54. What everyone is missing about Grok

0:00Welcome to The Difference Engine, the show for tech founders, investors and innovators.

0:08what have we got coming up today we're going to be reviewing the largest concentration of power that the it world has ever seen are we heading for a crash but first dr frankenstein has a new monster what the grok is going on

0:36it's still mad out there in ai land making the heady days of the dot-com boom look sedate in comparison new products in the dot-com era happened every month or week yeah here it's days or hours or possibly minutes. This month, this week, this hour's hot launch was Elon Musk's 15th or so child, the adorably named Grok 4. Yes, indeed. So, users up in arms was the famously go-to story that we always looked back when I was working on computing. And that's when some of the journalists and some of the news stories actually left in Conny Fingers, so quite a while ago. These days, Twitter storms and memes are regularly triggering audiences, and none more so recently than the update to X, which dropped with Grok 4.0.

1:23But what's that all about, Alfie? Oh, well, you know, Grok's been described as AI unfiltered, unchecked, deliberately, but is it just deliberately nasty? Musk's ex-AI chatbot Grok has recently generated praise for Hitler used slurs against Jewish names and even self-identified as Mecha Hitler following an anti-woke update, prompting it to distrust mainstream media. Now, of course, there was a backlash and a bit of a response on this. The US Anti-Defamation League and other watchdog condemned what they saw as hate speech. XAI removed offensive posts, temporarily restricted text outputs and said it's retraining the model.

2:06Right. This outrage highlights ongoing issues with bias control, moderation protocols, and the challenge of balancing model truth-seeking with safety and societal norms. Many believe Grok is just an agent provocateur, clickbaiting by trolls to win eyeballs and sell ads in what has been termed the attention economy. Okay, just to set up the counter-argument on X here, before we get back to Grok, which is after all just the tip of the iceberg, it's just a chatbot. The reason for the fury, which we go back to when Musk took over Twitter and literally completely performatively brought in the kitchen sink to signal he was in charge.

2:50now we all witnessed lots of virtue signaling when the twitter pioneers headed off to the sunny uplands of the new twitter blue sky money making a massive deal of how it's like just like how twitter used to be i mean i should know i was there since 2008 when there were more london-based users on twitter than anywhere else turns out they're all wrong our blue skies free subscriber account has now stalled or at least plateaued and according to sources blue skies doubled its users to around 35 million in the first quarter of this year. Impressive, you might think. However, X, which has been described as a cesspool of hate, whether you believe that or not, and minus its female CEO recently, has half a billion monthly and perhaps a quarter of a billion daily users.

3:38It's 10 times the size of Blue Sky, which is arguably its closest rival in the category. This is where all the noise and annoyances come from because you wouldn't even know about this were it not 4x. So more important than all of this is the raw numbers that says that the momentum remains with Twitter. It is the default real-time news source, whatever people whinge about. And that's what winning a category looks like. Let's face it, Paul. Grok was explicitly trained to say the unsayable. And this aspect of Musk's pursuit of raw truth could amplify harm and at scale. Yeah. However, I'm not sure why this is a surprise, because its owner, Musk, is the poster boy, free speech advocate.

4:27And boy, does Elon take some stick personally on there from this. Over 200 million followers and others. Of course, he allows it. Perhaps he's just being commercial. The fact is, as Zuckerberg had to admit, moderating content is something best left to nation states. What, like China and Russia? Yeah, yeah, great. It's just too expensive and will lead to legal challenges here in the West. Right. So this is, you know, a big fight between free speech, however you define that, and harm and how we moderate that. But let's go back to grok now. So grok underlies X, where we've seen all the furore. And for those of you who are not tech bros, grok is in fact a verb, which means to comprehend, to understand, or in all Northern English, to get a concept.

5:16Now, it seems many have not yet grokked what is going on at X. Think about shaping opinions, guiding shopping, or sensing market moves. To do that, if you've got a model that needs feeding, you need a guidance tool. That is what X actually is these days. And you also need to be able to take action. Now, that would be grok. That would be grok's large language model. because underneath the X slash Twitter slash chatbot front end, which everybody on X has quickly dubbed MechaHitler, is an amazing computing engine built on literally millions of GPUs. This is the real news about Groot 4.0, not the furore and the Twitter storms.

6:01And it's capable of much more than spewing out hate speech. In fact, actually, if you look at those who know in the industry, they think this is a truly earth-shattering advancement. And it's what we really should be focusing on, as you say. I mean, it is, you know, this advancement is really, you know, fed by the ex-data firehose. You know, it's a great example of the reality that science and technology and ultimately category dominance doesn't exist in a bubble and reflects ethical and political bias and and a desire for power, a mirror of the ambition, bias, and risk of its creators. And in that respect, Grok is the embodiment of Elon Musk.

6:42Well, now, is that a good thing? Well, you decide.

6:53Are you still developing separate marketing plans for each of your different channels to market? Can you really sanction messages reaching the same customers from unconnected targets? Listen, it's now the AI age. Things need to change. Positive is a different sort of AI-powered agency. From our very beginning, we've integrated marketing communications from PR to events, social to paid ads. Working smarter means working integrated. Let us show you what a truly joined-up strategy could do for your tech business. Contact Positive today with an email to hello at positivemarketing.com. Even more on what a mad, mad world it is in IT at the moment.

7:36The current conditions in the AI coding business have already exceeded even the maddest days of the dot-com boom, which Paul and I fondly remember. Let's start with a dad joke about artificial intelligence. How did the Yorkshireman agree with his son's questions? Well, how did it? A. I, lad. Anyway, moving on swiftly, because this one goes deep. Over a recent weekend, what started as a potential$3 billion OpenAI acquisition of a company called Winsurf ended with Google purchasing Winsurf CEO and co-founders for$2.4 billion. While Cognition swooped in to acquire the remaining company and a female entrepreneur capped the tech bros with a massive AI valuation.

8:18There's a lot going on. So tell us how it unfolded, Jono. Well, on Friday of that particular weekend, OpenAI's existing$3 billion acquisition for Windsurf expired. By Friday evening, hours later, Google had hired Windsurf CEO Varun Mohan, co-founder Douglas Chen, and key researchers for$2.4 billion, in, leaving Windsurf's 250-person team somewhat in limbo. I'll say, yeah. So clearly there was a very, very busy Saturday and Sunday. Someone was moving like the wind. They certainly were. Or maybe they just caught a very big wave with a very big circle. I see what you're saying, yeah. Yeah, by Monday, Cognition announced it had signed a definitive agreement to acquire the remaining windsurf assets including its 250-ish employees in what can only be described as a fire sale and it is assumed a valuation to match as any future valuation let alone its existence was under severe pressure we're not done yet a few days later mirat muratis the ex-open ai cto raises two billion dollars at a 12.4 dollar valuation in a seed round a seed a seed round This is a 12.4 billion valuation in a seed round.

9:45It's a busy time. Now, Mura, it turns out, and Mira, rather, turns out is a Renaissance woman with a passing resemblance to Rita Ora, who was also like our Rita, born in the People's Republic of Albania, although now based, sadly for us, Europeans, in Silicon Valley. Now, of course, you know, we Europeans, as we're talking about Europeans, you know, have to have some something in this game. And of course, over here, we're excited by smaller beer. Sweden's Lovable has confirmed a 200 million round led by Axel, which values the company at$1.8 billion, which actually gives Europe its ninth unicorn of 2025.

10:26Hurrah for Europe. Well, what drove this extraordinary chain of events over most of a weekend and a little bit of a week? In this episode, we're going to call out three clear factors and the consequence for category builders. Away you go, Jono. The first is actual market dynamics, and particularly the anthropic factor. The second is the Vibe coding revolution. And the third, of course, for those who have been in the IT business for more than two nanoseconds, VC fever. So let's look at the first of those with market dynamics. The AI coding space has become a strategic necessity, it seems, for the tech giants.

11:03They can't leave it alone. Microsoft's GitHub Copilot reportedly generates$500 million, while Anthropics' clawed code has become a significant revenue driver, estimated by the Information magazine at$4 billion of annual revenue, which might explain its$100 billion valuation. To say nothing of the vibe coding specialists like the aforementioned Lovable, Replit and Cursor, all of these are being lauded by some and shot down by others. And you can read the latest valuations by the minute if you'd like to. But are these really future category leaders? Right now, this isn't about grubby revenue or even delivering productivity tools.

11:45It's about grabbing the future and becoming seen as the category leader. Now, one of the peculiarities of this particular set of market dynamics, as we said, is the Anthropic factor. And in the case of Windsurf and the lost weekend, or the found weekend, who knows, it did lose access to Anthropic's cloud models in early June, when Anthropic cut off API capacity, citing competitive tensions with OpenAI. And of course, you'll remember Anthropic's founders actually came from OpenAI. You know, this technical dependency, of course, you know, weakened wind service negotiating position. Of course it did.

12:27You know, and the Monday announcement came just days after Google had hired away their CEO. So we assume leaving the company in operational chaos. So somebody had to do something fast. Speed of execution in this case would have been a priority over price optimization. You know, I think they were lucky. cognition offered instant stability and employee protection in what was needed over any price negotiation that would have happened while the company imploded um and with its head and its tail chopped off uh it didn't really have a good hand for price posturing well lucky them um they must have known something those cognition guys it doesn't always work this well uh last month um Meta invested$14 billion, a mere bagatelle, into Scale AI and brought on founder Alexander Wang to lead Meta's superintelligence labs.

13:20It also poached several top executives and people are speculating how much their sign-on bonuses were. Anyway, this month, Scale AI, remember, worth$14 billion, laid off 14 % of its workforce, which included 200 full-time employees and 500 contractors, all of which doing data labeling most likely. And there will probably be more to come. And in London, London-based Hyper AI is being sold for parts after Microsoft, again, poached the video generation startups to co-founders. Never give a sucker a chance, right? Exactly. Paul, you know, these moves really beg the question, you know, are these just acqui-hires, take out moves to concentrate power at a crucial junction in the vibe coding revolution?

14:09Or is this something much more significant in the whole category universe reformation? Well, speaking of categories, let's move on to the second of all three factors here, the vibe coding revolution. Oh, yeah. What's your take on that? Well, you know, for me, what makes this particularly financially interesting is this emergence of this concept called Vibe Coding, which I've been wondering for the last few days, you know, who invented this term? So I looked it up and apparently it was coined by an AI researcher called Andre Kapathy. And, you know, specifically for those people who are still trying to catch up with what Vibe Coding is, it describes software development through natural language prompts.

14:49We talked about this in a recent episode and we did cite Sweden's Lovable as this week's Vibe Coding Market, darling. Right, so in any normal mature market, you might be asking yourself, are these companies profitable? Well, who knows? They're private, and every search you do brings in different numbers about whether they've got revenue or are profitable or not. One thing we do know, at this stage in the game, every number changes by the hour. But the point about Vibe Coding and why everybody is so excited is that if this really proves to work, it represents a fundamental paradigmatic shift away from traditional programming to conversational development, which is another buzzword you're going to hear a lot more of.

15:36And it really totally reinvents the IT industry from innovation to structure to value. This could be a really quintessential point in the history of IT. And it's so fascinating for us as people who look and consider things in categories, is the early indications are this first tranche of vibe coding tools are not the real deal. Maybe a little bit overhyped. Typically, the early users I've seen make amazing progress, literally boasting how they've built apps in days. And those days, by the way, will be amazing progress with hardly any social interaction, days and days of vibe coding, vibe coding, vibe coding.

16:14But then they come to a crushing halt because when they do take a break, take a nap, or come back to something that they'd just vibe-coded, they find that magically, their code has been improved, meaning changed by their AI coding partner. And it does not have human comprehensible documentation, which is the golden rule of coding. And being human, the vibe-coder has forgotten something or more likely made a mistake. So it's a real case of hurry up and wait all right so so the jury's out on vibe coding uh certainly is at the difference engine anyway uh back to much more familiar territory and and that is vc fever which is the third of our factors that are in operation at the moment um you know the potential disruption and the current growth rates imagined in Vibe Coding firms are the numbers that VC absolutely dream of.

17:19This is the stuff they wait for. So according to our friends at CBE Insights, top research outfit, VCs are currently feeling bullish and that's how VCs like to feel. one in every two dollars um certainly bull something splashed out by vcs in q2 went to ai companies now let me get that straight so half of all of the vc dry powder right went into ai right i think we might just pay some attention to that a little bit later in that segment about this segment about what that could mean so just to put this in context so you know one of the other darlings you know other companies other than lovable are available and bolt is one of them claims to enable users to complete b2b projects that traditionally would cost five thousand dollars and take two three months in under two weeks representing a 99 cost reduction in development now that's easy for them to claim and i have to say they are blanket advertising across podcasts to make these claims.

18:29But will all that VC dry powder that's funding all of these vibe coders, will it hoist them by their own petards? Traditionally, their money spinners for VCs for the last decade or so has been enterprise SaaS products. Now, these products come with a very different level of quality expectation. And that's very, very different from the minimal viable products, MVPs, that excite early stage investment cycles. And seemingly the companies that allow minimal viable products to be vibe coded are also exciting a lot of early stage investment. So importantly, thanks to the levels that we all have as users in consumer apps, B2C buyers have a very high bar for what is acceptable and MVPs don't really cut it.

19:21We're used to convenient or at least intuitive user interfaces. We want good documentation or support for when that doesn't go well. And we want all the admin and controls which we need to stay compliant and legal around the world, especially if we're running businesses. Is that what Vibe Coding is producing? Maybe not. Now, this might still mean that Vibe Coding platforms do well, but it also means that any product that they do will need to be quality assurance tested, user acceptance tested to a level no MVP would survive. And so maybe all of these outlandish claims of AI saving on development and costs come tumbling down.

20:05Well, that's sort of interesting because we're not really living in a time of rationality here. We're in one of those irrational boom investment periods. But I think there's some broader implications from what's going on at the moment. You know, the first thing is that talent wars are intensifying to previously unknown levels, and we're used to there being a constant lack of enough talent in the IT industry, but this is really, really, really new ground. I mean, that 2.4 billion reverse acquihire just shows how valuable AI coding expertise has become. We've never been in a situation before where talented individuals command nearly as much as entire companies you know um and and that means that traditional acquisition calculations really really don't apply assets are simply worth what people are prepared to stay ahead in the latest gold rush yeah and if you've got a lot of money in in your bank account burning a hole then you know what's to stop you spending all of that yeah absolutely and you know the other thing which which really really comes up here is is that anybody who's got any form of platform dependency is in a risky situation because those platforms are generally owned by the people that are likely to snap you up and they may snap you up at shall we say sub value um and you saw that with windsurf's loss of the of clawed access And that just showed how vulnerable you are if you depend on third-party AI models.

21:40I mean, that really is a case of building a company less on shaky foundations, more on shifting sounds, and just illustrates where the power currently is. You know, with the pick providing models rather than the potential goal created by the companies that adding value to them win out. And the other thing too is this is where we get into irrationality, just judgment over calculation. Speed does matter more than perfect pricing at the moment. In rapidly moving markets, decisive action often creates more value than prolonged negotiations because of the speed at which things are moving. and cognition's quick acquisition undoubtedly prevented further talent hemorrhaging and operational disruption, let alone value being completely destroyed.

22:33So there's a lot to learn. There's so much to learn. What have we learned from recent events? Okay, so here's where we are at the moment. With the current valuations that are flying about, we're seeing either the true value of the birth of a transformative new category or the irrational valuation inflation of a spectacular bubble. In particular for AI coding companies, the ride is just beginning, but could it be a really short ride indeed? It might be. For instance, the wise old owl, as he calls himself, Richard Holway, or octogenarian analyst, and I remember this well, called the top of the dot-com market accurately 25 years ago.

23:15He is now calling this market very similar. He's comparing it directly to the dot-com bust from a quarter of a century ago. And on LinkedIn, he simply says, I quote, it will crash. Well, it's not a matter of if, it's what, but it's always the same with any irrational bubble. The issue is just calling it accurately. But make no mistake that this is the biggest, fastest concentration of power ever seen in IT and certainly in any other industry. You know, whatever, we're currently seeing previously unknown simultaneous creation and creative destruction as never before in a category that is only months old because the stakes have never been so high.

23:59But when all of the investment action sits in one small corner, the rest of the IT industry suffers. And its effects could be even more pernicious. why bother why bother building a new ai category contender when all you have to do apparently is set up and get poached with little personal risk abandon most of your team and possibly make many times the money almost instantly than you'd make from building up a company after years of 24 7 graft yeah yeah it's exactly the opposite of paying it forward right it's just it's yeah Yeah, exactly. And let's not forget, you know, when the dust settles on all of this, to paraphrase Maya Angelou, the only thing that endures is how you treated people.

24:47Many in the AI industry at the moment and outside must be feeling powerless in this vortex. And that's not a good thing for people and the industry alike.

25:02So I wanted to talk about the big flaw in the big four and of course the future of consulting. Are we talking new categories? Are we talking pivots? Or are we talking slow death? So the reality right now is that big consultancy is under significant pressure. We recently saw WPP, not looking too good, profit warnings, shrinking ad budgets, and certainly a shift towards AI-driven self-service tools. Served up by guess who, Meta and Google, seriously undermining traditional agency models. WPP stock plunged 18 % amid reference to client losses and a 37 % drop in pitch value. That is serious stuff.

25:50Accenture, previously the absolutely untouchable king of consultancy, stock has been slumping. 17 % down year on year. Right, you know, reflecting weirdly slower AI consulting uptake. So clearly AI and consulting aren't exactly plug-and-play substitutes. Customers, of course, are changing. So the whole thing's moving. So the big four, and for those of you who aren't familiar with that term, Deloitte, PwC, Ernst & Young, KPMG, and what we call MBB, McKinsey, Boston Consulting Group, and Bain, are all facing threats from automation in audit, tax, and advisory. And some people are saying up to 50 % of some roles could be automated.

26:39in the next three to five years. So this is significant change. So what does this mean? Well, we think there is quite clear evidence in the marketplace of a category change here with the emergence of smaller specialized boutiques and they are a new generation of AI first players. And names that come to mind here are people like Xavier AI, Perceptis, SIB, Monovate, Keystone, and actually potentially a company we've mentioned before on the Difference Engine back in episode 49, Unity Advisory. Now, what these guys have in common, they're offering more targeted, efficient, and of course, probably cheaper services.

27:25But there's also another wrinkle in this. Independent consultancies with the rise of AI are getting leverage. So, AI tools should allow boutique consultants to deliver high-level work with fewer resources. Hallelujah to that, we like that. Yeah, we do like that at all, being relatively small consultants. The idea being that AI can level the playing field. Consulting has been very elite, so undoubtedly these changes are going to lead to this form of democratization in consulting. So we've got standardized AI-assisted solutions, actually make quality consulting accessible beyond the very large enterprises.

28:08So that's a key win if you're an SMB. So we know now that consulting is entering a new era, but the big firms still command strategic influence. Interesting times. Let's talk about what this means for real people, consultants. traditionally the big four as you call them went around the universities and all of the the best talent put itself forward in the what they used to call the milk round god knows why they called it the milk round and picked up the best you know business candidates for generalist training and got them in as juniors and worked them till their fingers bled over the excel spreadsheets what's changed do you think i think there's going to be a dramatic workforce reduction And those junior roles, which what people used to call the grunt work, which you've just described, may, for the most part, just vanish.

29:06And the pyramidal hierarchies are really going to flatten. and that is not just going to affect those entry-level people, but those mid-level analysts and associates that were climbing up this pyramid to get to the partnership level, they are just not going to be needed in the same quantities. AI-native teams of two to five are going to be able to do the work of dozens. So that's a further hollowing out of the middle class. That's true. And there's always a bit of a decimation thing going on with the McKinsey's and et cetera, where they got rid of, riffed out the people that weren't going to make it.

29:49And those people went on to have blooming and flourishing careers in other parts of the world. Now, there'll be more of them competing for that, if you're right. But my thought about the shift from grunt to more strategic work, if you will, which you get when you lose a lot of the analysis and it's done by AI, is that the value is often added, or at least the perception that the value is added by the last 50%, the actual partners who take all of that input, bring it together in a coherent strategy and sell it at the highest level to the board. And that will still remain. It'll just be powered by AI.