Garry Tan: "Personal AGI Is How You Stay Under Your Own Power"
Y Combinator
2,854 views • 1 month ago Save 33 min 9 min read
Video Summary
Garry Tan argues that "personal AGI"—intelligence tailored to an individual's context and owned by them—is the next frontier, offering unprecedented leverage and autonomy. He draws a parallel to the 17th-century philosopher Baruch Spinoza, who was excommunicated and ostracized for his radical ideas but persisted in his work, grinding lenses by day and writing dangerous philosophy by night. Spinoza's "conatus," or striving, is presented as the engine for this personal AGI.
Tan contends that AGI is not a singular event but a diffusion into personal agents that run on an individual's infrastructure, learning from their unique data. This contrasts with "corporate AGI"—rented services that reset and can be altered by company decisions. He illustrates the power of personal AGI with his own experience, claiming a 400x increase in productivity through AI agents, and highlights how companies built on this principle are achieving exponential growth. The core idea is that owning your cognitive abilities, encapsulated in "skill files," is crucial for maintaining personal power and autonomy in an increasingly agent-driven world.
Short Highlights
- Personal AGI is the Future: It's not a singular event but an agent running on your infrastructure, learning from your unique context, and owned by you.
- Spinoza's Legacy: The philosopher Baruch Spinoza, excommunicated for his ideas, embodied "conatus" (striving) by building tools and writing dangerous books, serving as a model for personal AGI.
- Own Your Skills: "Skill files"—executable instructions in English—represent your cognition, and owning them is crucial for autonomy, unlike corporate AGI which is rented.
- Productivity Multiplier: AI agents can provide a significant productivity boost across all knowledge work, not just coding, enabling individuals and companies to operate at a vastly different scale.
- The Power of Context: While models improve, the true differentiator lies in the unique, owned context (your library) that personal agents leverage.
- Building Your Brain: Start by creating a personal library of your data and writing skill files for repetitive tasks to build an organization of one.
- Autonomy Over Renting: Personal AGI offers ownership and control over your cognitive abilities, preventing your judgment from being exploited by others.
Key Details
The Excommunication of Spinoza [0:00]
- Baruch Spinoza, a 17th-century philosopher, was excommunicated by his Sephardic Jewish community in Amsterdam for "evil opinions."
- He was offered a substantial salary to remain silent but refused, prioritizing truth over comfort.
- Spinoza's punishment was complete social and economic deletion, a ban that technically remains in force today.
"Cursed be he by day, and cursed be he by night, cursed be he when he lies down, and cursed be he when he rises up."
Spinoza's Response: Building and Writing [1:31]
- After his excommunication, Spinoza became a lens grinder, creating optical instruments sought after by European scientists.
- He also wrote a dangerous book, published posthumously, which influenced Enlightenment philosophers.
- Spinoza's life demonstrates building precision tools by day and writing profound works by night, alone and without permission.
"He wanted to remember what ideas cost."
The Engine of Striving: Conatus [3:14]
- Spinoza identified "conatus" as the inherent drive in all living things to persist and increase their power to act.
- This striving, not external validation like resumes or titles, is the core engine for personal growth and achievement.
- The talk's focus is on tools that amplify this fundamental drive.
"It means your striving, the drive in every living thing to keep going and to increase its power to act."
The Heresy of God and Intelligence [3:56]
- Spinoza's heresy was that God is not a distant king but is present in everything (God or Nature).
- Similarly, the current mistake is waiting for AGI as a singular, god-like event.
- AGI is arriving diffused, as personal agents running on individual infrastructure, not as a dramatic announcement.
"Everyone is watching the sky. And the thing they're watching for is already in the room."
Personal AGI vs. Corporate AGI [4:47]
- Personal AGI is an agent that runs on your infrastructure, uses your owned memory, and executes your procedures, compounding daily.
- Corporate AGI is a rented subscription service (like a chatbot) that resets, lacks true ownership, and is subject to company pivots.
- Personal AGI becomes an asset you build, while corporate AGI is a product you consume.
"One of these is a product you consume. The other is an asset you build."
The Productivity Multiplier [6:26]
- The speaker claims a 400x increase in personal output due to AI agents, even with conservative estimates.
- This multiplier applies to all knowledge work, not just coding, design, or product management.
- YC portfolio companies with high AI-generated codebases are showing remarkable growth, treating AI as a workforce.
"The number is large no matter how you torture it."
Joy as Increased Power of Acting [9:07]
- Spinoza defined joy as the feeling of one's power of acting increasing.
- The first time an agent completes a week's work in an afternoon, it feels like joy because your capacity has expanded.
- Sadness, conversely, is the feeling of power decreasing, akin to quiet quitting.
"It doesn't feel like a convenience. It feels like joy."
The Equation for the Next Decade [10:04]
- The core components are a rented frontier model, your owned and unique context (your library), and a "harness" to connect them.
- This creates an agent that acts as a fast version of you, amplifying your owned intelligence.
- Technology as an extension of man (McLuhan) or a "bicycle for the mind" (Jobs) becomes a "self-driving rocket."
"Model quality is rented, but your brain is owned ideally by you."
Working Memory and AI's Advantage [11:10]
- Human working memory is limited to about seven items (7+/-2).
- Institutions like org charts and checklists are prosthetics for this limitation.
- AI agents can hold millions of tokens (thousands of pages), allowing them to process and synthesize vast amounts of information instantly.
"Three Harry Potter books versus seven digits."
The Librarian Decides [13:04]
- The crucial question for an agent's effectiveness is who or what decides which information is prioritized (which three books are open).
- "G-Brain" is described as a "library plus the librarian," a personal knowledge system with 220,000+ markdown pages of the speaker's life.
- Agents use this comprehensive context to provide highly relevant assistance, turning an assistant into a colleague.
"The question that determines whether your agent is a genius or a goldfish is this. Who decides or what decides which three books are open on the desk?"
Skill Files: Markdown as Code [16:50]
- Skill files are simple pages of English instructions that agents can execute, like a "smart intern."
- If a human can follow the instructions, an agent can run them, making markdown effectively code.
- This empowers non-engineers (media, finance) to build automated workflows and manage agents.
"Markdown is actually code. If you can write clear instructions in English, you're a programmer."
Computation: Latent vs. Deterministic Space [20:11]
- Computation occurs in "latent space" (judgment, taste, model-driven) and "deterministic space" (arithmetic, database queries).
- Confusing these leads to agent failures; complex tasks require both, with markdown files orchestrating code execution.
- The example of scheduling 6,000 people shows the need for deterministic computation alongside latent judgment.
"The model fails where we fail. The fix is having the model compute the way humans compute."
The Compounding Library [22:56]
- The speaker's own talk was synthesized overnight by an agent from 1,500 pages of Spinoza biographies.
- This "compendium skill" demonstrates deep personal research capabilities far beyond corporate AI products.
- Personal libraries start small (a folder) and grow through daily compounding with agent assistance.
"Nobody builds the warehouse first. First you build one shelf."
Owning Your Skills: The Political Question [25:48]
- Skill files are pieces of your cognition; owning them means owning your expertise and career.
- If skill files are not owned by you, they belong to your employer, leading to "extraction" rather than career growth.
- This ownership is the key to staying "under your own power," as Spinoza protected his intellectual freedom.
"I believe skill files are yours. Own your skills."
Objections and Rebuttals [31:03]
- Objection 1 (Models improve): Better models increase the value of your unique library and context, not just the underlying weights.
- Objection 2 (Is it just RAG?): Retrieval is a primitive; the value is in what gets written, enriched, and curated – being "worth retrieving from."
- Objection 3 (Security/Leakage): Consolidating context on your own infrastructure with your keys is more secure than scattering data across multiple corporate clouds.
"Custody is the security model."
Renaissance Through Openness [34:41]
- The speaker open-sourced his personal OS because he believes powerful tools should be given away, fostering a "renaissance."
- Private leverage technology (like literacy, capital) creates priesthoods; open technology creates widespread empowerment.
- The gap between those using personal AGI and those not is widening rapidly.
"When it gets given away, you get a renaissance."
The Final Chapter: Personal AGI in Action [37:00]
- A father built a 80,000-markdown-file "brain" for his son with a rare form of epilepsy, pushing the boundaries of medical knowledge.
- This is personal AGI: a father, a laptop, and a library, driven by love and necessity, not permission or funding.
- The old barriers (team, funding, permission) are gone; the only limit is your striving and your ability to build.
"Nobody is coming to build that for him. So he built it. And nobody is coming to build yours for you."
The Collapse of Difficulty [39:50]
- "All things excellent are as difficult as they are rare"—Spinoza's closing words.
- The difficulty of building powerful personal tools has collapsed thanks to AI agents.
- The rarity is now determined by individual willingness to strive and build their own systems.
"The difficulty just collapsed. The rarity is now up to you."