and and that's a real fear.
00:04:33.519 --> 00:04:36.800
I I hear about it all the time when I'm talking to customers.
00:04:36.959 --> 00:04:44.399
They are afraid of making a huge commitment to one of these frontier models because next week there might be a new thing, and now they're stuck in this one.
00:04:44.639 --> 00:04:51.360
And from a competitive standpoint, I as a company trying to be more competitive than than my competitor.
00:04:51.680 --> 00:04:57.759
If I can move to the the better model faster, I can get ahead of where they are at.
00:04:57.920 --> 00:05:11.920
Now, if I don't have to think a whole lot about that, because I've set up you know my enterprise platform or go with a vendor that allows for that, that gives you the ability to pivot very quickly without disrupting your business.
00:05:12.079 --> 00:05:14.879
And I think that's going to be really important as I move forward.
00:05:15.279 --> 00:05:36.079
These agents that you you know, going back to the agents, these super agents, if you will, these bots that you know go off and spin up a you know a little computer, whether it's computer Kubernetes in the background or what have you, or some container in the background, and run these processes, install the code that you need to run those things, and then essentially complete projects.
00:05:36.319 --> 00:05:42.079
You know, this is kind of the next wave, and this is something that I'm very passionate about because it is what I work on.
00:05:42.959 --> 00:05:51.839
There, it's gonna be this, and we'll we'll be in this world of this evolving and and going back to your commoditization.
00:05:52.079 --> 00:06:04.000
This will become a commodity where there are going to be racing at who can have the best bot offering where it can go off and do tasks for you and then come back with the result.
00:06:05.680 --> 00:06:11.920
So on that, let me just jump in there because it's important we bring the audience with us on that.
00:06:12.000 --> 00:06:16.720
And and and what we're referring to, so obviously everyone's familiar with the large language model.
00:06:17.040 --> 00:06:19.680
Then if you remember, was it earlier this year?
00:06:19.920 --> 00:06:21.839
Must have been OpenClaw came out.
00:06:22.000 --> 00:06:24.399
Peter, who then moved to OpenAI.
00:06:24.639 --> 00:06:41.199
OpenClaw came out, and it was the first time you could actually just spin up your own agent, have it sat locally, or my mine was on a on a VM, but then just it it you could give it a task, and basically anything you could do on your desktop it could do too.
00:06:41.279 --> 00:06:48.000
And that was it sort of preceded the cowork functionality in the frontier models.
00:06:48.240 --> 00:06:48.480
Yeah.
00:06:48.720 --> 00:06:55.199
And even the cowork functionality in the frontier models, it's not what OpenClaw could do.
00:06:55.439 --> 00:06:58.240
OpenClaw could really, really help you out.
00:06:58.399 --> 00:07:04.319
It's genuinely like a virtual assistant that could do tasks.
00:07:04.959 --> 00:07:21.360
And it's when I really first learned about cron jobs, and and that's that's setting up a job, a specific job that an AI agent will go and do at a certain time, and then you could build these kind of like little workflows for it.
00:07:21.439 --> 00:07:22.480
And it was fantastic.
00:07:22.639 --> 00:07:31.600
And I I've mine is still running and and I love it, and I tell you what, it's much more sticky in terms of user.
00:07:31.920 --> 00:07:43.600
If you've invested some time setting all of that up, that's the sticky part, not the large language model, even with cowork capability, it's not that sticky.
00:07:43.920 --> 00:07:52.959
So then what's what you're referring to, Ben, is we've had a spate of these kind of like AI agents come out.
00:07:53.040 --> 00:07:56.560
We've because they kind of like block and tackle each other, don't they?
00:07:56.720 --> 00:07:57.439
These big companies.
00:07:57.920 --> 00:07:58.160
They do.
00:07:58.480 --> 00:08:14.000
We've had Grogbot, which is when you log into it and download it, you can see it's coming from Cursor, SpaceX bought XAI bought Cursor 60 billion dollars or some loose change they had knocking around.
00:08:14.399 --> 00:08:24.160
And and and so that came out almost immediately, hot on the heels, MetaMuse came out, very different product, and we can talk about that.
00:08:24.319 --> 00:08:32.720
And then the Microsoft, not Microsoft OpenAI dot has come out like literally the last couple of.
00:08:33.039 --> 00:08:35.279
I think I first saw that yesterday on X.
00:08:35.759 --> 00:08:40.639
So Microsoft has also released their own sort of version of this as well.
00:08:40.799 --> 00:08:47.759
So they're all they're all like they're not necessarily on parity yet, but they're going in the same direction.
00:08:47.919 --> 00:08:54.720
And we'll see we're gonna see this evolve for an enterprise for end users, it's a little easier to adopt right away.
00:08:54.799 --> 00:09:07.279
For enterprise, they have to figure out how to put the cardrails around it and as they go forward and make sure, make sure that they don't have things happen like what happened with the agents breaking into Australia Medicare.
00:09:07.440 --> 00:09:09.600
And that's completely possible and doable.
00:09:09.759 --> 00:09:11.840
It's just as they do that.
00:09:12.159 --> 00:09:24.879
My big thought on this is sure, each individual one of these model providers are are providing these bot abilities, but you got that their model.
00:09:25.679 --> 00:09:37.600
You know, if their model goes down, if their model is is under load, I mean, every morning, you know, we I monitor the health of the models, obviously, because you know, we we provide all of them.
00:09:37.679 --> 00:09:51.200
And the every morning you see you often see that, you know, as the West Coast spins up, you know, they you start to see constraints on the that's the ability to use some of these models for some of these model providers.
00:09:51.360 --> 00:09:56.399
Sometimes all of them well well well, Ben, try being in the UK.
00:09:56.559 --> 00:10:06.399
You have this lovely most of the morning, you get great model performance, and then who wakes up East Coast US?
00:10:06.879 --> 00:10:08.080
Yes, that's very true.
00:10:08.559 --> 00:10:24.639
Then then West Coast US, and it's like the model performance, it's just no nose diving, but you know, it's it's it's noticeable every day, so you want to kind of get your big job kind of like things done, like the big tasks done.
00:10:25.440 --> 00:10:30.000
Ben, take a look at your microphone as well, mate, because there was a little bit of crackle on there.
00:10:30.080 --> 00:10:31.600
I don't know if it's jumped again.
00:10:31.759 --> 00:10:33.519
Just make sure you're on the right mic.
00:10:33.759 --> 00:10:36.080
So and then and then okay, cool.
00:10:36.240 --> 00:10:46.159
And then you mentioned co-pilot, autopilot, and oracle's fusion claw, which I saw actually.
00:10:46.240 --> 00:10:47.840
I did I even see autopilot.
00:10:47.919 --> 00:10:49.279
Tell us about those, mate.
00:10:50.000 --> 00:10:53.519
So let me let me get the the details here.
00:10:54.000 --> 00:11:07.360
So if you've had I say let tell us all the if you've even had chance, I'm no, I was you're mentioning things to me I've never even heard of yet, and I I'm on top of this stuff all day, every day, right?
00:11:07.440 --> 00:11:12.559
So it's absolutely fine to go let me sign up the web page and I'll read it out to you.
00:11:12.799 --> 00:11:14.480
Nah, I got it right here in my notes.
00:11:14.639 --> 00:11:19.919
Last week, so Microsoft last week, Oracle yesterday, and then Microsoft's co-pilot, autopilot.
00:11:20.000 --> 00:11:23.360
So it gives each agent its own cloud computer identity.
00:11:23.600 --> 00:11:28.960
So here they are coming along with where other people are, and it moves that agent work to usage-based billing.
00:11:29.120 --> 00:11:31.360
So the seat price no longer caps up bill.
00:11:31.519 --> 00:11:32.399
So that's good.
00:11:32.480 --> 00:11:42.240
But you also want to make sure that you look at you know data usage in these various platforms, you got to be careful with because there's hidden costs, so just be careful there, potentially.
00:11:42.399 --> 00:11:58.240
And then so with that, also Oregon's fusion or Oregon, Oracle's Fusion Cloud is its own agents for finance, HR and supply chain, which that that's actually if you think about that, they're they're going into a niche there.
00:11:58.320 --> 00:12:00.720
So that's I kind of think that's a little smart.
00:12:00.879 --> 00:12:04.000
And so it runs runs it both on Gemini and OpenAI model.
00:12:04.240 --> 00:12:06.639
So there they're thinking about you know sorry, mate.
00:12:06.720 --> 00:12:07.679
Is that the Oracle?
00:12:07.919 --> 00:12:10.320
Is that the Oracle Oracle Fusion claw?
00:12:10.399 --> 00:12:10.639
Yes.
00:12:11.120 --> 00:12:12.720
Fusion core claw.
00:12:12.799 --> 00:12:13.840
Sorry, yeah, go on.
00:12:14.000 --> 00:12:16.000
So that's running running on Oracle.
00:12:16.159 --> 00:12:19.919
Sorry, uh, running on OpenAI and Anthropocet.
00:12:20.720 --> 00:12:22.480
No, OpenAI and Gemini and Gemini.
00:12:23.519 --> 00:12:25.200
Open AI and Gemini models.
00:12:25.440 --> 00:12:34.000
So so you're starting to see that a little bit, and I think we're gonna see that more and more as you know, Oracle, you know, stepping back.
00:12:34.240 --> 00:12:38.720
You see this with some of the big providers as well, like the HVEs and the Dells of the world.
00:12:38.799 --> 00:12:42.559
They've been thinking about this too with their platforms and what they're doing.
00:12:42.639 --> 00:13:00.799
They're they're stepping back, and they have you know the NVIDIA models and things that they offer as well, and they're stepping back and looking at their holistic environment and offerings to their customers, not only with the frontier models, but also with you know, bringing along the models that are that they provide, and then also that the customers are.
00:13:01.600 --> 00:13:17.679
And and speaking of the Dell and HPEs of this world, and IBM and IBM, you've got NVIDIA's Nemo Claw in the mix there as well.
00:13:22.159 --> 00:13:28.559
It's a lot lot lot broader market for enterprise actually when you start going through it.
00:13:28.960 --> 00:13:29.759
And it goes back to the back.
00:13:32.799 --> 00:13:35.759
Let me just ask one more, I'm I'm asking so many questions of you.
00:13:36.240 --> 00:13:39.919
I am assuming Oracle's targeting enterprise.
00:13:41.519 --> 00:13:46.879
Yeah, I I mean obviously finance HR supply chain, that makes a lot of sense to me.
00:13:47.039 --> 00:14:02.000
And they have a they have a very large customer base with lots of you know, people who know how to run their platform and you know so they have they have a very large you know community, if you will, yeah, in the enterprise.
00:14:02.159 --> 00:14:09.840
So you know it makes sense if they would actually focus on the HR supply chain and and those things as they go forward.
00:14:10.639 --> 00:14:18.480
I think we're gonna Can I ask you a question about the cost of these things?
00:14:18.960 --> 00:14:20.159
Oh yes, yeah.
00:14:20.480 --> 00:14:20.720
Right.
00:14:21.039 --> 00:14:23.120
When I set up my open claw, right?
00:14:23.600 --> 00:14:29.440
Anyone who's done this will know, but if you haven't, you have to kind of give it a an AI brain.
00:14:29.919 --> 00:14:39.919
And for my open claw, originally I set it up with an API into anthropic and it started doing like work for me.
00:14:40.159 --> 00:14:40.960
Do you know what I mean?
00:14:41.039 --> 00:14:47.039
It was like doing checking emails, doing daily this, that, and the others, nothing huge.
00:14:47.279 --> 00:14:55.440
And I started getting started getting my API invoices through through Anthropic, and the first one was for twelve dollars.
00:14:55.759 --> 00:14:58.000
I thought, whatever.
00:14:59.440 --> 00:15:01.440
$12, whatever.
00:15:01.919 --> 00:15:07.120
Then in my email, ding, another one for twelve dollars.
00:15:07.519 --> 00:15:10.799
About half an hour later, ding, another one for twelve dollars.
00:15:10.960 --> 00:15:15.360
I was like, switch, switch it off.
00:15:15.919 --> 00:15:28.000
Yes, and at that point, I that was my learning experience of actually using these frontier models, how much it costs.
00:15:28.080 --> 00:15:34.000
And for an enterprise, that is exactly they're probably getting a much better rate than I am, I appreciate.
00:15:34.559 --> 00:15:39.759
But but their usage is is is huge, and that's when, of course, for me it was easy.
00:15:39.840 --> 00:15:42.000
I just switched it to uh Moonshot.
00:15:42.559 --> 00:15:57.919
Yes, uh to the to the Chinese open model, and and and and and it's lived very happily as Moonshot for a long time now, and every now and again I pay them like five bucks, right?
00:15:58.240 --> 00:16:01.679
Like every few months, I oh, it's run out.
00:16:01.759 --> 00:16:04.159
Let me just put another 15 bucks on there.
00:16:04.799 --> 00:16:26.000
And I think that you bring up a really good point, and something that I think about a lot is how can you provide the just good enough great experience for the work that you're doing and not be out there using like Fable 5 all the time and paying tens of dollars for what you usually need to do versus pennies, right?
00:16:26.240 --> 00:16:35.360
So being able to route between those models in the in a not even have to think about you route route between those models is something that is, I think is important.
00:16:36.159 --> 00:16:40.960
As you're doing your work, you shouldn't have to think, oh, I need to change to this model.
00:16:41.279 --> 00:16:49.519
And often when you change to those models, they feed that context that you've had back into the new model, and that costs you more money to do that.
00:16:49.759 --> 00:16:55.279
So having an ability to route between those models seamlessly, I mean, that's what I do.
00:16:55.519 --> 00:17:05.839
But but also as enterprises are building this out, they need to think and think in that sense is okay, I've got this massive growing bill and it's gonna keep growing.
00:17:06.160 --> 00:17:18.160
And and especially when you look at the cost, so that$200 max subscription, it's not it's costing them a whole lot more per month to provide you that compute for that max subscription.
00:17:18.319 --> 00:17:22.799
I've seen numbers between five and fifteen hundred dollars, you know, in compute.
00:17:22.960 --> 00:17:27.039
If there's various things that you hear out there, somebody once has said like 2000.
00:17:27.440 --> 00:17:31.759
I don't know what like what they I've seen what's reported, but it does make sense.
00:17:31.839 --> 00:17:44.240
They are spending trillions of dollars on compute to in this horse race to get to to the front and get the installed customer base and get you know the hooks in.
00:17:44.400 --> 00:17:46.319
So then it's really difficult for you to move.
00:17:46.400 --> 00:17:54.720
And that's why I think it's really important, whatever you use, to think about your company and the money that you're spending in.
00:17:54.880 --> 00:18:14.319
And and also the big thing that I also like to think about is if you look at our trajectory on what we're using as a planet, and you think about the resources and the supply chain, those, the, the, the silicon that goes into it, you know, the GPUs, the the data centers.
00:18:14.559 --> 00:18:19.440
In the US, it's like 1.4, 1.5% of the US greatest all data centers.
00:18:19.519 --> 00:18:27.359
And they're they're saying that very here, very soon in the next, I think past a little bit, past 2030, it's gonna be maybe upwards of more than 10%.
00:18:28.240 --> 00:18:32.400
You think about that, that electricity has got to come some from somewhere.
00:18:32.640 --> 00:18:34.960
All those raw materials have to come from somewhere.
00:18:35.039 --> 00:18:47.119
So, what can we do as companies, as good stewards, is not always use the best model for the mundane work, like the work of moving files or creating emails and stuff like that.
00:18:47.359 --> 00:18:51.200
Should we shouldn't be using those fable fives for that?
00:18:51.359 --> 00:19:13.759
We should be able to deconstruct those prompts and use the pieces that need fable five, great, you know, as we need it, and then all the mundane stuff can be used on a Chinese model or another open source model that even like in the you know the on-prem world that could be an on-prem model that somebody's using with their current infrastructure.
00:19:13.920 --> 00:19:21.519
So yeah, companies are gonna have to figure out what their tolerance is here, but more one more thing on that.
00:19:21.680 --> 00:19:31.599
There's a reason why the stats are out there that say 95% of companies are trying to adopt AI do not have that uplift that they're looking for, and they're not what we call successful.
00:19:31.839 --> 00:19:33.680
Only about five percent are there.
00:19:33.920 --> 00:19:44.799
You see them saying, Oh, our adoption rate is this and this, but if you look at their PL and their feature set for their customers and their increase in you know, increase in sales and revenue, it it doesn't match up.
00:19:44.960 --> 00:19:49.200
There's a very few that out there that actually are getting there.
00:19:49.279 --> 00:19:52.079
I mean, the Walmart, you know, Walmart's a really good example.
00:19:52.240 --> 00:20:00.559
They they pivoted early when they started to see this and they came up with kind of their they don't call it a blueprint, but I have a newsletter on enterprise film notes you guys can go take a look at.
00:20:00.640 --> 00:20:04.400
I also have kind of a you know a little bit about that roof type that you can use.
00:20:04.480 --> 00:20:06.640
That's all free, so go subscribe.
00:20:06.960 --> 00:20:23.440
But that is a really good example of you know a company that would this big ship of a company moving forward with AI, having agents sprawl everywhere, and then be like, okay, wait a second, let's look at what we need to do, let's be proactive about this and then make it fit the business.
00:20:23.519 --> 00:20:31.920
And then they had that like 35% correlational, 35% uplift in in you know, for some of the some of the business.
00:20:32.079 --> 00:20:33.039
That's a big deal.
00:20:33.119 --> 00:20:36.240
That means that what they're doing, their customer experience is improving.
00:20:36.400 --> 00:20:39.599
That that's a good signal for them to continue on that road.
00:20:39.839 --> 00:20:45.359
And also they're controlling their, you know, they're they're really thinking about the holistic and controlling their costs.
00:20:45.519 --> 00:20:54.640
The VS, you know, Uber, who in the first quarter of this year said, we spent all of our budget, we have nothing to really show for it.
00:20:55.680 --> 00:20:57.119
Now we gotta rethink this.
00:20:57.279 --> 00:21:01.359
And and I I can only imagine what that looks like.
00:21:01.599 --> 00:21:03.759
Where how how does it happen?
00:21:03.920 --> 00:21:15.119
That happens because you know, you you think you go after AI because AI, and you don't go after AI because you're trying to achieve certain business results.
00:21:15.599 --> 00:21:17.519
I think I th I think you're you're right.
00:21:19.519 --> 00:21:22.559
It's packed full of information that's really useful there, Ben.
00:21:23.200 --> 00:21:36.640
I think also what kind of a realization was for me last week was so GPT Astra came out, big fanfare, usual thing, it's the best model, blah blah.
00:21:36.960 --> 00:21:52.880
And then I can't remember where I saw it one of the big podcasts, put a slide up just saying a chart, just showing that they had now overtaken the adoption in Enterprise for the first time from anthropic with Claude.
00:21:53.680 --> 00:21:56.319
I was like, but they just brought the model out.
00:21:56.640 --> 00:21:59.599
So wouldn't that take some time to move all?
00:22:00.240 --> 00:23:06.160
of the so if you're running services like customer services or whatever and that's you've got all that based on the anthropics models you're kind of gonna need some time to get those moved across and then I thought well no this is kind of like the tier top tier of all of the models it's not like this one of these like flash models Astra's the top tier and then it sort of dawned on me ah it's the coding isn't it so if I'm a hedge fund or I'm Arcley's bank or whatever I'm not talking where what's driving that growth in enterprise the massive spend where am I never gonna use a model that's slightly behind in the rankings it I might be wrong on this but I'd say it's coding and then it's faster shift to switch right yeah my coders are just sat there and they just go hey I just switched from using this to using Astra gave the feedback internally and then no one's looking at the bills.
00:23:06.400 --> 00:23:32.960
And and you mentioned hedge funds you know traders and things of that nature leveraging these to the nth degree I remember years and years ago when having the fastest network switch did where where milliseconds was not not fast enough and they they would they would sell these switches that were super fast because they wanted to make sure that that next trade was before was before the next guy.
00:23:33.200 --> 00:24:39.599
So absolutely where there's business use I I I agree with you there where there is a business reason to where they need to be out in front you're gonna see that adoption and and cost is not the biggest concern where they where the where that cost impacts companies is I think more so on the services side or on the retail side where they're looking at okay they need to keep costs reasonable and down you can't raise prices significantly because you decide that you're gonna use fable five for your entire retail you know yeah exact exactly exactly so so that's it it's kind of more of a sticky use case within your enterprise environment isn't it and governments too you know where there are different intelligence governments intelligence you saw we saw that China was China Claude reported anthropic reported it that they had got figured out that China was using using their model for allegedly using their model for intelligence and they had gotten rid of part of their intelligence team as a result.
00:24:39.680 --> 00:25:02.000
So you're like okay that's not great and see and that's the reason where they they have access to these great Chinese models and they're still using anthropic because they want to be they want to use the best of the best because that's how you you move the needle and get ahead of your gut you know your global competitor if you will so and they're not the only ones that are doing this.
00:25:02.079 --> 00:25:49.599
I'm not saying I'm picking on China but all the countries are that are have this stake are are more than likely doing these type of things where it matters the most you'll see the early adoption and that will that accelerate the growth where you know money comes into play and you're trying to make you know you don't have limited funds the for the rest of us in in the enterprise world being proactive and smart about it we we want to ensure that our bottom line and our profitability continues and that we don't eat up that profit layer you know that profit margin by spend on using the the top the the absolute top models when we can use much cheaper models to get the same job done.
00:25:49.839 --> 00:26:04.640
And I know that this is on the minds of lots of customers that I've talked to they think about this one of the biggest things that I've heard from even technology service providers is you know what we got to get our model cost down.
00:26:04.720 --> 00:27:28.880
You know we're providing services to our customers and we're providing these platforms to the customers but help us figure out how to get that model cost down because you know it's it's weading into their margins for what they're what they're doing it also slows them down because when the cost is there and the cost is so high you're you're gonna use less right whereas if you can get the cost down and this is the other big point I want to talk about by being smart about how your your AI spend is you get more productivity and more benefit out of the dollars that you're spending so you're you're going to accelerate your business profitability if you do it right right because you're not having as much spend per revenue per dollar of revenue that you had would have otherwise so then you can end up getting more revenue and accelerating those accelerating your business the same thing happens with code you look at code and what we're doing more code is being generated than ever than anytime ever because it's AI enabled the problems come with that from a quality standpoint which we still I don't think we've really figured that out yet like quality how do you handle quality when you're generating so much code and your resilience and reliability because having the human in the loop there becomes harder and harder as that as the number of lines of code that are generated increases.
00:27:29.200 --> 00:27:41.599
So companies are trying to figure that out as well but you know this is a we're on a rocket ship I would say and we just are all figuring out together and that's the whole point of us coming on here is to to share notes.
00:27:41.920 --> 00:27:56.400
Yeah yeah and look let's just thank you thank you Ben for for that insight and it's great for coming from somebody who's actually the layer of access your platform into all those model models for enterprise so that's that's awesome.
00:27:56.559 --> 00:28:29.440
So we've had so we've had meta muse I haven't I have not set up meta muse apparently it's much more aligned to kind of consumer whereas I'd say grokbot which I have set up by the way I've got grokbot kind of I'm just going through the motions kind of giving it little tasks here and there now but you can see it's more I don't know it's just more complex as a product meta I one of the benefits of MetaMuse that I like the sound of is you can communicate via WhatsApp through it.
00:28:29.519 --> 00:28:46.240
So when I set up my OpenClaw and I I communicate via Telegram through through open claw grockbot has its own very good actually very good kind of chat app if you like so I've got it on desktop and I've got it on my iPhone.
00:28:46.400 --> 00:29:34.720
That's been a really nice product actually but Muse I have not used I think we talked about last time so I won't go into it amazon immediately stopped the Muse shopping agent yes going on to its platform because they were like they didn't you they they said Amazon said that the meta people forgot to tell them that they were going to be sending this agent over so that was interesting and we'll do some digging around Microsoft's co-pilot autopilot I need to be co-pilot studio session so whilst I'm doing that I will check out autopilot and then look I'm keeping on time as well I what caught my attention from your newsletter then was Apple.
00:29:35.440 --> 00:30:53.279
I always love it I always love seeing when this is the Klaners and all the rest of it when a company says we're gonna replace the humans with the AI agents I'm I'm almost like grabbing my popcorn sitting back to watch the show so what happened with let's cover Apple and then let's start wrapping up so so Apple Apple was you know they were gonna lay off a bunch of people and and they talked about it earlier on and then they decided not to do that and and they won't actually comment on the why but there it's pretty obvious the you see you know Ford that I wrote about they they brought back their engineers you you you see this across the board where a company is like oh we're going ai okay get rid of all these staff and then they come back and say oh oops that was not the best thing to do because they just it's like they're they're they're using they're using AI to to go after you know to find a problem to solve instead of creating a solution that makes sense for the business.
00:30:54.079 --> 00:30:55.599
So who were these people in Apple?
00:30:55.759 --> 00:31:07.359
What do we know and it's all allegedly because obviously there's no press release on on this but who who was the affected group in Apple I think is the term we use these days.
00:31:07.680 --> 00:31:51.359
Yeah so Bloomberg reported that over the summer Apple drew up a plan to lay off about 5,000 Apple care support workers support and have AI phone and web agents take over some of their work it's now on hold it's on hold indefinitely and yeah that's that's what we know which it makes sense to me I do think that you know as we move forward and understand you know what the business needs are and what the customer needs are sometimes when you throw when you shotgun approach your adoption ai you you you get a rude awakening when you realize how that may impact your overall business.
00:31:51.519 --> 00:32:25.519
And so we just need to take you know a stepwise approach and and step back and and I think you know to to Apple's credit I mean I don't we don't know the internal details of other what's just been reported but to their credit they're probably making the right decision because they are like you know what either we're not ready and or or this isn't a good fit and and being able to step back and make that pivot is important especially for enterprises like Apple taking a misstep there could could have very impactful results.
00:32:25.680 --> 00:33:08.160
We see that you know as big companies stumble and if they stumble it it can have a significant impact in their business and their brand it's always kind of like after sales isn't it's like we've sold it now so Apple care you know let's just let's just take a look at that uh uh automating automating that I I do think the these things are connected as well I think there's a big push out there in these boardrooms to say well Elon Musk who's stood next to Donald Trump saying that in the next couple of years anything that you can is currently done by humans sat in front of a screen is going to be done by AI.
00:33:08.640 --> 00:33:42.240
Now right and so that creates pressure in the boardroom noticeably Dario and Sam Altman have stopped saying that they were saying the same things but they've stopped saying that and and then if you look at so that creates pressure in the boardroom I think that why aren't we being more radical are we going to get outstripped by our competitors here because we're not pushing hard enough on automation and well we said it a bunch on this podcast automation's not automating out the humans is not the goal.
00:33:42.559 --> 00:34:04.319
And Elon Musk and hey I'm an investor in both Tesla and SpaceX so clearly I believe in the in in the leader and innovator and I think the man's a genius but do I think he's genius level in his forecasting ability of when he says things are going to happen.
00:34:04.400 --> 00:34:25.360
No I don't Robo Taxi's taken you go back to when he says they're gonna deliver things no no that's fine and I think that's part of the management style I think he says and and so so that it's commonly known I think in in in the circles that I sort of am in as an investor it's called Elon time.
00:34:26.480 --> 00:34:39.920
So just because he's next to Trump saying this does not well you could argue it makes gonna make it less true and Elon's not like I don't think he's making it up I think he genuinely believes that's what's going to happen.
00:34:40.079 --> 00:36:21.599
But then you have to go yeah but is that going to happen in the next two years or is that going to happen in the next eight years from the management time horizon of most boardrooms which is probably three five years are you actually going to have automated out all the staff that are using a screen and I just think we know we can see the answers no look at the conversation we're having doesn't mean you're it doesn't mean you're not going to use ai to make that same group of people hopefully all of your people be able to be much faster and and and will there be some impact on new hires and all the rest of it but possibly so maybe you need kind of less people but hopefully then you are able to grow your business and do more and and that's a good point because the the good point and and I actually believe this and and and it might be something that's not thought of like I maybe one of the everybody's like oh you know we're adopting AI you know we're gonna have less jobs I don't think that's true if you look go back and look through all the paradigm shifts we've had since industrial revolution you know when electricity came on the telephone etc initially there is a disruption and we're in that disruption right now which means jobs are being lost the thing about it is when we have these generative generativity type platforms and not generative have a generativity type platforms that create more innovation we we end up creating more opportunities and accelerating what's going on so what does that end up in the end result?
00:36:21.920 --> 00:37:18.960
It ends up with not necessarily the same jobs or the same positions but there is more opportunities for jobs because there's more opportunities that are being created and and and let's look at where job so if you look at what what can we look at autonomous driving will that disrupt uber drivers jobs yes yes it will for sure disrupt uber drivers jobs and then the uber drivers are not all of them but a lot of them are driving for Uber not because they love driving for Uber if next time you're in an Uber ask your driver do they love driving for Uber as a company you're not gonna hear them saying loads of great things about Uber I I guarantee that and and so you're going to you're going to get that but that shift is going to happen over a period of time.
00:37:20.239 --> 00:37:58.559
It's not going to be that Waymo Wave and Tesla and there's a few others knocking around right are en masse gonna take all of those jobs overnight but will there be a shift in the workforce absolutely that's what humanity's done when we invented the wheel and the industrial revolution and so and so of course but we've always been like that Ben that's it's always always been that way we've been in technology for bloody 30 years.
00:37:58.880 --> 00:37:59.920
Yeah that's true.
00:38:00.159 --> 00:38:30.480
I and back to your Uber comment it's really funny you say that because it brings back memories of every time I'm at Uber I ask you know I I ask questions because I want to know like where'd you come from what do you do more often than not they you know I met a number of startup guys out there who are driving Uber just to get by while their startup is gets going so the entrepreneurial ship you know those Uber drivers you know they talk to lots of people they learn from lots of people as they're doing what they're doing and I think they'll be okay.
00:38:30.639 --> 00:38:34.960
I mean it it's but that goes back to the shape of things.
00:38:35.119 --> 00:39:00.800
You know we saw you know early on America was very much a farmer oriented country before the industrial revolution where you know we had lots of small farms and that was what our economy was you know really around and then that has evolved as technology has evolved and we're gonna we're seeing that evolution again globally and at an accelerated rate will the shape of what corporate looks like change?
00:39:00.960 --> 00:39:09.679
Yes it is changing I think we'll see a lot more small small businesses like we've had actually in the past which is good because that I think it's a good thing.
00:39:09.840 --> 00:39:10.320
That's good.
00:39:10.400 --> 00:39:48.960
It's the innovation comes out of that so the shape of what the world is from a career you know getting your corporate job and staying there for 30 years and retiring maybe you know you know this is the opportunity where people can you know chart their own way that but that hasn't been the case for I mean I have this conversation with my daughter who's who's gonna go going to go to university with that's part of that whole debate as well of course that the whole oh I go and get a job and stay in it for 30 years not doing anything different that's that's not how it's been for years and years and years.
00:39:49.360 --> 00:40:04.559
About 30 years you know you might obviously you might make a tangentio you don't have to make that decision.
00:40:04.719 --> 00:40:15.679
You have to make if you're a young person you have to make the decision of what's what's it really in of interest to me where can I go and earn some money that's important.
00:40:15.840 --> 00:41:59.119
That's why we all got into technology sales right it wasn't because well certainly wasn't for me that I was getting that I had a PhD in something yeah you go to go go there to earn some money and then you start learning and then you start that process and I totally agree and we'll make it the last point I totally agree with you that so many more young people are going to go to a company kind of get foundations in place and then they're gonna move out and start their own their own gig right their own startup because you can do that you can do that not for free but you don't have to borrow a hundred grand anymore to start a start a business and get a website and make it transactional all of that you don't have to do that and we we saw our you know billion dollar businesses that have that have come along in recent years that are very few people uh you know where where they take an idea and they with a very small staff made a a huge business out of it I think we'll see more of those I mean you see that reported in you know forecasts that we'll see more of those and I I agree I think not everybody's gonna be there but you know the average person that's out there you know trying to make ends meet trying to make a living they will find a way i and we're all very innovative and creative and I think the best thing that you know I think works there and is find what you're passionate about and I think like what you're talking about they start a job they almost to figure out what they like to do and what they're passionate about and then for a lot of people that I've talked to then they go off and build their own thing.
00:41:59.280 --> 00:42:05.360
Yeah but but yeah I I think it's a it's a wild world out there and it's exciting.
00:42:05.679 --> 00:42:54.960
Yeah I I do too we're we're optimists not doomers that is for sure so Ben that was a fascinating conversation I really learned a lot from you there I'm sure everyone else did who's listening show notes links uh sorry links are all in the show notes Ben Pickett go sign up for Ben's newsletter it's fantastic field notes benpicket.com and my daily is swar-ai.com we'll go check us out and let me know what you think yeah enterprise anyone in enterprise if you're thinking that was an interesting conversation and my token costs are getting a bit big these days go go reach out to Ben but Ben I've got a bail so great to see you again my friend I shall see you on the next one thank you bye bye bye bye bye bye
