The Deep Feed

01 — Lenny's Newsletter

The End of the Technical Gatekeeper

How non-coders are using AI agents to ship production software

By Lenny Rachitsky · 12 min read

Editor's note: The barrier to entry for software creation has collapsed, turning intent into the primary currency of development.

For decades, the ability to build software was guarded by a high wall of syntax and logic. You had to speak the language of the machine to make it do anything useful. That wall is crumbling. We are entering an era where the primary skill is no longer writing code, but orchestrating the agents that do. This isn't just about making coding easier for engineers; it is about making software creation possible for people who have never seen a terminal in their lives. The shift moves the focus from 'how' to 'what'. If you can describe a problem with enough clarity, the machine can now build the solution.

The Three-Step Dance

Take the case of Bryce Rattner Keithley. A recruiter by trade, she used a specific workflow to ship a fitness app to the App Store. Her method wasn't magic; it was a structured division of labour between different AI models. She used Claude as an architect to plan the logic, Claude Code as the engineer to write the lines, and the Terminal as the executor to run them. This cycle—plan, execute, deploy—allows a non-technical user to act as a high-level project manager. Instead of wrestling with semicolons, the user wrestles with intent and verification.

The human role has shifted from writing solutions to bringing taste and judgment to the process.

This change demands a new kind of technical expertise. The engineers who focus solely on finding the fastest working solution are becoming obsolete. The robots can find solutions faster. The new value lies in knowing which tools to combine, understanding when an AI's output is hallucinating, and having the vision to guide the process toward a useful product. It is a move from the micro to the macro.

The New Developer Framework

Architectural Planning: Using LLMs to map out logic before a single line is written.
Iterative Debugging: Using screenshots and visual references to correct AI errors.
Verification: Treating AI output as a draft that requires human validation.

The winners in this new economy won't be those who memorised the most libraries, but those with the humility to work alongside machines. They will be the people who can bridge the gap between a business need and a technical execution without needing to be the one holding the hammer.

Key Takeaway

Software development is shifting from a craft of syntax to a craft of orchestration.

02 — Cal Newport

The Silicon Valley Priesthood

Why tech leaders are trading pragmatism for prophecy

By Study Hacks · 8 min read

Editor's note: The industry is increasingly behaving like a religion, which is a dangerous way to run a business.

There is a strange, religious atmosphere settling over the AI industry. It is no longer enough for tech executives to build useful tools; they now feel compelled to invent the future. We see leaders attending religious services in the Vatican, not to find faith, but to 'keep in touch' with human values as they build something they believe will change life as we know it. This isn't engineering; it is a form of secular mysticism. They are acting as both priest and prophet, managing the transition to a new era of existence.

The Tower of Babel Problem

When tech leaders speak of inevitable job automation or the need for universal basic income, they are often performing a kind of hubris. They present a future that is already decided, a destiny that must be accepted. This mirrors the biblical myth of the Tower of Babel—an attempt to reach the heavens through sheer technical force. But this narrative ignores the fundamental purpose of technology: to be a tool that serves the common good, not a deity that dictates human terms.

Leave the religion to the Pope; I want my technology executives focused on building things people actually want.

Recently, we have seen a slight retreat from this stance. CEOs like Jensen Huang and Sam Altman have begun to push back on the more extreme claims regarding mass job displacement. Some call it PR damage control; others might see it as a return to reality. The anxiety spread by the 'p(doom)' crowd has created a climate of fear that may take years to dissipate. The industry needs to move away from the role of the solemn sage and back to the role of the builder.

Signs of the Tech-Prophet Shift

Focusing on 'inventing the future' rather than solving current problems.
Using existential risk as a way to avoid accountability for current product flaws.
Presenting technological inevitability as a replacement for political agency.

The danger is that in trying to appease the digital gods they are creating, these companies lose sight of the people they are supposed to serve. A tool that requires a new social contract just to function is not a successful product; it is a disruption that has failed to integrate.

Key Takeaway

Technology should be a tool for human agency, not a replacement for it.

03 — Not Boring

The American Electric Stack

Why manufacturing is the ultimate strategic advantage

By Packy McCormick · 10 min read

Editor's note: Design leadership is a hollow victory if you don't control the hardware that makes it move.

For years, the American strategy was simple: design the high-value software and let the rest of the world handle the heavy lifting. We exported intellectual property and imported the hardware. But in the coming electric revolution, this model is broken. If you want to lead in robotics and drones, you cannot rely on a supply chain that is easily disrupted or remotely controlled by geopolitical rivals. You need to own the motors and the actuators. You need to own the physical layer.

The Westmag Thesis

Westmag is a company betting on the idea that American manufacturing and design must be reunited. They aren't just making hats; they are building the electric motors that will drive the next generation of autonomous machines. This isn't about jingoism; it's about strategic coherence. When you manufacture the components, you learn how to improve them. You close the feedback loop between the digital design and the physical reality. This is how true innovation happens.

Maintaining design leadership without manufacturing leadership is not a coherent strategic position.

The argument against this is usually based on comparative advantage—the idea that it is cheaper to buy from China than to build in the West. But that argument ignores the cost of security and the cost of lost learning. Every time we outsource a component, we lose a piece of the ability to innovate that component. We become consumers of technology rather than creators of it.

Why Local Manufacturing Matters

Tightening the feedback loop between design and production.
Reducing vulnerability to geopolitical supply chain shocks.
Ensuring security in software-controlled hardware.

The future belongs to those who can control the entire stack. As AI moves from the screen into the physical world through robots and drones, the companies that own the hardware will hold the real power.

Key Takeaway

True innovation requires control over both the code and the components.

04 — The Marginalian

The Anatomy of Human Connection

A new classification based on our shared vulnerabilities

By Maria Popova · 6 min read

Editor's note: In an age of algorithmic sorting, we need a way to see each other that isn't based on data points.

We have an obsessive need to categorise. We sort people by wealth, by status, by political affiliation, and by data. We use these labels to create order, but they often serve to distance us from one another. We trade in preconceptions rather than genuine encounters. Our cognitive instinct is to sieve the world into manageable pieces, but in doing so, we often lose the very thing that makes life meaningful: the messy, unclassifiable essence of being human.

Hawthorne's Counter-Classification

In 1836, Nathaniel Hawthorne proposed a radical alternative to our standard social hierarchies. Instead of sorting by power or money, he suggested we sort by the things that truly unite us: sorrow, disease, and sin. He argued that everyone, regardless of their station, experiences loss. Everyone, regardless of their wealth, faces sickness. Everyone, regardless of their reputation, carries guilt. This isn't a cynical view; it is a deeply humanizing one.

Down where the spirit meets the bone, we are all facing the same struggle.

This classification strips away the artificial layers we use to protect our egos. It forces us to acknowledge that the person in the mansion and the person in the hovel are bound by the same fundamental vulnerabilities. It replaces the hierarchy of achievement with a hierarchy of shared experience.

The Three Human Classes

The Mourners: All who suffer the loss of loved ones.
The Afflicted: All who struggle with physical or mental malady.
The Guilty: All who carry the weight of human error.

If we adopted this lens, the way we view 'the other' would change. The stranger becomes a fellow traveller in the struggle to find meaning in a brief and often confusing existence. It is a way to find common ground when everything else seems designed to pull us apart.

Key Takeaway

Our shared vulnerabilities are more defining than our social distinctions.

05 — The Marginalian

The Trap of Self-Criticism

How our internal critics rob us of agency

By Maria Popova · 9 min read

Editor's note: Constant self-correction is often just a mask for a loss of creative will.

We often mistake self-criticism for a virtue. We believe that by being our own harshest judges, we are engaging in a form of intellectual honesty or self-improvement. But there is a fine line between critical thinking and a destructive, repetitive loop of self-flagellation. When criticism stops being a tool for growth and starts being a way to avoid action, it becomes a form of paralysis. It is a way to stay safe by never actually attempting anything worth judging.

The Masochism of the Ego

Psychoanalyst Adam Phillips suggests that our relentless self-criticism is actually a form of pleasure. It is a way of maintaining a sense of control. If we are the ones doing the criticizing, we are in charge of our own failures. It is a way to manage the discomfort of ambivalence—the fact that we can love and hate the same thing at the same time. By turning our attention inward to our inadequacies, we avoid the much harder task of engaging with the unpredictable world outside.

Self-criticism is a strikingly unimaginative repertoire of tirades.

This internal critic is often redundant. It tells us the same things over and over: that we are not enough, that we are behind, that we are failing. This loop doesn't lead to better work; it leads to a diminished capacity for joy and celebration. We become so suspicious of our own successes that we can no longer inhabit them.

The Costs of Excessive Criticism

The loss of creative momentum.
The replacement of action with rumination.
The inability to experience genuine satisfaction.

To break the cycle, we must learn to value celebration as much as we value correction. We need to allow ourselves the space to be imperfect without making that imperfection our entire identity. The goal is not to stop thinking critically, but to stop using criticism as a shield against living.

Key Takeaway

Self-criticism can become a way to avoid the risk of actually being alive.

06 — Lenny's Newsletter

The Agentic Shift

Moving from tools to autonomous collaborators

By Claire Vo · 7 min read

Editor's note: The next leap in productivity isn't faster tools, but tools that require less supervision.

The current way we use AI is largely transactional. We give a prompt, we get a response. It is a turn-by-turn interaction that requires constant human babysitting. But we are moving toward a model of 'agentic' AI—systems that don't just respond to instructions, but pursue goals. This is the difference between telling someone to 'write a letter' and telling them to 'manage my correspondence so I only see what is urgent'. One is a task; the other is an outcome.

The Power of the Goal

When you move from prompts to goals, the nature of the work changes. A goal defines what a successful outcome looks like and gives the AI the autonomy to find the path there. This allows for long-running processes where the AI can work for hours, verifying its own steps and correcting its own errors. It transforms the AI from a sophisticated autocomplete into a digital employee.

A prompt is an instruction of what to do; a goal is a description of what a good outcome looks like.

This shift requires a new kind of management skill. You cannot manage an agent the way you manage a traditional software tool. You have to define boundaries, constraints, and verification methods. You have to decide what the 'stopping condition' is. The skill is no longer in the execution, but in the definition of success.

Framework for Setting AI Goals

Outcome: Clearly define what should be true when the task is done.
Verification: Establish how the AI can test its own success.
Constraints: Set the boundaries of what the agent must not do.
Iteration Policy: Define how the agent should pivot when it hits a wall.

As these agents become more capable, the bottleneck in productivity will shift from the ability to perform tasks to the ability to define them. The most valuable people will be those who can articulate complex objectives with enough precision that a machine can execute them autonomously.

Key Takeaway

The future of work is not about managing tasks, but about defining outcomes.