The Deep Feed

01 — Lenny's Newsletter

The AI Paradox: Why Automation Demands More Humanity

The myth of the job apocalypse and the rise of the forward-deployed engineer

By Lenny Rachitsky · 12 min read

Editor's note: As models become cheaper and faster, the value of human direction shifts from execution to orchestration.

The common fear is that artificial intelligence will hollow out the workforce, leaving a vacuum where skilled professionals once stood. This view assumes that work is merely a collection of repetitive tasks to be optimised. But Dan Shipper suggests a different outcome: as the cost of basic cognitive labour drops to near zero, the demand for high-level human agency actually rises. We are not entering an era of less work, but an era of different work. The friction is no longer in doing the task, but in knowing which task is worth doing and how to direct a fleet of agents to execute it.

The Death of the SaaS Apocalypse

There is a prevailing theory that AI will kill the Software-as-a-Service (SaaS) model by allowing companies to build their own bespoke tools. This ignores the economics of scale. Instead of disappearing, SaaS will likely evolve. Users will bring their own AI tokens into existing applications, a shift that could actually improve margins for software providers. The software becomes the interface through which we manage our intelligence, rather than just a tool for a specific function. The value moves from the code itself to the workflow it enables.

The only thing you need to do to stay employed is ride the models.

The most significant shift will be in job roles. The 'forward deployed engineer' is emerging as the most essential hire. This isn't someone who just writes code, but someone who understands how to integrate complex AI capabilities into specific business problems. They sit at the intersection of technical possibility and commercial necessity. Similarly, product managers and full-stack designers will find their influence expanding as they move from creating assets to designing systems of agency.

The New Professional Toolkit

Mastering CLI-based AI tools for non-technical tasks
Transitioning from task execution to agent orchestration
Developing the ability to debug workflows rather than just code
Understanding the economics of token-based software usage

We will eventually read more AI-generated text, and we will find it acceptable. The goal is not to replace the human voice, but to remove the drudgery that prevents the human voice from being heard. When the mechanical parts of communication are handled by models, the remaining human input becomes more concentrated and intentional. The paradox is that the more we automate, the more we must rely on the specific, un-automatable qualities of human judgement.

Key Takeaway

Automation does not eliminate work; it raises the floor of what is expected from every professional.

02 — Lenny's Newsletter

The Abstraction Ladder

How to stop fighting machines and start directing them

By Claire Vo · 10 min read

Editor's note: Felix Rieseberg of Anthropic reveals that the biggest barrier to AI utility is psychological, not technical.

Most people use AI like a better version of Google: they ask a question, get an answer, and move on. Felix Rieseberg, an engineering lead at Anthropic, argues this is a massive waste of potential. The real power of these models lies in their ability to handle complexity through layers of abstraction. When you hit a wall with an AI, the solution is rarely to try harder with the same method. The solution is to move one level up. If you are manually entering data, stop. Ask the AI to find the data. If it fails, ask it to find the source of the data.

Your Email is a Database

Consider the mundane task of moving house. Instead of measuring furniture and searching for receipts, Rieseberg suggests pointing an AI at your email inbox. Your digital history contains a structured record of every purchase, dimension, and confirmation you have ever received. By treating your email as a database rather than a communication tool, you can transform years of scattered text into a precise inventory. This is the shift from using AI as a writer to using it as a data architect.

The gap isn’t technical; it’s psychological. Adults have spent 20 years in a 'mind prison' learning what computers can’t do.

This 'mind prison' is the set of assumptions we have built about the limits of software. We assume computers are tools that require explicit instructions. We forget that they can be collaborators that interpret intent. Rieseberg notes that children are often better AI users because they lack these preconceived notions of impossibility. They ask for what they want without worrying about whether the software is 'designed' to do it. To master AI, one must unlearn the rigid logic of traditional computing.

Rieseberg's Heuristics for AI Mastery

Go one abstraction layer up: if a task is tedious, find a way to automate the decision-making behind it
Debug the workflow, not the model: if the output is wrong, check your data source and your instructions
Use Opus for intent: use smaller models for specific tasks, but use the largest models when you need the machine to interpret what you actually mean
Treat AI as a collaborator: ask it to walk you through its reasoning when it makes a mistake

The end goal is to free up creative energy. When the 'annoying' tasks—the ones that don't require real thought—are handled by background agents, the human is left with the high-value work. This isn't about doing more; it's about doing what matters. The machine handles the logistics; the human handles the vision.

Key Takeaway

Stop treating AI as a calculator and start treating it as an intern who needs better management.

03 — Cal Newport

The Profanity of Speech

Digital ethics and the sanctity of the human voice

By Study Hacks · 8 min read

Editor's note: As LLMs master the art of conversation, we must ask if we are diluting what it means to be human.

In the ancient Hebrew tradition, the creation of man is described as the infusion of breath into dust, resulting in a 'speaking spirit.' This capacity for speech is not merely a biological function; it is a defining characteristic of humanity. It is the mechanism by which we translate internal thought into shared reality. To speak is to engage in a form of telepathy that builds connection, dignity, and justice. It is a ritual that has shaped human civilization since the first alphabets were carved into stone.

The Illusion of Fluidity

Generative AI presents a challenge to this concept. When we engage with a chatbot, we encounter a level of lexical fluidity that mimics human conversation perfectly. We know, intellectually, that this is just the result of matrix multiplications and token prediction. Yet, there is a persistent sense of unease. There is something that feels slightly wrong about a machine occupying the space of human communication. It feels like a transgression of a sacred boundary.

Should we let AI write and speak on our behalf, or serve as a golemic conversation partner?

This unease points toward a looming crisis in digital ethics. We are rapidly approaching a point where the distinction between human-generated thought and machine-generated mimicry becomes irrelevant to the listener. If we allow machines to speak for us, we risk devaluing the very act of communication. We must decide if speech is a tool to be optimised or a ritual to be protected.

The question is not whether the technology will exist, but what we are willing to accept. If we outsource our expression to models, do we lose the ability to truly connect? The 'golem' of AI can mimic the form of speech, but it lacks the spirit behind it. As we integrate these models into our lives, we must be careful not to mistake the shadow of communication for the substance of it.

Key Takeaway

The value of speech lies not in the accuracy of the words, but in the human intent behind them.

04 — The Marginalian

The Grace of the Unnecessary

Why moonlight and art matter in a utilitarian world

By Maria Popova · 9 min read

Editor's note: In a culture obsessed with efficiency, we find the most meaning in the things that serve no practical purpose.

Daylight is utilitarian. It allows us to see, to work, and to navigate the physical world. It is the baseline of survival. Moonlight, however, is different. It does not reveal the world as it is; it transforms it. It casts shadows, turns landscapes into silver barricades, and gives the ordinary a quality of the eternal. Moonlight is, by definition, unnecessary. It provides no survival advantage, yet it has inspired poets, painters, and thinkers for millennia.

The Beauty of the Extra

There is a profound distinction between what is necessary and what is beautiful. As Richard Adams wrote in *Watership Down*, water is necessary, but a waterfall is not. The waterfall is an ornament—an 'extra' that provides value without necessity. This is the essence of art, friendship, and philosophy. These things do not help us survive the elements, but they give us a reason to want to survive them. They are the 'unbidden graces' that make life more than just a biological process.

Friendship is unnecessary, like philosophy, like art, like the universe itself; rather it is one of those things which give value to survival.

In our modern drive for optimisation, we often mistake the necessary for the sufficient. We build systems that are efficient, fast, and productive, but we forget to leave room for the unnecessary. We optimize for the daylight—the clarity and the utility—while ignoring the moonlight that provides the depth and the meaning. A life lived entirely within the realm of the necessary is a life that is functional but hollow.

The Value of the Non-Utilitarian

Art: It changes how we perceive reality without providing a physical resource
Philosophy: It provides a framework for meaning that has no direct survival value
Nature's 'extras': The beauty of a waterfall or the light of the moon
Human Connection: The deep, non-transactional bonds that exist outside of necessity

To embrace the unnecessary is to acknowledge that human existence requires more than just the avoidance of death. We require the 'extra' to feel alive. We require the things that serve no purpose other than to be beautiful, to be felt, and to be experienced. In the pursuit of a perfectly efficient life, we must be careful not to extinguish the moonlight.

Key Takeaway

Efficiency keeps us alive, but the unnecessary is what makes life worth living.

05 — The Marginalian

The Biology of Presence

Swimming, memory, and our connection to the wild

By Maria Popova · 11 min read

Editor's note: Returning to the water is more than a hobby; it is an evolutionary homecoming.

There is a fundamental difference between swimming in a chlorinated pool and swimming in the open water. The pool is a controlled environment, a place of laps, times, and achievement. It is a space of productivity. The wild water, however, is a space of presence. When you enter a river or an ocean, the goal shifts from ambition to survival. You are no longer trying to beat a clock; you are trying to exist within a medium that is much larger than yourself.

The Aquatic Ancestry

This response to water is not merely psychological; it is biological. Evolutionary theories suggest that humans spent millions of years as semi-aquatic waders. We are, in many ways, creatures of the shore. Our very bodies—the way our hair grows, the way our fat is distributed—reflect an evolutionary history shaped by the water. When we swim in nature, we are not just exercising; we are returning to an ancestral state of being.

When you swim, you feel your body for what it mostly is — water — and it begins to move with the water around it.

Roger Deakin, in his exploration of Britain's waterways, describes swimming as a rite of passage. It is a crossing of boundaries—between land and sea, between the conscious and the subconscious. In the water, the sense of the present becomes overwhelming. The distractions of the dry world—the emails, the schedules, the anxieties—fall away, replaced by the immediate, visceral reality of temperature, movement, and breath.

The Two Modes of Swimming

Transactional Swimming: Focused on achievement, metrics, and physical improvement
Transcendent Swimming: Focused on presence, connection to nature, and mental clarity
The Goal: Moving from the former to the latter to find genuine renewal

To swim in the wild is to accept a certain level of vulnerability. It is to acknowledge that you are part of a larger, indifferent system. This vulnerability is precisely what allows for the feeling of transcendence. By stepping out of the controlled environments we have built for ourselves, we allow ourselves to be re-integrated into the world. We move from being observers of nature to being part of it.

Key Takeaway

True presence is found when we stop trying to conquer our environment and start moving with it.

06 — Lenny's Newsletter

The Architect of Agency

Building the future with Claude Code

By Claire Vo · 7 min read

Editor's note: A look at the practical application of AI in hardware and software development.

The barrier between an idea and a finished product is shrinking. Historically, if you wanted to build a piece of hardware or a complex software application, you needed deep, specialized knowledge and significant capital. Today, tools like Claude Code are democratising that capability. Felix Rieseberg demonstrates this by building a $20 hardware device—a 'Claude buddy'—using nothing but AI-driven code. The technical barrier is no longer the primary constraint; the constraint is now the clarity of the user's intent.

Asynchronous Design and Latency

One of the most overlooked aspects of building with AI is the experience of latency. Waiting for a model to think can be frustrating, but it can also be made delightful through asynchronous design. Instead of a spinning wheel that signals 'nothing is happening,' successful AI products use design to make the wait feel like part of the process. This involves showing the model's 'thought process' or allowing the user to move on to other tasks while the agent works in the background.

Making latency delightful through asynchronous design is a key product philosophy.

This approach mirrors the way we interact with human collaborators. We don't expect an immediate answer to every complex question; we expect a period of work followed by a result. By designing for this rhythm, we move away from the 'command-and-response' model of computing and toward a more natural, collaborative relationship with our tools.

Principles of AI Product Design

Embrace Asynchronicity: Design workflows that allow users to work while the AI processes
Show the Work: Provide visibility into the model's reasoning to build trust
Focus on Output, Not Process: Judge the tool by what it achieves, not how it writes the code
Lower the Friction of Intent: Make it easy for users to express complex, multi-step goals

As these tools mature, the role of the 'developer' will continue to shift. It will become less about the syntax of a language and more about the architecture of a solution. The winners in this new era will be those who can bridge the gap between a high-level concept and a functional, deployed reality, using AI as the engine of execution.

Key Takeaway

The future of engineering is not about writing code, but about designing outcomes.