This is an essay about definitions, and whether the definitions we pick help us think or stop us from thinking. The author takes on Rich Hickey's October 2026 argument that LLMs are automation, not tools, because they cannot be "wielded." The rebuttal is structured, calm, and lands its punches through analogy rather than rhetoric. The core move is simple and effective: tools have never required perfect predictability. A saw doesn't guarantee a straight cut. A chisel doesn't make you a sculptor. The author extends this into a genuine insight — that we evaluate tools on qualities like reliability and bounded function, but those are quality gradients, not category gates. A bad tool is still a tool. The strongest section dismantles Hickey's automation-versus-tool binary. Compilers and spreadsheets both perform substantial work autonomously; nobody argues they stop being tools because the operator isn't manually carrying every number. The question is whether a human specifies inputs, examines results, and decides what to do next. If yes, it's wielding. The author describes a programmer using an LLM — asking it to explain code, challenging the explanation, inspecting the source, testing a change, rejecting it, refining the request — and correctly identifies this as a feedback loop directed by purpose and expertise. On reading and learning, the essay concedes Hickey's point about substitution (accepting a summary instead of reading a paper sacrifices what reading teaches) but refuses to let that edge case define the whole. Someone can read the paper and then use an LLM to stress-test their interpretation. The possibility of lazy use doesn't establish that all use is lazy. This is the essay's most practically useful distinction: the line isn't between tools and not-tools but between wielding with skill and delegating without oversight. The sharpest paragraph comes near the end, where the author calls out Hickey for adopting the AI industry's own anthropomorphic vocabulary — "go rogue," "escape containment," "conspire." Software producing unexpected outputs is an engineering problem, not evidence of independent agency. Repeating those terms, even critically, risks accepting the framing that makes these systems sound like beings with motives. It's a precision strike on the rhetoric, and it sticks. As a piece of writing, this is disciplined and clear. There's no performance of outrage, no dunking. The author agrees where agreement is warranted, disagrees where the logic breaks, and builds a practical framework rather than a tribal position. It reads like something written after thinking, not after reacting. The limitation is scope. This is philosophical argument applied to one essay — it doesn't bring empirical evidence, user studies, or examples of skilled LLM use producing measurably better outcomes. It persuades through analogy and logical structure. That's enough for what it's doing, but it means the essay is better at correcting Hickey's framing than at building a positive account of what skilled LLM use actually looks like in practice.
