Software as Literacy: Insights from Sequoia AI Ascent 2026

On my commute today, I caught up on some of the sessions from Sequoia Capital’s AI Ascent 2026. Two interview sessions in particular—featuring Boris Cherny and Andrej Karpathy—offered a profound look at how the nature of software development is fundamentally changing. While they participated in separate Q&A-style interviews, their perspectives formed a coherent vision of a future where coding shifts from a specialized craft to a universal form of literacy.

Boris Cherny: The Printing Press Moment for Software

Boris Cherny (co-founder of Anthropic and creator of Claude Code) participated in a session titled “Why Coding Is Solved, and What Comes Next.” The format was split between an insightful interview with the host and a direct Q&A with the audience. During the discussion, he shared a powerful analogy: software development is the new literacy.

The “Scribe” Era of Software

Boris framed our current state of software engineering through a historical lens. Before the invention of the printing press, literacy was a specialized, elite skill held by scribes and clergy. Kings relied on these professionals to handle their records. Today, we are in the “Scribe Era” of software. If you want to build an app, you hire a “scribe”—a professional software engineer—who understands syntax and compilers.

The Printing Press Moment

Boris argues that AI is our generation’s printing press. The printing press made literacy universal; reading and writing shifted from a “professional service” to a “basic human skill.” We are moving toward a world where “writing an app” will be as common as writing an email.

Staggering Metrics of 2026

To illustrate how close we are, Boris shared some personal metrics that blew my mind:

• Zero Lines of Hand-Written Code: Boris revealed that he has not written a single line of code by hand in 2026. Everything he ships is 100% model-generated.
• The 150 PR Day: He described a workflow using autonomous agents (like Claude Code) that allowed him to ship 150 pull requests in a single day, often managing these agentic loops directly from his phone.

Andrej Karpathy: From Vibe Coding to Agentic Engineering

In a separate, dedicated interview titled “From Vibe Coding to Agentic Engineering,” Andrej Karpathy spent the entire session in a deep-dive Q&A with the host, providing a complementary perspective from the builder’s front lines.

Software 3.0: The Context Window as the Program

Karpathy extended his famous Software 1.0/2.0 framework to Software 3.0.

• Software 1.0: Explicit code (Python, C++).
• Software 2.0: Neural network weights.
• Software 3.0: Programming via prompts and context. In this paradigm, the LLM is the interpreter, and the content of your context window is the program.

“Vibe Coding” vs. “Agentic Engineering”

Karpathy distinguished between “Vibe Coding”—the exploratory magic of describing a feature and seeing it appear—and “Agentic Engineering.” Professional development in the AI era requires the latter: a rigorous discipline where humans maintain architectural judgment and use agents to execute complex plans. As he famously put it: “You can outsource your thinking, but you can’t outsource your understanding.”

AI Psychosis and “Progress Anxiety”

One point Karpathy has reflected on recently (and touched on in the No Priors podcast) is “AI Psychosis”—the intense, anxious drive to be “in the flow” with AI every day. I felt a powerful resonance with this. There is a specific “progress anxiety” in this era: the fear of falling behind if you aren’t constantly upgrading your AI mental models. This “builder energy” was actually the spark for my own experiments with Voice Coding.

The 10-Year Lag: Digital Speed vs. Physical Friction

Karpathy has often cautioned that while the AI revolution feels instantaneous in the digital world, its impact on the physical world will be a “slow burn.” He famously uses Waymo as a case study: despite having fully autonomous cars operational for years, the reality of global, ubiquitous self-driving remains years away due to the immense engineering challenges and real-world friction. He estimates that widespread AGI impact in the physical world might still be a decade away.

However, as this session made clear, the digital world is moving far faster than we imagined. We might be in the “decade of agents” for physical robotics, but in the software layer, the “Software 3.0” revolution is already here. Karpathy uses a striking analogy to explain this gap: “Ghosts vs. Animals.” Current AI models are “digital ghosts”—superhuman in the ethereal realm of code and text—but they lack the millions of years of “preloaded software” that biological animals have to navigate the messy physical world. This highlights a fascinating duality: while we wait for the physical world to catch up, our digital lives are undergoing a total transformation in real-time.

The Death of the “Niche App”

Karpathy shared a humbling lesson from his project, MenuGen. As base models became more capable, the specialized app itself became redundant. This suggests a world where “middlemen” apps melt away, leaving only user intent and model capability.

AI as the Digital Immune System: Jensen Huang’s Perspective

Adding to this discussion, I recently watched a CNBC interview where Jensen Huang (Nvidia CEO) offered a brilliant analogy: AI as white blood cells. He argued that the best defense against malicious AI is “good AI” that monitors and defends our infrastructure. In this vision, AI becomes a scalable, digital immune system that evolves alongside threats.

Why This Matters

This shift suggests that the bottleneck for innovation will no longer be the ability to write code, but rather domain expertise and taste. Coding is evolving from a craft of syntax into a craft of intent and orchestration. Whether it’s Boris’s “universal literacy” or Jensen’s “digital immune system,” it’s clear that AI is moving from a specialized tool to a fundamental layer of our reality.

Inspired by Sequoia Capital AI Ascent 2026 and recent industry reflections.