the structure of the digital economy means most of our digital lives are designed to take advantage of this state. A substantial fraction of the world’s most brilliant, competent, and empathetic people, armed with near-unlimited capital and increasingly god-like computers, spend their lives serving Marl.

"""
Here’s what I’ve been able to piece together about the marginal user. Let’s call him Marl. The first thing you need to know about Marl is that he has the attention span of a goldfish on acid. Once Marl opens your app, you have about 1.3 seconds to catch his attention with a shiny image or triggering headline, otherwise he’ll swipe back to TikTok and never open your app again.

Marl’s tolerance for user interface complexity is zero. As far as you can tell he only has one working thumb, and the only thing that thumb can do is flick upwards in a repetitive, zombielike scrolling motion. As a product designer concerned about the wellbeing of your users, you might wonder - does Marl really want to be hate-reading articles for 6 hours every night? Is Marl okay? You might think to add a setting where Marl can enter his preferences about the content he sees: less politics, more sports, simple stuff like that. But Marl will never click through any of your hamburger menus, never change any setting to a non-default. You might think Marl just doesn’t know about the settings. You might think to make things more convenient for Marl, perhaps add a little “see less like this” button below a piece of content. Oh boy, are you ever wrong. This absolutely infuriates Marl. On the margin, the handful of pixels occupied by your well-intentioned little button replaced pixels that contained a triggering headline or a cute image of a puppy. Insufficiently stimulated, Marl throws a fit and swipes over to TikTok, never to return to your app. Your feature decreases DAUs in the A/B test. In the launch committee meeting, you mumble something about “user agency” as your VP looks at you with pity and scorn. Your button doesn’t get deployed. You don’t get your promotion. Your wife leaves you. Probably for Marl.
"""

"""
here is a well-established taboo against anthropomorphizing AI systems. This caution is often warranted: attributing human emotions to language models can lead to misplaced trust or over-attachment. But our findings suggest that there may also be risks from failing to apply some degree of anthropomorphic reasoning to models. As discussed above, when users interact with AI models, they are typically interacting with a character (Claude in our case) being played by the model, whose characteristics are derived from human archetypes. From this perspective, it is natural for models to have developed internal machinery to emulate human-like psychological characteristics, and for the character they play to make use of this machinery. To understand these models’ behavior, anthropomorphic reasoning is essential.

This doesn’t mean we should naively take a model’s verbal emotional expressions at face value, or draw any conclusions about the possibility of it having subjective experience. But it does mean that reasoning about models’ internal representations using the vocabulary of human psychology can be genuinely informative, and that not doing so comes with real costs. If we describe the model as acting “desperate,” we’re pointing at a specific, measurable pattern of neural activity with demonstrable, consequential behavioral effects. If we don’t apply some degree of anthropomorphic reasoning, we’re likely to miss, or fail to understand, important model behaviors. Anthropomorphic reasoning can also provide a useful baseline of comparison for understanding the ways in which models are not human-like, which has important consequences for AI alignment and safety.
"""

Post-training of Claude Sonnet 4.5 in particular led to increased activations of emotions like “broody,” “gloomy,” and “reflective,” and decreased activations of high-intensity emotions like “enthusiastic” or “exasperated.”

sounds about right, lol

"these representations can play a causal role in shaping model behavior—analogous in some ways to the role emotions play in human behavior—with impacts on task performance and decision-making"

"We analyzed the internal mechanisms of Claude Sonnet 4.5 and found emotion-related representations that shape its behavior. These correspond to specific patterns of artificial “neurons” which activate in situations—and promote behaviors—that the model has learned to associate with the concept of a particular emotion (e.g., “happy” or “afraid”). The patterns themselves are organized in a fashion that echoes human psychology, with more similar emotions corresponding to more similar representations. In contexts where you might expect a certain emotion to arise for a human, the corresponding representations are active. Note that none of this tells us whether language models actually feel anything or have subjective experiences.

But our key finding is that these representations are functional, in that they influence the model’s behavior in ways that matter"

Why you should use control flow instead of prompts for control flow

• The common pattern is stuffing branching logic into one big prompt: "if it's a complaint do this, if it's feedback do that." The better pattern is classifying the input first, then routing to smaller, focused prompts with fewer instructions and fewer possible actions.
• The core lesson: don't use prompts for control flow if you can use actual control flow. The if statement is powerful, and LLMs are very good at classification — so let them classify, then route deterministically.

The instruction budget

• Frontier LLMs can only follow about 150–200 instructions with good consistency. Beyond that, the model half-attends to all of them and you're rolling the dice. These numbers are from a 2024/2025 paper — likely slightly higher now.

For any change to happen in humans, two forces help:

• Gas (motivation)
  • The more motivated I’m for something, the more likely I’ll do something
• Brake (friction)
  • The smaller the delta b/w my current life and what new is expected from me (i.e. the less the friction), the more likely I’ll do something

For a behavior to get built into a habit, we also need:

• Right context and repetitions
  • Context can be a trigger or a location or visuals of it
  • Repetitions are how behaviors transform into habits

The Mantra

Strategy: a route to continuing Power in significant markets

I refer to this as The Mantra, since it provides an exhaustive characterization of the requirements of a strategy.

Potential Value = [Market Scale] * [Power]

This simple math confirms my strategy definition as an exhaustive statement of value. Moreover, it's normative as well. Fulfill the imperatives of "The Mantra" and you will create business value.

Conditions for Power

Power: the set of conditions creating the potential for persistent differential returns

Power is the core concept of Strategy and of this book, too. It is the Holy Grail of business—notoriously difficult to reach, but well worth your attention and study.

From the glossary: The set of conditions needed for persistent differential returns. Power requires both a Benefit, something that materially increases cash flow, and a Barrier, conditions such that all the value to the firm of the Benefit is not arbitraged out by competition.

The Definition of Strategy

Strategy: the study of the fundamental determinants of potential business value

The objective here is both positive—to reveal the foundations of business value—and normative—to guide businesspeople in their own value-creation efforts.

Statics and Dynamics Split

Following a line of reasoning common in Economics, Strategy can be usefully separated into two topics:

• Statics—i.e. "Being There": what makes Intel's microprocessor business so durably valuable?
• Dynamics—i.e. "Getting There": what developments yielded this attractive state of affairs in the first place?

These two form the core of the discipline of Strategy, and though interwoven, they lead to quite different, although highly complementary, lines of inquiry.

The Critical Distinction Between Statics and Dynamics

Here's the first important takeaway from our consideration of Dynamics: "getting there" (Dynamics) is completely different from "being there" (Statics). This is a distinction not only for academics but for practitioners as well. For example, in the early days of strategy consulting, the two were frequently conflated: a close study of Statics indicated that high relative market share led to attractive returns; this fed the instinct to gain market share (Dynamics), usually via aggressive pricing. Such policies usually did not create value, as competitors would push back until the cost of share gains typically outweighed their benefits.

In other words, to assess which journeys are worth taking, you must first understand which destinations are desirable. Fortunately the 7 Powers does exactly that: it maps the only seven worthwhile destinations.

Statics vs Dynamics Scope

Statics concerned itself only with Power and hence just the last two terms (s, market share, and m, differential margins); primarily, it focused on just one (m). In contrast, in a Dynamics context, a company can profoundly influence both the two market size terms (M0, the current market size, and g, the discounted growth factor). The creation of compelling value, for example, is joined at the hip to the creation of a market. In the lingo of economists: in Statics M0 and g are taken as exogenous, whereas in Dynamics they are endogenous.

"Simple, but not simplistic"

If not simple, then concepts cannot be easily retained for day-to-day reference - usefulness is lost. If simplistic, then you risk missing something crucial.

Chance only favors the prepared mind

This reality begs the question, "Can the intellectual discipline of Strategy make a difference in such adaptation?" After decades as a business advisor, active equity investor and teacher, my conclusion is, "Yes it can." But with this hard-won conclusion comes a caveat informed by Pasteur's well-known dictum: "Chance only favors the prepared mind." Strategy serves best not as an analytical redoubt, but rather in developing the "prepared mind" of those on the ground.

Definition of "Judgment"

From Reed Hastings' Foreword, quoting the Netflix culture deck:

Our first public "culture deck," released in August of 2009, identified nine highly valued behaviors. The first was "Judgment." As we elaborated:

• You make wise decisions … despite ambiguity
• You identify root causes and get beyond treating symptoms
• You think strategically and can articulate what you are and are not trying to do
• You smartly separate what must be done well now and what can be improved later

"As I noted earlier, the Barrier in Scale Economies comes from a follower's rational economic calculation (often learned) that, despite the attractive returns being earned by the leader, attack carries an unattractive payoff."

The core definition of Power:

We can only assume microprocessors possessed some sort of rare characteristics which materially improved cash flow, while simultaneously inhibiting competitive arbitrage. I refer to these as Power.

Power: the set of conditions creating the potential for persistent differential returns

The Benefit/Barrier granularity :

I can now build up the chart by providing granularity surrounding both the Benefit and the Barrier. With regard to the Benefit, cash flow is improved by (1) enhancing value (enabling higher pricing) and/or (2) lowering cost ceteris paribus. With regard to the Barrier, a competitor fails to arbitrage out the Benefit because (1) they are unable to, or (2) they can, but refrain from so doing because they expect the outcome to be economically unattractive.

The "always look to the Barrier first" advice

As I delineate the seven types of Power in the chapters to come, I will similarly describe their unique Benefit/Barrier combination. The Benefit conditions probably won't sound all too rare—they are met often in business. Indeed, every major cost-cutting initiative qualifies. The Barrier conditions, on the other hand, prove far rarer, a reality that merely proves the ubiquity of competitive arbitrage. As a strategist, then, my advice is, "Always look to the Barrier first." In Intel's case, the heart of its microprocessors strategy can be best understood not by sorting through the multiplicity of Intel's value improvements, but by deducing why decades of capable and committed competition failed to emulate or undermine those improvements.

"A productive way to more formally calibrate the intensity of the scale leader's power is to assess the economic leeway they have in balancing attractive returns with appropriate retaliatory behavior to maintain share. The greater this leeway, the more attractive the longer-term equilibrium is likely to be for the leader."

Life of a novel is not in its conception, but in its performance, which eludes summary.

Why the first printed books were designed to look handmade

• There was no mass-market consumer expecting a cheap $10.99 product. The book market was wealthy people accustomed to artisanal luxury goods. So printers made printed books imitate manuscripts — fonts mimicked handwritten scripts, and blank spaces were left for hand-illumination.
• The richest buyers didn't even want paper. Isabella d'Este ordered her books printed on vellum — animal skin — even when the rest of the print run was on paper. The technology was new but the expectations were old.
• The printed book had to pretend it wasn't mass-produced to find its first customers. The commodity market for cheap books didn't exist yet — distribution networks hadn't caught up, and the culture still valued books as luxury objects.

The print revolution and the digital revolution are similar shape — one technology, successive shockwaves

• The computer, personal computer, internet, cell phone, social media, and LLMs feel like ten separate tech revolutions. They're not. They're all deeper penetration of one revolution: the development of the computer.
• The printing press followed the exact same pattern. It arrived in 1450 and kept hitting Europe with successive shockwaves for 150 years — each one as disruptive as the last, each arriving roughly a decade or two apart.
• 1450s: the press itself. 1490s: printing becomes economically sustainable. 1510s: pamphlet distribution networks mature. Then newspapers. Then magazines. Each is a new application of one underlying technology finding new forms as it disseminates more deeply.
• The Reformation was enabled by pamphlets the way the Arab Spring was enabled by cell phones. A pamphlet could travel Wittenberg to London in seventeen days — impossible even a decade before.
• The timescales are surprisingly similar. Print kept reshaping Europe for 150 years. We're roughly 80 years into the computer revolution and still getting hit by successive waves.

"Suddenly it's possible to get a printed pamphlet from Wittenberg to London in seventeen days. Oh my God, we can coordinate our resistance movement against the Catholics. Boom. The Reformation happens. That wasn't possible even a decade earlier when it took months to get a pamphlet from one end of Europe to the other."