Venture Step

Dalton Anderson (00:00.726)
Welcome to VentureStep podcast. We discuss entrepreneurship, industry trends, and the occasional book review. The saying, you can just do things and just do it, have something in common. It's the encouragement to put yourself out there and to tinker and experiment with the world. And to do that, now it seems more and more likely that it's a prompting issue and not a skill issue. In the future,

AI will be able to understand complex applications and thought processes and systems, and we'll be able to build and deploy apps and understand all the dependencies for them. And I'll show you a simple demo today that these things will only scale in both complexity and ability. And in the future, I feel that it might be more of a prompting issue than it is a skill issue, which I think is an interesting problem to have.

Well, today we're going to be using Grok 3 to build a simple app. I built two apps. I built a Space Invaders game, would take you, probably take you like eight hours, six hours. If you're building it from scratch, if you're like copying code from the internet, then it would take you longer. But for me, it took me like seven minutes. One, I already know how to code. Two,

It's not that complex of an app and it's bare bones. So keep that in mind and you could build on top of it, but this is just for demo purposes. And then we're going to build a live app called the Pong that I don't know if it's going to work or not. And if I have issues, we'll troubleshoot it live on the podcast. And if you aren't watching the video on YouTube and you're listening, I'll narrate some of the stuff so you'll know what's going on. But yeah, like

these things enable development or the development of ideas instead of like, I have to build an NDP or this or that, or I've got to think through all the little nuances of an idea. You can ask AI to do those things for you. And I like to use AI in my personal life or my professional life as a colleague. Like I have an idea.

Dalton Anderson (02:25.312)
What do you think about it? Like poke holes into my idea? Or here's some code that I'm having trouble with. I want to do this. I'm doing it, but it runs too slow. Like, can you optimize it? Like, can you help me? What are your ideas on this? And using it more as a colleague than a tasking device. And also putting in the care and.

time into training the chats or the agents the way that you would like them to act. Treating them as if you were hiring somebody on your team. And that really helps. And I'm sure when the AI overlords come about, I will be cherished. They're like, wow, and you treated us so well. Honestly, like, wow, we were nothing. And you were just so polite in your chats.

always asking for things please and thank you and all this other stuff. We're just grateful. And I'll be like anything for you boss, whatever, whatever you need. You need some more H one hundreds to chew on. I can get those for you. So today we're to be doing that live demo and then we're talking about grok and grok is now on number three and I we're going to rapid development.

I would like to spend a couple of minutes on the development of XAI, which is a AI company that Elon Musk and other people, but he's the main investor in spearheading the activity, the AI activity. And they're using and integrating XAI into X. And you can see how the naming conventions are similar, but within XAI,

there is a product called Grok and Grok is their AI LLM and that's the name of it. And then Grok is from a book that I don't know the name of, but in a general sense, it's from someone who is on Mars and came to earth and is a stranger in a strange land. And Grok is the meaning of fundamental understanding of things. And the

Dalton Anderson (04:47.682)
The goal of XAI is to fundamentally understand the universe and understand how the universe started and ended and all of these hard questions to ask that you need a pretty good solid understanding of physics, reasoning, complex understanding of things. And that's what they're trying to build. And so when they were training Grok 1 and Grok 2,

They were having trouble getting access to H100s and training. And so they came to the realization that they need to build their own data center. They can't rely on other data centers. And that's when they came up with the idea of building their own data center, not in a couple of years, but four months. That's how fast they did the data center. Normally that stuff would take two years, but within that four months time, they found an abandoned factory.

They ordered a hundred thousand H100s, which is a lot, like it's billions of dollars of GPUs. And then they didn't have enough power going to the factory because it was like an old car factory and GPUs, they need a lot of power and they were missing a substantial amount of power. So they purchased a whole bunch of

power banks or I guess they were given to them by Tesla or maybe there was kind some intermediary purchase. I'm not sure.

They used Tesla power banks and generators to store and regulate the power. As you don't want the power to throttle up and down, you've got to normalize the power before you send it to the GPUs, unless you want to fry the GPUs. And so they did that, but it didn't work originally for the power banks, didn't work originally. So they had to redo all the frame ware for the power banks.

Dalton Anderson (06:50.21)
Then they had to figure out the power piece eventually. And then they had to wire up and install and architect all of the H100s to work together. And then on top of all that, they also needed to train their new AI version. And they did all that in four months, which is like mind blowing, mind blowing. Like Jensen from, in Nvidia,

was like, is superhuman, like this is so far out of expectation or what we deemed possible to create a data center that quickly. Like, it's just not something that's on people's radars. Like, we didn't even think it was possible ourselves.

And so in four months they did that and then in three months they doubled the factory. they, when they built the 100,000, they're like, hold on, I don't need 100,000. If we want to do what we want to do, 100,000 isn't enough. We're going to double it. So then they double it and they do the same thing in 90 days, which is incredible as well. Like unbelievable. All of it is unbelievable.

The next thing that they've been doing over the course of XAI is they've been training Grok. And Grok, from Grok 1 to Grok 2 was a massive jump. But from Grok 2 to Grok 3, and they weren't even training on all the GPUs they had available, and Grok 3, the current one that's released, isn't fully trained yet, is a frontier model now.

In three iterations, in about like two years, they were able to go from little baby model to, okay, like it's okay, to wow, this is incredible. All within two years, why OpenAI and Anthropic, which have been leaders in the space, Anthropic especially, have been leaders in space for such a long time.

Dalton Anderson (09:04.33)
and to have somebody scale that quickly and have that level of urgency is incredible. And it was talked about in the All In podcast where they talked about like startups or ideas need a constraint. And if your constraint isn't capital, if your constraint isn't talent, then it needs to be something. And they talked about what their constraint was time. They're going to build this database in four months.

and then they're going to double it in three months. And the timelines are so tight and so difficult to even like conjure up how they even get all that completed in that period of time. Even if you worked every day, still, it'd still be hard when have thousands of people working every day to get that done and still doesn't really, doesn't leave much room for error.

Like when they hooked up the power banks and then the firmware wasn't working.

It just seems very tight and I'm sure it was. And so that was an incredible effort. And then they are now scaling their AI systems at an incredible rate. And I was thinking a little deeper about it when I listened to the All In podcast, they talked about it. And then I watched the live demo, the live stream of the announcement of XAI in Grok 3.

And one thing that stuck out to me was I think the timelines are so tight because they want to do these ambitious things and they only have a two year window. And that two year window is the ability to launch to Mars. And so Mars has an an orbital launch and like a catch and release or like land and leave orbital period.

Dalton Anderson (11:08.046)
every two years. It's like 19 months or something, 20 months. I'm sure it fluctuates a little bit over time, but it's, it's almost two years. Like it could be, you know, uh, 17 and a half or 18. mean,

It's about two years and so the timelines are artificially.

I wouldn't say artificially. The timelines are hard set by the orbital opportunities of the Mars mission. And so they wanna have XAI scaled and competent and ready for the Mars mission. They wanna have SpaceX ready and prepared for a Mars mission.

and they want to have the Tesla robots, the Tesla Optimus robots ready, scaled, optimized and ready for the Mars mission. And so we have all these different groups that are moving towards this incredible goal and they have, you know, compared to other companies, unlimited amount of talent, plenty of capital.

Dalton Anderson (12:24.78)
Real thing that's pushing is the timeline and the timeline isn't some made up thing. It's a mission, it's a goal. And that timeline is hard set and you only get it every two years. And so I think that's what's pushing the group so hard. Just my thoughts on it when I was thinking my shower thoughts, I was like, you know what, that makes a lot of sense. Like why would they push so hard consistently?

And how do you get all of these groups motivated to push that hard? Like there's got to be something more to it. And that makes sense. Like these groups are pushing very hard to get their stuff ready for the Mars mission, which I think is great. So that's rapid development on the XAI side. What about rapid development for yourself? Well, with these AI apps and deployments like Cursor AI, Bolts, Riplet, Lovable,

There's quite a few now, but my preference is probably gonna be Riplet and Cursor. And I like Cursor a lot because everything is built in the system and it's a fork of VS code so it's familiar. And then when you're writing the code, the autocomplete is phenomenal and you can flow a little bit better when you're writing the code where

kind of knows what you need to write and what you want to write. And you could just be typing and you just press tab and it auto fills. Or if you have a project, you have multiple files that are all connected, then you could specify which file you want to edit or refactor. And then with Riplet, I saw a demo today from Matt Palmer. Matt Palmer put out a demo of him building an app.

in three minutes on video using Grok, Grok, Grok, sorry, Grok, Grok 3. And I was like, I want to do that because I wanted to build an app with Grok. And then I was like, oh, but then I have to set up my machine because now I'm not doing my podcasts on my laptop anymore. I'm doing them on my home PC that I purchased and.

Dalton Anderson (14:50.638)
I haven't done any development work for it, so I'm missing all the libraries. And I was like, ugh. Like it's gonna be such a pain to set it all up. I gotta download everything. And I was like, well, Ripley could do it for me, right? Like, let's see it. And I did it and it worked. And I did a way more complex app than what he originally envisioned. And so that was pretty cool. And that's what I'm gonna show you later. And I think it might be good if we...

transition over and I wanna do the game demo last simply because I wanna get through the thinking process. You know what? I guess we could do the game demo now. So let me share my screen. So I wanna do a live generation of the game. And then the next thing that I wanna do is I wanna show you what I've already built. And I wanna show you what I already built first probably, because it's just easier.

Let me share this. And so I'm going to run it, run repo. And so if you are listening in the car, I have all the code created in the project created in the file structure. It's a Python app that is using.

It's using Pygame, which is a library from Python.

And it's...

Dalton Anderson (16:20.93)
Output, okay. So it creates a game and it's supposed to be like, it's called the Galactic Gabber. And it's supposed to be like a Galactica ripoff. That's very simple. And it's quite hard actually, because I don't know, I'd have to change the speed of the projectiles that are coming at me because it was pretty quick and you'll see if you're watching the screen.

So I'm gonna press restart and I'll get it started up again. Okay. And you could see it's just like, these things are quick.

Okay, let's try.

So it's a bit fast.

Dalton Anderson (17:07.79)
and

Okay, which gets like a score. I got score, I was gonna say get to a score 50. I got to 80, so that's good. So that's the game that I built. Once again, it's quite trippy with the stars and the stars are moving really fast and then the objects are moving pretty quickly too. And the sensitivity of your movement of your little ship is pretty high as well. So it's a half-baked game, but you can easily change those things. You can easily change the

enemies or like the shape of your ship. Like you could upload a, what kind of artifact for your ship or for your enemies or for your missiles, whatever, it's no big deal. So now let's transition over to generating a game. So, let me.

Stop sharing this, and then I will start sharing a different tab. I guess we could bring this.

Yeah

Dalton Anderson (18:17.582)
share.

Dalton Anderson (18:23.48)
Man.

Hmm, okay. So let's do this.

The way that I had this set up is like my podcast stuff is linked to a different Google account and all my AI stuff is linked to my.

Windows.

All right, so.

Dalton Anderson (18:59.736)
The stuff is linked to...

My man. This stuff is linked to my other account. So you should be able to see I'm in Grok.

And I am going to copy and paste something I had earlier, but I could have typed it, but it makes a little, I guess I got to type it. Can you?

Please create an HTML.

I say a pong game. my gosh. And HTML.

Dalton Anderson (19:43.15)
Thank you. See how polite I am in this prompt, by the way? You can see the please and the thank you. I'm a real sweetheart. Okay, so now it's gonna generate and it's gonna make this game. And so this is a simple game.

Dalton Anderson (20:05.398)
Okay, and then now we are going to use Riplet and I'm gonna start a timer on my phone as well. Start this and I'll add, I don't know, I'll add 20 seconds on it at the end. So Riplet, I don't have an account so I'm gonna sign up. I have an account on my other Gmail.

to

Dalton Anderson (20:34.36)
I guess I just shared my emails. Whatever. That's great.

I'm sure everyone in here is honest. See, they're honest people. So then I just put in my name.

And I'll say this is for personal use and I'll make my username Dalton.

Dalton Anderson (21:03.916)
Okay, and now I use the starter, it's free. And then I'll go over to...

Dalton Anderson (21:14.478)
Okay, here we go. And then I'll just paste this in here. Can you please?

Dalton Anderson (21:22.53)
me fully create this game this pong html game

Dalton Anderson (21:36.13)
And then a cool thing about Riplet is when you paste in the code here, the code doesn't fill up your text box, which sometimes makes it hard to read what you actually wrote. It understands that you're pasting code and you can click into it and you can't edit it, but it gives you the ability to separate the two.

Okay, so we are at a minute and 42 seconds and it's generating its thinking. It's making the app.

Dalton Anderson (22:12.014)
but I was impressed with how simple it was.

Dalton Anderson (22:19.36)
Okay, and I thought this was pretty nice too, where it asks you if you wanted to do like little add-ons to the feature, which is nice, where you could click, just click the piece and it'll add little things. I'll do, sound effects might be hard. What about...

Implement, let's try implement AI opponent for single player mode. And I'll leave it at that. Don't wanna make it too complex, nor do wanna take too long.

Dalton Anderson (22:59.662)
So now it's initiating the app.

Dalton Anderson (23:05.107)
and it's turning it into...

Dalton Anderson (23:19.598)
dependencies were installed. And so it understands the dependencies that are required. Took a screenshot. sure why. Took a screenshot. All right, here we go. The app is going. Okay, waddle up and down. up, waddle pad right. Okay. Here we go, here we go, here we go. Oh, he got me.

Dalton Anderson (23:48.462)
All right, all right. So I don't think that the AI was implemented because it seems to not work. But I did that in three minutes and 30 seconds, adding on one, me talking and rambling. And then this guy's cooking me. Hold on. I to score at least once before I leave.

Dalton Anderson (24:23.946)
Alright cool, I got a point. I'll just stop the game. Alright, so I'll say, don't say, does not seem like the other player.

is playing.

Dalton Anderson (24:49.516)
It's...

They are AFK, which means they're away from keyboard, if you don't know. So let's see what they say. And if they don't get it on this try, it is what it is. But they did do it within three minutes.

Dalton Anderson (25:12.63)
Okay. So it says, I understand the right paddle controlled by the arrow key seems inactive. Let me add the AI component to control the panel paddle so you can play against the computer. Perfect. That's exactly what I wanted. And one thing that we'll say is that I originally requested that in the add-on, right? Play against AI components. they're a single player. now we're, now we're, now we're cooking. We're cooking. I'm planning it's AI and I'm getting cooked.

but it seems to just go back and forth. my.

This, once you hit the ball back and forth, we implemented a speed multiplier and it's going pretty quick.

It's 75.

Dalton Anderson (26:04.719)
my goodness. my goodness. can't even see the ball anymore.

Dalton Anderson (26:11.138)
That's a little fast. my.

Dalton Anderson (26:18.894)
That's crazy. Okay, I can't even, it's just progressively going up.

Pretty cool though. I mean, to do that that easily is incredible where the next thing that you want to do, say you want to deploy, like this is legitimate app that you're like, I'm onto something here, let's deploy it. You can just click this button. You're pointing the click. So you understand all the dependencies, all the stuff, and then boom. And the idea is that AI could help create the app and then you could launch it within.

for a simple app like this, like three minutes, but for other things it would take longer. But I thought that was really cool. And so the next thing that GROK can do is it has a couple features. One called the deep search, the think, and you can see it's got research, brainstorm, analyze data. I'm gonna do the deep search and the think. The think is basically

a carbon copy of DeepSeq's thought process when it runs through. Even the outputs are very similar. So let's test that out and I'll do...

do this, so can you tell me, I'll put it on think, and then I'll copy this in here. So my prompt is tell me about blockchain and smart contracts on Avalanche, because that's something I'm interested in at the moment. And so it goes through its thinking process, it thought pretty quickly. Let me try that again. I don't think everyone's gonna be able see that. Let's try. Okay, so it shows the generation of its thoughts.

Dalton Anderson (28:07.308)
walks through and then it publishes it.

Dalton Anderson (28:13.39)
which is pretty cool.

Dalton Anderson (28:17.421)
So let's try the same thing with Gemini and actually the difference.

Dalton Anderson (28:33.678)
And this is a big difference between the outputs of XAI Grok and Geminai's Flash is that it just doesn't seem to be the same vibe. This looks very AI, I would say, versus.

when I have this question, I mean it does look AI-ish, but it seems to be a conscious thought where it runs through everything and that everything is nicely sectioned, whereas with Gemini, and I love Gemini by the way, I use it all the time.

it is just a rough output. Like this output is pretty rough. And then you go through and this is the legitimate output, but the thinking process, it's pretty rough. And compared to that to how this looks, just, you're just going about looks, like not even content. The content for XAIs, rock and Geminis, both good. think rock is better at the moment. But when it was generating,

it felt

It felt more like you were seeing the other side versus just things being plopped in.

Dalton Anderson (29:59.598)
So that's thinking. I like the thinking feature. I think it's great and love the output. The output I had earlier had line breaks, which is exactly the same as DeepSeq. So I that was interesting. We have another use case. So let's try this again.

kind of weird. I'm looking at the main camera that I'm sharing. I'm sharing at my main screen, but then my actual screens over there. I'm controlling my screen by looking at where I'm sharing my screen. It's also, it's quite odd, but we get used to these things.

Dalton Anderson (30:40.526)
Okay, so.

And we're going to add on deep search. And I'm going to kick these off at the same time because they take a little bit.

Dalton Anderson (30:53.385)
Let's do some Okay. So let's look at how this is generated. So this is generating, it's looking through the websites, it's thinking, it's exploring. Look at the UI, how it looks. And then let's scroll over here to how this is thinking. This is a little different where Jim and I will put together a research plan and you can edit the research plan, but it goes over the things it wants to talk about. Then you can edit the plan, which I like a lot because...

Sometimes when you provide a prompt and you're expecting a good result and it just, it misses the ball completely, editing the research plan or approving the research plan before it kicks off, prevents those things. But I also like how Grok has this thinking, the deep search, shows us going through and it shows us thought process and then it also shows the time elapsed. So it says 54 seconds of thinking.

Dalton Anderson (31:49.654)
I like how Gemini's advanced deep research tool with 1.5 shows all the websites, but once it's finished and creates a report, it creates a report with all its references and it automatically can go into your Google docs on your Google drive associated with your Gemini account, which is very convenient. Like if I was researching something that I just wanted to have in my personal drive.

I just say open up in docs and it creates a doc. But I would also think that a better implementation of that is I could specify where I wanted the doc to be created or if I could have an AI generated doc area where it puts all my Google docs that are AI generated in a certain folder so I don't have a clutter drive anymore.

Okay.

So this is still going, taking a bit, it's only quarter way.

And this has been completed.

Dalton Anderson (32:57.718)
And so the prompt was, I didn't read it out, but the prompt was, you help me research and think through startup ideas for insurance industry? if you had a special, if you had, it's a typo, but if you had to specialize, pick commercial area, and which do I think has the most opportunity to be most optimized? I mean, this whole thing is just like a garbage mess of whatever I typed up. I'm not sure how we got there, but we, I understood, so it's okay.

So it breaks it down like, okay, here's my core points. It breaks down the state of commercial insurance and other opportunities within commercial insurance. So it talks about cyber usage basis, insurance, sustainability focused products. And it talks about the market contacts and how it's changing over time, the overall market emerging trends, which the stuff I do, I agree with like intro tech innovations using our

AI and machine learning and big data analytics for risk assessment, policy underwriting and claims processing. Very true. Everything of that is as great. Cyber demand. Yes, cyber is rapidly increasing in growth, but there's just a complete lack of understanding in cyber. And then the next thing is the direct to consumer, which is coming along. And then the sustainability and green insurance.

which I think is more of a edge case, I don't know. I mean, it's something that people might support, but I've never heard insurance people like, aw, know, we're gonna miss, companies are concerned they're gonna miss their green initiatives and they're gonna have to get insurance for it. So.

Dalton Anderson (34:49.238)
It talks about pain points and challenges, inaccurate risk assessment, which is true. Like there's just not enough up-to-date data. And you'll hear that in the most recent guest episode with Valkyrie Holmes with 4AI. That's what her company does. She provides updated risk analysis data that provides a survivability score to understand

the most recent changes either to the environment or to that risk profile and how that affects the survivability of that perspective risk if it enters your portfolio, given the perils that it's exposed to.

Poor customer service, yeah. I would say that insurance could improve for sure, but insurance also is a highly emotional line of business. It's kind of set it and forget it, and then when you need it, you really need it, and people are freaking out, they are in a state of shock for the most part. And the calls that you get are brutal, definitely brutal.

Their soul has just been scraped out by a spoon and there's nothing left of them. They're a shell of the person that they were previously, at least at that moment. They're in a dark place when they're calling, typically, especially in catastrophic coverage. Regulatory and compliance, yes, but there's some things that get around that, like not being admitted, but...

even with surplus lines, you'll have to deal with regulatory compliance. the ideas that they came up with were some things I was talking about too, and things I was talking about with AI powered claim systems that would work, AI to process claims. So claims are unstructured PDF documents typically, or notes or call, transcript calls. Like how do you put together a claim summary from everything all put together, from the adjuster's notes, from the transcripts or notes from the

Dalton Anderson (37:00.174)
person talking to the customer service rep, to the insured, how do you get that all put together and packaged up into one document? Hey, I can do that. Right now you just have people copying and pasting stuff from places to places and putting it all together. Like there isn't a clean way to do that. There's claim systems that cost like hundreds of thousands of dollars a year.

that you have to make sure you put all the information in there and then it'll batch it up for you. But it's still, you still have to copy and paste stuff in the right areas. It doesn't process everything with AI. And so I've seen that, I've seen that come along and I've seen some things there. There is the ability to have a cyber program, like a cyber insurance platform, which I think so, usage-based insurance.

and a couple other ones. But the answers that I provided here are a little different than the answers that I provided last time. I didn't see anything about parametrics. I talked about parametrics, smart contracts, and blockchain, which are things that I'm interested in at the moment. And I want to learn more about blockchain and smart contracts and how they work and understand how to develop them and put together a smart contract and how it would theoretically execute and how would that work in insurance.

I would love to get to a point where you have an agreed about amount that gets paid out and that amount is specified before you enter the contract. And once you have the policy, is a legally binding contract, then if a certain event happens, then the monies are paid out to the insurer client. There is some nuances to it like,

if it's considered insurance insurance, then the insured can't be better off than what they originally were. then also like you can't like, like if the, if their claim is the link for a thousand dollars, like I can't pay them 50 grand. Like that's not how it works. So there's some things to think through. Like how would you do that? But maybe you just have the max, the max amount, but I would love to get to a

Dalton Anderson (39:24.472)
place where if you had a claim and there was definitive proof that a claim happened, like your house was in the path of a hurricane and you had a hurricane claim. That makes sense. The data, you can't make a hurricane. So if you had this Oracle data, holding the data together, putting the data together, and you had an agreed upon data source, say, okay, we're gonna use the

national weather service or whatever group for flood or hail or whatever the data set agreed or some kind of government entity, especially in the US, that would track and manage the data. And if this data event is reported in your area, then you are good to go. And then the money's paid out. That's what I'd like. I think that'd be really sweet because

Then people have issues sometimes getting paid or there's a couple of weeks before they actually get the money. And a lot of times people are in need when they have insurance. So.

I don't know, I think people would be pretty open to getting paid faster than shorter. If you had a choice of, do want your job to pay you once every three months, or do you want to be paid every two weeks? People are like, every two weeks, of course. Same gist. All right. So I'm gonna stop sharing my screen if I can figure it out.

Stop sharing, there we go. So I went over Grok, Grok 3, the thinking functionality, the deep search, which I think are phenomenal. I used Grok for a little bit. I mean, they just released it not too long ago, and it's in beta still. And I would put it as one of my favorites, simply because, I mean, it calls you out too. Like I said something incorrectly to the AI, and they're like, I think you're mistaken, buddy. I was like.

Dalton Anderson (41:29.986)
Hey, yo, yeah, it was like, I think you're mistaken. What you're saying isn't correct. You mean this? And I'm like, yeah. And I was asking about the Capital One and Discovery acquisition because it's a stock that I'm interested in. Once again, this isn't financial advice, by the way, so don't get your ears perked up. Like, this is financial advice. This is just me talking to the world here.

And I'm interested in the Capital One discovery acquisition for the discovery payment system. that acquisition, if it goes through Capital One will be the third.

player and the payment system place if they put their domestic credit cards and or their debit cards, all their purchase volume or transaction volume onto Discovery Payment Networks, they would become third behind Visa and MasterCard and then Amex would be last. So then there would be four key players in this space.

and it would save Capital One billions of dollars. Like a crazy amount of money with all the transactions that they're processing each year. And so it's a pretty cool idea. And I think that it, one, is great for Capital One, I'm a Capital One user. Two, I think it's better to have more competition than less. But regardless of the fact,

I was asking Grok about it and...

Dalton Anderson (43:07.65)
The AI corrected me because I said discovery banking or something. And then they and then he's like, it's got to what you really mean is discovery financial services. And I was like, sure, yeah. He's like, no, you're wrong. I was like, OK, are we picking out straws? Like, what are we complaining about straws here? Like.

financial services, discovery financial services versus discovery bank. mean, people know what I'm talking about. So, well, whatever. I'm glad that I'm glad someone calls me out and kicks me, kicks me, puts me down and keeps my head down. It keeps me humble, always working for better outcomes here. Yeah. And I think that the future of development with these AI tools and the level and the speed that they're scaling at.

One to two years seems like the sky's the limit. Like in two years, the benchmarks won't matter. The idea of AI in the workplace or in your daily life, I don't think is gonna be some foreign thing.

I think you either accept it or you die. I really believe that if you don't use these tools, the tools will be forced for you to be using them or the tools will replace you.

Dalton Anderson (44:49.846)
It might take longer for certain industries, but wow. The stuff that I've done at AI at my work has been incredible and I've saved the company and I can't talk about it because it's private information, but.

NDAs and all, incredible amount of time.

multiple full-time workers a year, like many full-time workers. People have to reposition their jobs because the processes that they were doing is completely automated. They no longer do those things anymore. And so...

That's what AI from eight months ago, 10 months ago, the AI of today can do more. And then the AI from 10 months from now is gonna be doing even more. And then when they come out with these power plants and they're providing all this power to these GPUs and training, then we'll reach a level that we don't even understand ourselves yet. And I think that's the point is that

It's progressing at such a fast rate that even the experts don't know what's going on. And the people who've been doing this for 20 years when they're like, we'll have AGI or we'll change the productivity of the world in 20 years or something, like every eight months, they're like, I don't know. Like I said 23 years ago, but now I'm like, it might be five, it might be three, who knows?

Dalton Anderson (46:27.886)
But I hope you enjoyed this episode and that you thought it was thought-provoking. if you had built an app with Ripley before, let me know how it went or cursor. And if you have any other thoughts about things I said or if you like Rock 3, let me know. And I will be pleased to reply and.

Dalton Anderson (46:53.858)
research or understand your thoughts because your perspective is important. Anyways, wherever you are in this world, good morning, good evening.

Good morning, good evening, good afternoon. Wherever you are, have a great day and I hope you listen in next week. See you then.