When designing a consumer product, you should consider every tap by a user to be a miracle. The motivation to stop using a new app will always be stronger than to use it To signup & complete a profile on Gas, we got it down to 15 taps total—with no keyboard required at any stage

They also don’t realize that yea I could keep going through thier onboarding flow, but timewise it’s competing with me looking at more dancing/cat videos on Tik Tok, and that is some seriously compelling content to compete with!

There’s some weird cognitive bias where product creators vastly overestimate how likely people are to do or pay attention to something based on how much the product creator wants them to, and they vastly underestimate how lazy people are.

To put it simply: you can look at a screen and intuitively predict the percent of users who will convert to the next screen within a 10% margin of error.

A dumb PM will stare at an onboarding funnel for hours, fabricate problems, create months of work for engineers, and ultimately improve conversion by 5%. A smart PM will look at funnel in a minute, make a button bigger, and improve conversion by 20%.

Counter positioning is an avenue for defeating an incumbent who appears unassailable by conventional wisdom metrics of competitive strength

What does the word "should" have to do with it? It’s not a matter of permitting something or forbidding something. Let him suffer, if he’s sorry for his victim . . .

Suffering and pain are always mandatory for broad minds and deep hearts. Truly great people, it seems to me, should feel great sadness on this earth.

.. My thought for today is something which I found in Epicurus (yes, I actually make a practice of going over to the enemy’s camp – by way of reconnaissance, not as a deserter!), ...

the tendency of Stoicism was always to exalt man’s importance in the universe rather than to abase him before a higher authority

"it is no part of the business of philosophy to turn people into better persons. His tremendous faith in philosophy as a mistress was grounded on a belief that her end was the practical one of curing souls, of bringing peace and order to the feverish minds of men pursuing the wrong aims in life"

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.