The Fourth Competency

Four competencies

In 2025, Anthropic — the company behind the language model Claude — publishes a teaching framework titled AI Fluency, together with the academics Rick Dakan and Joseph Feller. It describes the ability to work with “AI systems”: “effective, efficient, ethical, and safe.” The materials are freely available under a Creative Commons license; the framework is taught at universities.

Four competencies carry the frame. Three of them order the work with the machine: Delegation — deciding which work goes to it. Description — telling it the task with precision. Discernment — judging its outputs critically. The fourth aims at something else: Diligence — answering for what is made this way.

The framework divides this fourth competency into three parts. Transparency Diligence: “Being honest about AI’s role in your work with everyone who needs to know.” Deployment Diligence: “Taking responsibility for verifying and vouching for the outputs you use or share.” Creation Diligence: care in choosing the systems themselves. This note is about the fourth competency: about the confirmation it contains, and about the one word in the framework that does not survive it.

The confirmation

The maker of the Sprachsubstrat — the language substrate: what a language model supplies in statistically probable formulation — declares honesty about its role a competency. This is the same diagnosis on which Maschinenschrift is built, stated from the other side of the desk: not by an author writing with the substrate, but by the company that makes it. The account of the machine’s role belongs to writing itself. It is not a confession and not a weakness. It is part of the craft.

The second part converges as well. “Vouching” — standing behind what one shares — is the double character of self-declaration: whoever declares gives an account and at the same time stakes their reputation on that account being true. The declaration is statement and pledge in one act.

The third part — the choice of systems — lies outside what Maschinenschrift addresses. The project states its limits; this is one of them.

The missing form

A competency is a property of the person. An account is a property of the work. The framework trains the person; the work parts from her — it is sold, lent, passed on. What is missing is the form in which the author’s honesty becomes the work’s account.

“Everyone who needs to know” — for a work of literature, that is not the colleague on the project and not the client. It is the reader, including the one thirty years from now who finds the book secondhand and wants to know how it was made. The author can no longer be asked, and her good intention is not enough. What suffices is a notation that stays with the work: nameable, comparable, durable.

This is where Maschinenschrift begins. The seal gives the required honesty a form: three axes — Wortlaut (wording: how much of the linguistic surface the substrate supplied), Geist (German: spirit, intellectual presence — how much of the ideas, arguments, structures), and Beleg (the record: whether the work is bibliographically documented). The framework demands the account. The seal is a language in which it can be given — and which lasts.

The difference

One word in the framework does not hold. The middle of its three interaction modes, Augmentation, describes human and machine as “thinking partners” who collaborate.

The Maschinenschrift position is a different one: no second thinker sits at the desk. The substrate has no standpoint, no intention, no will to utterance — it supplies the statistically probable continuation of everything ever written. It is a medium, not a partner; closer to the printing press than to the editor’s desk. For a teaching framework, the partner image is understandable: it lowers the threshold; it speaks the language of those still learning the practice. But it has a price. Whoever writes inside the partner frame distributes authorship — and with it, unspoken, responsibility.

The framework itself does not sustain the image. The fourth competency assigns responsibility undividedly to the human: “taking accountability for AI-assisted work.” A partner to whom no responsibility can be transferred is none. Diligence disavows the vocabulary of the first three competencies — and this disavowal is the framework’s most honest moment. What remains is what Maschinenschrift has said from the start: a medium, and a human who stands behind what is written.

The interface

AI Fluency orders the process: how to delegate, describe, and judge while the writing happens — the first three competencies, addressed to the writer. Maschinenschrift records the trace: what of the substrate stands in the finished work, legible to everyone who needs to know — addressed to the reader. The two are compatible. The interface is the fourth competency.

It is also the point at which the framework depends on something outside itself. Honesty about the machine’s role needs a form in which it can be said. Maschinenschrift is such a form.

More on the form and the lineage: Idea. Why the project says Sprachsubstrat rather than machine or LLM is derived in a note of its own: Substrate, not Machine.

Self-declaration of this note: W 5 · G 3. Draft and most of the linguistic surface substrate-generated, the comparative argument substrate-mediated; occasion, selection, review, and responsibility with the editor. A note on the maker’s demand for transparency carries its own account.