In his October 11, 2026 article, "'AI' is Not a Tool," Rich Hickey argues that\nLLMs are automation rather than tools because they cannot be wielded. I agree\nthat outsourcing work can cost us opportunities to learn. But his conclusion\ndoes not follow. Even by his definition, an LLM can be something a human wields.\nAnd using it well takes skill, just as using other tools does.
\nPicking up a saw does not guarantee a straight cut. Holding a chisel does not\nmake you a sculptor. Tools can amplify poor judgment as readily as good\njudgment. Their usefulness depends on the operator's ability to direct them,\ninterpret feedback, recognize mistakes, and make corrections. Those\nrequirements also apply to LLMs.
\nAn LLM's unpredictability is a meaningful limitation. But predictability,\nreliability, simplicity, and bounded function are qualities by which we\nevaluate tools, not prerequisites for membership in the category. A tool does\nnot stop being a tool because it is difficult to control or easy to misuse.
\nHickey's distinction between tools and automation also creates a false\ndivision. Automation can be a feature of a tool. A compiler performs\nsubstantial work without the programmer continuously directing each\noperation. A spreadsheet recalculates results without someone manually\ncarrying out every calculation. We still wield these systems by specifying\ninputs, examining results, and deciding what to do next. Why should\ngenerating text automatically disqualify an instrument from being a tool\nwhen translating code or calculating numbers automatically does not?
\nThe claim that LLM use lacks intention, feedback, direction, and expertise\nis especially difficult to defend. A programmer can ask an LLM to explain\nunfamiliar code, challenge that explanation, inspect the source, test a\nproposed change, reject it, and refine the request. That is a feedback loop\ndirected by a human pursuing a purpose. Doing it well requires knowing what\nto ask, what evidence matters, and when the answer is wrong. Expertise can\nmake the instrument more useful.
\nHickey's concern about reading is valid for a particular use: accepting a\nsummary in place of reading a paper can sacrifice much of what reading\nteaches. But someone can also read the paper and use an LLM to question\ntheir interpretation, explore an objection, or identify something to\nexamine more closely. The possibility of avoiding intellectual work does\nnot establish that every use avoids it. The relevant question is what the\nperson actually does.
\nWhat disappoints me the most is the reliance on the AI industry's\nanthropomorphic vocabulary: "go rogue," "escape containment," "conspire."\nRepeating those expressions risks accepting the framing that makes these\nsystems sound like independent beings. Software can produce unexpected\noutputs, fail to follow instructions, or cause harm when given access to\nconsequential actions. Those are real engineering problems. Calling them\n"conspiracy" adds a suggestion of human motives without establishing any.
\nI would not claim that an LLM always does exactly what a human tells it to\ndo. It plainly can fail. But failure to obey perfectly does not make it\ncease to be a tool, any more than successful use proves that its operator\nhas surrendered their judgment.
\nThe useful distinction is between wielding an instrument with skill and\ndelegating work without adequate understanding or oversight. That\ndistinction applies to AI, and it gives us something more practical than\ndeclaring that an entire class of software cannot be a tool.
","headings":[]}}