The Deep Feed

01 — Lenny's Newsletter

The 15-Step Reality of the Big Deal

Why your CRM is lying to you about how enterprise sales actually works

By Lenny Rachitsky · 12 min read

Editor's note: Most sales leaders mistake a forecasting tool for a strategy. This is the actual mechanics of high-stakes closing.

Most sales professionals operate under a comfortable delusion. They look at their CRM, see five stages—Discovery, Qualification, Proposal, Negotiation, Closed—and believe they are managing a process. They are not. They are merely recording a history of guesses. For deals north of $100,000, the standard five-stage pipeline is a blunt instrument used for forecasting, not a map for winning. It tells you where the money might be, but it tells you nothing about how to actually get it. Real enterprise sales is a messy, non-linear sequence of tactical maneuvers that requires more than just a good pitch; it requires a deep understanding of organizational psychology and power structures.

The Pincer Movement

Winning a major contract requires attacking the organization from two directions at once. Jen Abel calls this the 'pincer model.' You cannot simply target the decision-maker at the top; by the time you reach them, the ground has already been won or lost by the people they trust. You must simultaneously engage the executive level and the 'N-minus-one' level—the managers and directors who will actually live with your product. If the executive wants it but the team hates it, the deal dies in procurement. If the team loves it but the executive sees no strategic value, the budget never clears. You are running two parallel campaigns that must eventually converge.

The standard CRM pipeline is a forecasting tool, not a sales process.

The first contact is rarely about a demo. In fact, showing a demo too early is a common amateur error that kills momentum. Instead, the goal of the initial engagement is intelligence extraction. You are looking for 'alpha'—the specific, high-value information that allows you to tailor your approach. You aren't selling features; you are selling a solution to a problem the client might not have even articulated yet. This requires a shift from being a vendor to being a consultant who understands their internal politics, their technical debt, and their specific definition of success.

The Enterprise Checklist

Identify the economic buyer and the technical gatekeeper
Run an intelligence-gathering call before any demo
Define success metrics jointly with the client
Structure pilots to be short, intense, and measurable
Prepare for the redline negotiation before the contract is sent

Once you move into the pilot phase, the stakes change. A pilot should not be a free trial; it should be a structured proof of value. If you give it away for nothing, you signal that your product has no inherent worth. A well-structured pilot has a clear beginning, middle, and end, with pre-agreed success criteria. If those criteria are met, the transition to a full contract should be a formality, not a new negotiation. The final hurdle—procurement and legal—is where many deals go to die. You must anticipate the redlines and the security reviews long before the final signature is requested.

Key Takeaway

Enterprise sales is a game of multi-level influence, not a linear progression of stages.

02 — Simon Willison

The High Cost of Intelligence

Why Anthropic's best models are losing the volume war

By Simon Willison · 8 min read

Editor's note: As AI models become commodities, the gap between 'the best' and 'the cheapest' becomes a chasm.

In the race for artificial intelligence supremacy, there is a growing tension between raw capability and economic viability. Anthropic's latest flagship models are undeniably powerful, yet they are struggling to capture the mass market. While their annualized revenue is climbing—reaching $65bn in July—the actual adoption of their most advanced models tells a different story. The market is bifurcating. On one side, you have a small group of high-spending enterprise customers; on the other, a massive sea of users opting for cheaper, 'good enough' alternatives. The premium for intelligence is becoming harder to justify for the average developer or business.

The Economics of Adoption

Data from the Ramp AI index reveals a stark reality: users are gravitating toward models that offer the best balance of speed and cost. Anthropic's Opus 5, despite its advanced reasoning, holds only a tiny fraction of the market spend compared to older, more efficient models. This isn't because the newer models are bad; it's because the marginal utility of the extra intelligence doesn't outweigh the increased cost for most tasks. For a company running millions of API calls a day, the difference between a model that is 5% smarter and one that is 50% cheaper is the difference between profit and loss.

The market is deciding that 'good enough' is better than 'perfect but expensive'.

OpenAI is navigating this same terrain, seeing massive revenue jumps following the release of GPT 5.6. The pattern is clear: the winners in the AI era won't just be those with the smartest models, but those who can integrate that intelligence into a cost-effective workflow. We are moving away from the era of 'model worship' and into the era of 'model orchestration,' where the goal is to use the cheapest possible model that can still complete the task. This shift places immense pressure on companies like Anthropic to prove that their premium tier offers more than just a slight edge in reasoning.

Market Adoption Breakdown (July 2026)

Opus 4.8: 28.0% (The reliable workhorse)
Sonnet 4.6: 8.3% (The middle ground)
Fable 5: 8.0% (The high-cost outlier)
Opus 5: 3.5% (The struggling flagship)

This economic reality forces a rethink of AI development. If the most advanced models are relegated to a niche of high-value users, the massive R&D costs required to build them become harder to recoup. The industry is approaching a plateau where the cost of incremental intelligence gains may soon exceed the market's willingness to pay for them. For agency owners and tech leaders, the strategy is no longer about finding the smartest model, but about building the most efficient pipeline of models.

Key Takeaway

In the AI economy, efficiency is a more powerful driver of adoption than raw intelligence.

03 — The Marginalian

The Tenacity of Old Blue

How one bird saved a species from the brink

By Maria Popova · 10 min read

Editor's note: A lesson in how extreme biological fragility can be overcome by a single, persistent force.

The history of the black robin is a study in near-total collapse. Introduced to New Zealand, these small, yellow-soled birds were decimated by invasive cats and rats. Within a century, the population had plummeted to just seven survivors. This wasn't just a decline; it was an extinction event in slow motion. The species, which evolved in a world without mammalian predators, had no defense against the sudden, violent introduction of new threats. When a species reaches such a low number, the math of survival becomes almost impossible. Genetic diversity vanishes, and the simple act of finding a mate becomes a statistical miracle.

The Failure of Surrogate Parenting

Conservationists attempted to intervene with scientific precision, but nature is rarely moved by human logic. They moved the survivors to Mangere Island and tried to facilitate breeding through surrogate parents. The results were a series of failures. Warblers couldn't provide enough food, and Tomtits were too effective—the black robin chicks raised by Tomtits grew up believing they were Tomtits, refusing to mate with their own kind. These biological errors nearly ended the species. It demonstrated that you cannot simply engineer a recovery; you have to respect the specific, often strange, requirements of a species' social and biological identity.

Mysterious are the ways of even a bird’s heart.

The turning point was Old Blue. She was an eight-year-old female, an outlier in a species where such age was rare. She was the only survivor capable of laying fertile eggs. Instead of the artificiality of surrogates, the chicks were returned to her care. In her presence, they thrived. Old Blue did not just survive; she became a biological engine. She lived to fourteen and raised eleven chicks. Today, the 250 black robins in existence are all descendants of her lineage. The entire survival of the species rests on the reproductive success of a single individual.

Lessons from the Brink

Complexity cannot always be replaced by simple substitutes
Biological identity is tied to social behavior
Resilience often comes from unexpected, singular sources
Recovery requires more than just habitat; it requires connection

The story of Old Blue is a reminder that while damage can be massive and systemic, recovery can be equally concentrated. It is a testament to the fact that a single act of persistence—in this case, the biological persistence of one bird—can undo an entire century of destruction. It is not a story of grand, sweeping ecological shifts, but of a single, tenacious thread holding a whole tapestry together.

Key Takeaway

Survival often depends on the stubborn persistence of a single, vital element.

04 — The Marginalian

The Art of Letting Go

Finding meaning in the inevitable end of things

By Maria Popova · 9 min read

Editor's note: A meditation on the necessity of loss in the construction of a meaningful life.

There is a specific kind of cruelty in the way life operates: we only recognize the significance of a moment once it has become a 'last.' We don't realize we are experiencing the last time a loved one touches our hand, or the last sip of coffee in a shop that is about to close, until the moment has passed into memory. These 'lasts' are the currency of loss. They are the price we pay for the ability to care about anything at all. To love something is to accept the eventual necessity of losing it. This is not a pessimistic view, but a functional one; the transience of experience is what gives it its weight.

Taxonomies of Loss

In her work, Maija Hurme categorizes these endings, moving from the trivial to the existential. There are 'late lasts,' like the final rays of a sunset, and 'hopeful lasts,' such as the final migrating bird of a season. There are even 'lasts that are firsts'—the final wobbly moment of a child before they walk. By naming these moments, we do more than just mourn them; we acknowledge their role in the cycle of living. We begin to see that life is not a collection of permanent possessions, but a sequence of temporary states.

What exists, exists so that it can be lost and become precious.

The ability to let go is a skill that must be practiced. It is an apprenticeship to the reality of change. When we resist the end of things—whether it is a project, a relationship, or a phase of life—we create a friction that prevents us from moving into the next state. Learning to lose is, paradoxically, the only way to learn how to truly live. It requires a shift from a mindset of accumulation to a mindset of appreciation. We are not here to keep things; we are here to experience them.

Categories of the End

Late lasts: The fading light of a day
Hopeful lasts: The final chance for a connection
Lingering lasts: The slow departure of a season
Lonely lasts: The final remnant of a system

Consider the light from a star that died millions of years ago. When we see it twinkle in the night sky, we are witnessing a 'last' that has traveled across time to reach us. The star is gone, but its impact remains. Our own lives function similarly. The moments we lose are not truly gone; they are integrated into the person we become. The practice of letting go is not about forgetting, but about making room for the next inevitable arrival.

Key Takeaway

Meaning is derived from the very fact that nothing lasts.

05 — The Marginalian

The Trance of Near-Living

Breaking free from the mental control of existence

By Maria Popova · 11 min read

Editor's note: A critique of the modern tendency to intellectualize life rather than actually living it.

Modern existence is often characterized by a strange kind of paralysis. We spend our lives trying to solve, control, and conquer our circumstances, treating life as a series of problems to be managed rather than a process to be experienced. This is what Anaïs Nin and D.H. Lawrence identified as a state of 'near-living'—a trance where we are physically present but mentally detached, attempting to use our intellect to bypass the raw, often uncomfortable, reality of being alive. We try to 'will' warmth or happiness into existence, but mental conjuring is a poor substitute for actual vitality.

The Error of Division

D.H. Lawrence argued against the division of the human experience. We tend to separate our 'rational' minds from our 'emotional' or 'physical' selves, creating a hierarchy where the mind attempts to police the body. This creates a state of constant resistance. When we try to force ourselves into a specific state—such as constant happiness—we are essentially committing a form of psychological death. We stop being a process of becoming and start being a static, controlled object. True livingness requires the integration of the whole self, including the parts that hurt, struggle, and fail.

All real living hurts as well as fulfils.

Happiness, in the shallow sense, is merely a respite—a holiday from the demands of existence. Real happiness, as Lawrence described it, is found in being 'used by life.' It is the state of being driven, goaded, and replenished by the very forces that challenge us. It is an active engagement with the world's imperatives, rather than a passive attempt to avoid its friction. To live fully is to accept the 'strange and subtle imperatives' of the soul, even when they conflict with our desire for comfort or control.

The Imperatives of Aliveness

Obedience to internal urges over external control
Acceptance of pain as a component of growth
Integration of mind and body
Embracing the process of becoming

Breaking the trance requires a surrender to the unfolding of life. This is not a passive resignation, but an active, courageous opening to the unknown. It is the recognition that the most vital parts of ourselves are often the ones we try hardest to regulate. By letting go of the need to master our experience, we finally become capable of actually inhabiting it.

Key Takeaway

Control is the enemy of vitality; true living requires the courage to be moved by life.

06 — Simon Willison

The End of the Free Lunch

Why the era of effortless AI coding is closing

By Simon Willison · 5 min read

Editor's note: The shift from 'model-first' to 'task-first' engineering in the age of LLMs.

For a brief window of time, it felt as though the fundamental constraints of software engineering were dissolving. With the advent of increasingly capable large language models, the focus shifted away from the structural integrity of code and toward the immediate gratification of generation. Why spend hours perfecting a coding harness or a complex context strategy when a new, more powerful model could simply 'paper over' your mistakes for a few cents? It was an era of the 'free lunch,' where intelligence was cheap enough to be treated as a disposable commodity.

The Fable Disruption

That era has ended with the arrival of models like Fable. While Fable represents a massive leap in capability, it also brings a brutal economic reality: the cost of high-end intelligence is too high for indiscriminate use. When a model is incredibly capable but also incredibly expensive, the engineering strategy must shift from 'generate everything' to 'orchestrate intelligently.' Developers are being forced to return to the fundamentals of work allocation—deciding exactly which tasks require the heavy lifting of a top-tier model and which can be handled by cheaper, more efficient alternatives.

Prior to Fable, it felt silly to waste time improving your coding harness. Then Fable landed.

This shift marks the professionalization of AI-assisted development. We are moving away from the novelty of 'magic' code generation and into a disciplined era of model management. Engineers must now consider the cost-per-token as a primary architectural constraint, much like latency or memory usage. The goal is no longer just to get the code right, but to get it right at the lowest possible cost. This requires a more sophisticated understanding of model capabilities and a more rigorous approach to software design.

The New Engineering Priorities

Granular task decomposition
Cost-aware model selection
Robust coding harnesses
Optimized context management

The 'free lunch' was a useful training period, but it obscured the importance of good engineering. As the cost of intelligence scales, the value of the structures that manage that intelligence increases. The engineers who thrive in this next phase will not be those who can prompt the best, but those who can build the most efficient systems for deploying intelligence where it actually matters.

Key Takeaway

As AI intelligence becomes more expensive, engineering discipline becomes more valuable.