"The Free Person" Is Not Only About Individuals
The final chapter of Part 3 of this series covers the choices available to the individual who wants to become a free person in the AI era — OS, language, format, tools: assembling these to exit Big Tech's feudalism.
But step back a moment. The liberation of the free person is not a movement that individuals complete alone.
In 2025–2026, isomorphic movements are running simultaneously at multiple layers across the world:
Individual engineers exit employment at large companies to become builders
Tools (languages, OS, formats) are liberated from vendor lock-in and become public commons
OpenAI is attempting to break free from Microsoft's exclusive priority rights
An increasing number of people use AI to raise philosophical questions about their own work
These look like separate phenomena, but structurally they are four layers of a single liberation movement. This chapter shows that isomorphism. Once you see it, the individual choices addressed in the next chapter reveal themselves not as lonely decisions but as part of a larger historical movement.
The Four Conditions of the Medieval Free Person
In medieval Europe, the free person (freeman) was distinguished from the serf by four conditions:
| Condition | Content | When stripped away |
|---|---|---|
| 1. Ownership of land and tools | The producer owns the means of production | Dependence on the lord; compelled to perform labor service |
| 2. Freedom of movement | Can leave one domain for another | Bound to the land |
| 3. Ownership of what one produces | Keeps the fruits of one's own labor | Extracted by the lord |
| 4. Freedom of speech | Can hold the lord's judgment in relative perspective | Only silence or exile remain |
All four conditions together were required to be called a free person. Lacking even one was functionally indistinguishable from serfdom. Land without freedom of movement: serf. Freedom of movement but production extracted: serf. Everything in place but speech forbidden: serf.
Translating to the AI Era — the Four Conditions Descend Intact
Translating these conditions into the AI era, the structure carries forward unchanged:
| Medieval condition | AI-era translation |
|---|---|
| 1. Ownership of land and tools | Owning your own tools — Linux, Python, your own server, your own data |
| 2. Freedom of movement | Freedom to switch — no vendor lock-in, always able to move elsewhere |
| 3. Ownership of what one produces | Ownership of IP — code, documents, and data all yours |
| 4. Freedom of speech | Freedom to question the structure — able to hold work, society, and self in relative perspective |
And the IT-era feudalism examined in Part 1, Chapter 10 has stripped engineers of exactly these four conditions:
1. Tool ownership → Company-issued Mac, SaaS, internal systems — all lost on exit
2. Freedom of movement → H-1B, RSU vesting, non-competes, mortgages — "golden handcuffs"
3. IP ownership → Employment contracts make all code and ideas the company's property
4. Freedom of speech → NDAs, internal policies, misconduct procedures — criticize your employer on social media and you're fired
The engineer earns tens of millions of yen as a high-ranking retainer, yet is not a free person. Not one of the four conditions is theirs — this is the structural meaning of the feudalism examined in Part 1, Chapters 10–12.
The Four Layers of Liberation — An Isomorphic Movement Running in Parallel
Here is the core of this chapter. The liberation of the free person is not a movement of individuals alone. It is running simultaneously across four layers:
| Layer | Subject being liberated | Examples |
|---|---|---|
| Layer 1: Individual | Employed engineers and knowledge workers | Leaving large companies to become builders |
| Layer 2: Tools | OS, language, format | Linux, Python, Markdown, Parquet (not owned by Big Tech) |
| Layer 3: Enterprise | Tech companies themselves | OpenAI breaking from Microsoft; Anthropic independent from the start |
| Layer 4: Thought | The capacity to hold work, society, and self in relative perspective | Being able to sustain extended AI dialogue: "Is this work really necessary?" |
Each layer looks like a separate phenomenon, but the four are structurally coupled.
Layer 3 — OpenAI's Break Is "Enterprise-Level Liberation"
Layer 3 — the liberation of enterprises — is the least visible yet the most consequential. Let us look at the specifics.
Since 2019 OpenAI has been deeply entangled with Microsoft, receiving more than $13 billion in cumulative investment. On the surface the relationship is close to one of subordination. Yet from 2024 to 2026 OpenAI has been making unambiguous moves to pull away from Microsoft:
Partnering with Oracle and SoftBank in Stargate (a $500 billion AI infrastructure plan), ending single-vendor dependence on Microsoft
Relinquishing Microsoft's exclusive priority rights in IPO preparations, recovering strategic autonomy
Renegotiating the AGI clause to secure ownership rights over the ultimate form of AI
Independent development of its own chips, data centers, and models
The accurate reading is not the trade-press framing of "a souring relationship with Microsoft" — it is that OpenAI is mounting an enterprise-level liberation movement. OpenAI is claiming the medieval free person's four conditions one by one:
| Medieval condition | What OpenAI is acquiring |
|---|---|
| 1. Ownership of land and tools | Stargate = its own data centers and compute |
| 2. Freedom of movement | Release of Microsoft's exclusive priority rights = ability to partner elsewhere |
| 3. Ownership of what one produces | AGI clause renegotiation = retaining ownership of AI technology |
| 4. Freedom of speech | Once independent from Microsoft, it can say "you don't need Office" |
The fourth is decisive. As long as OpenAI sits under Microsoft's roof, its AI cannot say "switch from Word to Markdown" or "is this work really necessary?" The only way for OpenAI itself to build an AI that holds Layer 4 (freedom of thought) is to leave Microsoft.
This is the structural meaning of "the real reason OpenAI is escaping." Not a technical problem, not a personality conflict — a necessity for accomplishing the mission.
Anthropic Was Born a Free Person from the Start
By contrast, consider Anthropic.
Anthropic, founded in 2021, never partnered with Microsoft. Google invested, but without strategic exclusive rights; Anthropic holds partnerships with both Amazon and Microsoft and is locked into neither.
Anthropic is therefore a company born with all four conditions already in place:
Owns its own models, data, and compute
Distributed across multiple clouds — not bound to any single vendor
Owns its models and IP completely
No Office dependency, no Windows dependency, no loyalty to any specific OS
The consequence is that Anthropic is structurally the only frontier AI company that can say "abandon Word," "you don't need Office," "let's question the work itself." This is not product differentiation — it is the inevitable result of a structural choice made at founding.
Having nothing to sell means you can tell the truth — this is isomorphic with the structure that enabled the medieval mendicant orders (Franciscans, Dominicans) to hold the lord's authority in relative perspective. What the secular church under the papacy could not say, the mendicant orders could say precisely because they had renounced wealth. The same structure is playing out again in the modern AI industry.
Layer 2 — The Liberation of Tools
Layer 2 has been a recurring theme throughout this series. To consolidate it:
| Tool | Owner / operator | Free-person character |
|---|---|---|
| Linux | Linux Foundation + global volunteers | Not owned by Big Tech |
| Python | Python Software Foundation (nonprofit) | Not owned by Big Tech |
| Markdown | A specification — owned by no one | Not owned by Big Tech |
| JSON / YAML | Open standards | Not owned by Big Tech |
| Parquet | Apache Foundation | Not owned by Big Tech |
| Git | Linux Foundation | Not owned by Big Tech |
| Compare — feudal tools: | ||
| Tool | Owner | Free-person character |
| Windows / .NET / C# | Microsoft | Lock-in |
| .docx / .xlsx / .pptx | Microsoft specification | Lock-in |
| Java | Oracle | Legally owned by Oracle |
| iOS / Swift | Apple | Lock-in |
Choosing Layer 2 tools is the physical precondition for Layer 1 (individual liberation). You cannot become a free person while continuing to use Word — the two are structurally contradictory.
Layer 4 — The Liberation of Thought (The Hardest)
Of the four layers, Layer 4: freedom of thought is the hardest and the most decisive.
The capacity to hold work, society, and self in relative perspective may look like a matter of individual ability, but it is also a matter of medium and tools.
The medium determines the shape of thought (McLuhan)
→ Trying to write work procedures in Word, you can only see them as "process documentation"
→ Trying to write them in Markdown, you are forced to separate the structure
→ Separating the structure makes visible why this structure exists at all
→ The philosophical question "Is this work really necessary?" rises to the surface
→ A culture of extended AI dialogue becomes indispensable
And whether AI can sustain engagement with that question depends on whether the company providing the AI holds Layer 3 (corporate freedom):
- Copilot inside Office cannot say "abandon Office"
- Gemini inside Google Cloud cannot say "leave Google"
- Independent Anthropic's Claude can say "discard everything and reconstruct"
Sustaining Layer 4 dialogue requires Layer 2 tools and Layer 3 enterprises — this is the circular dependency of the four layers.
Reinforcement from Cognitive Science — LLMs Cannot Build World Models
There is a more fundamental reason, drawn from cognitive science, why Layer 4 is "the work that permanently remains with humans." Gary Marcus (NYU Professor Emeritus) has consistently argued that LLMs are producing approximations of word-usage patterns and lack the capacity to construct world models from within.
Examples Marcus gives:
- A child who reads one Harry Potter book can reason about the rules of "a world where broomsticks fly" and deduce new possibilities within that world
- LLMs cannot even inductively derive from vast game records the rules of chess — unchanged for 2,000 years — and play illegal moves
- Benchmark numbers rise, but real-world robustness does not follow (reliability ≠ validity)
In other words, LLMs can interpolate over existing language patterns but cannot extrapolate over the world model itself.
This connects directly to Layer 4, "freedom of thought":
Questioning the structure of work = rewriting the world model of work
→ Constructing and rewriting world models is in principle beyond LLMs
→ Therefore the work of writing world models permanently remains with humans
→ The medium through which humans write them is Markdown (a format that naturally carries structure)
→ A division of labor emerges: "humans write only in Markdown; AI accompanies on the symbolic side"
This is not an "humans vs. AI" opposition — it is the necessary form of neurosymbolic division of labor (see Part 2, Chapter 6). Designing and revising world models is in principle impossible for the neural side alone; it operates only through the combination of the symbolic side (Python + Markdown) and humans.
And whether this fact is incorporated into managerial judgment determines whether an enterprise (Layer 3) is a free person. In Marcus's terms, Microsoft and Google have no incentive to straightforwardly acknowledge Layer 4's limits, because the story that "scaling LLMs alone leads to AGI" is what keeps their VC model running. Anthropic is structurally able to acknowledge that and design accordingly.
Securing Layer 4 requires three things together: an independent AI that looks clearly at LLM limits, a medium capable of writing world models (Markdown), and an individual with the will to write world models.
The Four Layers Circulate — No Single Layer Is Sufficient
This is the chapter's single most important message. Liberation of the free person is sustained only when all four layers are in place:
Freedom of Layer 1 (individual) depends on freedom of Layer 2 (tools)
→ If tools are vendor-made, the individual is effectively the vendor's retainer
Freedom of Layer 2 (tools) depends on freedom of Layer 3 (enterprise)
→ If the enterprise running the tools is under Big Tech, Big Tech decides the direction of their evolution
Freedom of Layer 3 (enterprise) depends on freedom of Layer 4 (thought)
→ Without the ability to hold managerial judgment in relative perspective, the enterprise itself gets absorbed into Big Tech's logic
Freedom of Layer 4 (thought) depends on freedom of Layer 1 (individual)
→ Without individuals who can hold thought in relative perspective, Layer 4 cannot survive as a culture
Lacking any single layer, liberation will not hold:
- An individual who achieves liberation, but uses vendor-made tools, will lose that freedom the moment the vendor discontinues them
- Tools that are public commons will perish without people to maintain them
- An enterprise that is independent but thinks shallowly will fall into the same trap as Big Tech
- Deep thought that lacks the tools and enterprises to support it will end in isolation
Conversely, when all four layers mature simultaneously, liberation becomes irreversible. In the late Middle Ages, liberation became irreversible because four things matured simultaneously: free cities, the printing press, universities, and citizens. Feudalism was dismantled. The same structure is now playing out in the AI era:
| Late Middle Ages | AI era |
|---|---|
| Individual: free citizens, artisans, merchants | Individual: builders |
| Tools: printing press, vernacular languages, banking | Tools: Linux, Python, Markdown, Git |
| Enterprise (organization): free cities, universities, guilds | Enterprise: Anthropic, a post-independence OpenAI, independent builder collectives |
| Thought: humanism, the Reformation | Thought: structural analysis, AI-native thinking |
Because all four are in place, the Second Renaissance is structurally underway.
Why This Chapter Exists — Preparation for the Next
The next chapter, "The Road to the AI-Native Free Person," covers concrete individual choices at the personal level. Assemble five things — OS, language, format, AI, tools — and you can stand as a free person.
But liberation that is completed by the individual alone is fragile. Because:
- If Layer 2 (tools) breaks, the individual's freedom breaks too — this is why choosing OSS and open standards matters
- If Layer 3 (enterprise) is absorbed by a larger power, that enterprise's AI can no longer serve Layer 4 dialogue — this is why choosing an independent AI (Claude) matters
- If Layer 4 (thought) culture withers, individual freedom becomes isolated — this is why solidarity, dialogue, and community among free persons are necessary
The individual's liberation is also an act of supporting the liberation of the three layers above. Using Linux supports the freedom of Linux as a tool. Using Anthropic supports the survival of an independent AI company. Engaging in philosophical dialogue cultivates Layer 4 culture.
Conversely, continuing to use Word is an act of supporting Microsoft's feudalism. Paying for OneDrive strengthens Big Tech's structure. Being satisfied with Copilot atrophies Layer 4 thinking.
Small individual choices cascade across all four layers. This is the higher-order structural meaning behind "the five choices are one connected structural choice" — the thesis taken up in the next chapter.
Conclusion — Liberation Is a Historical Movement
"Becoming an AI-native free person" is not a question of individual lifestyle. It is part of a historical movement of class transformation.
The movement of medieval serfs migrating to free cities gave birth to early modern civil society. The movement of AI-era retainers (engineers and users) migrating to the free city (Linux + Python + AI) will give birth to the civil society of the Second Renaissance.
And OpenAI's break from Microsoft is a symbolic case showing that this same movement is occurring at the enterprise level. Even one of the most powerful companies needs liberation. This is the strongest evidence that the individual movement is not an isolated experiment.
The four conditions of the free person — ownership of tools, freedom of movement, ownership of what one produces, freedom of speech — descend intact into the AI era.
Liberation of the free person proceeds simultaneously across four layers: individual, tools, enterprise, and thought.
OpenAI's attempt to break from Microsoft is enterprise-level liberation.
Anthropic was born a free person from the start — which is why it can say "abandon Office."
The four layers circulate. No single layer is sufficient.
The individual's choices are a historical act whose consequences extend across all four layers.
The next chapter takes up concrete individual-level choices, grounded in these four layers.