The Deep Feed

01 — Lenny's Newsletter

The HTML Pivot: Why AI Agents Need More Than Text

Moving beyond Markdown to build richer, interactive intelligence

By Lenny Rachitsky · 12 min read

Editor's note: As AI agents take over more of the heavy lifting, the way we communicate with them is shifting from simple text to interactive interfaces.

For years, Markdown has been the lingua franca of the developer and the writer alike. It is simple, lightweight, and easy for both humans and machines to parse. But as we move into the era of autonomous AI agents, Markdown is hitting a ceiling. When an agent like Claude generates a thousand-line technical specification, a wall of text becomes a barrier to engagement. You don't read it; you skim it, or worse, you ignore it. This is where the shift toward HTML begins. By using HTML as the primary medium for AI communication, engineers are turning static plans into living, interactive artifacts. This isn't about making things look pretty; it is about information density and the ability to actually interact with a plan before a single line of production code is written.

The Rise of the Compute Allocator

The role of the software engineer is undergoing a fundamental change. We are moving away from being pure code writers and toward becoming 'compute allocators'. When an AI agent can run for hours on a single complex task, the engineer's primary job is to decide how to spend that computational budget. You are no longer just typing syntax; you are defining boundaries, setting constraints, and deciding which problems are worth the $500 of compute required to solve them. This makes the planning phase—the specification—the most important part of the workflow. If the plan is a mess of text, the execution will be too. If the plan is a structured, interactive HTML document, the engineer can audit the logic with precision.

The engineer's job is no longer writing code, but deciding what is worth building and how to spend the compute to get there.

This new workflow encourages a 'throwaway' mindset. Instead of building massive, monolithic tools to manage a project, engineers are building micro-UIs—small, bespoke interfaces designed to solve one specific problem, like editing a complex data table or visualising a dependency graph. Once the task is done, the interface is discarded. This abundance of tokens allows for a level of bespoke tooling that was previously too expensive or time-consuming. We are seeing a world where 99% of the AI's output is dedicated to the scaffolding, the interfaces, and the communication, rather than the final product itself.

The New AI Workflow

Use HTML for interactive, high-density specifications
Build bespoke micro-apps for specific planning tasks
Focus compute on planning and verification rather than just production
Maintain living design systems in code-readable formats

Ultimately, the goal is to keep the human in the loop. The danger of AI is not that it replaces us, but that it makes the work so opaque that we lose the ability to critique it. By moving from text to interface, we ensure that the complexity of the machine remains legible to the person directing it. We aren't just making better software; we are making better ways to think about software.

Key Takeaway

The value of engineering is shifting from the ability to write syntax to the ability to allocate compute and manage complex, interactive specifications.

02 — Cal Newport

The Dark Side of Efficiency

Understanding the Jevons Paradox in the age of AI

By Study Hacks · 8 min read

Editor's note: Efficiency doesn't always lead to more free time; often, it just raises the ceiling of what we are expected to produce.

In 1865, William Stanley Jevons observed something counterintuitive about the industrial revolution. He noticed that as steam engines became more efficient at using coal, the total consumption of coal actually increased. This became known as the Jevons Paradox. The logic is simple: when a resource becomes more efficient to use, the cost of using it drops, which in turn spikes the demand for it. We see this today in almost every sector of technology. We didn't use faster computers to finish our work earlier and go home; we used them to run more complex simulations, more massive datasets, and more demanding software.

The AI Labor Market

There is a widespread fear that AI will lead to a massive contraction in the labour market. The assumption is that if one programmer can now do the work of five, companies will simply hire one instead of five. However, the Jevons Paradox suggests a different outcome. If the cost of creating software drops precipitously, the demand for software will explode. We may see a world where every small business, every individual, and every niche organisation begins building their own bespoke tools. The total number of programmers might not shrink; the total amount of software being built might simply increase by orders of magnitude.

Increasing the efficiency of a resource often leads to greater consumption, not less.

But there is a darker dimension to this paradox that we often overlook: the cost of increased interaction. Consider the rise of email and Slack. These tools were designed to make communication more efficient, yet they have resulted in a state of constant interruption. Because it is so easy to send a message, we send more of them. The result is a fragmented workday where the average knowledge worker is interrupted every few minutes. Efficiency in communication has led to a crisis of attention.

The Risks of Efficiency

Increased demand leading to higher workloads
Fragmented attention due to communication abundance
The 'soot' of the digital age: mental fatigue and burnout

As we integrate AI into our workflows, we must be vigilant. If AI makes us ten times more productive, will we be rewarded with ten times more leisure, or will we simply be expected to produce ten times more output? History suggests the latter. The challenge for the modern professional is not just to become more efficient, but to build the boundaries necessary to prevent that efficiency from consuming their entire existence.

Key Takeaway

Efficiency is a double-edged sword that often increases the total volume of work and interruption rather than reducing it.

03 — The Marginalian

The Power of the Margin

How being an outsider provides a unique vantage point for truth

By Maria Popova · 10 min read

Editor's note: To see the whole, one must often stand outside the centre.

There is a common assumption that privilege is found in the centre—in the mainstream, the powerful, and the well-connected. But there is a different, often overlooked advantage found in the margins. To live on the edge of a society, whether due to race, gender, or belief, is to develop a dual perspective. Those in the margins must understand the centre to survive, but they also see the centre from the outside. This creates a way of seeing reality that is both inward and outward-looking, a perspective that is often more accurate than the one held by those comfortably situated in the middle.

The Dual Vision of bell hooks

The late bell hooks articulated this brilliantly in her work. She described the margin not just as a place of deprivation, but as a site of radical possibility. Because those in the margins have little to lose by thinking outside the established norms, they are often the ones who translate unseen realities into ideas that move the world forward. They possess a 'mode of seeing' that acknowledges the existence of the entire universe—the main body and the edge—simultaneously. This awareness is a survival mechanism that evolves into a profound intellectual strength.

To be in the margin is to be part of the whole but outside the main body.

This perspective is essential for any meaningful change. A revolutionary ideology cannot simply be imposed from the centre; it must incorporate the experiences and insights of those on the periphery. When the knowledge of the margin is brought to the centre, it provides the necessary friction to challenge complacency and stagnation. The most visionary ideas often emerge from individuals who can navigate both worlds, using their outsider status to question the 'obvious' truths of the mainstream.

In a world that increasingly demands conformity and centralisation, the ability to maintain a marginal perspective is a form of intellectual resistance. It allows us to see the cracks in the system and the potential for something new. To find truth, we should perhaps look less at the consensus of the crowd and more at the observations of those standing just outside the circle.

Key Takeaway

Existing on the periphery provides a unique, dual-perspective that is essential for seeing reality clearly and driving meaningful change.

04 — The Marginalian

The Architecture of Friendship

Antoine de Saint-Exupéry on the inevitability of loss and the gift of connection

By Maria Popova · 11 min read

Editor's note: Friendship is a beautiful, heavy responsibility that we must accept knowing it will eventually end.

Friendship is one of the few human experiences that requires a total surrender of the self. As Seneca advised, once you decide to admit someone into your soul, you must do so with your whole heart. This level of intimacy is rare and precious, but it carries an inherent, devastating risk: the inevitability of loss. Whether through death, distance, or the slow erosion of loyalty, every friendship eventually faces an end. To love deeply is to knowingly accept the eventual sorrow of losing that connection.

The Bitterness of the Garden

Antoine de Saint-Exupéry, the author of *The Little Prince*, wrote about this with a clarity that avoids sentimentality. He viewed the loss of friends not as a sudden shock, but as a gradual process—a garden being locked away, bit by bit. He understood that old friends cannot be manufactured; they are built through shared trials, quarrels, and reconciliations. The 'shade' they provide is something that cannot be replaced by new acquaintances. This realization brings a specific kind of bitterness, a recognition that certain chapters of our lives are permanently closed.

It is idle, having planted an acorn in the morning, to expect that afternoon to sit in the shade of the oak.

Yet, in his most famous work, Saint-Exupéry offers a way to reconcile this loss. He suggests that the connection we forge with others changes our very perception of the world. For the Little Prince, the stars are not just distant lights; they are places where a friend's laughter lives. This is the ultimate consolation: that the people we have loved become part of the fabric of our reality. They change how we see the sky, how we hear the wind, and how we interpret the silence.

The Nature of Connection

Friendship requires wholehearted admission
Shared history creates irreplaceable value
Loss is an absolute inevitability
Love transforms our perception of the mundane

We live in an era of 'low-stakes' connection—social media follows and superficial interactions that require little investment and offer little risk. But Saint-Exupéry reminds us that the true value of friendship lies in its weight. The sorrow of loss is the price we pay for the richness of the connection. To avoid the pain of mourning is to avoid the profound joy of truly knowing another person.

Key Takeaway

The inevitability of loss does not diminish the value of friendship; rather, the risk of sorrow is the necessary price for the depth of connection.

05 — Lenny's Newsletter

The Curation of Community

Why high-signal, low-volume gatherings are the future of leadership

By Lenny Rachitsky · 6 min read

Editor's note: In an age of digital noise, the most valuable commodity is a room full of the right people.

The modern professional landscape is saturated with 'events'. There are conferences, webinars, summits, and networking mixers, most of which are designed for scale rather than substance. They offer a high volume of information but a low density of actual insight. For senior leaders, these gatherings often feel like a tax on their time—a necessity for staying relevant that rarely delivers a return on investment. The fatigue is real, and it stems from the lack of peer-level connection. It is hard to be vulnerable or truly strategic when you are surrounded by hundreds of strangers in a sea of superficial networking.

The Return to Intimacy

There is a growing movement toward 'intentionally small' gatherings. The idea is to pivot away from the mass-market conference model and toward handpicked, high-signal summits. These events prioritize depth over breadth. Instead of a hundred mediocre talks, you might have five profound discussions. Instead of a thousand random contacts, you have ten meaningful connections with people who occupy similar levels of responsibility and thought. This is not about exclusion for the sake of elitism, but about creating the psychological safety required for real learning and honest exchange.

The highest-ROI networking event is not the one with the most people, but the one with the most relevant peers.

For product and AI leaders, this shift is particularly critical. The challenges they face—navigating the ethics of automation, managing rapid technological shifts, and building sustainable growth—are not solved in a lecture hall. They are solved in roundtables, in workshops, and in the quiet moments between sessions where leaders can admit what they don't know. The value of an IRL (in real life) event in the digital age is its ability to facilitate the kind of unscripted, high-stakes human connection that a Zoom call cannot replicate.

Characteristics of High-Signal Events

Handpicked, peer-level attendance
Interactive workshops over passive listening
Small group formats for deep discussion
Focus on tangible, actionable takeaways

As we move forward, the premium will be placed on curation. The ability to filter the noise and bring the right people into a concentrated space is becoming a vital skill for community builders and leaders alike. We don't need more information; we need better context, and we find that context through the people we choose to stand alongside.

Key Takeaway

True professional growth happens in high-density, low-volume environments where peer-level vulnerability is possible.

06 — Lenny's Newsletter

The Human in the Loop

Maintaining agency in an era of autonomous systems

By Lenny Rachitsky · 9 min read

Editor's note: As machines take over the 'how', humans must become masters of the 'why'.

The central tension of the AI revolution is the struggle for agency. As agents become more capable of executing complex, multi-step tasks, there is a natural tendency for the human to drift into a passive role—a mere observer of a process they no longer fully control. This is the 'black box' problem. If we simply give an instruction and wait for a result, we have surrendered our ability to intervene, to correct, and to steer. To maintain agency, we must change how we interact with these systems. We cannot treat them as magic boxes; we must treat them as collaborators that require constant, structured engagement.

Verification vs. Testing

This requires a shift in our technical mindset. Traditional software testing—writing unit tests to check if a specific function works—is being supplemented, and in some cases replaced, by 'verification'. Verification is about checking outcomes and rubrics rather than just lines of code. It involves managing agents that check other agents, and using visual evidence (like video or interactive demos) to ensure the work meets the intent. This is a more holistic, higher-level form of control. It is less about checking the plumbing and more about ensuring the house is actually livable.

The goal is not to automate the human out of the process, but to use automation to elevate the human to a higher level of oversight.

To do this effectively, we need better interfaces. This brings us back to the importance of HTML and interactive specs. If the AI's work is presented in a way that is easy to critique, the human remains an active participant. If the work is presented as a massive, unreadable text file, the human becomes a spectator. Agency is maintained through legibility. The more transparent and interactive the AI's planning and execution become, the more effectively we can direct its power.

Strategies for Maintaining Agency

Prioritise interactive interfaces for AI outputs
Shift from code-level testing to outcome-based verification
Focus on high-level strategic constraints rather than low-level syntax
Use visual artifacts to audit complex plans

The future belongs to those who can direct the machine without being subsumed by it. We must become the architects of the intent, the curators of the outcome, and the final arbiters of the value. By mastering the tools of communication and verification, we ensure that as the machines get smarter, we get more capable, not less.

Key Takeaway

Agency in the age of AI is maintained through the creation of legible, interactive, and verifiable workflows.