Garry Tan: Personal AGI Is How You Stay Under Your Own Power
Our read
The coming decade is not an arms race between corporate AI gods, but a class war between those who rent their intelligence from OpenAI and those who own their local infrastructure.
What happened
At Y Combinator's Startup School, Garry Tan delivers a defiant manifesto for the age of AI agents. Drawing a direct line from the seventeenth-century excommunication of philosopher Baruch Spinoza to modern startup building, Tan argues that the ultimate divide of the coming decade is between renting corporate chatbots that get lobotomized on someone else's schedule, and owning a 'Personal AGI', a compounding, private library of markdown 'skill files' running on your own infrastructure. By externalizing your unique cognitive processes into tools you control, you unlock a 400x productivity multiplier and retain custody of your own intellectual labor.
The brief
The tech industry is sleepwalking into a digital feudalism where you rent your brain back from corporate landlords, making local cognitive ownership the only real survival strategy.
Key findings
The 400x productivity multiplier of 2026 is realized only by engineers who run automated agents and tasks locally while they sleep.
Leaving your troubleshooting judgment in a corporate repository allows an employer to extract and run your mind long after you resign.
The sides
- The Spinoza Precedent 0:08
True innovators must build without permission and resist corporate or societal pressure to conform or silence their work.
Evidence: Baruch Spinoza rejected a massive salary to stop building, ground lenses by day, and secretly wrote the most dangerous book in Europe by night.
- Ownership vs. Rental 6:45
Renting AI models is a trap that resets your progress, while owning your context creates a compounding personal asset.
Evidence: Corporate chatbots are slightly better autocompletes tied to company pivots, whereas a Personal AGI runs on your infrastructure and learns your unique context daily.
- The 400x Productivity Multiplier 8:01
Personal AGI and coding agents amplify individual output to a degree that makes traditional team structures obsolete.
Evidence: Tan went from shipping 14 useful lines of code a day in 2013 to achieving a 400x output multiplier in 2026 using automated agents.
- The Extractive Risk of Corporate Repos 29:05
If you do not own your skill files, your unique cognitive labor will be permanently extracted by your employer.
Evidence: A support engineer who leaves her skill files in a corporate repo leaves with nothing, while the company continues to run her judgment without her name in the commit history.
Quotes
“AGI isn't arriving as an event. It's arriving diffused. As your agent. Running on your context, doing your work.”
Garry Tan · 5:57
“Markdown is actually code now. The compiler is a language model.”
Garry Tan · 17:47
“Own your skills, or your job becomes a skill file.”
Garry Tan · 30:59
“A brain nobody curates is a garbage dump with great search.”
Garry Tan · 23:39
Why now
The prevailing narrative of Artificial General Intelligence is a centralized, corporate-controlled singularity, a god-in-a-box that you will eventually rent for $20 a month. Garry Tan's address at Y Combinator's Startup School 2026 completely dismantles this passive consumer model.
Tan frames the immediate reality of AI as Personal AGI: a highly diffused, private infrastructure consisting of a terminal window, a folder of markdown files, and automated jobs that run while you sleep.
This shift from centralized utility to personal leverage is a modern repeating of the struggle for intellectual autonomy.
Tan invokes the historical precedent of Baruch Spinoza, the seventeenth-century philosopher who was excommunicated, survived an assassination attempt, and rejected corporate bribes to stop building.
Spinoza ground optical lenses by day to maintain his independence while secretly writing his philosophy by night. The modern equivalent of grinding your own lenses is maintaining custody of your AI context.
If you do not build your own private library of automated workflows, you face a quiet, digital dispossession. Tan illustrates this with the story of a hypothetical support engineer.
If she documents her unique troubleshooting judgment into 'skill files' stored in her company's repository, she leaves with nothing when she resigns; the company simply continues running her automated mind without her.
If she keeps custody of those markdown files in her private repository, her personal leverage compounds everywhere she goes. In the age of agents, you must own your tools, or your cognitive labor will be permanently extracted.
Questions
What is the difference between Personal AGI and corporate AI?
Personal AGI is an intelligent agent system that runs on your private infrastructure, reads your unique personal memory, and executes your specific procedures. In contrast, corporate AI is a rented chatbot subscription that lacks personal compounding, does not retain your unique context, and is subject to being lobotomized or altered on someone else's schedule.
How does markdown function as code in the age of AI agents?
Markdown functions as code because modern large language models act as compilers. By writing clear, step-by-step instructions in plain English within a markdown file, you create a executable 'skill file' that an AI agent can read, interpret, and run to automate complex, multi-step digital workflows.
Why is owning your AI context more important than the underlying model?
The underlying AI models are rapidly becoming commoditized utility engines. As model capabilities improve, the primary differentiator shifts entirely to context, your unique, private library of data, history, and custom skill files. Retaining custody of this context ensures that every new model release acts as a free, seamless upgrade to a workforce you already own.
What is the risk of keeping your automated workflows in a corporate repository?
If your automated workflows and skill files are stored in a corporate repository, your unique cognitive judgment is externalized and owned by the company. When you leave the organization, you leave with nothing, while the company continues to run your automated decision-making processes indefinitely without paying you or crediting your name.
How does Personal AGI solve the limits of human working memory?
Human working memory is biologically limited to holding roughly seven items at once, which is why organizations rely on complex management hierarchies and org charts. An AI agent can hold a million tokens of context in its working memory simultaneously, equivalent to three Harry Potter books open at once, allowing a single individual to manage a massive, automated workforce without institutional overhead.
Receipts
Related dispatches
- Patrick Collison on the Death of the Lean Startup and the Rise of Cognitive L1 CacheThe VC-endorsed obsession with launching fast and breaking things is a luxury reserved for companies building toys. To construct actual, load-bearing infrastructure for the global economy, founders must embrace massive operational schleps and treat their own minds as high-performance hardware rather than outsourcing basic reasoning to a thin-client AI prompt.
- The Dumb Database Trap: Why SaaS Giants are Gatekeeping the AI Agent RevolutionSalesforce is locking the API because an agent that reads your CRM makes the seat-licensed UI look like a tollbooth.
- The Safe Path Inversion: Y Combinator on India's New AI LeverageThe credentialed corporate ladder has inverted into a career death trap, leaving prestige-chasing graduates exposed to automation while dorm-room hackers run high-volume token loops to build and sell enterprise software directly to the West.
- Jensen Huang on the Brutal Art of Corporate SurvivalNVIDIA did not scale to a multi-trillion-dollar empire through a flawless master plan. It survived because Jensen Huang had the humility to burn his original proprietary architecture to the ground, beg Sega for a life-saving bailout after failing to deliver on a contract, and run the company in an uncompromised, hands-on Founder Mode for over three decades.
- Karpathy's 2026 Playbook: Build Agent-First or PerishMost of the AI apps you're building right now? Dead on arrival. Andrej Karpathy's 2026 playbook is a brutal obituary for 'vibe coding,' demanding builders ditch their flimsy 'Software 1.0 plumbing' and embrace 'agentic engineering' in verifiable niche domains, or get eaten by the next LLM release.
- Gusto Co-Founder: Solving the Blank Canvas Problem with Agentic DatabasesSmall business operators do not want custom app builders or open-ended prompt windows; they want their existing database of record to quietly automate its own transactions via text message.
Visual-only receipts
- 0:51 - Black and white image of Albert Einstein with the quote: 'I believe in Spinoza's God.'
- 1:37 - On-screen text of Spinoza's harsh 1656 excommunication decree from Amsterdam.
- 8:02 - Log scale graph showing the 400x productivity multiplier from 2013 to 2026.
- 11:20 - On-screen equation defining an agent as: frontier model (rented) + your context (owned) + a harness.