The Deep Feed

01 — Lenny's Newsletter

The Great Tech Schism

How AI is splitting the workforce into the energized and the resentful

By Lenny Rachitsky · 12 min read

Editor's note: A vital look at the internal emotional state of the industry's engine room.

The tech industry has long prided itself on a shared sense of mission, a collective belief in building the future. But as we move through 2026, that cohesion has vanished. A massive new survey of thousands of workers across product, engineering, and design reveals a workforce that has split almost exactly in half. On one side, there are the 'Energized'—those who have integrated AI into their workflows to achieve a level of output previously thought impossible. On the other, there is a growing cohort of the 'Resentful' and the 'Disoriented,' people who see the tools not as assistants, but as existential threats to their craft and their livelihood.

The Four Archetypes of the AI Era

The survey identifies four distinct emotional states defining the modern tech worker. The 'Energized' are riding the wave, using large language models to automate the drudgery of coding and documentation. The 'Conflicted' are caught in the middle, using the tools while fearing they are eroding their own long-term value. Then there are the 'Disoriented,' who find the pace of change so rapid they can no longer find a stable footing in their roles, and the 'Resentful,' who view the automation of cognitive tasks as a degradation of professional standards. This isn't just a difference in tool preference; it is a fundamental disagreement on what it means to be a professional in a world of automated intelligence.

The workforce is splitting in two: one half is thriving, the other is shaken.

Burnout has not been spared by this transition. The data shows an 11-point jump in burnout rates in just twelve months. This isn't the traditional burnout caused by long hours or poor management; it is a new, more corrosive type of exhaustion. It is the fatigue of constant adaptation. When the fundamental requirements of your job change every six months, the cognitive load required to stay relevant becomes a heavy, permanent tax on your mental energy. Workers are no longer just doing their jobs; they are constantly re-learning how to do them.

The Tech Sentiment Breakdown

The Energized: Leveraging AI for massive productivity gains.
The Conflicted: Using tools while fearing skill atrophy.
The Disoriented: Struggling to find a stable career path.
The Resentful: Viewing automation as a threat to professional dignity.

Perhaps most telling is the collapse of industry advocacy. Almost no one in tech would recommend their career to a newcomer today. The prestige of the sector is being replaced by a sense of precariousness. The fear isn't just about losing a job to a machine; it is about the loss of agency. When a model can write your marketing copy or your boilerplate code for pennies, the question isn't just 'will I have a job?' but 'what will be left for me to actually do?'

Key Takeaway

The primary challenge for tech workers in 2026 is not job replacement, but the psychological exhaustion of constant reinvention.

02 — Simon Willison

The Access War

Anthropic's hesitation and the rising cost of intelligence

By Simon Willison · 6 min read

Editor's note: A case study in how compute constraints dictate the competitive landscape of AI.

In the race for AI supremacy, the bottleneck is no longer just algorithms; it is the raw, physical availability of compute. Anthropic’s recent decision to repeatedly extend the availability of its Fable/Mythos class models in Claude Max plans is a clear signal of this tension. By limiting access and using usage credits, Anthropic is attempting to manage a demand that threatens to outstrip its hardware capacity. They are playing a cautious game, trying to gauge the true appetite for their most advanced models before committing to a cheaper, unlimited subscription model.

The OpenAI Advantage

While Anthropic manages its scarcity, OpenAI is playing a different game. By appearing more confident in their ability to provide consistent access to GPT-5.6, OpenAI is capturing the users who value reliability over everything else. In a professional environment, uncertainty is a productivity killer. If a developer doesn't know if they will have access to the high-end model they need to finish a task by Tuesday, they will simply switch to a provider who can guarantee it. Anthropic's caution, while economically sensible, is creating a massive opening for their competitor.

OpenAI are winning users simply due to the uncertainty that surrounds Fable access.

This tension highlights a critical shift in the AI market. We are moving away from a phase of pure capability competition—where the goal was simply to make the smartest model—into a phase of reliability and distribution competition. It is no longer enough to have the best model; you must have the best ability to deliver that model to a paying customer without friction. The winner won't just be the company with the most parameters, but the company with the most stable compute pipeline.

The Compute Constraint Reality

Demand for high-end models is outpacing current hardware availability.
Usage limits and credits are being used as a management tool for scarcity.
Reliability is becoming a primary competitive advantage over raw intelligence.

For the power user, this creates a fragmented experience. You are constantly calculating your 'usage budget,' deciding when to use the premium model and when to settle for a lesser one. This constant decision-making is a tax on creativity. If the goal of these models is to reduce cognitive load, the current model of restricted access is doing the exact opposite.

Key Takeaway

In the AI economy, reliability and guaranteed access are becoming more valuable than marginal gains in intelligence.

03 — The Marginalian

The Danger of the Safe Life

Why avoiding heartbreak is a form of living death

By Maria Popova · 10 min read

Editor's note: A philosophical meditation on the necessity of risk and vulnerability.

There is a pervasive, quiet temptation in modern life to seek shelter from the very things that make us human. We build institutions—religion, therapy, marriage—to shield us from the inherent instability of existence. We want to be saved from the disappointment, the heartbreak, and the inevitable collisions between our hopes and a reality that is often indifferent to them. But as Maria Popova explores through the poetry of Mario Benedetti, this desire to 'save ourselves' is perhaps the most dangerous impulse we possess. To seek safety is to seek a distance from life itself.

The Catatonia of Comfort

When we prioritize certainty, we enter a state of emotional catatonia. We automate our inner lives through habit and convenience, choosing the halogen-lit safety of the predictable over the fertile doubt of the unknown. This is not living; it is merely surviving in a controlled environment. Benedetti's poem, 'No Te Salves' (Do Not Spare Yourself), serves as a fierce indictment of this tendency. He warns against the 'quiet corner' we try to claim from the world, and against the heavy eyelids of judgment that allow us to sleep through our own lives.

To spare ourselves from the despair by which we contour hope is to live a safe distance from alive.

The poem is a call to remain 'bloodied' by experience. To be alive is to be vulnerable to the touch of love and the sting of loss. When we petrify our joy or desire with reserve, we become brittle. We lose the ability to be surprised by ourselves. The tragedy is not that we suffer, but that in our attempt to avoid suffering, we end up losing the very capacity for joy that makes the suffering worth enduring.

The Costs of Self-Preservation

The loss of self-surprise and personal discovery.
The replacement of genuine connection with safe, predictable interactions.
The atrophy of emotional resilience through the avoidance of friction.

If we succeed in sparing ourselves, the consequence is a hollowed-out existence. We become 'bloodless' thinkers and 'lipless' feelers, standing motionless by the side of the road while life passes us by. The invitation is to reject the comfort of the quiet corner and to embrace the mess, the doubt, and the uncertainty that define the human condition.

Key Takeaway

True living requires the willingness to be broken by reality; to seek total safety is to forfeit the capacity for joy.

04 — The Marginalian

The God of the Laws

Stephen Hawking's rejection of the personal deity

By Maria Popova · 9 min read

Editor's note: An analysis of Hawking's attempt to reconcile the scale of the universe with the concept of divinity.

For centuries, the question of God has been the primary battlefield between science and superstition. For many, the existence of a creator was the only way to explain the order of the cosmos. But as our understanding of physical laws has deepened, the need for a 'personal' God—a being who intervenes in human affairs—has become increasingly difficult to defend. Stephen Hawking, in his final reflections, addresses this not with hostility, but with the clinical clarity of a physicist. He moves the conversation away from the emotional and toward the structural.

The Impersonal Definition

Hawking’s position is a form of scientific naturalism. He argues that if one believes in science, one must believe in the unchangeable, universal laws of nature. These laws govern everything from the flight of a tennis ball to the motion of galaxies. For Hawking, these laws are so powerful and so consistent that they render the idea of a human-like deity—one with whom we can have a personal relationship—implausible. When you consider the vastness of the universe and the accidental nature of human life within it, a God who cares about individual human morality seems like a mismatch of scale.

One could define God as the embodiment of the laws of nature.

However, Hawking does not necessarily reject the word 'God' entirely; he simply redefines it. Much like Einstein, he uses the term in an impersonal sense. To him, 'God' is not a person, but the sum total of the mathematical elegance and physical necessity that allows the universe to exist. This is not a 'proof' of existence, but a definition of the term that aligns with what we can actually observe. It is a way to respect the awe of the cosmos without resorting to anthropomorphism.

Hawking's Framework for the Divine

The rejection of a personal, intervening deity due to cosmic scale.
The recognition of universal laws as the true governing force of existence.
The definition of 'God' as the embodiment of physical laws.

By stripping the concept of God of its human-like qualities, Hawking removes the conflict between science and spirituality. The 'hymn of praise' that scientists feel when discovering a new law is not a prayer to a person, but a recognition of the profound order of the universe. In this view, science does not disprove God; it simply clarifies what we mean when we use the word.

Key Takeaway

The universe is governed by laws so absolute that any concept of God must be understood as the laws themselves, rather than a personal actor.

05 — Simon Willison

Coding with Codex

The automation of the developer's workflow

By Simon Willison · 7 min read

Editor's note: A practical look at how AI is being integrated into the very tools developers use.

The integration of AI into software development is moving from high-level chat interfaces into the granular, command-line tools that developers use every day. Simon Willison’s recent updates to `sqlite-utils` provide a clear example of this. By using an AI agent—specifically Codex—to review open issues and suggest implementations, the development process is becoming a hybrid of human intent and machine execution. This isn't just about writing code; it's about using AI to bridge the gap between a developer's idea and the technical implementation of a feature.

The New Development Loop

The new `--code` option in `sqlite-utils` allows users to pass blocks of Python code directly into CLI commands. This level of flexibility is essential in an era where we expect our tools to be programmable on the fly. Furthermore, the use of AI to identify 'obvious' missing features—like the ability to drop an index by name—shows a shift in how software evolves. Instead of a human developer spending hours triaging a backlog of minor requests, an AI can scan the issues, identify the low-hanging fruit, and propose the solution.

I had Codex review all open issues and highlight the easiest ones!

One of the most effective ways this is being used is in the testing phase. Willison describes a prompt that tells the model to 'manually exercise' new options to find edge cases. This is a critical distinction: the AI isn't just writing the code; it is being instructed to act as a skeptical tester, looking for the bugs that its own automated tests might miss. This creates a recursive loop of improvement that significantly accelerates the release cycle.

AI-Driven Development Patterns

Automated issue triaging to find low-complexity tasks.
Using AI to generate 'on-the-fly' logic for CLI tools.
Instructional prompting to force models into 'adversarial testing' modes.

This represents a fundamental change in the role of the maintainer. The job is moving away from the manual implementation of every feature and toward the orchestration of AI agents that can handle the routine aspects of software maintenance. The developer becomes an editor and an architect, rather than just a writer of syntax.

Key Takeaway

The future of software maintenance lies in the orchestration of AI agents to handle routine tasks and adversarial testing.