Property AI Report
00:22:59 · Matt Goddard
[00:22:59] S1: It's positive first and foremost, isn't it? That as you say, that it's not just the the the American labs that are kicking on and to to see Mistral still going. I think sometimes there is a question when we're talking about all these new models, all these things, and like people saying, Well, when do I use this and how do I use it and and such like it's a case of you know sometimes I just focus in and only use codecs and and chat GT because there's a lot of context switching. But um I do like the fact that A, you did a French accent when you said Mistral and then LeChonk had to go away and just fact I had to fact check, so it was generally called LeConk. Um that in itself might mean that I I want to use it. But I think that the main reason that seeing it is just the competition, isn't it? And the fact that we're not getting hemmed in and that if suddenly the American government says, right, we're gonna restrict everything. Um because those cybersecurity breaches we talked about last week, you know, people were responding to those quickly if they had access to Mythos or Astra 6.1, but us, me and mortals who didn't, people had to go to GLM 5.3 and others where they could use these open weight models to to actually respond to the attackers. So um, you know, where those cybersecurity barriers are in place, great, but we need something that helps us, and the fact that this is open open weight's effects of the open source, right? So we can go in and utilize this. So it's it's good to see more and just keep pushing that, um, pushing that competition is a is a good thing in my mind.
00:24:21 · Mal McCallion
[00:24:21] S0: 100%. Yep, and keeping on the uh kind of hacking slash challenging security issues. Um one of the big stories last week is Binance has a s had a 6.3 billion crypto theft um because of some holes in its um its overall structure. Um again, you know, just kind of highlights the different world we're in, right? Suddenly, where essentially um yeah, everything is check and double check because there are things happening that uh that the AI is discovering. In this particular instance, I think it's been linked to some um elite-level North Korean hackers. I think this is uh this story.
00:24:55 · Matt Goddard
[00:24:55] S1: I think probably worth um for those that don't know, Binance is the world's largest crypto exchange, and it's like 6.3 billion in cryptocurrency was stolen from a DevOps engineer's compromised laptop. So basically a user's laptop was compromised and they lost 6.3 billion, which is a pretty bad day at the office. But um, I sort of joked off air, maybe tongue in cheek, that we should start having a hack of the week, given that you know, when we were talking last week about Black October saying it's it's gonna be pretty much continuous now, and at least one really bad thing happening every week. And I seriously do think that's gonna be the case. You know, there's so much going on, and the the good news is the good guys have got as powerful AI as the bad guys at the moment. Uh, this is quite a severe one. It is, yeah, it is an arms race. And I think what we'll start to see more and more is tokens being stealed, and I know you stealed, stolen, and I know you um you talked about that last week, you're like token mining, and um in this one they've just gone straight for the jugular, right? And got the crypto straight off.
00:25:51 · Mal McCallion
[00:25:51] S0: It is, I mean it was somewhere else to describe it as a heist, which I think is great, uh, just in terms of because then it conjures up, you know, the Italian job or you know, all the time. Yeah, the Genie world models create it. Which is eleven, but it but it wouldn't make a very good film because it'd be somebody sitting in front of a laptop. You're trying to hack somebody else who's sitting in front of a laptop. Don't know what kind of films you watch. I was saying Ocean's 11. Anyway, um, all right, that is our um tech news, AI news for this week. Um, let us move on to our tool of the week. All right, this week um we were looking at a granola, which um more than just a very healthy breakfast, is also something that records your um I said records your meetings, and it's now kind of broadened itself out into um something a little bit more. What is it, man?
00:26:35 · Matt Goddard
[00:26:35] S1: So, so yeah, as you say, granola, widely known for its meeting being a meeting note taker that sits there quietly in the background. And I'd I'd say it's probably one of the best or most popular ones. I I appreciate you said it had gone maybe a little bit quiet when we were chatting off air, but um, I've certainly seen it pop up more so maybe than Otter and some of those other ones that that tend to email you and spam you after meetings, say someone in the meeting has has used it. But um, I think this is quite a clever play, and it comes shortly after Wispr, W I S P R, which is known for its dictation tool, launched a meeting note taker. So whether Granola's like, right, you lot, we're gonna come in and do yours. Um I do I generally think these sort of dictation tools, I use Wispr Flow, um, are really powerful because we've we've spoken about literally rather than trying to do prompt engineering or context engineering, all these terms that we we were talking about maybe a year ago or two years ago, where you're sort of trying to think about exactly what you want to be pushing into the large language models to start your your engagement, just jabber at it and get something like granola or whisper flows or even you know codex direct now is brilliant. Just chat to them, don't try trim what you're saying because a lot of things that you edit out are actually really useful context when you're trying to work through. Yes. I think granola coming in is quite a cheeky thing as well because sometimes these people say, Oh, this is free, as as granola's uh seems to be doing at the moment. But when it says it's free and it's unlimited words, it's not free, is it? When when something's free, you become the product. And um we think that's the case here as well, right? Exactly.
00:28:04 · Mal McCallion, Matt Goddard
[00:28:04] S0: Every word, every cough, every kind of somebody talking to you while you're on it, um becomes part of their training database, and that that helps them to make the product better, which you know you might say is a fair exchange. You know, having that sort of uh two-way thing for what you get back might be might be good, but yeah, I I 100% agree with you. I remember going out and then and doing the tour and talking to agents. We speak three times faster than we type. So in terms of data exchange, it is the most effective interface, as we as humans have known forever, because that's why we develop talking. So yeah, you know, I I now on my runs in the mornings, I will I will actually you know be talking into, in this case, Chat GPT, um, about what's happened during the day, blah blah, what have I forgotten about, what do I need to do, blah, and it's just so much more effective when you kind of get in and then you kind of look at it and it's got a to-do list and stuff, and it's just makes you, yeah, it's it's it's so much more effective in terms of getting your getting your day started. So, yeah, no, I'm hugely excited to see more and more of these vocal tools because I think it's gonna be we're gonna see everybody just chatting and walking down the street chatting, and you'd be like, Yeah, this is are they mad or are they just talking to as long as they're not doing it, doing it out loud and with microphone. [00:29:10] S1: And isn't Andy Burnham trying to ban that, which is you know, go burn on the bottom.