Anthropic’s Claude Code leak

19:45moving on to our story number two. Uh Anthropic accidentally leaked part of Claude codes' internal source code. Okay. Uh so, this for those who may not know, Anthropic is one of the sort of primary AI model creators along with Google's DeepMind, OpenAI, Facebook's Meta, whatever they're doing over there. There's a European player that no one ever talks about. But, you know, these foundation models, the brains behind the chat-based AI generative AI revolution we've seen have been crown jewels and is basically what the entire US economy in terms of GDP growth in the last year or two is effectively based

20:25on. So, whether you like it or not, our financial futures are kind of pegged to this stuff being successful in some way. Um but, Anthropic just had a pretty embarrassing AI security issue, uh which has been a little bit overhyped in the headlines. >> Okay. But, it's still somewhat significant. Um so, Claude code, so if we think about the structure we have Claude is the model, the actual brains that responds to you when you type into, you know, claude.com and and get it back right? >> is an LLM? >> That's LLM, right? That's the where where all the model weights are. That's where all they do all this really expensive training to make these things super smart. They're basically growing, right? The the the the

21:06intellectual response that we get when they train these models. But Claude code is simply a interface, an orchestrator for an architecture for you, specifically initially for coding, software coding, but is now evolved into what they call co-work as well for people who do things like financial modeling, deck creation, other just business-related functions. They all live within this thing called the Claude code application. You can either have it as a terminal or on the Claude desktop app. Okay. Now, initially this leak was reported on March 31st. Um, and