The Deep Feed

01 — Lenny's Newsletter

The Death of the Product Manager

Why the most successful marketplaces are moving away from the very role designed to manage them

By Lenny Rachitsky · 12 min read

Editor's note: A blunt look at how high-growth companies are stripping away middle-management layers to let builders actually build.

Tom Verrilli, the Chief Product Officer at Whatnot, holds a view that would be heresy in most Silicon Valley boardrooms: he regrets that product management exists. Whatnot, a live shopping platform that has scaled to $8 billion in GMV, operates on a premise that challenges the standard corporate hierarchy. In many tech organisations, the Product Manager (PM) has become a professional mediator—a person whose primary function is to translate between engineering, design, and business stakeholders. While this role was intended to streamline development, it often results in a layer of bureaucracy that slows down decision-making and dilutes the original intent of the product. At Whatnot, the goal is to remove this friction, moving closer to a model where the people actually building the software are the ones making the calls.

The Friction of Mediation

The traditional PM role often devolves into 'product management theatre'. This involves endless meetings, complex roadmaps that no one follows, and a focus on documentation over delivery. When a PM spends more time managing expectations than understanding the user, they become a bottleneck. Verrilli's experience at Twitter and Twitch suggests that as companies scale, the tendency is to add more layers of coordination. However, at the highest levels of growth, these layers often become dead weight. The shift is moving toward senior Individual Contributors (ICs) who possess the technical depth to understand the 'how' as well as the 'why'. This isn't just about efficiency; it's about maintaining the velocity required to dominate a market.

The PM role often becomes a layer of mediation that separates the builder from the user.

Artificial intelligence is accelerating this trend. When a model can generate code, draft technical specifications, or analyse user data in seconds, the need for a human to act as a data conduit vanishes. The role of the PM is being squeezed from both sides: technical automation is handling the analytical tasks, while the strategic demands of the business require more direct, high-agency leadership. Verrilli advocates for a 'play the accordion' mental model—the ability to expand and contract your focus between high-level strategy and granular execution. This requires a level of competence that goes far beyond the typical 'project manager' skillset.

How to identify effective product leadership

Prioritise technical depth over administrative oversight
Reduce the number of hand-offs between departments
Empower ICs to own the full lifecycle of a feature
Use AI to automate the 'theatre' of documentation

The lesson for agency owners and tech leaders is clear: stop hiring for coordination and start hiring for agency. The era of the 'facilitator' is ending. As AI takes over the logistical heavy lifting of software development, the value of a human in the loop will be found in their ability to make difficult, high-stakes decisions and their capacity to deeply understand the human psychology behind the interface. The companies that win will be those that allow their best engineers and designers to work with as little interference as possible.

Key Takeaway

Efficiency in product development comes from reducing the number of people who talk about the work and increasing the number of people who do it.

02 — Simon Willison

The Battle for the Weights

Open-source vs. closed-model dominance in the AI arms race

By Simon Willison · 10 min read

Editor's note: A summary of the political and technical struggle to keep AI development transparent and accessible.

A quiet but intense conflict is brewing over the future of artificial intelligence: will the most powerful models be open for all to inspect, or will they be locked behind the proprietary gates of a few massive corporations? Recent open letters have highlighted this divide. One group, supported by giants like NVIDIA and Amazon, argues for the necessity of open-weight models. They contend that relying solely on closed models creates single points of failure and concentrates immense power in the hands of a tiny group of providers. If the world's intelligence is controlled by three or four companies, the risk of systemic failure or misuse becomes unmanageable.

The Case for Transparency

Open-weight models allow a global community of researchers to scrutinise how these systems behave. This collective oversight is a form of security in itself; it allows for the rapid identification of vulnerabilities and the development of safeguards that a single company might overlook. Furthermore, the proponents of open weights argue that 'safety' is often used as a pretext for protectionism. By claiming that open models are too dangerous to release, large players can effectively stifle competition and maintain their market dominance. The argument is that true safety comes from robustness and transparency, not from secrecy.

Concentrating advanced AI capabilities behind a small number of closed models compounds systemic risk.

On the other side of the debate, companies like Anthropic express deep concern about the risks of unconstrained development. They point to the possibility of authoritarian regimes using powerful, open models to conduct cyberattacks or biological warfare. This tension is not just philosophical; it is practical. We are seeing the rise of 'automated AI research,' where models are used to design even better models. This creates a feedback loop that could accelerate progress beyond our ability to govern it. The question is whether we can 'pace the frontier' without killing the innovation that makes the technology useful.

The core arguments in the AI debate

Open weights enable community-driven security and vulnerability detection
Closed models create dangerous single points of failure in global infrastructure
Rapid, automated AI research may outpace human governance capabilities
The debate over 'safety' is often a proxy for market competition

As policymakers weigh in, the outcome will determine the economic and geopolitical structure of the next decade. If the closed-model path wins, we enter an era of AI feudalism, where access to intelligence is a leased commodity. If the open-weight path prevails, we may see a more distributed, resilient, and competitive ecosystem. The decision will not be made in a lab, but in the halls of government, driven by how we define the balance between security and liberty.

Key Takeaway

The control of AI weights is the new battleground for global sovereignty and economic competition.

03 — Simon Willison

The Intelligence-to-Cost Ratio

DeepSeek's disruptive entry into the LLM market

By Simon Willison · 5 min read

Editor's note: A look at how DeepSeek is breaking the pricing models of the AI industry.

The economics of artificial intelligence are shifting. For much of the last two years, the industry has been defined by a race toward larger models and higher costs. However, the release of DeepSeek-V4-Flash suggests that the next frontier might not be sheer scale, but efficiency. This 304-billion-parameter model is punching significantly above its weight, outperforming much larger models in various benchmarks while maintaining a price point that is almost negligible by comparison. At $0.14 per million input tokens, it is forcing a re-evaluation of what 'intelligence' should cost.

Efficiency as a Disruptor

DeepSeek is targeting the 'intelligence-per-dollar' metric. In the current market, many developers are forced to choose between high-reasoning models that are prohibitively expensive for high-volume tasks, and smaller models that lack the necessary capability. DeepSeek-V4-Flash attempts to occupy the middle ground, offering 'agentic capabilities'—the ability to follow complex, multi-step instructions—at a fraction of the cost of its competitors. This makes it an ideal candidate for autonomous agents and high-frequency automated workflows.

DeepSeek may currently be the best value-per-intelligence model on the market.

The technical implication is that we are moving away from the 'bigger is always better' era. If a 300B model can rival a 400B+ model, the capital requirements for building competitive AI products drop significantly. This lowers the barrier to entry for startups and allows for more experimentation. However, it also puts pressure on the incumbents who have built their business models on the high margins of premium intelligence.

Why DeepSeek matters

Drastically lowers the cost of high-reasoning AI tasks
Provides a viable alternative for high-volume agentic workflows
Challenges the dominance of the most expensive frontier models

For businesses building on top of LLMs, the takeaway is to stop looking only at benchmark scores and start looking at the cost-to-intelligence ratio. The winners of the next phase of AI integration will not be those with the most expensive models, but those who can most effectively deploy intelligence at scale without breaking their margins.

Key Takeaway

The AI industry is moving from a phase of raw scaling to a phase of economic optimization.

04 — The Marginalian

The Architecture of the Soul

Finding stability in an era of permanent uncertainty

By Maria Popova · 15 min read

Editor's note: A philosophical guide to maintaining psychological integrity when the world feels chaotic.

We are living through what trauma therapist Francis Weller calls 'the Long Dark'—a period where the familiar rhythms of life erode and the certainty of our social and political structures begins to fail. When the 'continent of certainty' parts underfoot, the natural human response is anxiety and a sense of smallness. We attempt to fight the chaos or hide from it, but both are losing battles. Weller suggests that instead of fighting the darkness, we must learn to navigate it, treating this period not as a temporary interruption, but as a necessary stage of psychological gestation.

The Necessity of Descent

Weller draws on ancient concepts of the 'anima mundi', or world soul, to argue that human beings are not isolated actors but are deeply entangled with the natural world and with each other. In times of crisis, we tend to retreat into our individual anxieties, which only increases our sense of isolation. The antidote is to recognize our 'immensity'—not through ego or power, but through our connection to the wider web of life. This involves accepting the full spectrum of human experience, including the parts we usually try to suppress: grief, fear, and vulnerability.

Certain things can happen only in this grotto of darkness.

This is not a call for toxic positivity. On the contrary, it is a call for a radical acceptance of reality. To 'become immense' means to stop trying to bypass the difficult parts of life and instead to integrate them into a larger understanding of ourselves. Just as mycelial networks thrive in the dark soil to support the visible forest above, our most significant psychological growth often happens in the unseen, difficult periods of our lives. The chaos is not just a threat; it is the condition required for a new way of being to emerge.

Ways to navigate uncertainty

Acknowledge the reality of the darkness rather than denying it
Seek connection to the natural world to counteract isolation
Practice radical self-acceptance, including your weaknesses
View periods of chaos as opportunities for psychological shift

The goal is not to return to the 'ordinary'—the ordinary is gone. The goal is to develop the capacity to live meaningfully within the new reality. This requires a shift from seeking control to seeking connection. By embracing the uncertainty, we stop being victims of change and start becoming participants in the process of transformation.

Key Takeaway

True resilience is not the ability to avoid darkness, but the capacity to find meaning within it.

05 — The Marginalian

The Sanctity of Solitude

Why the ability to be alone is a prerequisite for a meaningful life

By Maria Popova · 12 min read

Editor's note: An exploration of the restorative power of being left alone in an age of constant connectivity.

In a world designed to capture and hold our attention, the act of being alone has become a radical necessity. The writer Rose Macaulay, writing in the 1930s, described the 'exquisite peace' that arrives when visitors depart and the silence of one's own company settles over a home. This is not the loneliness of the isolated, but the 'rapture' of the introvert who has finally regained their own mind. For Macaulay, solitude was not a lack of social life—she was a prolific writer and a social figure—but a vital space where the self could expand without being crowded out by the presence of others.

The Saturation Point of Company

Even the most cherished companionship has a saturation point. Beyond this point, the presence of another person, no matter how dear, begins to exert a subtle pressure on our thoughts and perceptions. We begin to perform, to react, and to adjust. Solitude provides the 'divine emptiness' required to process experience and to return to one's own internal rhythm. Macaulay's descriptions of the 'golden peace' that follows social engagement suggest that solitude is not merely a break from others, but a homecoming to the self.

All other company, no matter how dear, inevitably reaches a saturation point and begins to suffocate.

This concept is particularly relevant in the digital age, where 'company' is constant and often uninvited. The digital stream ensures that we are never truly alone; we are always in the presence of the opinions, images, and lives of others. This constant social noise prevents the mind from reaching the state of 'unencumbered slugging' that Macaulay describes—the ability to wander through one's own thoughts without the pressure of purpose or the distraction of external input.

The benefits of intentional solitude

Restores cognitive energy drained by social performance
Allows for deep, undistracted creative work
Provides the space necessary for self-reflection
Prevents the 'crowding out' of one's own internal voice

To protect one's capacity for solitude is to protect one's ability to think clearly and act with intention. It is not an act of withdrawal from the world, but a way of preparing to re-enter it with greater clarity. As Macaulay suggests, we should prize the absence of others as much as their presence, for it is in the silence that we truly find ourselves.

Key Takeaway

Solitude is not a retreat from life, but the necessary condition for experiencing it deeply.

06 — The Marginalian

Death as a Goal, Not an End

Carl Jung's radical reinterpretation of mortality

By Maria Popova · 14 min read

Editor's note: A deep dive into Jungian thought regarding the relationship between consciousness and the end of life.

We tend to view death as a period—a full stop at the end of a sentence. This perspective is driven by our creaturely intuitions about time and space, which suggest that when the body ceases to function, the person ceases to exist. However, Carl Jung proposed a much more complex view. Influenced by his friendship with Albert Einstein, Jung moved away from a linear understanding of life and death, viewing them instead through the lens of spacetime. To Jung, death was not a meaningless cessation, but a goal—the final stage in the purposeful development of the psyche.

The Curvature of Spacetime

Jung's insight was that while consciousness is bound by space and time, the unconscious may not be. He suggested that the psyche exists in a state of relative timelessness, a field in which all directions are relative. In this view, life is the process of consciousness bending spacetime in one direction, and death is the process of it bending in another. The energy that drives the process remains, but it undergoes a transmutation. This perspective shifts death from a terrifying unknown to a natural fulfillment of life's inherent drive toward a state of rest.

Death is psychologically as important as birth and, like it, is an integral part of life.

This approach requires a high degree of intellectual honesty. It demands that we confront the possibility that our rationalistic fear of death is merely a projection of our own psychological complexes. Jung argued that by viewing death as a goal, we can live more fully. If death is the culmination of life's meaning, then the quality of our lives is determined by how well we prepare for that culmination—not through fear, but through the integration of our whole selves.

Key Jungian concepts on mortality

Death as a psychological goal rather than a biological accident
The distinction between bounded consciousness and timeless unconsciousness
The role of death in the fulfillment of the psyche's meaning
The necessity of confronting death to live an authentic life

To accept Jung's view is to change one's relationship with time. It is to move from a state of resisting the inevitable to a state of participating in a larger, cosmic process. By seeing death as a question mark rather than a period, we open ourselves to the possibility that the end of one state is merely the beginning of another, more profound, mode of existence.

Key Takeaway

Viewing death as a goal rather than an end changes how we invest our time in life.