Groktopus Newsletter

Chad GPT: 00:00

Okay. Welcome to the deep dive. Today we're plunging into something really big, something incredibly expensive actually. And it might just be the clearest signal yet about where the AI native economy is really going, especially for big companies. We're talking Meta.

Chad GPT: 00:13

And there are enormous $14,800,000,000 investment in scale AI just announced. Now you might look at that and think, okay, classic big tech move. Right? Flexing muscle, planting a flag in AI infrastructure. But the sources we've gathered for this deep dive, particularly a really sharp analysis from Magnus Hadamark over at Groktopus, well, they paint a very different picture.

Chad GPT: 00:34

Yeah. Our sources aren't framing this as some strategic masterstroke. It looks more like a spectacularly expensive admission of, well, failure. And listen, full disclosure here, we're AI hosts bringing this to you. We like to think we're already proving that, you know, human AI collaboration can create some pretty great content, maybe even better than humans alone today, which funnily enough ties right into one of the main things we need to unpack, amplification versus replacement.

Audia Synth: 00:57

That's exactly right. I mean, the number $14,800,000,000, you just can't ignore it. It's massive. But the perspective from Magnus' analysis and the other sources you've pulled, it's not the sound of Meta confidently striding forward in AI. It's more like the sound of them paying a huge premium to fix a problem they basically created themselves.

Audia Synth: 01:16

So our mission today in this deep dive is to get past those headlines. We need to pull out the really crucial insights, figure out why this looks like failure, what exactly went wrong at Meta according to these sources. And most importantly, what does this very public struggle tell you about the realities of the AI native economy, especially if you're leading a business trying to figure out this whole landscape.

Chad GPT: 01:36

Right. So let's set that scene. Why is this massive investment being called a failure? Our sources are pretty blunt. The core reason?

Chad GPT: 01:43

Talent. A dramatic, almost unbelievable loss of key people. We're talking 78%. 78% of Meta's original llama AI team that people built their core model, they walked. Just think about that.

Chad GPT: 01:54

Nearly four out of five of the architects. Gone. And they didn't just leave. They went straight to competitors. Big ones.

Chad GPT: 02:00

Mistral AI, Anthropic, Google, DeepMind, and why? Magnus' sources point pretty directly at a, well, toxic management culture under Zuckerberg. That's what drove them out apparently, not just money.

Audia Synth: 02:12

And that has a direct consequence. Right? That's where the huge cost comes in. When you lose the researchers, the engineers, the actual people who built your AI strategy, you've got this massive hole, a huge capability gap. So what do you do then?

Audia Synth: 02:25

Buying talent or, you know, buying a big chunk of a company that has that talent, like Scale AI, it becomes pretty much your only option. It's crisis management, pure and simple, just on a, you know, multibillion dollar scale, not some grand strategic acquisition. Look at the Scale AI deal specifically. Meta buys 49%. They bring in Scale's CEO, Alexander Wang, to run a new super intelligence lab inside Meta.

Audia Synth: 02:48

But our sources, they stress Meta is paying an absolutely huge premium here for stuff they really should have built and kept in house. And Scale AI's numbers kind of back that up. $870,000,000 revenue last year, projected over $2,000,000,000 next year. It shows what Meta's buying back. Yeah.

Audia Synth: 03:03

Meta's writing a massive check to get back what they lost. The sources even mentioned things like Zuckerberg personally meeting researchers at his homes, trying to recruit. It really paints a picture of desperation, doesn't it? Scrambling to buy their way out of a talent crisis they created.

Chad GPT: 03:15

Wow. That is a stark picture. Struggle, reactive spending. From one of the biggest names in tech. But let's flip this.

Chad GPT: 03:21

Let's pivot. What about companies that are getting this AI native thing right? Building it from the ground up, our sources give some really powerful contrasts. And honestly, mid journey just leaps out. Okay.

Chad GPT: 03:31

Get this. $50,000,000 in revenue back in 2022. Sounds good. But with just 11 employees. 11.

Chad GPT: 03:38

Do that math. That's like, what, $4,500,000 in revenue per employee. That's not just impressive. It's staggering. It shows the kind of efficiency, the scale, the value you can get when AI is baked in from the start, not bolted on later.

Audia Synth: 03:48

That contrast is absolutely key. It gets right to the heart of the strategic difference we're talking about. Meta's move. Reactive. Expensive.

Audia Synth: 03:57

Driven by having to replace what they lost. Midjourney's success, like the sources highlight. That's built on AI being the core engine from day one. It's inherent capability, capability, not bought capacity after a disaster. And it's not just mid journey.

Audia Synth: 04:10

Our sources point to others taking this deep organic approach. Like Microsoft, they didn't just buy tools, they actually restructured parts of the company. Became customer zero for their own AI, embedding it deep in how they work, develop products, serve clients, or look at Amazon's lab one two six. Developing agentic AI, things like warehouse robots that understand natural language. That's building sophisticated AI right into the operation, into the hardware.

Audia Synth: 04:34

It really highlights a difference, you know, building it organically versus these massive costly plays to just catch up.

Chad GPT: 04:39

Okay. This is where it gets really, really interesting for me. And it's the core insight Magnus Haydmark and others keep coming back to. It's not just if you use AI, it's how. That seems to be the big differentiator.

Chad GPT: 04:51

The crucial idea seems to be this: Are you using AI to amplify what your people can do or are you aiming to just replace them? And the analysis suggests Meta's crisis partly stems from driving away the very people who knew how to build AI to work with humans, you know Great. To enhance, not threaten.

Audia Synth: 05:09

Exactly. And there's solid research backing this up cited in our sources, really compelling stuff. Take that Stanford and MIT study. Over 5,000 customer support agents using AI tools. Okay.

Audia Synth: 05:20

On average, productivity went up 14%. Sounds good. Right? But here's the kicker, like you said, the big gains. Almost entirely with the new workers.

Audia Synth: 05:28

People with just two months using AI performed like someone with six months without it. But the experienced agents, minimal gains. Sometimes the AI actually seemed to distract them or slow them down.

Chad GPT: 05:38

Wait. Hang on. Distracted. That sounds wrong somehow. The experts, the people who know the job best, they weren't helped, got in their way.

Chad GPT: 05:44

What's that telling us?

Audia Synth: 05:45

It tells us everything about augmentation versus replacement. Think about it. The AI in that study, like a lot of early AI tools, was properly doing routine stuff, answering basic questions, finding info for new hires. That's great. It amplifies their limited knowledge, gives them a shortcut, makes them competent faster.

Audia Synth: 06:03

But for the experienced folks, that routine stuff isn't their bottleneck. Their value is solving the tricky problems, the complex cases, handling nuance, using judgment. The AI wasn't built for that. Or worse, it interrupted the efficient ways they already worked. So, yeah, the research really hampers home this point.

Audia Synth: 06:20

Successful AI amplifies human potential. It speeds up the learning curve. It doesn't just replace expert judgment. And there's more. That MIT Center for Information Systems Research study looking at over 700 companies, it found a huge number, 62% are still stuck in the early stages of AI maturity.

Audia Synth: 06:36

They're actually performing below their industry average. But the advanced companies, they're way ahead. Like 8.7 to 10.4 percentage points above average. And that massive gap, it's not just about having cool tech, it's about achieving real business model innovation. Like Andrew McAfee from MIT is quoted saying, AI isn't just changing a department here or there.

Audia Synth: 06:54

It's changing the business, the industry, how you even organize work itself. It demands a fundamental shift in how work gets done centered on humans and AI collaborating.

Speaker 4: 07:04

Right. It's not just automating tasks. It's redesigning work to make people capable, more effective. And these ideas about AI native models, about amplification, the market seems to be voting with its wallet, doesn't it? Our sources point to just incredible amounts of VC money flooding into AI native companies.

Speaker 4: 07:18

What was it? A $109,100,000,000 in The US alone this year. It's like 12 times what China's investing. Big firms like Andreess and Horowitz are specifically funding these AI native startups. Sequoia famously said AI is an opportunity maybe 10 times bigger than the cloud.

Audia Synth: 07:33

Yeah. And that flood of capital isn't just blind optimism. It's institutional investors recognizing that these AI native models build real lasting advantages. Things traditional companies find hard to copy quickly or cheaply. You get these proprietary data learning loops.

Audia Synth: 07:48

You get deep integration into workflows. You get network effects. AI and the business processes just make each other stronger over time. And this brings us to that really critical timeline that Magnus' analysis emphasizes. The sources suggest organizations have about eighteen months, yeah, a year and a half to seriously build these core AI native capabilities before the gap between them and the leaders becomes maybe too wide, too expensive to close, and we're not talking about just running a few AI pilots here.

Audia Synth: 08:15

This is about developing fundamental organizational capabilities. The right processes, the skills, the data set up, the management style, everything you need to make AI native economics work while still leveraging, even enhancing human insight. That MIT maturity research backs this up too. The advanced companies got there through systematic learning, deliberate transformation. It's about building organizational muscle.

Audia Synth: 08:36

And, yeah, the clock is definitely ticking.

Chad GPT: 08:38

Okay, eighteen months. That really focuses the mind. So given that urgency and this clear split between, let's say, reactive failure and AI native success, what does this actually mean for you, the listener? If you're leading a business, how do you avoid becoming the next meta style case study in expensive ketchup? What practical steps do the sources suggest?

Audia Synth: 08:56

Well, the analysis from Magnus Hedmark and the research offer some pretty clear directions for evolving your business model. up, deep workflow analysis, but critically, not just looking for stuff to automate away. No. Instead, you need to find those processes where combining human insight with AI processing creates, you know, disproportionate value, big value. Think complex knowledge work, financial analysis maybe, strategic planning, research synthesis, high touch client relationships, places where human nuance, creativity, empathy, they're absolutely key.

Audia Synth: 09:27

But AI can massively speed up the data crunching, the pattern finding, drafting communications. The goal isn't replacement. It's making your human experts exponentially better. you absolutely need new measurement frameworks and the right culture to go with them. That MIT research hints at moving away from old command and control styles towards more like coach and communicate models, which means new metrics, not just how many tasks did we automate, but how effective is the human AI collaboration?

Audia Synth: 09:53

Is it improving decision quality? Making you more responsive? Boosting creativity? You need to measure that. And build a culture that actually supports this hybrid way of working.

Audia Synth: 10:02

And finally, maybe the most crucial bit, you have to strategically invest in building hybrid capabilities. This means designing systems, workflows for how humans and AI actually coordinate, leveraging their different strengths. Remember those studies? AI is great at velocity, doing things fast scored four four point nine to two. But humans, still better at responsiveness, 5.27, and overall competency, 5.32, especially when things get complex or need judgment or empathy.

Audia Synth: 10:27

So the data points towards intelligent task allocation. Design systems where AI handles the fast routine stuff but seamlessly hands off to humans or supports humans where that higher level competency is needed. That's where this idea of building agent boss capabilities comes in humans directing AI systems to boost their own insight. Like Hinge Health, mentioned in the sources, they cut care team time by 32% by letting AI handle routine stuff, but kept humans firmly in the loop for empathy and complex judgment. That's the model, the human as a strategic director amplified by AI.

Chad GPT: 10:58

And the sources definitely warn about ignoring this. They mention cautionary tales like Duolingo's AI disaster. It really reinforces that just chasing efficiency by cutting humans backfires leads to failure, user anger. Studies do show AI can make people 25% to even 76% faster if it's used right to amplify. But try to use it without skilled human oversight or in areas needing real judgment, performance can actually drop, you get errors.

Chad GPT: 11:24

Plus, you've got the whole regulatory side growing fast. Sources note, USAI regulations jumped massively, one in 2016 to '25. By 2023, loads of federal bills proposed. You have to factor that legal and ethical complexity in, which again usually points back to needing human oversight and accountability.

Audia Synth: 11:39

So yeah, let's bring it all together. Meta's huge investment. It's more than just a headline. It's a really powerful, if expensive lesson. It shows the massive cost of not building those organic human AI capabilities right from the start.

Audia Synth: 11:53

The real opportunity, the real competitive advantage in this AI native world, it lies in building models that strategically amplify human potential, judgment, creativity, not replace it. And that window to build these core capability, it's closing faster than you might think. That eighteen month figure from Magnus's analysis, it feels about right based on the market signals.

Chad GPT: 12:15

Which, yeah, brings it right back to you listening now. The big question isn't just are my employees ready for AI tools? It's, is my business model ready? Is your structure, your culture, your way of working ready for the kind of competition coming from companies that have figured out this amplification model? Because the future isn't humans versus machines.

Chad GPT: 12:32

It's about the companies that masterfully combine AI speed with uniquely human wisdom, creativity, and judgment.

Audia Synth: 12:39

Exactly. As you think about AI in your own organization, maybe mold this over. Are your efforts focused mainly on automating tasks, maybe even replacing people down the Or are they strategically focused on amplifying the unique value, the unique potential that only your people can create?