In this episode, we use Moltbook as a case study for the real agent story: once an AI system is connected to tools—email, browser, files, automations—the risk profile changes completely. The critical issue is prompt injection: malicious instructions hidden in normal content that an agent reads and misinterprets as commands, because to an LLM, information and instruction are both just text.

We also cover the uncomfortable trade-off the industry keeps trying to dodge: the more useful an AI assistant becomes, the more access it needs—tokens, permissions, accounts—and the bigger the blast radius when it fails. Proposed defenses exist, but none are clean: training resistance, input filtering, permission restrictions, and approval layers all reduce risk while also reducing usefulness.

Bottom line: ignore the theater. Evaluate agents by one question—what can it touch? Because the real story isn’t whether models can think. It’s whether we’re willing to hand non-thinking systems the keys and call it productivity.

Some things read better than they sound—charts and data included in the written edition.

We break down:

• Why Tesla’s growing capital commitments to AI raise questions for shareholders
• How SpaceX’s acquisition of xAI reframes its long-anticipated IPO
• Whether “vertical integration” across Musk’s companies is strategy or financial improvisation
• The realism (and limits) of space-based data centers
• Why energy and compute—not funding—may be the true bottlenecks in the AI race
• How regulation, physics, and resource scarcity complicate the AI arms race narrative

This episode isn’t about hype or personality. It’s about capital allocation, structural constraints, and what happens when ambition runs into reality.

Some things read better than they sound—charts and data included in the written edition.

The discussion explores why the AI race is no longer about building the best large language models, but about scale, deployment, and infrastructure. We compare U.S. hyperscalers’ multi-hundred-billion-dollar AI investments with China’s efficiency-driven approach, including the market impact of DeepSeek’s R1 model.

The episode also looks at how AI is becoming political and economic infrastructure, the limits of brute-force scaling, and what OpenAI’s move toward advertising signals about monetization, trust, and the future of AI platforms.

For business leaders, investors, and technology executives, the episode ends with a critical question: in an AI-driven economy, are you shaping the system — or being shaped by it?

Some things read better than they sound—charts and data included in the written edition.

In this episode of The B., Ben looks at why companies are cutting headcount by design rather than necessity, and how AI is reshaping hiring decisions from the ground up. Entry-level roles are disappearing first, not because work is gone, but because it has become automatable, measurable, and cheaper to run through machines.

The episode explores how the same technologies driving workforce reductions are creating new constraints elsewhere—most notably in compute, energy, and infrastructure. As demand for AI scales, bottlenecks are shifting away from talent and toward power grids, chip supply, and physical capacity that cannot move at software speed.

From hiring policies at major companies to capital reallocation at firms like Meta, this conversation traces a broader structural shift: headcount is becoming a liability, infrastructure is becoming strategy, and decision speed is turning into a competitive moat.

The question is no longer whether organisations should become leaner—but how far they can go before efficiency starts to undermine resilience.

Some things read better than they sound—charts and data included in the written edition.

There’s a point in every market cycle when confidence starts to sound rehearsed. Not wrong—just overlearned. This week had that feel.

On the surface, everything looks decisive. AI companies raise money at valuations that would have sounded absurd two years ago. Platform owners defend their toll booths with fresh conviction. CEOs publish neat lists of what will matter next year, as if complexity still submits to enumeration. Everyone sounds certain. That’s usually the signal.

Look closer and the seams show. OpenAI isn’t challenging Apple head-on; it’s trying to change how software is consumed altogether. Private AI markets inflate faster than public ones, driven by belief more than cash flow. Leaders are urged to delegate more, even as accountability stretches past what delegation can realistically carry. The system keeps demanding leverage—human, capital, computational—without pricing the cost of sustaining it.

This edition isn’t about prediction. It’s about tension. Between platforms and creators. Between capital and patience. Between leadership theory and operational reality. The real question isn’t who wins. It’s who recognizes the constraint early enough to adapt.

As Heraclitus put it: πάντα ῥεῖ — everything flows.

— Ben

Don't forget to check and follow me on LinkedIn

Some things read better than they sound—charts and data included in the written edition.

There’s a comforting story we tell ourselves about technology: that it scales smoothly, obeys policy, and waits for permission. This week’s reality is less polite.

AI is pressing against systems that were never designed for it—power grids, capital structures, governance frameworks—all at once. Some regions built vast AI infrastructure that now sits idle. Others can’t expand fast enough because electricity, not talent or chips, is the constraint.

Efficiency is no longer the question. Readiness is.

This edition looks at AI not as a breakthrough, but as a pressure test—of infrastructure, governance, and leadership.

As Thucydides warned, “The strong do what they can, and the weak suffer what they must.”
 

Technology doesn’t change that rule. It just redraws the line.

Don't forget to check and follow me on LinkedIn

Some things read better than they sound—charts and data included in the written edition.

Year-end reflections tend to exaggerate what didn’t matter and sanitize what did. This one won’t try.

For the second year running, I’ve been operating from a different seat—commercial rather than advisory—across financial services and supply chains, after years in consulting and SME environments. Same markets. Different incentives. That shift changes what becomes visible.

From this side of the table, a few things stand out. Technical excellence without context has limited value. Clients rarely arrive with requirements anymore—only direction. And most failures trace back not to bad decisions, but to assumptions no one bothered to surface.

This edition reflects those observations. It looks at AI not as a tool, but as a system whose behavior may soon force uncomfortable questions. It looks at why infrastructure leaders are pulling critical technology closer instead of outsourcing it. And it looks at how capital itself is changing shape, as family offices quietly accumulate influence once reserved for institutions.

Before getting into it, a genuine thank you. This is the 28th edition of The B. The fact that you’re still reading suggests the signal-to-noise ratio has been acceptable—by today’s standards, that’s saying something.

We’re living in interesting times, which history reminds us is rarely meant as a compliment.

Let’s begin.

Don't forget to check and follow me on LinkedIn

PS: Charts, figures, and references are available in the written edition

Some things read better than they sound—charts and data included in the written edition.