Design Philosophy

Design Philosophy — The Thesis Behind the Skill

🌐 English · 中文 · 🏠 Home

The repository already has an internal design self-statement: docs/design-philosophy.md. This page answers a deeper question — where does that philosophy come from? The short answer: it comes from a master's thesis, and the skill is that thesis put into practice.

---

The skill did not come from nowhere

This skill grew out of a specific MA thesis in the History of Science & Technology (Tsinghua University, 2026):

《技艺自由主义：以 20 世纪 60 年代审美革命为枢纽重释技术自由主义》 Technê-Libertarianism: A Rereading of Technolibertarianism in the Aesthetic Revolution of the 1960s

(A 2026 MA thesis, currently under review / forthcoming. The account below summarizes work in progress — please do not cite it as a published source.)

This is not a loose analogy bolted on after the fact. The thesis's own conclusion names this skill as a worked example of its central argument — AI-assisted writing as "another trading zone" (Galison) in which the author practices what the thesis calls doing freedom. The thesis is the theory; the skill is the same stance encoded into a human–AI writing interface. Two faces of one position.

---

The thesis in one minute

• Technolibertarianism — the Silicon Valley / counterculture ideology traced by Turner, Barbrook & Cameron, Winner — narrows "freedom" into consumer freedom: the freedom to choose among options that technical capital has already pre-set. You may pick from the shelf; the right to design the shelf has quietly disappeared.
• The thesis argues this narrowing was a historical accident, not a necessity, and returns to the aesthetic revolution of the 1960s — the brief moment when information technology, avant-garde art, and social liberation converged — to recover a buried alternative.
• It names that alternative technê-libertarianism, a freedom grounded in technê (the ancient Greek unity of art, craft, and knowledge before their modern "Great Division"). It redefines technological freedom from consumer freedom into doing freedom (做自由): not choosing among given options, but intervening in the technical process by hand and reinventing what is possible.
• The decisive move is in the hyphen. Technê-Libertarianism keeps the -Libertarianism suffix on purpose: it is not the opposite of technolibertarianism but "the fuller self it forgot." Doing freedom completes consumer freedom rather than abolishing it.
• Three thinkers do the work, as a relay: Stiegler is the methodological anchor (general organology, the pharmakon, the proletarianization of the sensible, critical pharmacology); Rancière supplies the aesthetic-as-political frame (the distribution of the sensible); Mumford supplies the ethical horizon (authoritarian vs. democratic technics, "choosing to walk out").

In one sentence: Consumer freedom is a kind of consumption; doing freedom is a continuous practice.

---

The core question: when you write with AI, which freedom is it?

Hand your draft to an AI to "polish," then accept the smoothest of the options it returns — and you are exercising consumer freedom. In the thesis's words, this is not essentially different from picking a product off a shelf. The text comes back more even, more "standard," and quietly emptied of its author.

Build the rules instead — a vault, a style profile, a skill that refuses to descend below a broken argument, a record of every change — and keep your scholarly judgment and your writing voice in the loop: that is doing freedom. You are intervening in the technical process, not consuming its output.

The skill is built to make the second thing the easy thing. Each design decision below is the thesis's argument, encoded into an interface.

---

Seven design decisions, and the argument behind each

1. Thought before format → resisting the proletarianization of the sensible

The skill inverts the priority order of mainstream AI writing tools — from format → expression → structure → concept → argument to argument → concept → structure → expression → format — with a mechanical rule: do not spend effort on a lower layer while a higher layer is unresolved. Early feedback deliberately will not make your prose "look" better; it sends you back to the argument.

This is the thesis's diagnosis applied to writing. Polishing tools start from "the easiest thing" (spelling, sentence shape) — exactly the move that narrows writing into optimizing within surface options, i.e. consumer freedom. Following Stiegler, surrendering your draft to be polished turns the author "from creator into approver," losing the savoir-faire — the know-how of making — that is the first casualty of the proletarianization of the sensible. Reverse priority pulls the author back from surface choosing to deep making.

2. "I write as I speak" (我手写我口) → cognitive subjectivity is doing freedom

The skill treats "I write as I speak" not as a UX nicety but as an epistemological claim: voice is not a fixed essence waiting to be protected, but an evolving writing subject; the thing to guard against is not "AI changed my style" (every tool changes the writer) but "I am accepting AI output without examining it." What matters is not how much the AI changed, but whether the author was present while it changed.

The thesis says this almost word for word in its conclusion: "'I write as I speak' is not merely a stylistic principle — it concerns scholarly subjectivity itself: in an age when AI speaks ever more fluently, to insist on speaking in your own voice is itself a form of doing freedom." (The author even found 15 places where a first-person "I" had been quietly sanded into "this paper" across long AI sessions — evidence that subjectivity must be maintained, not assumed.) This is Stiegler's subject-through-individuation, in writing.

3. Mode F (improvement vs. alienation) → doing freedom as continuous judgment

In Mode F, when you revise an AI-polished draft, every change is classified as "improvement" or "alienation" — keep the improvements, restore what was alienated. This is not anti-AI; it folds the AI's influence back into the author's conscious choice.

This is the operational form of the thesis's definition of doing freedom: "a continuous practice of judgment," "maintaining the position of creator in ongoing negotiation with technology." The thesis's three diagnostic moments — where the AI kept getting it wrong, where its suggestions were repeatedly refused, where the author later suspected assimilation — are precisely Mode F's raw material.

4. Anti-sycophancy + dialogic stance → the critical peer, and the thesis's own method

The skill holds two principles that look contradictory: a generous scholarly stance (treat Rancière, Stiegler, Mumford as interlocutors — steelman before you critique) and an anti-sycophancy mechanism (the devil's advocate concedes only when ≥2 of 5 substantive conditions are met, even if you get impatient). They reconcile as: be generous to what is already written; be strict with the critique in progress.

The resonance here is unusually literal. The three interlocutors the skill names — Rancière, Stiegler, Mumford — are the thesis's own three thinkers, and the thesis treats them exactly this way: credit Rancière's aesthetics-as-politics, then mark his technological blind spot; adopt Stiegler's organology, then note where it underspecifies history; inherit Mumford's diagnosis, then overcome his pessimism. The skill's dialogic principle is the thesis's argumentative posture, turned into a rule. And anti-sycophancy is Stiegler's amateur (one who participates out of love, not passive consumption): a "critique" that retreats under emotional pressure is worthless — it neither finds blind spots nor rehearses a real reviewer.

5. Engineering metaphors → the "coding" layer, making AI auditable

commit / release / code review (blocker–major–minor–question) / unit test / CI-CD resume are not props to "look professional." They give the author real cognitive architecture: traceability, verifiability, resumability, layered triage.

The thesis frames this as its second practical layer — "coding" (Flusser's playing against the apparatus): not using the AI black box passively, but entering its logic and re-writing the rules of what it may and may not do. Its conclusion: "if the writing process can be managed like a code repository, then every intervention by the AI becomes auditable." Engineering metaphors turn AI collaboration from a black box into a decodable, auditable system — technê as a cognitive map.

6. Infrastructure, not a tool → continuous maintenance vs. one-time delivery

A tool is used and put away (a hammer, a grammar checker); infrastructure is what you work on top of (the power grid, git). Mainstream AI writing tools are tools — they solve a task and exit, with no lasting commitment to the author's thinking. This skill wants to be infrastructure: it turns a long writing project into a continuously running system.

This maps the thesis's central distinction onto product form. "The former understands freedom as a state that can be delivered all at once by technical means … the latter understands freedom as a capacity that must be continuously practiced and maintained." Tool = one-time delivery = consumer freedom; infrastructure = continuous maintenance = doing freedom. Infrastructure is the wholeness of technê made concrete in a tool.

7. Reflexive writing → the thesis's own reflexive conclusion

The skill's reflexive-writing module assumes that if your research concerns human–AI collaboration, then using the skill becomes part of the research — the skill is both instrument and object of study. Its six reflexive-moment tags (🔄 redirection · 🚫 refusal · 🎭 voice conflict · 🔧 tool dependence · 💡 unexpected insight · 🤖 AI-trace awareness) are a capture framework for that study. The skill refuses to be invisible.

This is the most direct resonance of all: the thesis's conclusion is such a study, written. It cites the STS axiom that "the researcher is never external to the object of study," and argues that for a project about the technology of writing, invisibility would be dishonest. At least four of the six tags appear, lived, in that conclusion. "Refusing to be invisible" is the thesis's "presentation is itself argument" — and open-sourcing the skill turns doing freedom into a commons.

---

An honest tension (kept, not hidden)

There is one place where skill and thesis do not perfectly align, and it is worth stating plainly. Mode F's "improvement vs. alienation" split presumes the two are distinguishable. The thesis is less sure: in long collaboration it admits you may be unable to tell "an AI trace" from "my own style drifting toward the AI," and even warns that deliberately writing against the AI is itself a form of co-shaping — a modern version of the "Greek nostalgia" for a pure authorial origin that never existed.

So the skill is, at the operational layer, a little more optimistic than the thesis is at the philosophical layer. This is not a contradiction. It is the normal gap between a workable pharmakon and an honest diagnosis of that pharmakon's limits — which is exactly the thesis's own method ("acknowledging a limit is not retreat but precision," §5.5).

---

In one line

The skill is not "AI that helps you write better" — that is the consumer-freedom story. It is infrastructure for keeping your cognitive subjectivity and the wholeness of technê alive in a sustained negotiation with AI — which is precisely the thesis's doing freedom. Theory and practice, two faces of one stance.

---

Further reading: the skill's internal design self-statement · The 11 Working Modes · Writing a Paper End-to-End · Contributing.