Kimi K3: AI for Business Growth

**Jayden** (0:00)
If you want to get access to this episode and my next 30 episodes, all ad free, so there'll be no ads on them, go check out my podcast AI Chat. You can go search for that on Spotify or Apple. It's AI Chat. I'm gonna post all of these news episodes, and I'm also posting interviews, like I just interviewed the CEO of Cohere. They've raised over a billion dollars for their AI model, talking about what they're gonna be spending the money on and the direction of the AI industry, along with all of this new stuff. So if you wanna go check it out with no ads for free, it is AI Chat. Today on the podcast, the very hyped moonshot Kimi K3 model is forcing OpenAI Google Antithropic to rethink closed AI. There's so much drama in Silicon Valley from this. But also, there is a ton of money that you can save in your own business by using some of these open source models instead of cloud and OpenAI to run really repetitive, heavy tasks. So we're gonna talk about how to get some of the cost savings with these AI models, what they're useful for, what some of the alarm bells are, what people are concerned about, all the drama. There's so much to get into. Before we do, Jamie, would you want to tell them about this school community?

**Jamie** (1:04)
Yeah, if you've ever wanted to learn how to actually make money online using AI or grow your business, you're going to want to check out our school community called AI Hustle. We have over 600 members, and each week we release bonus content over there. We show you exact strategies and ways we're making money. This week, Jayden talked about his...
Well, actually, we talked about our podcast and how we're making money there. But he talked about his exact Facebook ad breakdown. He spent over $2,000 on Facebook ads promoting his AI tool called AI Box. And he went through the results and how he optimized it and whether or not it was profitable and made sense for him to keep doing. So broke all that down. And so it's really just a great place to learn that kind of thing. We go over marketing, vibe coding, how to become an AI consultant, and then also how to make money with AI content and music. So if you're interested in any of that, we'd love to have you be a member. It's free to join. You can get our weekly live streams and also some bonus episodes. But if you want access to the full vault, it's $19 a month.
Well, let's talk about Kimi K3. This is another pretty funny named software, if you ask me, AI company.
DeepSeek, I remember, was really mysterious sounding, and this one even to me sounds more funny. But the big thing with a lot of these open source AI things is security. People are concerned about the security of their data and also national security, all those kind of things. Jaden, is that something to really be concerned with and make your decisions based on that? Or, you know, we're going to talk about, yeah, what are the reasons you would want to use Kimi K3 over Claude or OpenAI?

**Jayden** (2:59)
I mean, the biggest reason just comes down to economics. And even when DeepSeek came out, it was the exact same thing. It's so much cheaper. So for so many tasks, you just get way more bang for your buck.
Kimi K3 in particular is interesting because in a bunch of different benchmarks, it was actually better than Opus, I think 4.8 on some tasks, worse than others. But it was really close up there. And Moonshot actually used something that Anthropic calls Distillation Attack, where basically they go and ask Claude a million questions and they get the responses back and they take the question responses from Claude. And they train a model to just be like, answer like Claude answers basically. And sometimes I've seen with these models where you ask them, what is your name? They're like, I'm Claude Opus 4.8. So like, it just like straight up. Anyways, Anthropic doesn't like this because they're spending billions of dollars to train their models. And if you go run a distillation attack, you get very similar sounding results and responses. You do have to have underlying like data sets and like information and stuff, but you can kind of scoop up the whole internet and then you can fine tune or you can distill Anthropic for how to answer things. And then you have to be clever on how you do like logic and reasoning. There's a bunch of the pieces that Anthropic does not like being stolen. So they're trying to get it really legislated. They're trying to make it illegal to do.