Real Builders Are Quietly Winning the AI Era

It's it's crazy like content lead you can use in CHP and it just shows up in CHD. It's amazing. Um I mean AI voice I mean we've got I think one of the most advanced AI voice applications out there probably anywhere right now like built on top of 11 Labs and um a few other uh um uh Twilio and a few other technologies. Like this is just in one year. It's just insane. absolutely insane, you know, and that's how I think I think you and I can talk with a little bit of authority. We have a like a lot of backing like because we're actually doing it. Like we can see we can see it because we're actually doing it. You know, there is it's so funny like when um I go on YouTube and there's um there's quite a lot of like famous AI people now who say you know Gemini comes out and they're like Gemini just killed the thousand startups or something like that. like just click clickbait things, but like I don't think they've actually built anything meaningful. Like they're just sort of like, you know, the newscaster of the AI age. But, you know, I think we have a bit more depth. It might not get as many clicks, but it feels like we have a lot more depth because we're actually like right in the weeds of doing actual real stuff, you know? — Um Yeah. It's funny because the only stuff I watch on YouTube now is the people who are actually building things. — Yeah. You know, that's why I like Nate Jones so much because he both has the big picture sense, but he also he gets in and the stuff he builds like the skills he's building are way better than the ones I'm building. — You know, it's like, you know, his PowerPoint stuff is just lights out. And yeah, I'm learning a lot from — somebody who can build at that level. And um — but again it all comes back to doing you know it's like you know taking apart what you know somebody's built and saying how did they make theirs so much better than mine. — Yeah. — And you know we're using the same tools. Um [clears throat] — and then you can really see the difference where the human skill matters. — Yeah. — I am this close to starting a new YouTube channel called Full Stack AI. I've been thinking about it for a couple of months now and you know I didn't I didn't really want to get back into the content creation game but I'm just so close I'm so close um to doing it because I just need to find the right format. I format because I don't think anyone wants to sit through 2 hours of me developing with AI and yelling at AI and you know getting annoyed and then having some progress. So, but some I think there is something there for like what you can do at the edge of these tools, you know, but I'm just trying to think how you sort of wrap it up. like Nate Jones has a really good format, but in a lot of cases he's sort of, you know, he's ideating. he's ideating and I kind of maybe want to just show something like more tangible and so m maybe it's like maybe the idea is that you need to film yourself for a couple of months and then you sort of mash it all together at the end on specific projects or something you know like you actually have to do some quite heavy duty um uh film editing or video editing of like lots of different things. I mean, some people do this, like, you know, Mark Robber might do a project for a month or something or a couple of weeks and he'll sort of mash it all together. Maybe maybe not. Yeah, the other thing the other thing that um you know that I would tell you is um the Gemini Ultra um version has some expanded capabilities even into things like Notebook LM and — what you can do in terms of just throwing some really highlevel stuff in Notebook LM and saying make me a five minute clip on how to do this precise, you know, really specific difficult thing. it'll do it. Um, — and you the visuals are getting really good. The um the voice is great. Um, and so in a sense it's um I'm kind of I' I've kind of got my own custom YouTube channel that makes me videos for the stuff I need to know. — And um you know that — Do you still post on your YouTube channel? — I have never been that. I mean, here's not the way to sell subscriptions on your YouTube channel. Um, I often don't even bother putting my LinkedIn videos on my YouTube channel. Um, I should and I think what I was thinking about is building a skill that'll do it for me. I actually you know I would not be surprised in the next 10 to 20 years that these big tech companies become like energy companies like energy providers because if you think about it they need so much energy and energy used to be funded by um governments local

governments but those lo local governments and governments are not going to benefit from well maybe they would but they're not going to uh develop the energy fast enough I don't for what they need. And so the I wouldn't be surprised these big tech companies actually become energy companies. They become energy suppliers to America to the to to the globe. I don't know. But um I wouldn't be surprised. I would not be surprised. — You hear you hear crazy thing about that. — Yeah. — The first job I had my graduate school internship was with a um the trade organization for the electric power companies in the US. And um and at the time we're working on the project I worked on summer was um it was called wheeling regulations and what it was it was the power generators would have the ability to sell their power back to the grid and so you know I haven't worked in that area in you know 40 years but I know the wheeling stuff happens all the time here and that you know there's constant you're a power consumer, but you can also the next stage turn around be a power producer if your um consumption is higher — than your um than your intake or your production is higher than your consumption. I would bet I mean my guess is right now that they're running at a negative for most of them that their consumption is much higher than their production. But I think you're right. I think it'll probably get to a point where these data companies have um you know kind of peak load covered and so in non- peak load they I bet they sell it back to the grid. I think you're right. — Yeah. And I just think they're going to be the only ones with the money to actually spend on on energy. I mean they are they already are they they already could buy basically any industry right now. Like they're so big. Um and I think energy is just a natural extension of them. They need it. The world can do with unlimited energy. So like there's almost unlimited demand in there. So I sometimes um cuz you you do hear a lot of push back regarding AI sucks up a lot of energy. — My take on that is yes it does and it would be horrific if consumer prices did um elevate significantly because of that. But I do believe that it will encourage the production of more energy um plants and energy providers, right? Which is a positive thing, which is a real positive thing because I mean you hear a lot particularly in the states because it's well studied like it's notorious for like um low investments in this area over the last like few decades. So I think it's like overall long-term gonna be quite a positive thing. Well, I mean, one of the things too that you have to take into account is that cost per token has fallen incredibly. I mean, I remember that Mary Mer report earlier in the year. Let me see the other um the other interesting thing I heard on a podcast today is that Anthropic and Microsoft have done a deal for like for $30 billion of Azure credits or something over the next 5 years. So Microsoft are now um basically in partnership with everyone I think. — Yeah, — it's a smart way to be, you know, because it's like everybody — everybody's doing, you know, something that Okay, so here we go. Um this isn't the one I was thinking of, but um let me just share screen real quick. Um — as the cost per token has gone down has gone up as much you know like or more I don't — and you just look at that in the last 18 months flagship model prices have reduced by 97% cost per million — in the last 18 months. Wow. — Yeah. — It's even just 18 months ago is just just feels so far away. Just honestly — might as well be 18 years. — Yeah. We were using GPT 3. 5 back then maybe I don't know were we or was it four? — Yeah. So I mean you look at this and it's like the you know the usage is going through the roof but the costs are you know at the same time you know falling through the floor. Um so you know kind of who which one races to the end first.