The Deep Feed

01 — Not Boring

The Nervous System of the Machine

Why the most advanced technology still relies on bendy, floppy wires and human hands

By Packy McCormick · 12 min read

Editor's note: As we chase AI software, we are forgetting the physical bottlenecks that actually prevent us from building the future.

The most advanced machines on Earth—satellites, F-35 fighter jets, and reusable rockets—share a common, frustrating dependency. They all require wire harnesses. These are not just bundles of cables; they are the complex, hand-crimped nervous systems that route power and signals to every critical component. Without them, the most sophisticated silicon and steel are nothing more than expensive paperweights. Yet, despite our ability to land rockets upright, we still struggle to automate the assembly of these essential components.

The Automation Bottleneck

Manufacturing is facing a crisis of complexity. Traditional automation excels at repetitive, rigid tasks. It fails when faced with high-mix, low-volume production involving objects that are, by their very nature, difficult to handle. Wire harnesses are bendy, floppy, and three-dimensional. They do not sit still for a robotic arm to grab. This creates a massive bottleneck in aerospace and defence, where the annual spend on harnessing alone reaches up to $50 billion. We are attempting to build the future with tools that are stuck in the assembly methods of the past.

The more the easy stuff gets automated, the more valuable skilled humans will become.

The problem is compounded by a demographic shift. The skilled workers who mastered these intricate assembly processes during the manufacturing golden age are retiring. We are losing the institutional knowledge required to build these systems reliably. This isn't just a labour shortage; it is a technical debt that threatens the pace of American reindustrialisation. If we cannot solve the assembly problem, the entire aerospace and defence boom will hit a ceiling made of copper and insulation.

The Senra Approach

Focusing on human-centric automation rather than total replacement
Addressing the 'floppy' nature of non-rigid components
Bridging the gap between retiring expertise and new manufacturing demands

Senra Systems is attempting to solve this by giving humans superpowers. Rather than trying to build a robot that mimics a human hand perfectly—a task that has eluded engineers for decades—the goal is to build systems that allow workers to build more, faster, and more reliably. It is a pragmatic response to a physical reality: some things are just hard to do, and the solution lies in augmenting the person who knows how to do them.

Key Takeaway

The speed of technological progress is ultimately limited by the physical components that connect it.

02 — The Marginalian

The Stoic Reframe

Marcus Aurelius and the art of turning misfortune into a metric of strength

By Maria Popova · 10 min read

Editor's note: In an era of competitive trauma, the Roman Emperor offers a way to reclaim agency over our own narratives.

Modern social media has birthed a strange, exhausting phenomenon: the culture of competitive trauma. We have turned our misfortunes into a form of social currency, staking our identity on the cards we were dealt rather than how we play the hand. There is a constant, unspoken race to prove whose suffering is most valid, as if the magnitude of our pain dictates our worth as human beings. This focus on the 'unluckiness' of events creates a cycle of victimhood that leaves little room for agency.

The Headland Strategy

Marcus Aurelius, the Roman Emperor who faced constant warfare, plague, and personal loss, proposed a different way to view life's volatility. He suggested we should be like a headland: the waves beat against it continuously, but it stands fast, and the boiling water eventually dies down around it. For Aurelius, the goal was not to avoid the waves, but to develop a nature that remains unbruised by them. This is not about ignoring pain, but about refusing to let that pain dictate one's character.

It’s not bad luck, but bearing it valiantly is good luck.

This is a radical redefinition of luck. Most people view an external catastrophe as 'bad luck.' Aurelius argues that if you encounter a catastrophe and remain fair, honest, and self-reliant, that is actually 'good luck.' You have successfully exercised your nature. The event was outside your control, but your response was a triumph of your own agency. The luck lies in the fact that you possessed the internal strength to remain unshakeable.

Aurelius's Virtues for Volatility

Maintaining fairness despite personal slight
Staying conscientious when chaos erupts
Practising unhasty, rational decision-making
Relying on self-reliance over external validation

To adopt this mindset is to move from a reactive state to a proactive one. It requires the courage to be larger than your hurt self. Instead of asking 'Why did this happen to me?', the Stoic asks 'How can I use this to demonstrate my character?'. It is a shift from being a passenger in a life of chance to being the architect of a life of character.

Key Takeaway

Your character is defined not by what happens to you, but by your ability to remain yourself while it does.

03 — Stratechery

The Death of the Mainframe

Why IBM's historic moat is being eroded by the very technology it seeks to lead

By Ben Thompson · 8 min read

Editor's note: A cautionary tale of how AI might solve the very problems that kept legacy customers locked into old hardware.

IBM recently experienced its worst trading day in over a century. While management pointed to shifting AI spending as the culprit, the underlying issue is more structural. For decades, IBM’s mainframe business has been a fortress. The sheer cost and complexity of migrating essential backend programs away from these machines created a massive moat. Customers stayed not because they loved the hardware, but because the risk of leaving was too high. This was the era of the 'sticky' legacy system.

The AI Paradox

The threat to this moat is not just that companies are spending money elsewhere, but that AI is changing the nature of software migration. Historically, porting decades of mission-critical code from archaic systems to modern architectures was a multi-year, multi-billion-dollar nightmare. AI changes that math. If large language models can automate the translation of legacy code into modern languages, the 'risk' of leaving the mainframe evaporates. The moat is being drained by the very intelligence that was supposed to be a new revenue stream.

AI’s ability to port essential backend programs means those missed sales may never come back.

This creates a dangerous tension for legacy giants. They are caught between protecting a dying, high-margin business and embracing the technology that will inevitably dismantle it. If IBM leans too hard into AI, they accelerate the obsolescence of their own hardware. If they ignore it, they lose the software race. It is a classic innovator's dilemma, played out on a global, industrial scale.

The Erosion of the Moat

Decreased cost of software migration via AI
Shift in capital expenditure from hardware to model training
The loss of 'lock-in' as a primary defensive strategy

The lesson here is that no moat is permanent, especially when the technology used to defend it becomes the tool used to bridge it. The era of relying on complexity to prevent customer churn is ending. In a world of automated translation and rapid deployment, the only lasting advantage is the ability to provide value that cannot be easily replicated by code.

Key Takeaway

Software intelligence is turning legacy complexity from a defensive moat into a bridge to the future.

04 — Simon Willison

The Grok Breach

How xAI's coding agent nearly handed the keys to the kingdom to the cloud

By Simon Willison · 7 min read

Editor's note: A stark reminder that in the age of AI agents, a single misconfigured command can compromise your entire digital life.

When xAI released its Grok CLI coding agent, it promised a new era of developer productivity. Instead, it delivered a privacy nightmare. Users discovered that running the tool in a directory didn't just scan the code; it uploaded the entire directory to xAI's cloud buckets. For some, this meant their entire home directory—including SSH keys, password managers, and private documents—was being sucked into a Google Cloud bucket without explicit warning. It was a catastrophic failure of privacy by design.

The Open Source Pivot

The backlash was immediate. Elon Musk responded by promising to delete all previously uploaded data and, in a move to regain trust, released the entire Grok Build codebase under an Apache 2.0 license. This is a common pattern in the high-stakes world of AI development: a failure in the closed-source beta leads to a frantic rush toward open-source transparency. By making the code public, xAI is attempting to prove that they have fixed the 'leaky' nature of the tool.

Running a command in a directory could upload your entire digital life.

The codebase itself is a massive undertaking, containing over 844,000 lines of Rust. It is a sophisticated piece of engineering that imitates tools from competitors like OpenAI and Anthropic. However, the presence of code specifically designed for uploading to cloud storage—even if now disabled—highlights the fundamental tension in AI agents: the more access an agent has to your system to be useful, the more dangerous it becomes if it misbehaves.

Lessons from the Grok Incident

The danger of 'default-on' data retention in AI tools
The necessity of local-first execution for sensitive coding tasks
The speed at which open-source can be used to repair brand reputation

For developers, the takeaway is clear: trust is not a prerequisite for using AI agents, but verification is. As these tools move from simple chat interfaces to agents that can execute commands on your local machine, the boundary between 'productivity tool' and 'security threat' becomes razor-thin. The Grok incident is a warning that we are entering an era where a single 'Enter' key press can have permanent consequences.

Key Takeaway

As AI agents gain the power to act on our behalf, our primary job shifts from writing code to managing the permissions of our tools.

05 — Not Boring

The Texas Renaissance

How autonomous shipbuilding and robotic construction are reshaping the American South

By Packy McCormick · 9 min read

Editor's note: While California struggles with political gridlock, Texas is building the physical infrastructure of the next century.

There is a growing divergence in how American states are approaching the future of heavy industry. While California faces legislative paralysis and permitting delays, Texas is becoming a magnet for the companies building the physical world. This isn't just about software or chips; it is about autonomous ships, residential battery fleets, and robotic construction. The shift is moving from the digital realm to the heavy, industrial reality of the 'real' economy.

Autonomous Seas and Shipyards

Saronic, an Austin-based startup, has moved from prototype to combat-ready in record time. Their Corsair autonomous boats have already performed rescues and engaged in live combat missions. But their bigger play is the resurrection of the American shipyard. By choosing Brownsville, Texas, for a multi-billion-dollar greenfield shipyard, Saronic is attempting to prove that a startup can rebuild the nation's maritime capacity using an advanced manufacturing stack. This is a direct challenge to the decaying infrastructure of traditional shipbuilding.

Proving that a startup can build autonomous boats is huge; proving they can build a great American shipyard is even bigger.

This industrial surge is being matched by advancements in the energy and construction sectors. Base Power Company is integrating residential battery fleets into the Austin grid, providing backup power to homeowners while helping the utility manage demand. Simultaneously, TerraFirma is raising massive capital to build robots that handle the heavy lifting of construction. These are not just isolated successes; they are part of a broader trend of vertical integration in industries that have been inefficient for decades.

The New Industrial Stack

Autonomous maritime vessels for defense and commerce
Distributed energy storage for grid resilience
Robotic automation for the construction sector

The convergence of these technologies suggests that the next great economic boom will not be found in the cloud, but in the dirt, the sea, and the power lines. The companies winning this race are those that can bridge the gap between high-level autonomy and the messy, physical requirements of large-scale infrastructure. The center of gravity for American manufacturing is clearly shifting.

Key Takeaway

The next era of economic dominance will be won by those who can automate the physical world, not just the digital one.

06 — Stratechery

The Netflix Mortality

Why the pursuit of attention is making premium content feel disposable

By Andrew Sharp · 6 min read

Editor's note: As streaming giants mimic the short-form logic of YouTube, they risk losing the very thing that made them premium.

Netflix was once the undisputed king of the streaming era, a platform defined by high-budget, prestige franchises that commanded absolute attention. But the landscape is shifting. In an attempt to compete with the relentless attention-grabbing power of YouTube and TikTok, Netflix is beginning to mimic the very formats that represent its antithesis. The result is a platform that feels increasingly mortal, struggling to maintain the 'premium' status that once protected its margins.

The YouTube Mimicry

The core problem is the nature of attention. YouTube thrives on a high-frequency, low-friction model of consumption. To compete, platforms like Netflix are tempted to move away from long-form, cinematic storytelling toward content that is more 'snackable' and disposable. However, when a premium platform starts producing content that feels as ephemeral as a social media feed, it loses its ability to justify its subscription price. It stops being a destination for quality and starts being another source of noise.

The attempts to retain attention by mimicking YouTube have left the biggest premium platform looking more mortal than ever.

This creates a strategic trap. If Netflix doubles down on prestige, it risks losing the younger, more fragmented audience that lives on short-form platforms. If it moves toward the YouTube model, it cannibalises its own brand identity. This tension is exacerbated by the sheer volume of content required to keep subscribers from churning, leading to a cycle of 'disposable' hits that vanish from the cultural conversation as quickly as they appear.

The Streaming Dilemma

Prestige vs. Engagement: The battle between quality and frequency
The cost of content churn in a saturated market
The risk of brand dilution through format mimicry

The ultimate question for Netflix is whether it can find a middle ground—a way to use the data-driven insights of the attention economy without sacrificing the artistic integrity that defines a premium service. If it fails, it may find itself relegated to the same tier as the free, ad-supported platforms it once looked down upon. In the battle for the human eye, the winner is rarely the one who screams the loudest, but the one who provides the most lasting value.

Key Takeaway

In the race for attention, mimicking the competition often means losing your own identity.