The Deep Feed

01 — Not Boring

The Economics of the Unrepeatable

Why venture capital is pivoting from code to icons

By Packy McCormick · 10 min read

Editor's note: As AI makes software infinitely replicable, the value of physical and cultural scarcity is skyrocketing.

The venture capital industry is undergoing a quiet, expensive transformation. For decades, the goal was simple: find a scalable software solution, build a moat around it, and watch the margins expand. But software has a problem. It is increasingly easy to replicate. When a new model can write marketing copy for pennies, the value of the code itself begins to erode. This is driving a shift toward 'Scarce Assets'—things that cannot be coded, cloned, or scaled through a server farm. We are seeing smart money move away from the digital ether and toward the bedrock of human culture: the sports team, the historical institution, the physical landmark.

The Lakers Premium

Consider Thrive Capital’s recent move with Thrive Eternal. They aren't just looking for the next SaaS unicorn; they are hunting for iconic franchises. The San Francisco Giants are a start, but the real prize is the Los Angeles Lakers. A $12.5 billion price tag for an NBA team isn't just a sports transaction; it is a hedge against the infinite. You can build a thousand new AI startups, but you cannot build a new Lakers. You cannot manufacture the history, the tension, or the cultural gravity of a courtside seat. In a world of digital abundance, the only way to maintain high margins is to own something that is fundamentally finite.

People keep getting richer, but they stay the same amount. The scarcity of culture is the new frontier of capital.

This scarcity extends to the physical world, where heavy-lift drones are beginning to rewrite the rules of logistics. DARPA’s recent Heavy Lift Challenge saw the Katana uncrewed aircraft carry nearly four times its own weight. While the immediate application is military, the long-term consequence is the decentralisation of heavy industry. If we can move massive payloads across difficult terrain for a fraction of current costs, we change how we build, how we rescue, and how we supply. It is another form of breaking a physical constraint, much like how software broke the constraint of information.

The New Asset Classes

Cultural Icons: Sports teams and historical institutions
Physical Infrastructure: Heavy-lift autonomous systems
Environmental Control: Weather modification and seeding

Even the weather is becoming a sector for differentiation. Companies like Recast Systems are moving beyond simple observation into active intervention. By using AI-driven models and efficient seeding materials, they are attempting to direct the sky. This is the ultimate high-stakes game: moving from reacting to the environment to actively managing it. It is a massive, expensive, and highly differentiated way to play the future.

Key Takeaway

As digital goods become infinitely reproducible, value migrates to the physical and cultural things that cannot be copied.

02 — Stratechery

The Capital Constraint

Why the AI revolution might stall on a pile of money

By Ben Thompson · 8 min read

Editor's note: The bottleneck for AI isn't just chips or electricity; it's the actual cash required to build the future.

The prevailing narrative around AI is a struggle for compute and power. We talk about H100 shortages and the massive energy requirements of data centres. But there is a third, more quiet constraint: capital. To build the next generation of frontier models, companies need hundreds of billions of dollars in upfront investment. This is a massive gamble. If the revenue from these models doesn't arrive fast enough to service the debt or satisfy the equity holders, the entire buildout could face a liquidity crisis.

Financial Engineering as a Bridge

We are seeing the emergence of new financial mechanisms to keep the train moving. Nvidia is exploring ways to help its customers tap long-duration capital, and Google is leading the charge in using equity to fund the massive CapEx requirements. This is essentially financial engineering designed to bridge the gap between today's massive spending and tomorrow's uncertain revenue. It works, for now, but it expands the blast radius of any potential bubble. If the AI revenue doesn't materialise, the fall will be much harder than a simple tech correction.

The risk isn't just that the models fail, but that the money runs out before they become profitable.

This tension is also playing out in the regulatory space. The EU's mandate for AI watermarking is a prime example of a policy trying to solve a technical problem with a blunt instrument. While intended to protect human creators, it creates a massive friction point for developers. It forces a choice: comply with a system that is easily bypassed, or fight a regulatory battle that could stifle innovation in the region. It is a classic case of policy lagging behind the speed of technological change.

The Three Pillars of AI Growth

Compute: The hardware required to train and run models
Power: The energy required to keep the lights on
Capital: The cash required to fund the massive upfront costs

Ultimately, the AI era will be defined by how these constraints are managed. If the capital flows continue, we will see a rapid expansion of capability. If the revenue fails to match the spend, we will see a massive consolidation where only the most capital-efficient players survive.

Key Takeaway

The AI revolution depends as much on sophisticated financial engineering as it does on silicon and electricity.

03 — The Marginalian

The Measure of a Life

Rejecting the shallow metrics of success

By Maria Popova · 12 min read

Editor's note: In an age of quantified self and public metrics, Arundhati Roy offers a different way to judge a life.

We live in an era of metrics. We track our steps, our followers, our net worth, and our productivity. We have been conditioned to believe that success is a trajectory—a line that should always point upward. But this belief is a trap. It assumes that wealth and fame are the mandatory components of a meaningful life. When we adopt these external templates, we risk turning ourselves into 'cold silver figurines'—polished, impressive, but hollow at the core.

The Dignity of Defeat

Arundhati Roy, the Booker Prize-winning author, provides a necessary corrective to this obsession. After the massive success of *The God of Small Things*, she found herself suffocated by the idea that her happiness had peaked because she had 'stumbled upon success.' She realised that the world’s definition of success—fame and fortune—is often at odds with the human need for depth and connection. She speaks of 'making friends with defeat' not as an act of surrender, but as a way to remain open to the world. To fail is to remain in the arena, rather than retreating to the safety of a trophy case.

Success is to love. To be loved. To never forget your own insignificance.

Roy points to her uncle, a Rhodes scholar who gave up a prestigious academic career to run a small jam factory with his mother. In the eyes of a metric-driven society, this is a regression. In the eyes of a life lived with intention, it is a triumph. He chose a life of tangible connection and simple purpose over the abstract prestige of academia. This is the difference between pursuing power and respecting strength.

Roy's Metrics for a Meaningful Life

Seek joy in the saddest places
Pursue beauty to its lair
Never simplify what is complicated
Respect strength, never power
Never look away

True success is not a destination or a number. It is a way of seeing. It is the ability to stay awake to the world, to feel its violence and its beauty without becoming numb. It is the refusal to let the templates of others dictate the shape of your own soul.

Key Takeaway

Success is not a metric to be achieved, but a capacity for connection and awareness to be maintained.

04 — Simon Willison

Hallucinating the New

Why LLMs should stop trying to classify and start imagining

By Simon Willison · 5 min read

Editor's note: The greatest strength of AI isn't its ability to sort existing data, but its ability to invent new categories.

Most people use Large Language Models (LLMs) as sophisticated filing clerks. We ask them to categorise text, tag emails, or sort data into existing buckets. This is a safe, conservative use of the technology. It treats the model as a tool for precision. But in doing so, we ignore the model's most interesting quality: its ability to hallucinate. In the context of classification, 'hallucination' is usually seen as a bug. But if we change the prompt, it becomes a feature.

The Vector Solution

If you try to force an LLM to pick from a list of 2,000 existing tags, it often fails. The context window is too crowded, and the model gets bogged down in the specifics of the existing vocabulary. A better approach is to let the model 'hallucinate' its own tags. Tell it to imagine the perfect labels for a piece of content, without any regard for what labels already exist. This allows the model to use its full creative reasoning to describe the essence of the data.

Don't force the model to fit your world. Let the model describe the world as it sees it.

Once the model has generated these 'imagined' tags, you can use vector embeddings to map them back to your actual, concrete taxonomy. This creates a bridge between the model's fluid understanding and your rigid database. You aren't just sorting; you are discovering how your data actually relates to the concepts it represents.

The New Workflow

Prompt the LLM to generate novel, unconstrained labels
Use the model's 'hallucinations' to capture semantic essence
Map these labels to existing tags using vector similarity

This shifts the role of the AI from a clerk to a cartographer. It doesn't just find where a thing fits; it helps define the boundaries of the territory.

Key Takeaway

The real value of AI lies in its ability to generate new conceptual frameworks, not just navigate old ones.

05 — Stratechery

The Network Effect of Cities

Why New York endures while Hollywood migrates

By Andrew Sharp · 7 min read

Editor's note: A study in how geography and density create moats that even political rhetoric cannot easily break.

There is a growing debate about the death of the great city. Critics point to rising costs, political volatility, and the rise of remote work as evidence that the urban centre is obsolete. But this view misses the fundamental power of the network effect. A city is not just a collection of buildings; it is a dense web of human interactions, specialized services, and spontaneous collaborations. This density creates a moat that is incredibly difficult to cross.

The New York Resilience

New York City serves as a primary case study. Despite caustic rhetoric from local politicians that might drive away some residents, the city's core industries remain anchored. Why? Because the value of being in New York is not found in the politics, but in the proximity. To be in finance, media, or law in New York is to be at the centre of a high-frequency human network. The cost of leaving is not just the cost of a new apartment; it is the cost of losing the serendipity that drives professional growth.

Networks are harder to move than people. You can relocate a worker, but you cannot easily relocate a network.

Hollywood, by contrast, offers a cautionary tale of how these effects can be broken. We are seeing the filmmaking industry being outsourced to neighbouring states and foreign countries. When the industry becomes modular—when production can be unbundled from the creative hub—the network effect weakens. If you can achieve the same result by moving the 'work' part of the work elsewhere, the geographic moat evaporates.

Conditions for Urban Survival

High-frequency serendipity (unplanned interactions)
Specialised talent density
Low modularity of core tasks

The lesson for business leaders is clear: if your value is derived from being part of a dense, interconnected ecosystem, your geography is your greatest asset. If your value is modular and can be unbundled, you are vulnerable to migration.

Key Takeaway

Geography is a moat only as long as the work being done cannot be unbundled from the people doing it.

06 — The Marginalian

The Twenty-Year Archive

Reflections on a life of curated wonder

By Maria Popova · 9 min read

Editor's note: A retrospective on how the pursuit of meaning can create a lasting digital legacy.

In 2006, it began as an email to seven colleagues. It was a search for meaning, a collection of book passages, neuroscience studies, and art. Twenty years later, *The Marginalian* has become a permanent part of the digital archive of the Library of Congress. This trajectory is rare in the attention economy. Most digital content is designed for the immediate, the shallow, and the ephemeral. To build something that lasts two decades requires a different kind of commitment: a commitment to depth over speed.

The One-Woman Labor of Love

The project has remained remarkably pure. No advertising, no AI, no assistants. It is a singular voice reckoning with the bewilderment of being alive. In an age where content is increasingly generated by algorithms to satisfy engagement metrics, a human-led, curiosity-driven project becomes an act of resistance. It proves that there is still a hunger for thought that is slow, considered, and deeply personal.

Wonder is a technology of the soul, used to prevail over the probable.

The upcoming celebration at the Saratoga Performing Arts Center—featuring poets, physicists, and musicians—is not just an event; it is a gathering of a community built on shared curiosity. It demonstrates that the 'fractal spread' of ideas can create real-world connections. When you offer something of value for free, you aren't just building an audience; you are building a constellation of people who value the same things you do.

Lessons from Two Decades of Curation

Depth is a sustainable competitive advantage
Curiosity is a community builder
Purity of intent attracts high-quality attention

The legacy of such work is not measured in clicks, but in the way it changes the reader's perspective. It is about providing the lens through which people can see the world more clearly.

Key Takeaway

In a world of instant gratification, the most enduring value is found in slow, deep, and human-centric inquiry.