AI Agents: Still Not There...

0:00With the American Express Platinum Card, you can access over$3 ,500 in annual value with benefits and eligible purchases across travel, entertainment, and more. There's nothing like Platinum. Learn more at AmericanExpress.com slash Explore-Platinum. Enrollment requirements, monthly, and other limits in terms apply. It is Ryan Seacrest here. There was a recent social media trend which consisted of flying on a plane with no music, no movies, no entertainment. But a better trend would be going to ChumbaCasino.com. It's like having a mini social casino in your pocket. Chumba Casino has over 100 online casino-style games, all absolutely free.

0:33It's the most fun you can have online and on a plane. So grab your free welcome bonus now at ChumbaCasino.com. Sponsored by Chumba Casino. No purchase necessary. VGW group void. We're prohibited by law. 21 plus terms and conditions apply. We know that right now one of the hottest areas in AI is AI agents that can go and complete tasks for you. Just this week, Jamie and I, in our AI Hustle school community, which if you're not part of that, we'd love to have you as a member basically make a video every single week that we share on there. We don't post it anywhere else, breaking down how to use a different AI tool.

1:05And this week in that we're breaking down ChatGPT's new agent tool, some interesting things that Jamie found, and he was kind of showing some tips and tricks for how to get the most out of that. And so if you want to go check it out, there's a link in the description to the AI Hustle school community where you can kind of see that. But this is one of the hottest areas in AI right now is kind of these agents that go and complete tasks for you. Sometimes they have their own virtual private machines. But this is not an easy technology. Just based off of the podcast I've made in the past about ChatGPT Operator, I'm sure you know this isn't a perfect technology.

1:37This week, we wanted to cover a really interesting story where basically there's a company that went to Y Combinator, right? One of the biggest AI startups. And it's got a whole bunch of like really high profile people talking about kind of what they're working on, but they've actually given up on this completely. So originally, this was a startup that was making AI agents for Windows, and they have given up completely on this mission and pivoted. So on the podcast, we want to talk a little bit about what they were doing, why they're pivoting, and why this is such a tricky area in AI. But Jamie, just, I guess, give us a little bit of like the overview of pretty much what happened with this story.

2:12Yeah, so the startup called Pig.dev had the, they kind of, they participated in the Y Combinator winter 2025 batch, and they had the goal of, you know, creating an AI agent for the Windows desktop. So, which is a great idea. But as, as Jaden had mentioned, there's, there's some hangups, at least so far, that kind of make it not super useful yet. The first being, or the major thing is that it's kind of slow. So I believe kind of how they work right now is your, you know, The agent will have basically a series of screenshots of what it's looking at and kind of click around and explore and figure out what's going on.

2:56And it kind of just takes a while to do. And it takes a lot of energy and compute as well. So that was kind of, I believe, what their hang up was. So they kind of pivoted to something different. They're calling it muscle mem. And so this is like a cache system for AI agents. So basically, once you can train your agent to do exactly what you want, that information will be stored in the cache. So then it can repeat that action much quicker. So, you know, for example, if you were trying to make, you know, a weekly newsletter for your company and you had a basic template all set up and you just need to switch out, you know, the text and the news, it could do that much, much quicker and be a lot more efficient.

3:42so I think that's I think that's really a great problem they're trying to solve because as again as Jaden had said and some of my current frustrations with the agent is the the slow speed so yeah so if you go over to their to their website you'll see that right at the top it says this product has been paused the team is now we're working on muscle mem so you can go over and click on that. And that is basically their GitHub that you're going to see. So they brought up a really interesting problem that Jamie and I have both seen, like both encounter while we're running into agents. I want to talk about their problem.

4:20But first, I would just want to say I kind of actually love this concept that they they had a great idea for a company. They really worked at it. They weren't able to get really product market fit or a lot of people using their product. And they found basically another problem. It's like while developing this company, they found another problem and they focused on that problem. I think this is kind of cool, Because a lot of times when you're starting a company, you have assumptions about what you need to do or what people need. And you realize another problem that's even more pressing in the process.

4:50And sometimes tackling that is the best thing to do. So they've done that. I have no shade to them on making a pivot for their company. But it's really interesting. So basically what they figured out, these AI agents, they do repetitive tasks, right? That's pretty much why you want an AI agent. I wanted, I got ChatGPT operator to replace my virtual assistant and some of the tasks that they were doing with, for example, my podcasts, I was, you know, getting it to help me write descriptions and make, you know, content for like my podcasting to help make it faster. But like, basically, it does the same thing every single time.

5:24So what they built is a caching system where if you if you see an AI agent that does the same task over and over again, they say, basically, you're wasting compute. it's really slow because these AI agents are not fast right now and it reduces variability. So this is one of the big issues. And in fact, this is a really cool, completely different angle on something that I'm personally working on for the next launch of AI box, my own AI startup, which if you haven't tried it already, it's AI box.ai. You get access to the top 40 AI models. You try them all out for 20 bucks a month. So I'll leave a link in the description.

5:57You can try that out, but it's actually a really interesting problem that we're tackling where we're basically building these like AI mini apps so that agents will be able to come and use those mini apps to complete an automation in the future to basically solve the same problem that they're solving. The AI agent, they're slow. They don't know how to do things perfectly every time. There's too much variability. So you might say like, hey, go and do X, Y, and Z task. And it might do it differently, you know, three different times. And so like, that's really not something you want. You want it to be predictable.

6:25You want to be able to do the same thing. So it's interesting. They're tackling this from a different angle, but I do love this. And they also said this eliminates token cost, right? So if you basically built an automation where every time your AI agent gets to like a certain part in the process and it can just zip through with this little automated code that they've written, you don't have to worry about, you know, spending all the token costs every single time that it has to do that exact task. So that I do think is very cool. Totally. Yeah. I mean, I think, you know, I, so I, again, kind of what our AI hustle call about was tonight was about, I have access to ChatGPT's new agent tool.

7:05And so far, it's great. You know, you can give it a task to do. And then it will also do research within that task. And it's very nice, but it is really slow. Like, it's still working a half an hour later on a task, a relatively simple task I gave it. So, but if I can, you know, work the kinks out, it would be great to next time just have it do it that same task a lot faster. Yep. A hundred percent. One interesting thing that I'll say in kind of their mission to build this company and to pivot what they were saying about it was basically he said that users in that were used to using kind of legacy app things basically just wanted automations.

7:50They basically wanted to say, hey, I just need the AI. Like, sure, I'll use your agent, whatever. But like, I don't want your agent just for the sake of using an agent. I have a problem and I just want this problem solved. I just want this automation done. And so he said he felt like he was getting kind of like shoehorned into building custom automations for people. She didn't really want to make these one off automations for companies. And so they that's kind of where they made their pivot. And I will also say I give them big kudos because they were featured on some really big podcasts. I believe that there was like a podcast or one of the one of the founders, I think, of like Y Combinator was.

8:24Oh, yeah. One of the partners at Y Combinator was talking all about them on this really high profile podcast with Mossad, who is the the co-founder of or the CEO of Replit. So this really big vibe coding companies that, you know, there's like these really high profile people. They were talking about pig, you know, pig.dev and were like, look, this thing's like really cool. you know, once you apply that to enterprise, it's going to be huge. The guy was like, yeah, once, as soon as that technology works, this company is going to be really, really big. So like you can imagine if you get these really high profile people in tech talking about your company and what it's doing and talking about how big it's going to be once you achieve, once you like arrive there and you still make a pivot.

9:04I actually think that has a lot of courage because you're giving up basically a lot of like quote unquote brand equity or value from, from a lot of these public appearances or what people might be thinking about you. and basically because you're not chasing vanity metrics or shout outs or recognition, you're really just trying to build a product that people can actually use that actually works. So I do give them big props for doing that and I'm excited to follow along on what they are building out. Hey, if you enjoyed this episode, be sure to leave us a review wherever you're listening. We really appreciate those and read every single one.

9:37And also again, be sure to check out the AI Hustle School community if you want to learn how to grow your business and make money on the side using AI. Thanks for listening and we'll see you next time. Are you really buying a car online on AutoTrader right now? Really? I can get super specific with dealer listings and see cars based on my budget. You can really have it delivered. Or pick it up. Mommy, look! I think your kid is walking up the slide. Really? AutoTrader, buy your car online. Really? Tyler Reddick here from 2311 Racing. Victory Lane? Yeah, it's even better with Chumba by my side. Race to chumbacasino.com.

10:11Let's chumba. No purchase necessary. VTW Group. Void where prohibited by law. CT &C's 21 Plus. Sponsored by Chumba Casino. You're listening to a podcast right now. Driving, working out, walking the dog. If you're into podcasts, chances are you have something to say too. With rss.com, starting your own podcast is free and easy. Upload an episode and we distribute it to Apple Podcasts, Spotify, Amazon Music, and more. Track your listeners, see where they're from, and start earning from ads just like this. If you've been thinking about starting a podcast, this is your sign. Start your new podcast for free today at rss.com.