Personal AGI Blueprint: Spinoza’s Conatus to Your AI Workforce

The speaker, self-proclaimed as one of the most "AI psychotic people online," draws parallels between the 17th-century philosopher Baruch Spinoza and the modern era of Artificial General Intelligence (AGI) and startups.

The Story of Baruch Spinoza

Spinoza, at 23, was excommunicated by his tight-knit Sephardic Jewish community in Amsterdam in 1656 for "evil opinions" and "forbidden thoughts." This ban, unique for its lack of a repentance clause, remains technically in force today. Before his excommunication, he was offered a substantial sum of 1,000 guilders a year to simply attend synagogue and remain silent, an offer he refused, stating he sought truth, not comfort. Shortly before this, he survived a knife attack by a fanatic, keeping the torn cloak as a reminder of the cost of ideas.

After his excommunication, Spinoza supported himself by grinding lenses, creating optical instruments so precise that Europe's leading scientists sought them out. By night, he wrote a book so controversial that it could only be published posthumously. Upon his death at 44, his manuscript, "Ethics," was sent to a publisher and went on to inspire many Enlightenment philosophers. Albert Einstein later famously stated his belief in "Spinoza's God," who reveals himself in the lawful harmony of the world, rather than a God concerned with human affairs.

The speaker highlights Spinoza's resilience: canceled by his community, offered a bribe to stop, and nearly killed, he responded by building precision tools and writing a dangerous book, alone and without permission. This spirit, driven by what Spinoza called "conatus" – the striving in every living thing to keep going and increase its power to act – is presented as a model for startup founders.

Spinoza's Heresy and Modern AGI

Spinoza's "heresy" was the idea that God is not a king on a throne but is spread through everything that exists, equating God with nature. The speaker argues that a similar mistake is being made about intelligence today, with people awaiting AGI as a singular, god-like event.

The speaker's updated version of Spinoza's heresy is that AGI is not a future event but is "already in the room." It doesn't look like a god but like infrastructure: a terminal window, a folder of Markdown files, a job that finishes while you sleep. AGI is arriving diffused, as a "personal AGI" – general intelligence for one person, you, running on your context and doing your work.

Personal AGI vs. Corporate AI

The speaker distinguishes "personal AGI" from marketing-driven "personal AI" products:

• Corporate AI (what it is not):

  • A chatbot you pay $20 a month for.
  • A slightly better autocomplete.
  • An assistant that only knows your calendar.
  • A subscription you rent, a corporate AGI you don't own.
  • Resets when you close the tab.
  • Knows what everyone else already knows.
  • Gets "lobotomized" on someone else's schedule when the company pivots.

• Personal AGI (what it is):

  • An agent that runs on your infrastructure.
  • Reads from a memory you own.
  • Executes procedures you wrote.
  • Compounds daily as it learns more about your life.
  • An asset you build, not a product you consume.

The speaker asserts that almost no one has this personal AGI yet, but everyone could have it soon. This kind of intelligence, they argue, should be owned, not rented.

The Multiplier Effect of Agents

The speaker shares a personal anecdote: in 2013, as a YC partner, they shipped about 14 useful lines of code daily. This year, running YC full-time with a child, their output is about 400 times higher, or at least 8 times higher even with extreme penalties for verbosity and scaffolding. This multiplier is not just for coding but for all knowledge work, including design, product management, and growth.

At YC, they observe this at portfolio scale. A year and a half ago, a quarter of companies in a batch had codebases that were 95% AI-generated. These companies now use AI agents for everything and are on track to be one of YC's fastest-growing and most profitable batches. The speaker clarifies that while correlation isn't causation, the fastest-growing founders treat AI not as autocomplete but as a workforce. The leverage comes not from the model's weights but from the context provided, its relevance, and its application at the right step.

Joy and Sadness in the Age of Agents

Spinoza defined joy as the feeling of one's "power of acting increasing." The speaker relates this to the experience of an agent completing a week's work in an afternoon, describing it as "joy" because one's power of acting (conatus) has increased. Conversely, sadness is the feeling of one's power of acting decreasing, like the "heaviness" of Sunday nights when one feels their ability to influence the world receding.

The Equation for the Next Decade

The speaker presents an equation for leveraging this new paradigm:

Frontier Model (rented, commodity, getting cheaper) + Your Context (owned, unique) + Harness (wires them together) = Agent (fast version of you)

The model quality is rented, but your brain (context) is owned. This combination creates a "self-driving rocket," extending human capabilities. Paul Graham's advice to "make something people want" and "do things that don't scale" still holds, but agents now provide a multiplier for the latter, enabling one founder to do unscalable things at scale.

Grinding Lenses for the Mind: The GBrain Architecture

The speaker describes their personal system, "GBrain," as a way to "grind lenses for the mind."

• Working Memory: Humans have a limited working memory (7 ± 2 items). All human institutions (checklists, org charts, meetings) are prosthetics for this limit. An AI agent, however, can hold a million tokens (about 1,000 pages or three Harry Potter books) in its "head" simultaneously, synthesizing information across them in seconds. This represents a different "operating regime."
• The Library + The Librarian: While 1,000 pages is a lot, a human life is a library. The critical question is what "three books" are open on the agent's desk at any given moment. GBrain is designed to be this "library plus the librarian."
• GBrain in Practice: The speaker's personal OpenClaw system has a knowledge wiki of 220,000 Markdown pages, diarizing 25 years of their life (emails, meetings, notes, photos, drafts). This is compiled, curated, and searched by agents. When a founder emails about a crisis, the agent pulls all prior conversations, relevant portfolio company experiences, and effective solutions. The agent acts as a colleague, knowing everything the speaker knows.
• A Day with GBrain:
  • Overnight: The agent processes the inbox, triaging emails, pulling context, and identifying what's truly being asked. The speaker wakes up to a briefing, not a pile of emails.
  • Meetings: Before every meeting, a prep document is generated, detailing who is being met, past discussions, changes, and key questions.
  • Research: Midnight curiosities are researched and filed by morning.
  • News: Interesting world events are read, cross-referenced, and filed before the speaker has coffee.
• GStack: On top of this library is GStack, a coding framework with 123,000 stars on GitHub. Its core is "skill files" and a browser that agents can drive.
• Skill Files: These are pages of English instructions, like "When a meeting recording lands from Circle Back, transcribe it with speaker labels. Pull out commitments, who made them, and deadlines. Cross-check named people against the library and link their pages. File the summary here, full transcript there. If anything contradicts something we already believe, flag it, don't override it." If a smart intern could follow it, an agent can run it.
• Markdown as Code: The speaker asserts that Markdown is code, and if you can write clear English instructions, you are a programmer, with the language model acting as the compiler. This enables non-engineers (media, events, finance teams) to build skill files and scheduled jobs.

Latent vs. Deterministic Computation

The speaker emphasizes the importance of understanding where computation happens:

• Latent Space (in the model): Best for taste, judgment, and interpreting vague human requests. Steered by Markdown files.
• Deterministic Space (code/databases): Best for arithmetic, SQL queries, and structured data. Markdown files can call this code.

Confusing these leads to agent failures. For example, seating five people can be done in latent space, but custom schedules for 6,000 people require the agent to write and use code to manage deterministic data.

The Spinoza Compendium Skill

The speaker illustrates the power of this system by describing how their talk's opening story about Spinoza was generated. Five days prior, the agent acquired and read three biographies (1,500 pages), synthesized them into a dated chronology, highlighted disagreements, extracted verbatim quotes, and identified the 10 most "tellable moments" with delivery notes. This "compendium skill" is a personal, mega-deep research tool.

The speaker notes that their own GBrain started as a simple folder of Markdown files, growing through daily compounding with agents handling the filing, much like building one shelf before a warehouse.

Managing a Workforce of Markdown

When using an agent, you are not coding but "managing a workforce made of Markdown."

• Skill File: An employee with one clear capability or job.
• Resolver: An org chart that decides which Markdown file (or "employee") handles an incoming task.

This allows an individual to run an organization of one plus agents, effectively becoming the founder and entire management layer of "you incorporated." This new paradigm is already producing companies with unprecedented revenue per person, breaking old math. These companies started with one or two people wired in this way.

Software as a Personal Tool

The new physics changes what software is. It no longer needs to be precious. You can build exactly the tool you need for an "audience of one" in a weekend. The old advice was to "scratch your own itch and hope it's the market"; the new version is to scratch your own itch because it's nearly free, and some of these personal tools will evolve into companies.

Caveats: Memory and Hygiene

A "brain nobody curates is a garbage dump with great search." Without hygiene, retrieval will surface stale facts with confidence, and bad skill files encode bad processes. The primitive is "memory plus hygiene," requiring provenance on facts, contradiction checks, and a "librarian" for pruning. Treat the brain like production infrastructure for compounding benefits; treat it like a dumping ground for confident but untraceably wrong agents.

How-To Guide for Personal AGI

1. Tonight: Pick a Harness. Run an agent on your machine (e.g., OpenClaw, Hermes agent with GBrain, Codeex, Claude Code). The concepts are more important than the specific tool.
2. This Weekend: Start Your Library. Create one folder of Markdown files. Export notes, emails, and write a page for each project and person, including what you know, what you're building, what you owe them, and past conversations. This is your unique context.
3. Write Your First Skill File. Choose a hated weekly task (e.g., expense reports, meeting notes). Explain it to your agent in plain English. Correct its mistakes, adding every rule and exception. This page becomes an "employee."
4. Wire It Up as a Recurring Job. Schedule the skill (e.g., "Every morning at 7, do this," "Every Friday, summarize that"). Waking up to completed work shifts your perspective on the unit of work.
5. Never Do One-Off Work. After every task, ask the agent to "skillify" what it did, turning it into a reusable Markdown file. If you have to ask for something twice, you've failed. This discipline ensures continuous learning and compounding.

The speaker predicts that after 12 weeks, you'll have a library that answers before you finish asking, a dozen skill files running dreaded tasks, and tools that others want to borrow, potentially leading to a startup.

The Political Dimension: Ownership of Skills

The speaker returns to Spinoza's definition of sadness and the political implications of skill files. A skill file is an externalized piece of your cognition, your judgment, extracted and executable. The future depends on "who controls it."

• Version 1 (Ownership): If Maya, a support engineer, teaches her agents 40 skills over two years, and these files live in her repo, they go with her when she changes jobs. She compounds her judgment and can even start a company based on her expertise.
• Version 2 (Extraction): If those files live in the company's repo, Maya leaves with nothing. The company continues to execute her judgment without her. She had an "extraction," not a career.

The doctrine is: "Own your skills, because if you don't, your job becomes a skill file." Craftsmen owned their tools; knowledge workers assumed their tools (cognition) were safe in their heads. Skill files end that, as cognition can now be extracted, stored, versioned, and owned. The question is by whom.

The "thousand guilders" offer (a comfortable arrangement where your judgment compounds in someone else's repo) is still present. Spinoza, when offered a professorship with "freedom of philosophizing" provided he didn't disturb religion, declined, protecting his "power of acting" under his own control. Personal AGI is about controlling your cognitive abilities and protecting yourself.

The speaker advises keeping your brain and skills in a repo you control from day one. Spinoza, who owned almost nothing, kept his "Ethics" locked in his desk drawer – his repo.

Addressing Objections

1. Models are improving too fast, making harnesses obsolete: The speaker argues that as models improve, context becomes the differentiator. A better model makes your library more valuable because a smarter reader extracts more from the same books. Every model release is a free upgrade to your owned workforce.
2. Is this just RAG (Retrieval Augmented Generation)? Yes, but RAG is a primitive, not the product. The hard part is everything around it: what gets written down, how it's enriched and linked, what's hot memory vs. cold reference, and how contradictions are arbitrated. Being "worth retrieving from" is the product.
3. Security and Privacy (leaks): The speaker argues that consolidating your life into one system under your control is more secure than the default, where your life is scattered across 10 clouds owned by companies with misaligned incentives, searchable by everyone except you. Custody is the security model.

Why Open Source?

The speaker open-sourced their personal operating system (harness, brain architecture, skills) because they can, and because they believe "tools of the powerful should be given away." Historically, leverage came from literacy, then capital. Today, it's this "harness, library, workforce made of Markdown." Keeping it private creates a priesthood; giving it away creates a renaissance.

The speaker's creed: "Say the things other people won't. Fund the people other people won't. Build the buildings other people won't. Write and give away the code that other people won't. Leave behind the institutions that other people won't."

The Adoption Curve and Real-World Impact

The speaker notes that building in public guarantees meeting the pattern: "First, they quote tweet you, then they git clone you." The "dunks" are just the adoption curve announcing itself.

The ultimate example of personal AGI's power is a friend whose son has a rare form of epilepsy. This father built a repo of 80,000 Markdown files – a "brain for one small boy" – pushing the limits of human knowledge about his son's condition. Every specialist visit, paper, seizure log, and drug interaction is indexed and cross-linked. When a new doctor has an idea, he knows in minutes if it's been tried. This is personal AGI: a father, a laptop, and a library, aimed at the one thing he loves most.

The speaker concludes that the traditional requirements for building (team, funding, permission, credentials) were workarounds for human cognitive limits and work hours. That fact has expired. You can "fly now," mechanically. Every problem where you wished you could hire someone, every archive too big to read, every "ocean you were told not to boil" – now you can boil the ocean.

The core message: "It's all made up, but you get to make it up." Institutions are made by people no smarter than you. Previous generations needed dozens of believers; you need a laptop and your own history. The machinery shown allows "striving to go straight to work" for one person, without intermediaries or permission.

Spinoza's "Ethics" closed with: "All things excellent are as difficult as they are rare." The speaker asserts that the difficulty has collapsed; the rarity is now up to you. "Go and build."