'AI' Is Not a Tool by Rich Hickey

TEXT START: First order of business - is 'tool' going to have an interesting, distinguishing and thus illuminative meaning or be a mere platitudinous stand-in for "anything useful"?

THE DISSECTION

Hickey is performing a legitimate category correction. He distinguishes an implement wielded by a human from automation that executes a request on the human’s behalf. His target is abdication: outsourcing reading, coding, writing, and judgment, then pretending the operator acquired the underlying competence. The epistemic cost is real. Users learn model quirks and prompt folklore instead of the task, the world, or the logic beneath it.

The article’s strongest point is that LLM interaction is probabilistic delegation, not ordinary tool use. Its weakness is that it stops at the level of craft and cognition. It diagnoses what humans lose while using automation, not what the economic system gains by making them unnecessary.

THE CORE FALLACY

The article treats the semantic truth—“LLMs are automation”—as if it were a defense of human productive relevance. Under the Discontinuity Thesis, it is the opposite. The decisive question is not whether AI deserves the noun “tool,” but whether it performs cognitive work cheaply enough to sever labor from wages. Automation is the kill mechanism.

Unreliability, non-determinism, and verification requirements are not durable moats if generating and checking an output costs less than employing an expert to produce it. The checker becomes a narrow Servitor role; the mass of producers lose bargaining power. Human direction may improve outputs, but competition will compress it into templates, workflows, and supervisory layers.

The article assumes that because LLMs do not build expertise, expertise remains economically necessary. That does not follow. Expertise can remain intellectually valuable while becoming economically disposable. DT separates those two facts.

It also mistakes a personal choice—refusing to abdicate reading or coding—for a system-level solution. When competitors adopt automation, refusing it does not preserve the old employment circuit. It makes the refuser more expensive. The drive-thru analogy is not a rebuttal to automation. It is an accurate miniature of its purpose: remove a human activity from the production chain.

HIDDEN ASSUMPTIONS

• Human mastery will continue to command wages even when automation produces acceptable outputs.
• Quality thresholds will remain high enough to protect human craft from substitution.
• Verification will require the same level of expertise as original production, at scale.
• Individuals can opt out without losing employment, status, or bargaining power.
• Skill atrophy is the primary danger, rather than the disappearance of the job category that rewarded the skill.
• Distinguishing tools from automation changes the strategic outcome.
• Institutions can preserve human-only cognitive domains despite competitive pressure.
• The value of reading, writing, and thinking will be paid for as labor rather than retained as a private cultural good.
• Current AI failure modes are permanent defenses rather than defects that engineering and workflow design can erode.

SOCIAL FUNCTION

Classification: partial truth, prestige signaling, and craft-preservation copium.

The article gives skilled practitioners a precise warning about cognitive abdication. That warning is valid. But it relocates the crisis from political economy to personal discipline: keep wielding real tools and remain competent. This preserves the identity of the expert while leaving P1, P2, and P3 untouched.

Its implied refuge is that genuine skill remains a defensible economic position. Under DT, that refuge is narrow. Human competence may survive as cultural virtue, verification labor, or Servitor specialization. It does not thereby survive as mass productive participation.

THE VERDICT

This is a sharp local autopsy and an incomplete systemic one. Hickey correctly identifies LLMs as automation rather than ordinary tools, and accurately describes the cognitive damage of delegation. He does not follow that insight to its economic endpoint.

The article identifies the machine’s nature but mistakes human craft for a durable moat. If LLMs are automation, they are precisely the machines capable of severing labor from wages. The corpse can retain expertise, taste, and judgment. That does not mean it is still alive.