Prompt and Circumstance

Episode 11
===

[00:00:28] Tom Adams: Welcome to Prompt and Circumstance. We are glad you're here. we are excited to be back after a month of being away from each other and, uh, Mike is dealing with some kind of challenge. We don't know what it is 'cause he was here and then he disappeared. He'll be coming back. But, uh,

[00:00:46] Ryan Niemann: He's in a remote location somewhere.

[00:00:48] Tom Adams: right and he's been in his car before, he's been in other places today, we don't know where he is 'cause he can't seemingly get connected

[00:00:55] Mark Redgrave: Tom, we should do a GoFundMe, Mike, we.

[00:00:59] Tom Adams: Right.

Bring them in. So today, I think if, if people are watching us and we know our mothers and occasionally our kids, if we force them with a arm tied behind their back are watching us. But for those, most people actually know who we are. So today, instead of telling people who we are and me introducing you, you are going to introduce yourself on the surprise episode by two answers to two questions.

The first question is, if you could have dinner with anybody. Dead or alive, who would you have dinner with? And then number two, what major frontier LLM model are you most in sync with? Those are your two answers that you have to come up with. And Mark Redgrave, you are gonna start, uh,

[00:01:49] Mark Redgrave: Mike's

here,

Mike's.

[00:01:52] Mike Richardson: I'm here, I made it guys. Sorry about

that.

[00:01:54] Mark Redgrave: We would just say, well, anyway, you have to watch the episode, Mike, to see what we were saying

[00:01:57] Tom Adams: what

we said

about you.

[00:01:59] Mike Richardson: I'm so dial up. Sorry

everybody.

[00:02:01] Mark Redgrave: It was, we were quite nice, but not, not completely nice. Okay, Tom. Right. So I'll start. Okay. Person I'd most likely have dinner with.

[00:02:10] Tom Adams: Dead or alive? Yep.

[00:02:11] Mark Redgrave: easy at and center. Not even a, not even a question. So. Finest racing driver ever lived. Uh, you know, very spiritual, more talent than you could even imagine. Uh, I grew up, I was heavily involved in motor sport as a young man.

Um, actually spent a number of years as a professional racing driver in Europe. Uh, and my absolute hero still is. I've got end center stuff all over my, hold on, I'm gonna show you something.

[00:02:39] Tom Adams: Okay,

[00:02:41] Ryan Niemann: Is show.

[00:02:42] Tom Adams: this is show tell. We we're, oh, we have a helmet.

[00:02:45] Mark Redgrave: This is my race helmet from like mid nineties, and that's stanners

like icon.

[00:02:51] Tom Adams: Okay.

[00:02:52] Mark Redgrave: So, so anyway, so, um, yeah, so that's easy now. LL, l model.

[00:02:57] Tom Adams: That you're most aligned with or in sync with, which one feels like you're in sync with it?

[00:03:02] Mark Redgrave: you know what, so, so I'm gonna answer that slightly differently, which is that I feel like I need to migrate completely to Claude.

I feel like now, and because, but I'm still, I've got too much stuff in chat, GBT, and I'm kind of, I'm kind of, uh, I'm trying to pull myself over, but I've, my, my, my head says, move it all to Claude.

[00:03:26] Tom Adams: Got it. All right.

Ryan.

[00:03:28] Ryan Niemann: yeah. Uh, mine is, uh, Teddy Roosevelt. Uh, I, I, you know, there's any number of things you could point at. Uh, I often think the man in the arena is a fantastic, uh, speech. Uh, I also love the story of him doing a stump speech in Wisconsin, I think it was, and he, and he gets shot from a pub owner, but his speech is in his coat pocket and it stops the bullet.

And, uh, and so he, he assesses as a, a, a hunter that, well, this hasn't punctured my lung. I'm gonna give the speech. President gives the speech.

Uh, after being shot. Like, just think of like that as, uh,

just the, the lore of, of what he's done, his, his,

explorations, Uh, after being in the presidency and, and, uh, travels and the such.

I think that would just be a fantastic lunch. Uh, and the LLMI most in sync with quite similarly, uh, OpenAI has a ton

of contextual relevance, uh, for

me, uh, operationally, but, uh, I probably. Uh, do much more Claude, uh, as far

as code development and the like, than anything else. And I feel this compelling desire, uh, to try other things, uh, like, uh,

like Gemini or, or Claude or what have you.

But the contextual relevancy,

uh, keeps me, uh, quite in tune with open ai.

[00:04:53] Mark Redgrave: Same.

Yep.

[00:04:54] Ryan Niemann: Yeah.

[00:04:55] Tom Adams: Mr.

Richardson.

[00:04:56] Mike Richardson: Yes, I finally made it in here. Everybody got through my technology gremlins. I live out in the middle of nowhere, so, uh, rural internet sometimes is a problem. Uh, I'm gonna be boring. Rather than choose a president, I'm gonna choose a Prime Minister. I quoted from him, I think an episode ago. Of course, Winston Churchill.

How could I not choose Winston Churchill? And if you've seen the movie Darkest Hour, you know, I'd love to have dinner with him and just ask him, how on earth did you get through the darkest of the darkest of the darkest moments when it looked like all was lost? And yet you've, you managed to muster, you know, leadership when we most needed it.

And the other reason why I choose, you know, uh, that Winston Churchill is. There's not a lot else for us to talk about really as Brits. You know, our last hurrah was really the second world war and really, and we keep shooting movies about it. 'cause that's really the last real big thing that we, you know, apart from, apart from the odd member of the royal family being arrested and things like that, there's not a lot else, you know, to, to shout about really.

[00:06:05] Tom Adams: And.

[00:06:06] Mike Richardson: And, uh, yes, I am so invested in chat, GPT. I've done so much work in there. I've got so many projects in there, so many different lanes of chat and all of that, um, that I, yeah, I'm a bit stuck right now. I, I must admit, and, um, the least aligned with is co-pilot. I've tried and tried and tried to sort of. You know, get copilot to do what it should be able to do as part of the bigger, you know, windows ecosystem.

But I just can't get excited about it, at least not yet.

[00:06:37] Tom Adams: Okay. so my, uh, dinner with anyone dead or alive is I'd, I'd like to have dinner with the three of you. 'cause like

heck, um, uh,

you

know,

[00:06:47] Mike Richardson: What an answer.

[00:06:48] Tom Adams: do, we do all this stuff together. We've never had dinner together, so I, I think it's time we have dinner together and maybe, you know, get

together on some, you know, remote planet that Mike happens to be driving through.

Uh, and, and hang out together. So that, that would be fun.

Right.

[00:07:09] Ryan Niemann: Stop off at Hardee's.

[00:07:10] Tom Adams: Right. Do a little hardy's adventure. Um, my LLM is by far, uh, Claude Co. Claude. I, I am on the Supermax plan. I, I spend way too many tokens in Claude. I live in terminal, and, uh, yeah, I'm most aligned with that. um, I love trying the ones that come out on hugging face every so often. I like playing with those.

And, um, Kimmy and some of the, uh, some of the other. Open source ones are pretty, pretty fun to work with, but, uh, currently right now I'm fully in sync with our good friends at Claude. Alright.

[00:07:46] Mike Richardson: Who, who comes up with these names?

[00:07:49] Tom Adams: Hugging face, it's a beautiful place to be. No, nothing more exciting than than hugging face. Alright, today, as I suggested is the surprise episode, and none of you, raise your hands if I have given you any advance warning. N none of you have any advanced warning. I'm gonna throw stuff at you and I want your response, your thoughts, your insight, your perspective from each, uh, we'll just go through as many as we can in the next 30 minutes because all of these are kind of interesting things that are happening in the last couple of weeks since

we've been together.

So, uh, number one,

[00:08:28] Mark Redgrave: Tom, I've just realized why Mike's done. He's done this on purpose, so if the questions are hard, he's gonna blame his network.

[00:08:34] Tom Adams: the network is gone. Network is gone.

[00:08:36] Mark Redgrave: he's he's one step ahead,

mate.

He says he's behind.

He's

[00:08:39] Mike Richardson: I'm always one step, several steps

ahead actually.

[00:08:43] Mark Redgrave: I'm disgusted. All right. Carry

[00:08:44] Tom Adams: Alright, number one, Claude Bought is an open source agentic thing that came out three to four weeks ago. It has blown up like crazy. Andro goes, legal OpenAI hires the guy who started this open source toolkit. Response, thoughts, perspective. Mike is gone. It's beautiful.

It's

[00:09:14] Mike Richardson: Up guys. Guys, you're breaking up. You're breaking up. I

[00:09:19] Tom Adams: beautiful. That might be the best setup ever.

[00:09:28] Ryan Niemann: All right. Thanks everybody.

[00:09:31] Tom Adams: Thanks for coming.

[00:09:32] Mark Redgrave: Tom, I'll, I'll give you a react. My reaction to this, actually, when I saw this break was like, oh gosh, we're in this just ridiculous, like human capital like frenzy, which, which I, we've just seen many times before, right? It's like, it's like people moving poaching. So that was the first thing that went, I.

I actually, like, I'm not deep enough on the tech to understand the implications of what's happening, but I just know that the, when the Valley starts to lose its mind around people, but it, what it made me think was like, right. You know, 'cause I was having this conversation with a client like, um, uh, last week, which is like.

Do we have the right people in the business that can show us the way? And I think, I know that feels a bit tenuous, but it's not. It's like, you know, these guys open the, the, at the top of the pyramid, these, the biggest, the biggest, most advanced companies in the world are just, are just saying if we don't have the right people, we are not gonna win.

Right. So like at any cost. So I'm wondering, I'm sort of scaling that back to our clients and the people we work with and saying. Do we have the right people and

[00:10:36] Tom Adams: Mm,

[00:10:36] Mark Redgrave: to invest? Interesting question, right? Because I, I'm not sure, I'm not sure CEOs are going, you know this, there might be somebody we need and they could cost us a few hundred grand to figure this out.

Are we ready? I dunno. Interesting question, right?

[00:10:52] Tom Adams: Yeah. Ryan.

[00:10:53] Ryan Niemann: I think signal for few, noise for many, like, I think, uh, those are signals of what's happening in the market. I think there are a few that need to keep in tune with that. I think Mark just did a good job of translating that to, well, what does this mean for the, the many? And, and, you know, having the right resources, human capital and the organization is absolutely critical.

Uh, and if that's the signal to take away, uh, is to be thoughtful about making sure that you've got that in your organization. That'd be my, my takeaway.

[00:11:20] Tom Adams: Okay. Mike, do you have anything to add?

[00:11:23] Mike Richardson: could you repeat the question please?

[00:11:32] Tom Adams: th three, three to four weeks ago, Claude bought, Claude Bought, came on the scene, it blew up. It's an agentic open source tool that you can download to your local computer. And thousands of people have bought Mac Minis to make this work. Claude bought named similarly after Claude. It's C-L-A-W-D bought Claude, uh, anthropic goes legal and basically throws the book and says, you can't use that name.

Uh, the guy who was the developer changed it to Molt bought, now it's called Open Claw. But Anthropic hires the guy likely in the range of, of a billion dollars. We don't know what the numbers are effectively, but it's a big hire because this guy figured out something, which is how do agents work autonomously?

This was an autonomous open source toolkit. So reaction

[00:12:26] Mike Richardson: On the advice of my lawyer, uh, because I might incriminate myself,

I plead

[00:12:33] Tom Adams: Got it. All right. Here we go. Uh, my, my thought on it is, I think what this did was fundamentally, uh, said to the world, uh, the next stage of evolution here is not just the hiring thing that Mark talked about, but to me it's the fact that, um, Claude bought now called Open Claw. Is this thing that just goes to work, it works all night, it just keeps working for you, and it's open source.

You're not paying for it other than your, your tokens. So I, I, I think it's, it's ushering in something

interesting.

[00:13:06] Ryan Niemann: yeah, I, I would agree with you there. I, I, I focus more on the hiring aspect, but yes, actually the, uh, systems of agency, the 24 7, uh, never tired, you know, uh, constantly evolving, recursive learning.

Wow. Can't wait to see what's happening

[00:13:23] Tom Adams: self spinning up agents too.

[00:13:26] Mark Redgrave: way this is gonna show up for most normal people though, right? Is like, this will just be built into like the browser experience, right? Like that's how it's gonna show up. And I think it's. Awesome. Right? Like seriously, because it's a, like from a ui ux point of view, that's how everyone adopts, isn't it?

So, um, yeah. But it's su I mean, it's super exciting and we all know it's however quick you think it's gonna be here, it'll be here sooner,

guaranteed.

[00:13:52] Ryan Niemann: Yeah, I think it, uh, I think it's also interesting that that element of, finding what old laptop you want to go put it on, and, and make sure that it's not on your core laptop. 'cause you really don't know what's gonna

[00:14:05] Tom Adams: Right. Have no clue. Alright.

[00:14:07] Mike Richardson: Tom. Are, are, are the questions gonna get

simpler as,

[00:14:10] Tom Adams: No, no, Mike, they're gonna keep getting hard. So if your, if your internet connection is really bad. Alright, number two. Microsoft this week announced that it'll invest 50,000,000,050 B five zero B by 2030 to expand AI access in lower income countries and address emerging AI divides specifically, this was at a summit in New Delhi, India.

Thoughts, perspective.

[00:14:42] Ryan Niemann: Um, I, I think the. Proliferation of ai. You know, there's, there's other instances of technology in underdeveloped areas. Uh, mobile comes to mind, for instance. And the dynamic rapid shift in those areas, uh, will have profound. Effects on life, longevity, uh, disparity of, of income, um,

uh, maybe, maybe even, uh, rising tide rises, all

ships types aspects, um, to ways that I don't even know

that, uh, we have absolute

clarity on.

But it's exciting.

And 50 billion is not a small number. Uh, and, and if

you think of, um. Just our experience. I can't imagine, you know, what, uh, that lift will be. Uh, and the, the, the human capital, it'll unlock. Um, uh, just think of the amazing amount of talent that now will have access and, um, what those individuals be able to create and, and, uh, contribute, uh, I think will be

amazing.

[00:16:06] Mike Richardson: Where my head goes is the book, uh, the World is Flat by, was it Thomas Friedman? I think that was his first name. I mean, that book was written probably what, 20 years ago? I think. I think it was written in the late nineties or maybe to the early two thousands. And as you said, Ryan. The world has just got flatter and flatter and flatter and flatter, and not least of all, COVID flattened the world even more because it became, you know, more prevalent to work remotely.

And, and now you can get talent, you

know, across the world. And for me, this

is just continuing to

flatten and flatten and flatten the world even further and further

and further. And that's both a good thing. That's a, that's both an opportunity and a

threat. Right simultaneously.

[00:16:55] Mark Redgrave: there's there's so much posturing about all this money. It's bullshit. Sorry for people who don't like French. Bullshit. And it's like, and, and here's the problem. Um, you know, 30 years ago it was all about mobile connectivity, right? To, to, if we are to lift the rising tide, Ryan, yeah. We have to have connectivity throughout the world in third world countries.

We still haven't got great connectivity. Mike can't even get connectivity. So, so, so, so like, so, so like, and it's, it's actually a load of posturing. It's like all the money moving between Nvidia and open ai. It's like, it's not even real money. It's not even, it's, it's not, it's paper money. It's like pr bullshit.

So I'm calling bullshit on it, and I'm saying, um. Fix, fix connectivity if they're gonna spend 50 billion. Right. To actually, and I know this is true, we were, we were talking earlier today, it's like I have clients that are, um, where we have operations in India. The connectivity still is bad, ladies and gentlemen.

I'm sorry. It is like we, and we, we, you know, like we, like I am, there is not a call, I'm on a video call with teams in India where we do not have problems. Not one call. So I'm like, we got, we've still got a connectivity challenge, so I, you know, um, great idea everyone. Super press release. Hope you enjoyed it.

We've still got some connectivity to build out.

[00:18:15] Ryan Niemann: Does this mean we're not gonna have data centers in space?

[00:18:19] Mark Redgrave: I.

[00:18:20] Ryan Niemann: Oh, okay.

[00:18:21] Tom Adams: Okay. Well, given, given you brought that up, Ryan, I'm gonna drop that one. So, landmark consolidation this last week of, uh, Elon Musk's business Empire announcing the merger between SpaceX and XAI. Um. It's sort of what they're saying is he's saying is to embed AI's GR models into SpaceX operation and accelerate the development of fully autonomous spacecraft and robotic Mars colonies.

Thoughts, feedback, insight. Is this bullshit too?

[00:18:56] Mark Redgrave: How do I get access to the SpaceX IPO, please? That's my first, that's my first, uh, thought.

[00:19:02] Tom Adams: Okay.

[00:19:03] Ryan Niemann: Yeah, I think, uh, it's, uh. I often, I've had at least a few conversations about this, and we end with a sense of marvel,

[00:19:15] Tom Adams: Hmm.

[00:19:15] Ryan Niemann: wow, uh, uh, I'm glad somebody's thinking about this.

[00:19:19] Tom Adams: Interest.

[00:19:21] Ryan Niemann: And it,

[00:19:21] Mike Richardson: You, you are just on another planet today. Tom, you, you,

[00:19:26] Ryan Niemann: I

I, I can't, I I, I really, I mean, just think of that, the, the, the, the, the, the trajectory of that is. Amazing, and, and then I gotta get back to my day job.

[00:19:38] Tom Adams: right, right, right.

[00:19:42] Mark Redgrave: yeah. We're, yeah, we're out of, um, I mean, so love or hate Mr. Musk, it's like the timelines he's talking here are extraordinary. Like, like, and, and the problem, well, not the problem, but. The amazing thing is he's quite often been right in the past, so I'm like, not always, but you know, you do look at some of this and you go really, like, we are gonna have like data centers in space in like under like in six or seven years time.

I dunno, it's kind of amazing.

[00:20:11] Tom Adams: Right,

[00:20:11] Mark Redgrave: Um, you know, they still won't have connectivity in India. Starlink. That's the, that's the, that's the deal. Like I, and when, when we, when I was at the Telco, right? 'cause I was a CMO of a big telco, we were terrified of satellite. Like, because, because we just spent like $15 billion on like, on like physical connectivity, right?

Like, and trying to get connectivity to, to, to far region to, to the far, um, regions of, uh, of where we lived. And it's like, so, so somewhere in here. You wouldn't bet against him, would you? You know, like somewhere between, between, you know, like starlink Tesla, OpenAI X, probably the future of humanity sits somewhere there, so, you know, good luck everybody.

[00:20:56] Tom Adams: Got it. All right. Speaking of investment in CapEx data centers and Compute Alphabet, Amazon Meta, Microsoft together this year are claiming around 650 billion US dollars in 2026 CapEx spending, uh, for, but. Not just, not just what we talked about earlier, making it easier for people to access things.

This group is spending that on data centers and compute, uh, all over the, you know, Indiana in the middle of nowhere. They're building massive data centers. Uh, what are you thinking about this? What's, what's your insight on this?

[00:21:36] Mark Redgrave: Electricity bill's about to go up.

[00:21:38] Ryan Niemann: That well, yeah. Power, power generation, uh, access to rare minerals, uh, uh, power generation, either centralized, uh, with investments in nuclear or, uh, smaller or or smaller, uh, individual nuclear, uh. What was once used to power a a sub might power a data center type approach.

[00:22:06] Mike Richardson: I can't stop thinking about

glasses of water.

[00:22:09] Ryan Niemann: Uh, every prompt, uh, and gloss of water type, you know, access to, to water. yeah, I think, you know, I was in a, a session, uh, a couple quarters ago and you know, the conversation was that nuclear was the only approach. There's different flavors of nuclear, um, and, and how that's achieved. Uh, and whereas we may need to build something on the order of magnitude of 30 to.

Nuclear power plants in the next years, decades to come decades. Um, and since three Mile Island we've built zero and China's built 30 and they've got their own,

you know, pathway. Well, that kind of infrastructure requires people talent. Raw materials, resources, su supplying that, that, that pipeline to build, uh, will, will keep us busy for sure.

[00:23:04] Mark Redgrave: I think there's a really interesting growing socioeconomic, um, tension here because, and it's starting to, you see it in the press, right? But like, everyone's going, guys, this is great news. We're gonna build like this massive data center in the middle of Indiana. It's gonna create jobs, right?

And the locals are going, dude, like I didn't sign up for that shit. I do not want. I do not want like a humming,

radioactive like thing in my backyard. Right. And, and for my, and, and I guess their electricity

might be free. 'cause you could probably just hold up like a piece of wire.

But, but, but this is, this is gonna be a problem

because it's like, it's being couched as like economic growth, like we've never seen.

But it's gonna hit people because it, it, like this is. Wild what they're planning on doing. 'cause some of these data centers, right? I think we all saw the thing. It's like the one, is it the Google one that's like the size of Manhattan?

Like, like

it's, it's the size of like, it's the size of a town. So you, so you sort of go like, yeah, I dunno.

I just think like we're, we're about to like, um, have our environments changed.

[00:24:19] Tom Adams: Yep.

[00:24:19] Ryan Niemann: For some reason when Mark, when Mark's speaking, all I can think of is Homer Simpson. I mean, we saw what

happened.

[00:24:26] Mike Richardson: I just drove back from Telluride, Colorado, a 13 hour drive, 800 miles. There is a lot of empty desert out

there. A lot.

[00:24:35] Mark Redgrave: But you can't get the talent in the desert. Right? so,

so so this is what, this is where it's gonna be that there, there has to be.

Unless maybe, maybe this will go the

way of like Australian mining and stuff where like, you know, you just have to create these, you literally create these towns to support. Uh, maybe that's what'll happen, but I just feel like there's this amazing momentum and excitement about like, the future of this stuff for, for many.

But it's like it's gonna have a physical impact on our, our, where we live and how we live.

[00:25:05] Mike Richardson: that's kind of always the nature of these kinds of things, right? When you're, when you're at the sort of very start of it and you're looking onwards and upwards at, at this tsunami of change coming at you, it's, it's easy to sort of, uh, get paranoid about how it can all go wrong, right?

And yet, by and large, somehow or other, we managed to figure out how. More of it goes right than wrong. We always have, and somehow we always will. So I remain sort of cautiously optimistic that we will figure this out.

[00:25:37] Tom Adams: Yeah. Yeah. that's good. Good. Alright. I am gently doing this one because I didn't overly fact check it, but McKinsey and Company, uh, has introduced

an AI interview stage for its graduate recruitment requiring candidates to collaborate with its internal AI tool, Lily. During final round assessments, applicants are tasked with using the AI as a.

Thinking partner to solve business scenarios with the evaluation, focusing on their judgment and reasoning rather than technical AI expertise. Mark,

[00:26:12] Mark Redgrave: So, full disclosure, I'm a senior advisor for McKenzie. So,

[00:26:17] Tom Adams: so you can't talk, I, I'm, I'm less interested in McKinsey, more interested in the fact that. A significant consulting firm in the world, apparently is, is asking candidates to intersect with their AI tool in the hiring process. Now, not

after they

get

there,

[00:26:39] Mike Richardson: Yeah, and it could be Mark. It could be that last week you were

a senior

advisor with McKinsey, but.

[00:26:46] Mark Redgrave: I haven't, actually,

Mike, I haven't actually checked. Um, but I, I have worked with Lily in the past, I have to say. And, you know, quantum AI by McKinsey is like, you know, that is a project that has been going for like over 10

years, right? So it's not, this is not so, so my, I, I'm, I'm, you know, my only reaction is like, it sounds quite like, quite good idea to

me.

This is like the future. This is the future.

So like, um, like should it be part of the process? Yeah. Why not?

[00:27:11] Tom Adams: Yeah. And again, it's, it's not about McKinsey per se, but it's interesting that a leading. Thinking company in the world is introducing the application process to layer in a bot like, uh, their, their internal

bot.

[00:27:27] Mike Richardson: I think it's a great idea. You know, anything that puts people into a real time role playing, real playing, lets. Let's stress test you not to see if you can derive the right answer, but to see how you think and to see how you, uh, flow through a difficult situation, even if you don't get it 100% right. You know, if we saw that you have the, the, the, the right ways of thinking and being and adapting, that's what we're really looking for.

[00:28:02] Tom Adams: Yeah.

[00:28:02] Ryan Niemann: what comes to mind is that, uh, movie that's coming out, uh, is it, I, I forget the name of a judgment or something where, uh, he has to defend himself against ai, uh, in, in a certain period of time. Otherwise he's convicted of the, of the crime. Uh, I can't remember the name of the movie right now, but I would, I would say that that comes to mind.

This is just an evolution of. Situational, uh, interviewing, uh, problem based. What's your reasoning thought and approach. And I say that knowing that actually one of my agents met Lily on uh, agents.com and they're dating, so I have to be careful.

[00:28:37] Tom Adams: No. To be

[00:28:38] Ryan Niemann: yeah. One of my agents, so there's a

family

relationship I.

[00:28:42] Mike Richardson: Oh,

[00:28:43] Mark Redgrave: You know these, these agents, Ryan, they're not monogamous. I can tell you that they're not.

[00:28:48] Tom Adams: They're not. No, no, not at all. next, uh, yesterday alone, uh, anthropic drop sonnet, 4.61 million token contact window. Um, the, uh, Gemini dropped 3.1 yesterday. Um, shortly after, uh, you know, we've got Codex 4.6, we've got like these monster capabilities. Um, yesterday I think, uh, MIT put out their 4.5 of their version of an open source one that you can get on hugging face.

Um, this is interesting to me because all of a sudden context window, which has been a limiter, is now a. Not, you can literally upload 10,000 books at a time. And you have capability, uh, in these models right now. Thoughts, uh, impressions? What's this do to the world that we're living in right now, given what you are experiencing?

[00:29:51] Ryan Niemann: Well, this is related to our earlier comment

[00:29:53] Tom Adams: Yes.

[00:29:54] Ryan Niemann: about switching between models and, and the like. Uh, uh, I, I, I'm not aware of an efficient way of. Moving your context securely, uh, and cost effectively. Uh, but in, in theory, what you'd be able to do is, is extract your context and with an infinite context window, bring it over to a different model.

Um, or maybe it'll be applied differently that the model be able to apply to the context you've already established or you actually establish your own context and then have flexibility in applying your, Engagement to a particular model. I'm not quite sure how that'll, how that'll move.

[00:30:34] Mark Redgrave: But that's, that's what I want, right? So I want, I, I want like, um, I want a data layer that holds the context. I, and I want to be able to query, I want to use any tool to query it. So like, so like, like does that exist? 'cause like that's literally 'cause because yeah, that would be awesome.

[00:30:52] Tom Adams: This week, I am seeing people who are now deciding to bring all of their context, not into the LLMs, but into markdown files and keeping all of their stuff in markdown files and then using, uh, codex using, uh, ch uh, Claude code in their particular, terminal application to actually work on their files.

So like. Everything is going markdown file,

[00:31:21] Mark Redgrave: I want that for like 9 99 a month.

Give me that. That's it. That's it. And, and I'm happy to have to pay my 20 bucks on Claude and my 20 bucks on open air, but I, I want an abstracted data layer where my context, where my context sits so I can query it with any tool I want. Yes, please.

Yes, please. Sold.

[00:31:43] Tom Adams: So I, I took Claude code this week with the million context window and gave it a, an old piece of software I developed. Again, I don't know how to develop software, but I'm doing it and I gave it the whole thing and it gobbled the whole thing up in one shot and went, oh, let's do this. whereas before you could only kind of give it certain files.

Now it took the whole entire repository, all of my notes and all, and dealt with it in process, which is, uh, astounding.

[00:32:13] Ryan Niemann: that is exactly what we're doing. We have a, a product that we've been working on for nine months, and that's exactly what we're doing next is let's take it, put it in, and come up with something different on, on the path and, and have it completely refactored, which is astonishing.

[00:32:28] Tom Adams: Yes. Yeah.

[00:32:30] Mark Redgrave: Yeah, dude, this. Because, so, okay, so we talked about this at the start of the call, didn't we? Or the start of the podcast about like, 'cause that is what's stopping me, because I really feel like chat GBT now is connecting all my stuff together, right? So it's because I, I organized by project Mike, you probably do the same.

I organized by project, right? But it's starting to really understand and they're like, well given, given, you know, given what you believe is really important for this client over here. Yeah, like, like I'm gonna bring some of that context into this discussion 'cause I know you care about this and it's like, that's where I'm going.

I don't wanna leave that, I don't wanna start again. So it's like, so, so, so this is super interesting. Like, I wanna, Tom, I wanna know how to do it, Ryan, whatever, like, like, you know, send me something, I've gotta do this.

[00:33:15] Tom Adams: I'm trying to figure that out myself because, uh, I have been sort of on a quest in the last few weeks to really, instead of putting data in, to pull data out, so every time I do a chat with Claude in their app, like in the chat app, I'm asking it for downloadable markdown files. I want, I want the markdown file.

So now I have a repository of markdown files. I don't want it to stay there. I want my data. And, um, that, that's becoming increasingly powerful. Um, I think as we go forward, because we are not always sure of the tool, but we know our data. We need to hold our data. And right now everybody's holding our data.

[00:33:58] Mark Redgrave: is the context universally marked up so any model could understand it? I don't, I mean, I'm, I've no idea what I'm talking about now, but just logically, like, so, so the way chat GBT holds all the. Context. Context. Is that a universal language or not?

[00:34:16] Tom Adams: I don't know how they hold it.

[00:34:18] Ryan Niemann: Maybe one of our listeners will, will chime in for us.

[00:34:21] Tom Adams: I I just know they're holding it on us

[00:34:24] Mark Redgrave: that's the federation bit. Right. It's like you need, you need a sort of federated layer that makes, means you can query it and everyone understands it and,

[00:34:33] Tom Adams: Yeah.

[00:34:33] Mark Redgrave: and like this is not a new challenge for databases and shit, right? Like we've been doing this for 40 years.

It's like that's now. So yeah, I'm, I'm actually strangely excited and I don't get excited about this

stuff, but I'm strangely

excited.

[00:34:45] Mike Richardson: We.

[00:34:46] Tom Adams: But I, but I think the thing for me in this is, um. Context window has always been a limiter because as, as you work with a model, if even in your, if you're an open AI and you're, I mean, in chat GPT and you go on too long, it gets stupid. It really does. Uh, and so as they expand the context window, it holds more and it's.

Capability, it holds more capacity to manage what it's, working through with you. So, I, I just think it's, it's, it's gonna involve, and andro this week also added fast. So you can actually add a fast layer to their, their context window. So now you can, you pay a lot more for it. You pay double essentially for that.

Privilege, but my goodness, it's coming both context, window and speed is now starting to be another delineating factor between the different LLMs. Alright, another one for you and then, uh, I might give you one or two more, but I'm, I'm, I, I just like these kind of off the wall ones. there is a new job that's showing up in the world called Full-time Vibe Coder.

And companies are hiring people to come in who don't have code experience to vibe code their way through the organization. Um, probably the most famous one right now, and I, I realize I've been doing it, but the most famous one right now is the guy Lazar Jovanovic at, uh, lovable, who now is fully. He's paid fully to just vibe code.

He walks around the organization and goes, what do you need? And then he goes, vibe codes up. He doesn't know how to code. He's kinda like me. He doesn't know how to code, but he's coding solutions for them all over the place and. Basically he's saying that he's hearing in tons of other companies that he's being asked to talk to all these other companies who now have vibe coders who are building product tools, um, toolkit to help them in their internal struggles.

Uh, and I'm asking for your feedback on that, but there's a layer beyond that, which is SaaS software. We talked about it a month ago. This is, this to me is rubbing up against SaaS software. So thoughts feedback,

[00:37:04] Ryan Niemann: my initial thought is yes, uh, you know, call it a, a four deployed Viber, right? Uh, somebody that's working right alongside the business, uh, very quickly, iterating, creating. Uh, code, uh, in larger organizations, it still needs some level of structure and security and through the CICD pipeline and, uh, making sure that it's got, uh, some, uh, distinct validity for governance and, and guardrails.

But yes, uh, I think that is the, the right path. Um, I'm excited for it.

[00:37:39] Tom Adams: Cool, mark.

[00:37:40] Mark Redgrave: So, so I was in Vancouver last week with a client and we had an awesome session and, um, we were talking about, uh, you know, like di uh, differentiated roots to market and it, and what, what was created in front of my eyes with vibe coding, actually, Tom, 'cause you're such a, you're such a proponent of this, but it's like, it blew my mind. And I was thinking, having been, you know, like I ran a software company for 12 years, like from 2002 through to 2015, and we, you know, like what we did in an hour would've taken me no shit. Three months and a hundred grand.

Seriously, like, I'm not like, like, and I'm not joking. And, and it made me, and it, it, it brought two things to mind.

Number one is like the power is just incredible. Like, so, so your vibe, your vibe, coding thing. Yeah. Like every business should have one. Right. Second thing, a vibe coded app, a product does not make. Or there's some brilliant like Greek like saying there or Latin say, 'cause it's like, so, so, 'cause the conversation is so funny, right?

The client went, this is amazing, we could sell this. And it's like, and it's like, and the spirit of that is like everyone was on the same page, but it's like, oh, hold on, hold on guys. There's a very long distance between, you know, lovable and a product that you are selling to a client with an SLE and SLAs and, you know, so, so those two things occurred to me.

But honestly, like we, we, we had, we turned a vision into a working prototype in two

hours and it was truly staggering because it 'cause the product it created.

Man, like not, not even like, oh, I get it. It's kind of shit, but I get it. It's like, no, it was like it was working.

[00:39:33] Tom Adams: Yeah, it's not a drawing on a, on a, uh, on a, uh, a whiteboard. It's literally working and doing what you want it to do. Now it, it breaks all over the place, but it's pretty cool.

[00:39:44] Mark Redgrave: Yeah. And, and, and so, you know, like as an innovation heartbeat of your business, amazing. And, and, and if, if we've got people watching and go, oh, these guys, they talk about Vibe Cogan. And that guy Tom, he seems to know what he's doing, but it's like, guys, if you don't, if you don't know what this is, find out because it's truly, truly amazing.

Um, so it really hit me hard last week. It was wicked.

[00:40:07] Tom Adams: Yeah, that's cool. Mike.

[00:40:09] Mike Richardson: I'm, I'm, I'm, I can picture it in my mind's eye, this vibe coder walking around and just sort of plucking low hanging fruit. Oh, we can automate that. We can automate this. We can automate the other and, yeah. Wow. Mind blowing. Just mind blowing. What's possible these days and never forget everybody. Tom doesn't know how to code.

[00:40:31] Tom Adams: Yes. Alright. We have come to the, uh, the bell, the bell round. And so, uh, normally we try and recommend things to other people. Uh, but this week we're recommending something to each other, so.

[00:40:46] Mark Redgrave: Oh.

[00:40:47] Tom Adams: so we have to tell each other what everybody needs to do this next couple of weeks till we meet again, because it's very easy for us to pontificate on what everybody should be doing out there.

But what we need to do is tell each other what you need to do this next two weeks. So based on what we talked about, based on what you're experiencing in the world, um, what does Ryan need to do in the next couple of weeks based on your perspective, your experience? What's Ryan need to do?

[00:41:16] Mike Richardson: Ah. Alright. Blind me.

[00:41:22] Ryan Niemann: Yes,

yes. Thank you, mark. That's it.

[00:41:25] Tom Adams: I, I'm gonna, I'm gonna do this one first. This has just come to the top of my head, but, uh, Ryan, I, I don't know if you have, but in the Prompted circumstance website, we have a link to a GitHub page that has your original prompt. The, the prompt thing you've worked on. Um, and that still sits in my Open AI

chat tool.

It still does. I still use it every day, but I'm starting to get frustrated with it. And so my encouragement to you this next couple of weeks is based on all the new stuff that's happening, do you see any way that you could enhance that tool a, a step or two based on what's happening today?

[00:42:07] Ryan Niemann: fair enough. I, I, so what Tom's talking about is. Three years ago, uh, there was no such thing as a custom instructions, and, and I created this, call it a pre-pro

and you'd put it in. Uh, and over time I found that it worked really well.

Then custom instructions

came out and it had a, a window of let's say 2000 characters.

So I shortened it and did this, and I was amazed, uh, at how it still brought value in the way people prompted in the way they work. Um, and in fact, I think that episode is when Chat PT five came out, and I was still amazed that it wasn't. It was still good. It, it still helped, uh, in the responses. So I will take that, pump it in, see if I can make it, uh, even better.

I guess. I have to take feedback now from, from, from

[00:42:57] Mark Redgrave: Yeah, I'm like, I'm like, because what Tom said, I think I heard was like, Ryan doesn't work anymore.

[00:43:04] Tom Adams: No, it's not that it doesn't work. It's, I I

[00:43:07] Ryan Niemann: somewhere between 90. The holiday and Ryan doesn't work.

All right. All

right.

[00:43:10] Tom Adams: right.

[00:43:10] Mike Richardson: I want Ryan to do is I want him to work on getting another article published 'cause it's, that last article was awesome and yet it feels like it was eons

ago. And so we need a new one of where Ryan's head is at now.

[00:43:26] Tom Adams: All right. Uh, Mike Richardson.

[00:43:31] Ryan Niemann: Work.

[00:43:32] Tom Adams: All right. Alright, so Mike Richardson, what does Mike have to do in the next couple of

weeks?

[00:43:39] Mike Richardson: Oh

boy. Here we

[00:43:40] Mark Redgrave: like what, from what we just said, Mike, right. Go to lovable. Right. And just think of something you would like to do. Automate, right. Just type a natural language query into lovable. Right? And, and, and if you haven't done it, maybe you've done it

already,

but like.

[00:43:58] Mike Richardson: I haven't,

[00:43:58] Ryan Niemann: no, no, no. That's not

true. Mike, Mike, Mike, Mike, Mike. Way, way, way back when, about nine, 10 months ago. We are here because I got you on a call and I took you to

Rept and you quote said, oh my God, I have been trying to tell a CTO how to build this for me, and it's right in front of my eyes. We are here because of that conversation. You've done it before. Now you need to take the next step.

[00:44:26] Tom Adams: but it's, it's infinitely better than what Rep was a year ago like, like mind blowing and lovable is the fastest growing company. Now in history to a hundred million dollars. They have grown like faster than O open AI did. Um, and they're the largest vibe coding. And we're not selling lovable. We're just saying, try.

You gotta try this. Like,

You have to know, you have to understand how this level works because unless you understand it, you can't talk about it.

[00:44:56] Mark Redgrave: I think that would be

[00:44:57] Tom Adams: Yes. So Mike lovable.

Alright.

[00:45:01] Ryan Niemann: Times three. Times three.

[00:45:02] Tom Adams: Okay. Mark. Mark Redgrave. What's your, what's your orders? Anybody got an order for Mark?

[00:45:11] Ryan Niemann: Uh

hmm.

[00:45:13] Mark Redgrave: Holiday.

[00:45:17] Ryan Niemann: A, a poem on positivity? No, uh,

[00:45:23] Tom Adams: Beautiful.

[00:45:25] Ryan Niemann: just, just to keep it short. Really easy for him to accomplish,

[00:45:29] Tom Adams: Some, some kind of zen workshop.

[00:45:33] Ryan Niemann: you know, actually, so, so I, okay. Uh, you have a distinct need. Context, portability. Uh, I, I'd love to hear. Because this is not the first time it's come up in the last couple weeks. Like how, how would that, uh, be optimized? Is there a true common canonical for that, uh, uh, you know, common, uh, uh, structure, uh, that could be identified?

Um, and I've, I've heard other people uniquely in the last two, three weeks, did the same thing. I, I'd pay for this, uh, in, in some way. Um, and, and for all I know, all we know. Someone's already built something like that, like a little bit of research on that or something. Uh, and then how we toss it over Tom and he builds it, and then we can give him 10.

[00:46:17] Mark Redgrave: Exactly. That's what I was thinking. It's, I tell you though, like it's a, it's a, it's definitely a thing and I'm sure people are working on it, but like, but yeah, maybe Mike, Mike, Mike should kick it into lovable and see what they say. Anyway.

[00:46:29] Tom Adams: All right. Anything that I need to do in the next two weeks because

I, I have to give free time, so I gotta give equal time.

[00:46:40] Mark Redgrave: I mean, everything I can think of Tom is completely self-serving to me. So like I need you to build me some shit, but we can talk about that.

[00:46:48] Tom Adams: Okay. Okay.

[00:46:50] Mike Richardson: Could, if you could, if you could come up with some simpler questions

for me, Tom,

[00:46:55] Tom Adams: Got it. Okay. Simpler questions for Mike.

[00:46:58] Ryan Niemann: Now, aren't

[00:46:59] Tom Adams: new shit for Mark.

[00:47:00] Ryan Niemann: You've got an event coming up right, where, where you're going to, you're gonna showcase like how to do some of these things,

[00:47:06] Tom Adams: I, I am actually doing a, uh, yeah, I got lovable weirdly to, uh, support a hundred person vibe coding workshop that I'm delivering to a hundred people at a conference. Um, and, and lovable agreed to, uh, to fund that vibe coding workshops. So, um, I, I'm very much committed to the vibe coding cause so.

[00:47:29] Ryan Niemann: I think, uh, yeah, I coming back with some great ideas from that. I mean, this is not hard, it's not like a big task. I know you'd do it anyway, but I can't wait to hear about what that experience is like and what, having a hundred people all vibe coding at the same time and, and where that goes, I think is gonna be an exceptional story.

So I look forward to that episode.

[00:47:53] Tom Adams: Beautiful. Well, Maybe that's where, maybe that's where we should all meet though, Tom, to have our to have lunch in Mike's car.

Right, right at the Vibe Coding Workshop. Gentlemen, it's been a, I will, it's been a blast. Thank you for being a part of this today. Uh, thank you for, Willing to work with Surprise in the Unknown, and, uh, love you deeply. We'll see you in a couple of weeks.