Speaker 0: apparently, anthropic is worth a lot more these days. Speaker 1: yeah. so this is interesting. and i have to say, i still love anthropic, but i think they're falling behind. they haven't released a new model in a long time, but this is great. so they are on the verge of closing or they might have, i don't know when this article was published, a massive two billion dollars round led by lightspeed venture partners. and it's set to value the company at sixty billion dollars so that's more than tripling its valuation from just a year ago. and the funding round is part of a larger six billion dollars funding effort following a four billion dollars investment from amazon in late twenty twenty four. so this capital injection, it kind of positions anthropic as more of a major player, but it does from a valuation perspective, but, i think they need to get their crap together like they had some great things and it's crazy that they had some great things mid twenty twenty four late twenty twenty four. and we're saying that they need new things now just because of how quickly, things have changed. but they're becoming the fifth most valuable us startup. so after spacex, openai, stripe and databricks, which is crazy, but they do need to release some new llms. what do you think about this, craig? Speaker 0: sixty billion is the last eval we saw for xai. yeah. was gonna say grok, but it doesn't recall that. Speaker 1: the oki, xai raised six billion. i don't know what the valuation was, sixty billion Speaker 0: when we were talking about it. Speaker 1: okay. Speaker 0: so, it's neck and neck here. anthropics has always had some interesting hybrid technology that everyone else is getting into now. so it feels to like the investors in this article that's part of that they're positioned to be a leader in this area. Speaker 1: yeah. Speaker 0: have you seen that open ai? sorry, anthropic is about to release a new model. i know you're talking about that, but they're about to release this new model. Speaker 1: yeah. they released that coding window or that additional window. that was really cool. that was recently, but it's crazy how long ago that was compared to everything else that's come out, especially with openai and everything they released around christmas. so, i love the race. i love this is great. so thing, one thing to note is that ai startups have accounted for nearly half of all venture capital investments in the us last year. so that's crazy. it just shows you that there is this, i don't know if it's a bubble for say, but we're getting some high valuations. just for those that don't know, antarctica's valuation was eighteen billion and it's going to skyrocket to sixty billion in just a year, which is crazy and great. i'm happy for them. they're my favorite platform to use for certain things, but it's crazy how half of venture capital investments have been in ai startups, which is not surprising, right? Speaker 0: yep. okay, this article corrects me. this article says that elon's sk ai is valued at fifty billion following their okay. Speaker 1: so i thought i was going to beat them. Speaker 0: and openai is at one hundred and fifty seven and perplexity is valued at nine billion. Speaker 1: yeah. and speaking of perplexity, they just released a free deep research, branch of their of their, llm. Speaker 0: yeah. and i think deepseek just released mo, which is a new tiny model, which is supposed to be really cool. i want to check that out. and the new anthropic model that's coming out is supposed to be an o3 minutei encoding, and it has a quick answers and a deep think mode and a slider that's gonna allow you to adjust between those. Speaker 1: cool. you mean open the Speaker 0: eye? no, anthropic. Speaker 1: oh, an o three? Speaker 0: i said it's better than three. this is what i Speaker 1: was saying. sorry, missed that. Speaker 0: anthropic is preparing to release its next major model. this is reported by techcrunch. the model tries a hybrid, can switch between deep thinking and quick answers. the company will reportedly introduce a dynamic scale that allows the user to flexibly customize the model for themselves. this new model will be available in the next few weeks, is expected to outperform openai's three minutei model in code writing. i just said encoding. in addition, the model shows excellent results in the analysis of large code bases and other business tests. Speaker 1: yeah. so as of right now with three point five sonnet, i can pick concise, explanatory, formal or normal as far as the responses. so i don't know if that's the same thing, but yeah, the one thing i find about claude is that i find it more consistent with its responses on a day to day basis compared to other models. Speaker 0: that might be the case. i don't use cloth very much, you know, you're always bugging me, but i'm trying right now. i'm trying gemini. i'm trying deepseek. o three mini and o three mini high has been very good. i was been very unhappy with four o, but o three mini and very unhappy with four o. o three minutei and o three minutei high, i've been very good. so i've been pretty happy with that. yeah. and i'm testing these other models. there's only so many i can work with. so i am pretty happy to try all these out. and it looks like groc just dropped grok three. i wanna be trying that. and it is great for image generation, but i haven't used it for much of interaction or work like i've done with the others. Speaker 1: yeah. yeah, no, i don't know why i've always found claude to be much better as far as how i use it. and i've tested all these other models when they get released and they still don't compete compared to what the output that i get. so maybe it's just my prompt is specific. it works great for claude compared to others, but i don't know. but, yeah, that i need to try gemini too. that's the one i wanna try next. Speaker 0: and the other thing, you know, this is a thing i'm starting to do because these models all output different types of stuff is they all right. i used to just, you know, have my prompts and change them up a bit, but the best way to do it now is give it examples. so once you have a good thing, this is a good format. i like this. output should look like this. yeah. and then i find that you'll get better results from all the models. yeah. so it's just the way it goes. you know, when i asked one thing to create an index of one of something, it does it one way and one thing does another way and they don't, they're not really compatible. so you got to really say, hey, this is how to do it. yeah. so you being that kind of stuff, using examples, fine tuning it a bit. examples are great. that was what i was going to say is, okay, if you're unhappy with the other models, you know what you do. do what gemini does takes the output from claude and feeds it into gemini and say, no, this is what i expect from you. and it says, oh, okay. and it performs a lot better. that i mean, that's one of the things that fifty dollars model did is it took that reasoning output. it took very specific input, took the output and the reasoning and fed it back into their cheap model and saying this is what i expect. then it started getting some amazing results. and it did that for fifty bucks with a thousand. i think it was a thousand queries. Speaker 1: oh, crazy. Speaker 0: so you can make these things do all sorts of stuff. so i'm excited about this race because i don't know any of these companies are making money. we always say that only the hardware companies are making money in the ai space. even sam altman with their two hundred dollars pro feature, two hundred a month say they're losing money on that. so yeah, know, these, the computational power of these things is so much. and that's why, you know, we see nuclear, we see all these other energy stuff happening. and exciting to see the world go to clean energy faster because of the demands. Speaker 1: yeah. and these companies or the larger companies are trying to design their own ai chips too, own gpu's, right? Speaker 0: yeah, and there's a huge push to start designing them and building them in the us. Speaker 1: of course there is. Speaker 0: intel is floundering, and there's just somebody talking about buying them and using them and doing different things. because they're they're breaking them up from a foundry to the ip part. yeah. and we'll see how that goes. but it's it's a shame this one's great chip making company has been reduced to what it is. Speaker 1: yeah, it's crazy. Speaker 0: so we'd love to hear your thoughts on are we in the middle of some huge ai bubble? is this another over evaluation or john said over, well, you said, i think you said over fifty percent were ai companies. Speaker 1: over fifty percent were ai startups as far as that got vc funding last year. Speaker 0: and it makes a lot of sense because again, if you're not using ai and automation, you're probably going to be crushed by your competition who is. so when you're investing, you want to see companies who are embracing that technology, never mind leading the way. yeah. so, let us know what you think. is this, again, are we in the bubble now? what part of this hype curve are we in for ai? is it, are we just around the corner from the drop or we still got a long way to go? and what do you feel about this race between anthropic perplexity, open ai, xai? is there going to be a winner in this race? and who do you think it is? let us know.