AI is Already Here: A Shift to Agent-Driven Productivity, Not a God-Like Event

A substantial portion of the Winter 2025 Y Combinator batch has reported codebases that are 95% AI-generated, signaling a significant shift in how startups operate. These companies are not just using AI for code; they are deploying AI agents across all facets of their operations. This high degree of AI integration is positioning the current cohort to be one of the most profitable in YC's history, demonstrating the profound impact of AI on startup efficiency and scalability.

The speaker introduces Baruch Spinoza, a 17th-century philosopher, as a historical parallel for understanding current disruptive shifts. Spinoza's radical ideas directly challenged traditional religious authority during his time. His philosophical pursuits led to one of the most severe excommunications in history, highlighting the profound impact of challenging established norms in pursuit of truth. The historical context provided frames the current discussion on AI as a similar paradigm shift.

"His name was Baroo Spinosza. And in case the philosophy elective wasn't your thing, here are the highlights you need to know."

Spinoza’s commitment to his intellectual pursuits was so profound that he rejected an offer of 1,000 guilders per year from his community, which was intended to silence his controversial ideas. He firmly declined this financial comfort, explicitly stating his preference for truth over security. This anecdote underscores Spinoza's dedication to intellectual integrity, a trait the speaker implicitly links to the drive necessary for current innovation.

"They've offered him a salary to stop building. He said no. Not for 10,000. He said he wanted truth, not comfort."

To sustain himself, Spinoza earned a living by grinding lenses, crafting optical instruments that extended human vision. While making these tools by day, he secretly dedicated his nights to writing revolutionary works, which were considered so dangerous that he could not publish them during his lifetime. This duality of building practical tools and pursuing profound, disruptive ideas in isolation serves as a metaphor for current innovators who develop powerful tools and concepts that challenge existing paradigms.

"He grinds lenses. By day, he makes optical instruments, tools that let human beings see further than their eyes allow. And by night, he writes a book so dangerous he cannot publish it while he is alive."

Spinoza coined the term 'Conatus' to describe the inherent striving within every living thing to persist, to keep going, and to increase its power to act. This fundamental drive, according to the speaker, is a crucial internal engine that propels individuals beyond external validation like titles or resumes. It is the core motivation that fuels continuous action and growth, an essential quality for innovators and founders in the current rapidly evolving landscape.

"He had a name for the engine that kept him going. Canadus it means your striving the drive in every living thing to keep going and to increase its power to act"

Society largely anticipates Artificial General Intelligence (AGI) as a monumental, 'god-like' event that will manifest in a single data center. However, the speaker argues that AGI is already present, not as an all-encompassing entity, but as diffused agents operating within individual contexts and data. These agents appear as everyday infrastructure, such as terminal windows, folders of markdown files, or background jobs that complete tasks while users sleep. This redefinition suggests that AGI is an integrated, ongoing process rather than a singular, future arrival.

"Everyone is watching the sky and the thing they're watching for is already in the room. It doesn't look like a god. It looks like infrastructure, a terminal window, a folder of markdown files, a job that finishes while you sleep"

The speaker quantifies a massive leap in personal productivity, noting a 400x increase in output compared to 2013, attributing this to AI-agent workflows. This significant efficiency gain is observed across various roles including design, product development, and growth. This surge in productivity is exemplified by the fastest-growing YC companies, which are reportedly building their products with 95% AI-generated code, highlighting the transformative power of AI in streamlining and accelerating development cycles.

"I did the math on my output, and I'm at about 400x what I did in 2013." — Speaker

The most successful founders funded by Y Combinator do not view AI as a mere autocomplete tool but rather as a genuine workforce. This distinction is critical for achieving substantial efficiency gains, moving from a 2x to a 100x improvement in operations. The key to leveraging AI effectively lies in providing relevant context to the agents, rather than focusing solely on model weights, enabling them to act as knowledgeable and productive partners.

"The fastest growing founders we fund are not treating AI as autocomplete. They are treating it as a workforce." — Speaker

Spinoza defined joy as the feeling of one's power of acting increasing, and conversely, sadness as the feeling of one's power decreasing. Applying this philosophy to AI, the speaker explains that when an AI agent accomplishes a week's worth of work in a single afternoon, the feeling evoked is not just convenience, but profound joy. This emotional response stems from the significant augmentation of one's capacity to act, making AI agents a source of genuine empowerment.

"In ethics, he defines joy as the feeling of your power of acting increasing, which is why the first time an agent does a week of your work in an afternoon, it doesn't feel like a convenience. It feels like joy." — Speaker

AI agents are fundamentally altering the physics of startup growth by enabling a single founder to perform tasks that were previously considered unscalable. This effectively multiplies the impact of Paul Graham’s enduring advice to 'do things that don’t scale,' transforming it into 'do unscalable things at scale.' Agents serve as a powerful multiplier, allowing founders to extend their reach and capacity far beyond traditional human limitations in the early stages of a startup.

"Agents are how one founder now does unscalable things at scale." — Speaker

Human beings typically can hold only about seven items in their working memory at any given time, a limitation often cited as seven plus or minus two. Historically, every institution, from libraries to corporations, has functioned as a prosthetic to overcome this fundamental biological constraint. In stark contrast, an AI agent can manage a million tokens, effectively possessing the capacity of an entire library, thereby transcending human cognitive limits. This vast memory capacity allows AI agents to handle complex tasks that are beyond individual human capabilities.

"You and I as human beings hold about seven things in our head at once. Seven plus or minus two... An AI agent though holds a million tokens." — Speaker

The concept of 'GBrain' is introduced as a comprehensive personal knowledge system, functioning as both a vast library and an intelligent librarian. The speaker's personal 'GBrain' stack consists of 220,000 markdown pages, accumulating data from 25 years of their life, including emails and meeting notes. This extensive digital repository enables the system to store and intelligently retrieve a lifetime of accumulated information, making it a powerful tool for personal knowledge management.

"That's what GBrain is meant to be. The library plus the librarian." — Speaker

The crucial distinction between an assistant and a colleague lies in an agent’s ability to leverage 'lived experience' to deeply understand context. When an agent acts, it does so with full knowledge of the user's entire history and past conversations, drawing upon a comprehensive understanding of what the user knows. This enables the agent to provide highly relevant and informed solutions, effectively operating as a knowledgeable colleague rather than merely following instructions like an assistant.

"When my agent does anything, it does knowing everything I know. And that's the difference between an assistant and a colleague." — Speaker

The architecture of agentic systems is remarkably straightforward, primarily consisting of 'skill files' and browser access that agents can drive. These skill files are essentially English-language instructions that guide the agents' actions and interactions with the digital world. The underlying principle is summarized as 'fat skills, thin harness,' emphasizing that the complexity resides in detailed, human-readable instructions rather than intricate magical code, making the system accessible and manageable.

"It's mostly skill files plus a browser that the agents can drive pages of English and a way to act on the world. Markdown, not magic." — Speaker

In a real-world application, an AI agent successfully synthesized 1,500 pages of philosophy books overnight into a stage-ready story, which the speaker could then edit. This output included detailed chronologies, identified disagreements, and ranked anecdotes. The speaker refers to this capability as a 'compendium skill' and uses it daily, noting that even the structure of the current talk was derived from this agent-run process, demonstrating its powerful utility in personal knowledge management and content creation.

"1500 pages became a stage ready story that I could edit. I call it a compendium skill and I use it daily." — Speaker

The role of a founder is shifting from coding to managing a workforce made of markdown, where a 'skill file' operates as a single-task employee. A 'resolver' acts as an organizational chart, assigning these tasks to appropriate agents, effectively coordinating their efforts. This new paradigm allows founders to operate an entire organization comprising themselves and a team of AI agents, thereby enabling unprecedented scalability and operational efficiency without traditional human resource overhead.

"you're not coding. You're man managing a workforce made of markdown." — Speaker

An AI brain that is not properly curated effectively becomes a 'garbage dump with great search,' highlighting the critical need for diligent management. Poorly constructed skill files can permanently encode inefficient processes, leading to suboptimal performance. Therefore, a librarian-like approach to pruning and organizing information within the AI system is essential to maintain its effectiveness and ensure it operates with clean, relevant data, preventing the accumulation of digital clutter.

"A brain nobody curates is a garbage dump with great search." — Speaker

A core discipline in leveraging AI effectively is the commitment to never performing 'one-off work.' This means that any task, if requested more than once, signifies a failure to properly 'skillify' it. The goal is to transform every task into a reusable skill, ensuring that knowledge and processes are permanently captured and automated for future use. This continuous learning and systematization are mandatory to prevent 'amnesia' in the workflow, allowing for consistent improvement and efficiency. By embedding tasks into reusable skills, individuals and organizations can build an ever-growing repository of automated capabilities, optimizing their operations and eliminating repetitive efforts.

"Never do one-off work." — Speaker