#138 Is Claude the future of agentic AI

"It didn't exist. They vibed it into existence, what, three months ago or something like that? Not even maybe two months ago." Dara

"Nothing has to change. That new model, a new model comes out and that's just instantly a better product." Matthew

Show full (AI-generated) transcript

[00:00:00] Lizzie: Hello and welcome to the Measure Pod by Measurelab, a podcast dedicated to the ever-changing world of data and analytics. With your hosts, Dara Fitzgerald and Matthew Husson. Between them, they've spent more years and they'd like to admit wrestling with dashboards, data quality, and the occasional Google Curve ball.

[00:00:32] Lizzie: So join us as we share stories about how analytics really works today and where it might be headed tomorrow. Let's get into it.

[00:00:41] Dara: Hello. Welcome back to the Measure Pod. I'm Dara, joined by Matthew. How are you today, Matthew? I'm good. Thank you. Yeah. Yeah. Getting there. How are you? Yeah, I'm not too bad. I'm not too bad. I thought you might point out my new background. What background? It's more of a, it's more of an absence of background.

[00:00:58] Dara: Yeah. I go from, I'm a kind of extreme guy, on one end or the other. So I went from the very visibly Spanish yellow decor to looking like I could be inside of. Time capsule. 

[00:01:13] Matthew: Yeah, it's just, it's a bit of a contrast now, isn't it? 'cause I've got the Cary mess that is my office behind me, and then we've just got the plain, white, sterile environment of you. 

[00:01:24] Dara: But you, you could probably guess this, but behind the camera, it's a very different situation. Basically, I set up the, I treat not only, not only have I moved away from the dining room, I actually treated myself to a proper desk, like an office desk. Not just a dining room table. Yeah, I've gone, I've gone crazy.

[00:01:46] Matthew: Huge steps. 

[00:01:47] Dara: I know. 

[00:01:48] Matthew: But unfortunately, it was sitting on a kitchen chair on a pile of books on your laptop, not cutting it anymore, or? 

[00:01:54] Dara: It's been 

[00:01:55] Matthew: How long, how long were you there for? 

[00:01:56] Dara: Seven, eight months, something like that. But it's remarkable what I'll get used to. I'm, I'm, I'm pretty, you know, low source kind of guy.

[00:02:04] Dara: I don't need these complicated setups. But I did decide it was probably time to have a proper dedicated space after eight months or whatever it's been. But what it did mean is everything that was in this room, I just piled up against the corner. So what you get to see is, yeah, the nice, overly clean, almost clinical as you said.

[00:02:25] Dara: but the Spanish sun is outside that window, so. I'm just not gonna share any of that with you or any of our listeners who also watch the YouTube videos. I don't wanna give too much away. Okay. Enough about you're very low maintenance, apart from your need for 15 AI subscriptions at any one time. Listen, apart from that, I'm a, I'm a contradiction.

[00:02:44] Dara: I'm we all, and look, that's a serious problem and I'm, and I'm, I'm sure I'm seeking help for it, so Please, yeah. We should have one of those disclaimers, you know, if you are also struggling with, AI addiction, if you're about it, please. The issues raised. Yes. No, I, I'm, I'm gradually reducing my number of subscriptions, although I say reducing them down, I tend to collect new ones as I, I reduce down the main ones.

[00:03:09] Dara: Like I moved from, nobody really needs to have three cloud accounts. So I finally, finally saw the error on my waist, but then I collect all these little ones where I see something. It could be anything like a, you know, like I did it with 11 labs and, you know, any, anything that I'm just playing around with, I'll tend to just go, oh, I'll just, I'll just sign up.

[00:03:26] Dara: Then I would think, why have I got no money? I, I, I remember say the amount of emails I get from companies that I've all but forgotten what they are or what they do. He's like, why are you emailing me? Oh, yeah, I paid you money six months ago. 

[00:03:38] Matthew: That's what you get to be, that's, that's the price you pay for being on the cutting edge.

[00:03:41] Dara: Yeah. When you respond, you know, when you unsubscribe and you give the reason of, I never signed up to this in the first place, and they reply back to you saying, why have you been paying us for six months? Then? They go, Hey, oops. Alright. Shall we, shall we move into the, the, that's enough. That's enough for rising, isn't it?

[00:03:59] Dara: That's enough jovial banter. Yeah. 

[00:04:01] Matthew: Whatever that means. 

[00:04:02] Dara: Whatever any of this means. Yeah. Let's get down to the serious, this is a serious podcast. Let's get serious. What do you wanna get serious with? I mean, we m we what? We normally do a bit of news, don't we? But it feels like not really anything's happening in the world at the moment.

[00:04:13] Dara: So should probably just skip the news. Yeah, it's, yeah. No, it's been pretty slow. An AI especially has been quite quiet. 

[00:04:20] Matthew: Yeah. Nothing's really happened since chat gPT-3 0.5 or whenever that was. No, there are some, quite a lot of things I'll, I'll, I'll be selective with what I say 'cause I think there's pages that we have an internal, channel on our Slack at Visual Lab that has, that we, everyone puts AI bits and stuff they found and it has been a particularly busy.

[00:04:44] Matthew: What would it be? Three weeks since we last recorded, our last podcast release? I can't remember now. Yeah, so let's, let me just reel off a few of those. so a few, a few big ones in Google and particularly in BigQuery. So they've released the Gemini agents to more to, to general release. I think they've already existed in preview, but they're never released to general release.

[00:05:08] Matthew: So this is what you can essentially create. Dedicated agents within BigQuery point it to specific data sets and tables. You can layer in a business glossary of terms and, and other contexts and things you want, and then you can, essentially, have a little conversation analytics agent that sits directly within your data warehouse that you can talk to, and, and do things with.

[00:05:35] Matthew: So we did our webinar on conversation analytics. Well, what was it, December a few months ago? 

[00:05:44] Dara: 1985 I think it was. Roughly, yeah. 

[00:05:47] Matthew: And a lot of what we were saying was like a lot of the complaints we had or the, the, the issues that we saw, I presume Google listened to the, to the webinar, but they've, that's a lot of what they've addressed.

[00:05:57] Matthew: It's that context and we were having to add in column names and things like that to get it to properly work. And now you've got all these inbuilt, the agents with that context sitting in there and golden queries. To be able to give examples of good SQL that it should be running, et cetera.

[00:06:13] Matthew: I like that. That's a good term, isn't it? Golden queries. Golden queries. yeah. That's cool. I've liked playing with it a little bit. Not as much as I wish I had, but if I played with everything, as much as I wish I had, I'd thought they'd do anything else. It'd just be playing, signing up to endless subscriptions with ai.

[00:06:30] Matthew: Yeah. We can't all do that, can we? No, we can't. Someone's gotta do something. 

[00:06:33] Dara: But, but, but also, you've gotta wonder as well.

[00:06:41] Matthew: Yeah, there is, there is def there's a, there's, that's a joke. But there was definitely a, 

[00:06:45] Dara: It wasn't, I was just sort of a naturally jokey persona. 

[00:06:49] Matthew: Yeah. But, then in two weeks things have moved on. Two weeks. This is impossible. 

[00:06:55] Dara: Yeah, it's impossible. It is. You just gotta get stuck in, don't you? And you've gotta see it. You've gotta see it as well as not being precious about what you've done. I think that's the, that's a, that's a psychological change that has to happen, isn't it? Get stuck in play with, play with it the way it is now, but be fully expecting for it to, it to completely change and, and something that took you, you know, weeks or days or hours or whatever could be just redundant tomorrow. And that's okay. 

[00:07:20] Matthew: Yeah, no, so definitely a seed change you're gonna get used to. Don't be precious about anything. We have released and until that end, we have released something like a writeup on the webinar that we did in December. So keep it sort of relevant. We've put in like a March, 2025 set of updates throughout it that talk about all the new releases that Google's put in since we did that, to try and keep it a little bit relevant to people who wanna have a read.

[00:07:45] Matthew: So go and have a look at that on the site. if you wanna read a bit more about our conversation on analytics and then go and have a look at the webinar, our favorite image generation nano B model nano B, there's now nano B two. So they've released a new model so it can do, it's got, it's got even more advanced subject consistency, which it was already pretty good at.

[00:08:12] Matthew: So, I mean, we, we talked, we have about half a podcast pretty much dedicated to that at one point, if I remember when, subsecond 4K generation, I know that some people have been playing with it internally and. Text has taken a big leap forward. 

[00:08:28] Dara: It's a lot better. Yeah. And it was, it had already gotten better, didn't it? But now it's like, really, it's really sharp. None of these random made up almost realistic, like merging of two words or swapping letters around or whatever. It doesn't seem to be doing that anymore, which is good. Yeah, 

[00:08:45] Matthew: It's interesting. I'm on the, I, I I on Reddit. I always see this. There's a, there's a channel called, is it ai?

[00:08:52] Matthew: And it's literally people posting pictures 'cause they're confused. They can't quite tell the difference between if it's AI or not. And the hallmarks have always been text in the background of things, nonsense text written on a beer bottle or written on this, that or the other. weirdness with fingers like being all sort of, it messed up and crowds didn't look right.

[00:09:16] Matthew: And over the last sort of few weeks I've noticed a trend where. No, there's not really been a consensus on the images that are being posted, that people are, people are posting things that, to me, look clearly, look like real folk. There's like, well, obviously that's not ai, but they're still questioning it. And however, that's a symptom of people just questioning absolutely everything in the world now. 

[00:09:37] Dara: Yeah. It'll, well, it'll reach a point where if you're good at it, it's not gonna be a 50-50 chance anymore. If someone says, is this ai you, you're better off saying yes. 

[00:09:45] Matthew: Yeah. I mean, it's gonna be, it's gonna be a hell of a lot easier to produce interesting images via AI than to go out and take a picture.

[00:09:53] Matthew: And, and there was, could you, did you see that? I dunno which video model I did this with, but I've noticed another trend going off on Molly here. But anyway, another trend of video models where like on YouTube and things, I've been seeing a lot of. There was one of Tom Cruise and Brad Pitt fighting.

[00:10:10] Matthew: I dunno if you, if you saw that. And it looks, it just, to me it was like, yeah, that looks like it's real. Yeah. It looks like it's real. And there was one, there's, there's a few people who've been making these absolutely illegally, but making these Disney, their own little mini Disney clips and their own sort of fan animation versions of anime and things, that are getting more and more convincing.

[00:10:36] Matthew: So we're definitely in some interesting world. How long, how long until you can generate an entire TV show on Netflix for yourself? Well, I mean, you could, I guess you could do it now. You could. Well, no, but I mean, like Netflix. Sell it What? Sell your show or No? You, you go into Netflix, you say, I wanna show about, an Irishman and an Northman Northman who were Renegade.

[00:11:00] Dara: Renegade cops. Yeah. Who, who fits. And, you know, you just described the whole thing and it just makes you a series that you can just sit and watch. Hang on, stop. We obviously need to cut this bit out 'cause we're giving away the gold here. Yeah. What gold we've, we, this is, this is a trademark to TV shows.

[00:11:21] Dara: This is gonna be the next hit on Netflix. Don't just give it away to our listeners. Yeah. I'll back out. I should cut that. Cut that, cut that, that. 

[00:11:27] Matthew: But anyway, the point. 

[00:11:29] Dara: Is Yeah, it's, it is, is that's, that's only a short time away. I thought you meant how long until you could do that for yourself and you probably could do that now.

[00:11:37] Matthew: Yeah. I've seen people making shot for shot. sort of like there's an anime that I sort of saw as a kid called Dragon Bull see, and people are making shot for shot, like. Making Leonardo DiCaprio one of the characters, and they'd like making shot for shot and like it's, yeah, it's mad, but I'm quite impressed.

[00:11:56] Matthew: I, there's a lot of lawsuits. I've got a reference for once. It's quite old, right? It's an eighties TV show. 

[00:12:01] Dara: Maybe that's why, maybe, maybe that's why it worked for me. Yeah, I'll just talk in the eighties and nineties and, yeah. That's more my, more my, my, my jam. you made a point the other day as well about how it's, well, it's not gonna reach a point.

[00:12:12] Dara: It's kind of there now. It's almost gonna flip the other way around where you were looking for things like the text being a bit funny to, to, to, to kind of, show that it's ai, it's gonna reach a point where it's gonna be the other way around. You're, if it's, if it's too perfect, you're gonna know it's ai.

[00:12:29] Dara: You're gonna look for those human mistakes. You're gonna look for typos. You're gonna look for things that are a little off as a sign of authenticity. Whereas if something's too perfect and too polished, you're just gonna think, well, that's obviously not. No, no human made that it's, it's, it's, it's too correct.

[00:12:45] Matthew: Yeah. I, I do, yeah. I intentionally leave spelling mistakes. I mean, I make a lot of spelling mistakes as I'm dyslexic, which I used to spend endless time correcting, but now, like I'll leave that in there to give it a bit of the old spice. Yeah. Although Matthew obviously wrote that. Yeah. Yeah, exactly.

[00:13:01] Matthew: It's becoming a superpower. Yeah. the other thing that came out, which looks interesting is. Google Code Wiki. Yeah. It's what it is in a nutshell. It's essentially AI generated documentation and diagrams and all of these things on any given GitHub repo and looking into it, it's essentially powered by notebook l.

[00:13:32] Matthew: So it's presumably pulling all these things in, generating all of the, the documentation, everything around it, and then building out diagrams and these things, using notebook LM. But I thought you'd just be able to point that at any repo you wanted to. But I couldn't do that. It said it needed it, and it was working on cataloging repowers.

[00:13:52] Dara: It's just a bit disappointing. Yes, I agree. And there's a sign up I think as well that you need to sign up to be able to do that. So it's just got a limited set of repos that you can look at now, but I also was just underwhelmed in general. 

[00:14:06] Matthew: Okay. Just please elaborate. 

[00:14:09] Dara: Well, I, I got the idea of it, but I didn't think it offered it. Maybe it was just the way they've done those examples, but it didn't really offer anything over and above what you would just get from Git. I didn't really see it, it wasn't interactive enough. It didn't have anything in there that stood out as being really wildly different from what you'd see in some of the diagrams and stuff.

[00:14:32] Dara: But you could obviously get them in a, in a reba, and if you're using something like obsidian, then you'll get those by default. So I just, I don't know. I, I think I expected more, maybe, I don't know, maybe that was wrong of me, but I just, I, it just looks like a slightly more stylish get repo. 

[00:14:51] Matthew: I, when I first saw it, I assumed, and this, I mean, this is what they say, a new perspective on development for the agent era, I assumed it was very much going to be generating AI, compatible documentation for everything else.

[00:15:10] Dara: So, so, so that agents can understand the things with a lot more, A lot more ease, but I don't know, maybe it will, I mean, it's probably just a classic case of like, they just put something out there to, to get, you know, put the feelers out and then it'll evolve. It's probably not the finished product that surprised me not to change the name of it.

[00:15:28] Dara: Well, they will in time. Yeah. It'll become, you know, some, something completely obscure and then it'll just become code Wiki again. 

[00:15:36] Matthew: Yeah, okay. It's a weird space, but yeah, it's interesting, but I kind of wanna see where they go with it. Yeah, it's been a long time as well since they'd first announced that as a bit of a Google habit.

[00:15:47] Matthew: Yeah, it was ages ago. They talk about it months and months and months ago, which, as we've alluded to, pretty much in every podcast for the last six months, months, and months, and months is a very, very long time. So you kind of forget about it. It just goes out of your memory entirely and then it pops back up.

[00:16:03] Matthew: Maybe if you'd have seen it three months ago, you'd have been a bit more in awe of it, but because of what you've seen since, you'd be like, meh. Google Workspace, CLI came out, there's been an interesting trend of, CLIs and, and people maybe recommending CLIs over CPS in some instances. I know that the CLI can go into an MCP, but yeah, Google's released a full workspace.

[00:16:30] Matthew: CLIs. You could create docs, templates, the whole nine yards, so I'm sure that's going to be. At an MCP near you or in Claude or wherever. and then will be a pretty powerful tool. 

[00:16:44] Dara: Yeah. Our, say our friends, we don't know the guy, but we're, we're open Claw fans and the guy who created Open Claw, he.

[00:16:52] Dara: I think a lot of this came from him, or he certainly stirred a bit of this up where he's kind of favoring CLIs over mcps, but I, it's another one of these things, isn't it, where some people latch onto these things and it becomes a bit polarized, but it's not really a one or the other. Is it? It's like they're both useful and it depends on them, it depends on what you're trying to do.

[00:17:11] Matthew: Yeah. I think I, I think, and it depends who you are and what your knowledge is a little bit, I think it can be a lot easier just to. I dunno, in Claude Cowork, for example, just to plug in a connector and away you go, rather than if you are going to Claude code or into terminal and adding in and installing CLI via that methodology.

[00:17:33] Matthew: It can be a bit, maybe, you know, pretty much instructions to take you through it step by step nowadays, but still, and I think, yeah, horses for courses, I think is the term I used, but yeah, it's not one or the other. What else? Somewhat. I'm trying to, I'm trying to tick through these, but this is my, this is, this is the boiled down list, by the way.

[00:17:56] Dara: Yeah. Just for our, just for our listeners' benefit, I'm looking at the full list that Matthew shared as part of our prep before this, and it's like a blindingly long list. So yeah, this really is just the, the, the, the super trimmed down version. 

[00:18:11] Matthew: There's some few interesting things that come out with cowork, so there's. Scheduled tasks were released, which I've been using quite a lot. What we've, we've recently sort of got cowork quite widely at Measure Lab and been using it and multiple people have been using it. Scheduled tasks are quite cool. Similar sort of thing that you're seen in things like Open Core and, and Open Eyes had a kind of version of this for a while, but not quite.

[00:18:37] Matthew: So I've got, I've got a few set up that look like me. Downloads and desktop folders in the morning and organizes them from the days worth downloading and, and recordings and things. I've got some looking up news, pulling things in. I've got a daily routine to look across Gmail and Slack and things like that.

[00:19:00] Matthew: I'm pulling new memory pieces and new tasks, et cetera. So, so what? I'm just kind of getting my head around what I can get it doing and working on, but it's kind of, you can imagine. Setting it off to do all sorts of things overnight while you are sleeping, it's that promise, a 24 7 worker working for you. Just trying to figure out what I want it to do. 

[00:19:22] Dara: Yeah. And are you finding as well, like, are you having to go more and more meta with it? 'cause it's like you, you're like, I'm too busy to do this thing, so I'll set up a, a, a process or get an agent or sub agent or something on it and then it's doing it, but then you're like, I'm too busy to look at the output so that you need another one to look at all the outputs.

[00:19:37] Dara: So then you just think, where does this stop? How many there are, I'm basically just building my own elaborate pyramid scheme, an AI pyramid scheme for myself, for content. I knew pyramid schemes had come back. I don't know why they ever went out. I don't know why they ever went outta fashion, to be honest.

[00:19:54] Matthew: No. Well, here we are, but, and, and yeah. Essentially NVIDIA's at the bottom of that pyramid. Yeah. Just collecting all the cash. yeah, so the schedule crashes and then there's also Claude Remote Control. Mm. Which. Again, a lot of the stuff I mentioned about c Claude, and I don't, can't remember if we mentioned this last week or not, I don't think we did, but it feels very reactionary to open claw.

[00:20:19] Matthew: Like the, a lot of the features of open core that people were finding really useful suddenly are rolling out very quickly on, on philanthropic, and, and, and, remote controls. Essentially you can. Set up a session to start doing some coding or something on your laptop, and then you can continue that session remotely via the app on your phone.

[00:20:39] Dara: And I guess cowork itself was kind of pre-pre-predicted open cloud, but it was, it's, it's gone a lot better since, like, when cowork came out. It was, I, maybe, maybe it was 'cause we were busy using load code and, and didn't initially see maybe the. The full benefit of it. But I all, I do think objectively it's gotten a huge amount better since they first released, and part of that will be the pressure from probably not just open clouds, but just the pressure from the other, the other solutions, the other alternatives.

[00:21:09] Matthew: Anyway. Yeah. Yeah. And we can, we can probably dive in a bit deeper to cowork after the legacy. The legacy news. Yeah. Yeah. There's only 750 items left to cover. We're nearly there, I'm sure. Last podcast we said GPT, five point threes come out and this podcast thing, I'd say GPT five point fours come out and I, I've not checked on this, but it feels like 5.1, two, three and four have all come out this year.

[00:21:43] Matthew: I might be wrong there. There definitely seems to be a change of pace with OpenAI. Well, everyone actually, I Google, I think we might have mentioned this last time. Maybe we didn't. Gemini 3.1 came out. We definitely mentioned code 4.6 coming out. 'cause 5.3 came out at the same time. But yeah, more releases, more like topping benchmarks or this, they seem to be just doing this slowly but surely, incremental gains over the next one.

[00:22:13] Matthew: And the next one comes out and says, I'm better. What else would they be doing with the a hundred billion or whatever that they've raised through all of their.

[00:22:21] Dara: I mean, you, you, the least they could do is put out a new version every couple of days. Every couple of weeks. Otherwise, where's all that money going? 

[00:22:28] Matthew: I'm, I'm assuming as well, I, I might be wrong here, but I'm assuming that five is the training. Like the, the compute and then the points are, are improvements on reasoning and, and behavior and post-processing stuff. That's the way I've always kind of, 'cause I can't see how they could possibly be doing 5.1, 2, 3, 4, 5 in a couple of months when they need to train.

[00:22:55] Dara: Yeah. I would've the same. I mean, I don't know, but that sounds reasonable to me. Yeah. but are you, I have to admit. I'm not, I'm not really keeping up with, open eyes, but I'm just not, again, I know we've said it quite a few times, kind of anthropic, fanboy, but even between that or Gemini, I, but I'm just not seeing it anymore with, You've just done a thumbs down for, for anyone not watching, for some reason you disagree with me, you just did a little emoji thumbs down there. I don't know why that keeps happening. I must have something on my body that isn't my fingers that looks like a thumbs up, down. Suggest a thumbs down. Yeah, you just got a thumbs up.

[00:23:37] Dara: You've just got a thumbs down energy about you. I have, yeah, right now. but 

[00:23:41] Matthew: yeah, I, I'm just, no, I don't, I don't use it either and I don't know. Used to.

[00:23:45] Dara: Yeah, same. Yeah. Anyway, whatever that means. That, that may or may not mean anything, but it's, it just is interesting with all of the, the continued huge amounts of money they're raising.

[00:23:56] Dara: But I just, and obviously they do still have the market share, the, the large, I think. But that's probably because of all the people who started using it basically like a search engine and just are still using it. 'cause why would they, why would they move to anything else? Maybe we covered this. I'm losing track of what we've said in real life.

[00:24:12] Matthew: We did talk about it a little bit, but Yeah. That we were unsure of if OpenAI were, were in trouble, whether we were just, were not in that bubble. Whether, yeah. 'cause we talked. Yeah, we did. Sorry. Business. We don't know what we're talking about anymore. 

[00:24:30] Dara: But did we add, did, did we, did we before? Is that a new problem? No, no. People still listen in and no, it's the merging. No, in, in our defense, it's the merging of, there's so much news happening and it's, it's, you know, and a lot of it all links together. even an example of which I just thought when we were talking about open air, this back and forth, or this war that was going on between the Pentagon and.

[00:24:58] Dara: And then OpenAI swept in, even though Sam Alman had said, we, we, we would stand behind Anthropics' decision. And then weirdly, 24 hours later or something, they had taken the contract. So that whole fiasco has all kind of happened in the last few weeks as well. 

[00:25:14] Matthew: Yeah, and, and it's weird because it came, there was one big news story that, oh, there was one big news story that came out that was saying, anthropology is dropping.

[00:25:22] Matthew: It's. Security posture to keep up, to keep pace and this, that and the other. And then literally the next day there was this big open letter they released that was what they had said to the Pentagon of like, you, you're not allowed to use it in autonomous weapons and this, that, and the other. So I couldn't quite tell what that, I can't quite remember what that original story was unless it was a preemptive story that they were going to allow the DOA to do that.

[00:25:50] Matthew: DOA. Pentagons. DOA is DOJ isn't it? In the DOJ? 

[00:25:57] Dara: That's a department of aggression in my head. Probably a better name. Yeah. 

[00:26:01] Matthew: Well, what, what, what is it? 

[00:26:02] Dara: They, yes. I dunno what that visual story was. No, I don't know either. And now they're taking them, they're taking the Pentagon to court. But I don't, we mentioned this before as well, it's a bit weird.

[00:26:11] Dara: It's like if you're gonna, if you're gonna do the deal in the, it's a little bit of, What would you call, not posturing, but there's a little bit of kind of like convenience, I think, around the anthropic position. Like they were happy to take the contract in the, for what did they, what did they think they were gonna do with Claude?

[00:26:26] Dara: It's just for clerical, it's just for organizing Christmas parties and, and the potluck supplies, you know, like organized and when the coffee needs to get ordered. It's certainly not for mass surveillance or, or, guiding drones or anything like that. I mean, what as if the Pentagon would use it for that?

[00:26:44] Matthew: I mean, to be cynical. I like anthropic messaging, everything resonates with me, but I'm pretty convinced a large amount of it is still marketing in one way, shape, or form. They just, that's their flavor, that's their, their brand, right? But the brand is there and ultimately to sell, sell more of this thing.

[00:27:07] Matthew: And if it stops working ultimately and they start falling behind, I wouldn't be surprised if they switched. That brand will change that posture. Yeah. Yeah. Happened with Google, it happens with, happens with all of them, doesn't it? Yeah. And OpenAI, I wasn't particularly surprised that they jumped in and because I feel like they need wins.

[00:27:28] Matthew: They need money. They're making a lot of big, expensive bets, like we said last time. 

[00:27:33] Dara: But it was just weird that Sam Alman, I dunno whether that was him getting ahead of himself. And then the board said, hang on a minute, we need to do this. This is an opportunity. But he had come out and said he was, you know, you, I, I don't know the exact wording, but he, he came out supporting philanthropic, and then had to backtrack, I think pretty much within 24 hours.

[00:27:55] Dara: But people forget that, don't they? It's like. There's a lot of flip flopping that goes on with these things, and 

[00:28:01] Matthew: Yeah, there is. And yeah, and so, and so philanthropic are suing the government because essentially Trump said, oh, well, you anthropic are not having any government contracts anymore. In this way he tends to make sweeping statements like that Toys outta the pram.

[00:28:17] Matthew: Yeah. So I think, yes, it's all, all gone. Matched. Yeah, let's, that's the correct technical term. Yeah, yeah, yeah. so that's pretty much, I think they're the biggie, aren't they? They're the biggie. There was a perplexity, perplexity computer, which was their attempt at sort of like a cowork type release. I looked at it. Then I closed the tab down. Yeah. 

[00:28:49] Dara: And, and you know, if Dara doesn't immediately sign up to it, then it's not worth it. I just, there wasn't an opportunity to give them any, well, there wasn't enough opportunity to give them my money. So it was like, yeah. I just thought, well, look, if, if there's not, there's no long term subscription that I'm not, you know, I'm not interested. It's at least three pages to a pay window and I'm just not, I ain't got time for that. Yeah. 

[00:29:10] Matthew: No, don't, don't, don't gimme that kind of friction to hand my money over. No, but yes, that looked interesting. But it was very much, seemed to be the way, the, the way the world is heading currently from these frontier models or these, these big AI companies to make these local computer based systems that can do things for you.

[00:29:31] Matthew: So that brings us, well it kind of brings us onto our subject. 'cause I think one of the big pieces that we've been using a lot internally recently at Measure Lab is. Claude Cowork and as we kind of had a bit of a Google focus special episode last week on where Google are at and what they're building and and their specific little piece of the pie, it feels like it might be a good opportunity for us to do the same for Claude.

[00:29:59] Matthew: 'cause I think, and I think this is probably true for you as well, Dara, the other models that we both use the most, I would say. Despite us both having access to Gemini, the, the forefront Gemini models, but yes, we do, we do use Clot quite a lot. Why, why do, why if, what is it? Why do you use it more than anything else?

[00:30:24] Dara: I think, and we did kind of, we, we, we share the cynical view, but I, I, one of the reasons, even if it is largely marketing, is just their approach to what they share and their approach to trying to expose weaknesses in their own models. And, you know, all the red teaming they do and they talk about all that openly and the vulnerabilities.

[00:30:46] Dara: You know, maybe it is mostly marketing, but you've gotta, you've gotta, you've gotta try and support the morals where you see them. And, and I think of the companies that are out there, I'm, I'm no fool. I know they're all trying to be profitable. They're all trying to make money. But I just like what Anthropics stands for.

[00:31:04] Dara: So that was the, that was the main reason. But to be honest, like since then that was kind of what got me to try them out. but it's the usability of their. Of the features they roll out. It just, it just is a little bit more, I dunno, everything just feels a little bit more considered. It all joins up a little bit better.

[00:31:22] Dara: I mean, go Google, like, they just don't make things easy and even though Gemini is woven through everything, it's just not, the Google ecosystem's just not as easy to navigate around. It's possibly, you know, it's, as we talked about last time, it's better for things like. You know, they're, they're obviously focusing a big part of their efforts on infrastructure and cloud and, and you know, like they, they, there's no arguing with them on that perspective.

[00:31:51] Dara: But when it comes to the AI models, I feel more comfortable with clauses. Definitely. And as we said about open air, I mean. I don't know. They were the first one I used 'cause they were the first one everyone pretty much used in terms of the mainstream. But I just don't see any reason to use it now.

[00:32:08] Dara: Even if their models were better, I might use them for certain things. In fact, I think I do use them. I do use some of the open AI models for some of the things that I do like. Yeah, exactly. Exactly, exactly for both of those. Yeah. but as, just as a general all purpose ai, I don't like what they stand for and I don't really see any reason to use it compared to using Claude or even Gemini.

[00:32:34] Matthew: Yeah, I think, I think I'm pretty much the same. I like it. I like the moral stance. It's not necessarily the, the, the moral aspect of it that gets me, I, I, I like the data. I like them, just the studies and the sort of sitting and reading about how they've gone about certain things and, and the economic index stuff they released to sort of show global adoption.

[00:32:58] Matthew: And there was one, there was a, there was a, we'll try and remember to share it in the, in the podcast notes, but there was, There was like a spider diagram they released the other day that shows like potential AI on the table versus utility actually being used. And it's just interesting data and, ultimately we're, we're data people, right?

[00:33:18] Matthew: And, and it's, it's nice to see. But so, so there's all of that. And then I just had I, I've not used full disclosure like the codex, the local Codex, 5.3, 5.4 stuff, but from a, pretty much from a. Coding perspective. Claude Code was just head and shoulders above everything else for, for a long time. Like I tried, I really consciously tried to use Gemini, CLI to do some of the coding stuff.

[00:33:49] Matthew: I tried to use anti-gravity, so like I've just sort of held them at arms length. 'cause I'm like, no, I, I only use Claude, but it is just. Better and it produces it, it, it, the experience is better and, and it just produces better code from, from what I can see. And I've, and I've had, I've talked to, you know, some people internally who've gone on that same journey.

[00:34:16] Matthew: They've really sort of said, no, I'll use Gemini. Use Gemini, and then coding purposes. They've ended up coming to the same realization that called code. It just works better. and I've, and I've got a few developer friends who are, who are the same. They've used various things like cursors. Oh, cursor, yeah.

[00:34:38] Matthew: I've got a few developer friends who have used Cursor previously, and like they've got company licenses for Cursor and then they've put DI Dipped into cloud code. Called CLI and have sort of had that, it's that, I think we mentioned the term Claude pilling, last, last time. But yeah, it was just that kind of awakening moment.

[00:34:59] Matthew: And then added to that, like we, I, I, we, we both used open CLO quite a lot and that was basically built around things like core skills. Which is a really neat and interesting way of getting l LMS to do certain things and their equivalencies in the different frontier models, but it felt like they kind of nailed it.

[00:35:23] Matthew: Cps, another Claude, anthropic thing that has pretty much embedded in as the standard across the industry. And then cowork, which. Has, I think you said it earlier, but it's really moved on in the past month to the, to the point where it is. I, I was u last, I was using it like on Friday to just work through a lot of busy work and, and cleaning and I was just, I had like six, seven different things running off and doing things for me at once and coming back and returning, They're sending messages for me on Slack and all sorts of stuff, and it, and it was going off onto, onto the Chrome browser and doing some research over there. And I, I got this real sense of the future, that I hadn't before. and so it seems again like another thing they've put out there that is. Bang on the money. Yes it is. 

[00:36:24] Dara: I think we maybe initially, because it, the whole idea with it is that it's like Claude code for non code tasks, but I think when it first came out, and it probably is a combination of it like improving a lot since then, but also maybe what you were already used to Claude code, like you were probably the same.

[00:36:45] Dara: The first couple of things I tried in cowork, I did them and then thought, why didn't I just do that in Claude code? But, now it's the, the, the, you know, improving the connectors and the plugins. And I think, I think it's clicking now, isn't it? It's like, okay, for all these other tasks where you're not, where you're not building something, the power of cowork to go off and, and integrate with all of these different systems and as you said, have to have these things happening in parallel and it's all automated. I mean, it's fright, it's frightening how capable it is. 

[00:37:20] Matthew: And I, I spent a lot of, I was kind of getting it set up for the company at the end of like, mid, mid to late last week. And I was a bit like, this is very buggy. It's very buggy, this, and then I realized that pretty much all of those bugs were, were I being an idiot.

[00:37:36] Matthew: So I've changed my tune on that a bit. Yeah. It does feel a lot more polished than it did. Like I've not, I've been using it pretty solidly for a good half a week now, and then it's been. It's been pretty bulletproof. and yeah, we, I built like a plugin, for example. So plugins are like con, they're almost like containerized sets of skills and CPS and instructions and things like that.

[00:37:58] Matthew: And I, like, I built one to gather up documentation and build manifests and regularly update those like that documentation and, and act on them and create calls. Potential calls, meeting notes and just loads of things. It is kind of just starting, it's like any of this stuff, you have to start doing it first, start playing with it, and then slowly but surely the use cases start kind of revealing themselves and you get on a bit of a role and away you go.

[00:38:29] Matthew: You kind of, you, you're building some really, really interesting stuff. I know you were playing quite a bit with, The data plugin, right? 

[00:38:37] Dara: Yeah. Yeah. And even that, like, I was saying this to somebody earlier, like the fact that that's one plugin and it uses whatever connectors you wanna use. So we were playing around with it with Big Query, or you, you know, you could have the GA four data in there, whatever.

[00:38:53] Dara: and that's one plugin and one or two connectors. And that in itself, just on its own is. I'm trying to think how to avoid this, like potentially that is a better, easier, more accessible way of working with data than some paid for enterprise solutions. Now, obviously there's caveats there because those enterprise solutions will offer all sorts of other things that you won't get straight outta the box with the data plugin, but just in terms of getting at data, visualizing it, and interrogating it, it just works.

[00:39:26] Dara: And it's crazy and you can be, because it's effectively clawed code, even though it's cowork, you could get it to you, you can build with HTML or it can write Python, it can write CS, s, whatever. And you're just connecting straight to it and, and working with it. And it's, it's crazy. And that's just one, that's just one plugin.

[00:39:44] Dara: That's, it's not like that's what it does, but it already does that better than some things that only do that. You could apply that probably to everything else, every connector we plug in and you, and the ability then to pull all those things together and then have, have the clawed model on top of that is just, it's insane.

[00:40:01] Dara: The possibilities are endless. Really. 

[00:40:03] Matthew: Yeah. It goes back to something, I think it was what Mark Edmondson said, it's like anything, anything you currently have, anything that's built like Cloud cowork right now and its ability and that plugin and its ability to go and pull data out and do certain things.

[00:40:19] Matthew: Nothing has to change. That new model, a new model comes out and that's just instantly a better product. 

[00:40:24] Dara: It's better. Yeah, yeah. Without, it's without a tweak. Yeah, yeah, yeah. Like it's, it's, it's the, the, the plumbing is there, the infrastructure's there, and every time a model comes out, it's just, it just ramps it up by another, well, who knows?

[00:40:36] Dara: 10%, a hundred percent, 200%. Like, who knows? 

[00:40:40] Matthew: Yeah. And, and it's, it's the first time I think I've seen so, so. We've had, you know, we've been, we've been messing around with AI and building out AI platforms and playing the conversation analytics for, for like, probably in the literal years at this point. But the, and, and we've been trying to sort of get that AI adoption wider across the company.

[00:41:05] Matthew: And, and we've had some successes, some failures there, partly because maybe some aspects of it, of AI and the, the maturity of it wasn't there yet. but I think partly because it was, it's, people have only ever really thought of it to who they are, who aren't coders have only ever thought of it as a chat interface of like, you just go and ask this box a question.

[00:41:29] Matthew: And it will return an answer to you. And then maybe you'll copy and paste that and take that somewhere else and blah, blah, blah. And while that's all, that is very impressive and we, we, we mustn't forget that that is still bonkers and magic the way this now just integrates and can just perform actions for you.

[00:41:48] Matthew: I've, I've seen that. Yeah. I've seen a much more enthusiastic sort of, I've seen a lot of light bulbs going off. Not just at Measure Lab, but everywhere. You know, like a bit, ah, okay. Like this is what it can do. 

[00:42:03] Dara: This is what it can do. Yeah. 'cause you're mo you're moving away from, as you say, like a standalone thing that you've gone and you've put a question in, it's given you an answer back and it's been great.

[00:42:13] Dara: And that's sped up a lot of processes. It helps with a lot of thinking and brainstorming and all the rest of it. You've still had to go. And then take that and take that into, whether it's a spreadsheet or a document or whatever, or into an analytics platform or an email marketing platform, wherever, wherever you're gonna do the thing then, and what this is allowing you to do is just like it's jumping you 10 steps ahead and saying, not only can you pull the data sources in and then run some analysis and ask questions on it, you can then do the final thing that you were trying to do in the first place.

[00:42:42] Dara: And you can do, like, there's a, there's probably a lot of. Like roles now that you could already just do everything within cowork. You wouldn't, you wouldn't need another, you wouldn't need a software license. You'd just need your cloud license. You wouldn't need to be switching between tabs all the time.

[00:43:01] Dara: You'd literally just have it connected to whatever, whatever data sources you're using, whatever context you're using, and then your outputs would all be there as well. Create me a spreadsheet, create me a graph, create me a report, whatever. And you could just do everything within, within cowork. It didn't exist.

[00:43:19] Dara: They vibrated it into existence, what, three months ago or something like that? Not even maybe two months ago. 

[00:43:24] Matthew: No, it is. It's such a simple id, I mean, all the things you just described kind of, and maybe this is why. You know, I, I I, I've been banging the drum a little bit at Measure Lab because I was in the cloud code world and I could see all the things you described, you could kind of see that happening in, in cloud code, right?

[00:43:44] Matthew: You could just go there, say do this thing. You see these files being created and the code being Red Theft Threat Center, it pops out local host URL, and you click it and it's like, fucking, that was an app there. so that sort of autonomous going off that agentic behavior, right? That, that, that agentic stuff.

[00:44:02] Matthew: Has been clearer to people who maybe have been in the CLI world and working in code, code and other, all the similar aspects, but not necessarily to a wider audience who aren't, that's not their natural world that they, they normally live in. And it feels like, yeah, that is now it's crossing that bridge and maybe they, it says it, I say it's obvious.

[00:44:21] Matthew: Why didn't they do it before? Maybe they just couldn't quit, but now they can. But whatever it is, it's, it's cool. And I think. I can see now playing with it for the past week. Why they've been trying to crack down so hard on open claws to, to close like the, the owo gap. So, so for those who don't know, you can, with open claw, you could set up, you could use your sort of OAuth connection to.

[00:44:52] Matthew: Claude to use as the models and everything with an open claw, which is significantly cheaper. 'cause if you use the API, it's Claude's very hungry, open claws, very hungry, and it would cost you an absolute fortune, like thousands of pounds probably even if I don't go down even. Yeah. It's not a subscription, is it?

[00:45:08] Matthew: You're not interested in, and, and I and, and played with it. I think I was, I was, I was doing something today, I can't remember what it was, but I was like, organizing all my files and creating like a memory place. Folder for it to work in permanently and organize files and things. And it was going off and creating Claude code instances and stuff like that.

[00:45:28] Matthew: And I thought this, this is what they want you to subscribe to their setup and for it to be a nice, cheap way of doing all this stuff, or relatively cheap. and they don't want some open source community to co-opt their OAuth and steal their thunder. Which is fair. 

[00:45:48] Dara: Have you, have you got to a point yet where you can see what would be lacking? 'Cause we've obviously both been playing around with Open Claw now you've spent a lot of time on cowork. I haven't spent as much time, but I am liking what I see. Are you, 'cause you're, because with some of the memory you've got like folder structures and you're setting up memory. So you, you, is that where you, is that where you're, you're heading for that? You're basically trying to replicate what you were previously doing with open law. 

[00:46:14] Matthew: Yeah, it's, yeah, essentially I think the main gap is, potentially the telegram, the, the sort of sending a message and then letting it go off and performing various actions. 'cause I, I don't know that. It might be possible. Well, one of my next thoughts is to talk, to, talk to coworkers, say like, right.

[00:46:35] Matthew: What could, how could I extend your capabilities? How could we add to you and see if it's forthcoming with those kinds of ideas. But yeah, I'm building out, really simple, really. But it's, it's, one file that it lives in. We've got like vaults, we've gotta work. We've got obsidian vaults with, for various different things like a management one and then a more wider company one, and all these different types of collections of MD files and thoughts and, and things like that.

[00:47:04] Matthew: So they all exist within it. And then it's got like a code instruction file, a memory file, a tasks file, and then it can kind of just work within there. It's got all the context from everything you can look through. Memories, update memories, add and remove tasks based on. Messages I've got in Slack and Jira and, and Gmail.

[00:47:23] Matthew: It can pull all that context out. It's kind of like doing an enterprise. 

[00:47:29] Dara: I'm finding the enterprise knowledge retrieval stuff better than Google, but she's mad 'cause it's, most of it's Google it It is, but it's, the interface isn't now that, that stops like. 

[00:47:43] Matthew: Yeah, maybe I, I'm not, and maybe it's worth saying that I've not, probably not used Gemini for a month, and that's a long time. So I don't know where it's, where it is now, but, but also I, 

[00:47:57] Dara: I think it's back to what I said earlier, it was just like, it, even, even Google, like Google's always so much more capable than what people can easily access as well. So it's not just about whether it can do the same things. It's how easy it is for you as the user to actually.

[00:48:12] Dara: Interface with that. And I think, you know, and maybe the, maybe an advantage Anthropic hub is they're not trying to do as many different things as Google and connect 'em all together. But it just, things just work better. They're just more intuitive than trying to, like, it's so, it's, it's not, it's not just about capability, it's about accessibility as well as, how you actually. How you interface with that and, and Google struggled with that. 

[00:48:37] Matthew: And I do wonder if, I can't remember if he said this last week in the last podcast or if this was a separate conversation we had, but Anthropic, you would imagine, I don't know how many, how big the company is. It's probably pretty large, but the distance between the frontier model makers and the CLO called team and the clo cowork team.

[00:48:59] Matthew: I don't imagine it's very big. I imagine there's like handfuls of people all sort of very much collaboratively working because ultimately the product is clawed. But Google is huge, massive, and the Deep Mind is a completely, almost a separate entity and company within Google that is creating the Gemini models and then they are being presumably passed out to all of these distinct teams.

[00:49:23] Matthew: Some of those teams might be bigger than the entirety of Anthropic. And it sometimes feels like there's no, there's not, there's not the cohesion or the, the sort of passing back and forth and, and pulling together of resources and ideas in Google, it feels like there's, you just don't have that luxury at that, at that size and scale.

[00:49:41] Dara: And it's that there, that's a disa you know, we talked a lot about their advantages last time, but that's a, that's a disadvantage in this current. The world's size could work against them. It's a disadvantage for a lot of them, the, you know, there's gonna be a disadvantage for a lot of large companies. 

[00:49:57] Matthew: Because if they are big, their size and, and everything has been an advantage for them in the past, in some areas, is going to start being a disadvantage because they need to move quickly.

[00:50:08] Matthew: If you've got 20, you know, 20 odd stakeholder groups or more than that, and your only recourse is to dictate. AI usage and tool usage from one up high and let that propagate down. That's so slow and with the way the pace at which things are moving, and it feels like if, depending on the industry, but if some upstart disruptor appears who can move quicker and sort of do something differently, there's a real danger there that there's a lot of disruptors start appearing in a lot of sectors because they're able to just work with this new stuff. Move. 

[00:50:42] Dara: Just move. 

[00:50:43] Matthew: Yeah. 

[00:50:43] Dara: Yeah, yeah. Exactly. Yeah. Yeah. 

[00:50:46] Matthew: So, I mean, if you are a massive company and you need help just come to measure that, we'll sell you out and knock the problem. 

[00:50:50] Dara: Exactly. Yeah. Yeah, exactly. That, that, that was the, that's what you're really saying there. Yeah. And going back to the, just for a second, going back to it, 'cause I just wanted to ask you on it, you were saying about how, you know, the, the, the kind of telegram and obviously you can use WhatsApp and this other channels as well for open club, but I don't know about you, but I've kind of become a bit frustrated with that.

[00:51:08] Dara: And actually I don't know how much of an advantage that even is. And with things like the remote, whatever they call it, the cloud remote access or whatever, some of those advantages that maybe Open Cloud had, I'm not sure how much they exist anymore anyway. 

[00:51:24] Matthew: I haven't used to be, to be completely honest, I haven't used the Telegram functionality probably in a couple of weeks.

[00:51:33] Matthew: I don't know why I, I, there's a, there's a, there's a, there's a certain amount of novelty to it, I think that you can sit on your Seti. I'd like messaging and it's going off and delivering and building things out, et cetera. And there's definitely use there. But yeah, I've not tried the remote Claude session yet, but if I, if I'm working on a project and just start cloud code on my computer, I could probably still just go and sit in my city and just be talking and getting it to perform new tasks from there.

[00:52:05] Dara: I dunno, how much are you still using it? Well, I'm less, I'm using it less. and I am, and when I am using it, I'm frustrated with it. So I know you did this too, but I went to build, mine's not as advanced as yours. but I went and built my own custom desktop app because I was getting frustrated with having to go to Telegram to do things.

[00:52:26] Dara: And then some of the limitations within Telegram. But then again, back to that problem we mentioned earlier, if you're basically using an application to then copy and paste text out of that and put it into whatever. You want to use it in, so, so Telegram was a novelty at the beginning, but it's not offering as much anymore.

[00:52:45] Dara: But then I think with Cowork, I guess that's one of them, at the moment, there's no way to work with cowork apart from on a cloud desktop. Right. 

[00:52:51] Matthew: That's the only way you can Yeah. 'cause I think it, I think that's, that's its big idea in a similar way to that, that's essentially what. Claw bot called, I can't remember the name of it now.

[00:53:05] Matthew: Open Claw o. Open Claw is it meant to be able to enact things on your files and, and your browser and things like that. It's meant to be like a local sitting system that sits on your machine. Yeah, it coworks, coworks the same. It's, you know, meant to work in your files and close the Chrome extension to go off and do things in Chrome and all that.

[00:53:28] Matthew: All that good stuff. So, no, I wouldn't be surprised if there's at some point a way for you to deploy it. Yeah, that's what I'm wondering if they're gonna head that way. Yeah. Yeah, because like, I mean, it's worth saying I suppose, that it can be a bit ram hungry, right? From what I've heard, I've not had that problem 'cause I, it depends, it depends how much ram you have.

[00:53:50] Matthew: I, like I said, I ran out of Rama like a year and a bit ago 'cause I was messing around with ai. So, you know, you're just too slow. Too slow, but yes, it can be a ram hungry. So I've not noticed that. 'cause I've got 32 gigs of 

[00:54:03] Dara: Ram. But, well, I told you, I told you Claude mocked me. Claude basically said, of course your Mac is struggling. You've only got, you know, you, you, you've only got the ram of a small child. 

[00:54:13] Matthew: Yeah. You got a Fisher Price laptop.

[00:54:14] Dara: .A fish Fisher Fisher Price laptop. Yeah. Basically mocked me. 

[00:54:17] Matthew: So, yeah. But no, it's really cool. It is cool. I'm gonna keep using it. I'm definitely getting more and more utility out of it and that kind of general.

[00:54:28] Matthew: Sector of always on assistance that I think open core is the beginning of. But I think it's gonna be rolling out heavily everywhere. And I'm sure with open AI soon, 'cause they literally hired the guy, you build up a version. 

[00:54:40] Dara: So, yeah. Well that's, yeah, that's actually a really good point. Like what's he doing?

[00:54:45] Matthew: Yeah. Building that or or advising on that, surely. 

[00:54:48] Dara: And is it gonna, is it gonna offer something over and above Cowork? Well, we'll have to wait and see, but it's gonna be hard. I think it's gonna be hard. It would be hard for them to pull. Avid users have clawed away now. I think it would really have to be offering something truly above and beyond.

[00:55:04] Matthew: We said this about, if you think about it, we, we kind of said this about chat GPT for a while. Well, it would've been, we would've said when chat BT first came out, like, yeah, what would, what would the reason be to move away from it? 

[00:55:18] Dara: Yeah. 

[00:55:18] Matthew: Google is dead and then, you know, Google brings, you know, slowly but surely claws back and starts to pull together.

[00:55:25] Matthew: You know, and I, like last week, we talked about all of Google's advantages. They have a lot more going for them other than just the models. 

[00:55:31] Dara: Yeah. They're spread, they're spread a lot more across different, different areas. 

[00:55:35] Matthew: It's so, and maybe, maybe what we'll have to do, we've done, we've done Google, we've done a, a, a deeper one on Claude.

[00:55:41] Matthew: Maybe we're gonna have to sign up to chat GPT and go and actually use it for a few weeks and then do a deeper dive on chat. G-P-T-I-I might be on holiday that week. 

[00:55:50] Dara: No, it might be amazing. And listen, I'll probably, I shouldn't say this so confidently on the podcast 'cause I'll probably end up in a couple of weeks being like, yeah, no, I'm, I'm totally ai. What was it doing? 

[00:55:59] Matthew: Claude 

[00:56:00] Dara: Claude's rubbish. Yeah. You never know it all. It's all, all in flux, isn't it? 

[00:56:06] Matthew: Yeah, it is. 

[00:56:08] Dara: This is true. All we should say is, this is true at the time of recording. Everything we say is vaguely true at the time of recording, as far as we know it, and we're not as far as we know it.

[00:56:19] Dara: So it may or may not be true at the time of recording. Right. That's probably a good point to stop and I look forward to preparing for the open AI dedicated episode. I'll start researching. Give you more things to subscribe to. You must feel like that, at least. It's a good point actually. Well, I do. I am paying them money for what I am using them for. Oh, well, there again.

[00:56:42] Dara: That's it for this week's episode of the Measure Pods. We hope you enjoyed it and picked up something useful along the way. If you haven't already, make sure to subscribe on whatever platform you're listening on so you don't miss future episodes.

[00:56:54] Matthew: And if you're enjoying the show, we'd really appreciate it if you left us a quick review. It really helps more people discover the pod and keeps us motivated to bring back more. So thanks for listening, and we'll catch you.