Garry Tan: "Personal AGI Is How You Stay Under Your Own Power"

Spinoza's Resilience and the Offer [00:00:00]

• The speaker begins by introducing Baruch Spinoza, a 17th-century philosopher excommunicated by his community for his ideas.
• Einstein famously believed in "Spinoza's God," who reveals himself in the "lawful harmony of the world."
• Spinoza's community offered him a substantial salary to remain silent, which he refused, prioritizing truth over comfort.

• "He wanted truth, not comfort."

The Cost of Ideas [00:00:00]

• Spinoza was nearly killed by a fanatic with a knife shortly before his excommunication, an event he chose to remember by keeping the damaged cloak.
• After his excommunication, Spinoza supported himself by grinding lenses, creating optical instruments sought after by scientists.
• He also secretly wrote a dangerous book, published posthumously, which influenced Enlightenment philosophers.

• "He wanted to remember what ideas cost."

Spinoza's Relevance to Startups [00:00:00]

• Spinoza's response to being ostracized and nearly killed was to build precision tools and write a dangerous book, serving as a model for startup founders.
• His driving force was "Conatus," the striving in every living thing to keep going and increase its power to act.
• The talk aims to explore tools that amplify this striving.

• "The striving itself."

Spinoza's Heresy and Modern Parallels [00:00:00]

• Spinoza's heresy was the idea that God is not a king but spread through everything (God or Nature).
• The speaker draws a parallel to the current understanding of intelligence, where AGI is anticipated as a singular event rather than a diffused infrastructure.
• AGI is already present, not as a divine entity, but as infrastructure like terminal windows, markdown files, or background jobs.

• "God is spread through everything that exists, God or nature, he wrote."

Personal AGI vs. Corporate AGI [00:00:00]

• Personal AGI is defined as general intelligence for one person, running on their infrastructure, owning their memory, and executing their procedures.
• This contrasts with corporate AGI, often marketed as chatbots or assistants, which are rented, reset when tabs are closed, and controlled by companies.
• Personal AGI improves daily with use, becoming a personal asset, while corporate AGI improves only when the company releases updates.

• "One of these is a product you consume. The other is an asset you build."

The Exponential Multiplier Effect [00:00:00]

• The speaker recounts a personal experience of a 400x increase in code output due to AI agents, even after applying conservative estimates.
• This multiplier effect applies to all knowledge work, not just coding.
• YC portfolio companies using AI agents for everything are showing unprecedented growth and profitability.

• "The multiplier for coding is not just for coding. It's for every piece of knowledge work."

Joy as Increased Power of Acting [00:00:00]

• Spinoza's definition of joy is "the feeling of your power of acting increasing."
• The first time an agent completes a week's worth of work in an afternoon evokes this feeling of joy because your capacity has expanded.
• Sadness is defined as the feeling of your power of acting decreasing, akin to quiet quitting.

• "In ethics, he defines joy as the feeling of your power of acting increasing."

The Equation for the Next Decade [00:00:00]

• The formula for the next decade is: Rented Frontier Model + Owned Context (unique to you) + Harness (e.g., OpenClaw, Hermes) = Personal Agent.
• This creates an agent that acts like a faster version of yourself.
• Technology like computers are extensions of the mind (bicycle for the mind), and this setup is a "self-driving rocket."

• "Model quality is rented, but your brain is owned ideally by you."

Working Memory and AI Agents [00:00:00]

• Humans have a limited working memory (7 +/- 2 items), leading to prosthetics like org charts and checklists.
• AI agents, however, can hold millions of tokens (thousands of pages) and synthesize information rapidly.
• The key question is who or what decides which "three books" are open on the agent's desk—the librarian function.

• "Three Harry Potter books versus seven digits. You could argue that's not quite AGI yet, but it is already a different operating regime."

G-Brain: A Personal Knowledge Library [00:00:00]

• The speaker describes building "G-Brain," a personal knowledge wiki with hundreds of thousands of markdown pages, indexing 25 years of their life.
• Agents compile, curate, and search this library, ensuring no question is re-asked and providing context for every action.
• When a founder emails with a crisis, the agent pulls up all prior conversations, relevant portfolio company data, and potential solutions.

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

Skill Files: Markdown as Code [00:00:00]

• Skill files are simple pages of English instructions that agents can execute, making markdown effectively code.
• These files define specific tasks, like processing meeting recordings, transcribing them, identifying commitments, and cross-referencing people.
• The test for a skill file is whether a smart intern could follow it; if so, an agent can run it.

• "If a smart intern could follow it, an agent can run it."

Computation: Latent vs. Deterministic Space [00:00:00]

• Computation occurs in two spaces: latent (taste, judgment, steering models with markdown) and deterministic (arithmetic, SQL queries).
• Confusing these leads to agent failures; for example, seating 6,000 people requires deterministic computation, not just latent judgment.
• The fix is having the model compute like humans, using markdown files to call code and databases.

• "The model fails where we fail. The fix is having the model compute the way humans compute."

The Compendium Skill and Building Your Own [00:00:00]

• A "compendium skill" can synthesize vast amounts of information, like biographies, into a usable story with key moments and citations.
• This personal skill is a "mega version" of deep research, far beyond corporate AI products.
• Building your personal AGI starts not with a warehouse, but with a single shelf—a folder of markdown files and a few skills.

• "1,500 pages became a stage ready story that I could edit."

Managing a Workforce of Agents [00:00:00]

• Skill files are like employees with one capability, and resolvers act as an org chart directing tasks to the right skill.
• This allows individuals to run an organization of one plus their agents, acting as founder and management layer.
• Companies built this way are achieving unprecedented revenue per person.

• "You are the founder and the entire management layer of you incorporated and the headcount under you is now whatever you decide it is."

The Double-Edged Sword of Skills [00:00:00]

• Skill files are pieces of your cognition externalized; they represent your judgment and expertise.
• If these files are owned by a company, your expertise can be extracted, leaving you with nothing when you leave.
• The doctrine is: "Own your skills," because if you don't, your job becomes a skill file owned by someone else.

• "Your name isn't even in the commit history. She didn't have a career. She had an extraction."

Personal AGI as Staying "Under Your Own Power" [00:00:00]

• Spinoza declined an offer of a professorship with conditions, choosing to protect his freedom of philosophizing.
• Personal AGI is about controlling your own cognitive abilities and staying "under your own power," not someone else's.
• The speaker advocates for keeping your brain and skills in a repo you control from day one.

• "Your power of acting exists either way. The political question in 1673 and in 2026 is who commands it."

Objections and Rebuttals [00:00:00]

• Objection 1 (Model Improvement): Better models make your unique library (context) more valuable, as a smarter reader extracts more from the same books.
• Objection 2 (Is it just RAG?): Retrieval is a primitive; the value is in what's written, enriched, linked, and arbitrated—being worth retrieving from.
• Objection 3 (Security/Leakage): Consolidating context on your own infrastructure with your keys is a security model; the default is your data scattered across insecure clouds.

• "A better model makes your library worth more because a smarter reader extracts more from the same books."

The Renaissance of Open Source Tools [00:00:00]

• The speaker open-sourced personal AGI tools because powerful tools should be given away, fostering a renaissance rather than a priesthood.
• This technology, the harness, library, and markdown workforce, is the private technology of leverage for this era.
• The gap between those who have it and those who don't is widening.

• "When something like that, that powerful, stays private, you get a priesthood. When it gets given away, you get a renaissance."

Building for Impact: A Father's Repo [00:00:00]

• A father built a repo of 80,000 markdown files for his son with a rare form of epilepsy, indexing every piece of medical knowledge.
• This personal AGI allows doctors to quickly check if new ideas have been tried, demonstrating the power of focused personal intelligence.
• Nobody was coming to build this for him, so he built it himself.

• "A father, a laptop, and a library. That is personal AGI, not a benchmark, not a demo."

The Collapse of Difficulty [00:00:00]

• The old limitations—needing a team, funding, or permission—have expired with personal AGI.
• Individuals can now "boil the ocean" with a laptop and their own history.
• The difficulty of building something excellent has collapsed, making rarity a matter of individual choice and action.

• "All things excellent are as difficult as they are rare. The difficulty just collapsed. The rarity is now up to you."