Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Rich Hickey’s point is that an LLM may be useful without being a tool in the sense he cares about: a tool is something a person wields, continuously directing it and receiving feedback while acting. In his October 12, 2026 LinkedIn essay, “’AI’ Is Not a Tool,” he classifies LLMs instead as automation: a person makes a request, then the model carries out work without continuous human direction. The disagreement is partly about what “tool” means, but Hickey’s definition also raises a practical question: when a model does the reading or writing for you, are you delegating the work and the learning together?
What does Rich Hickey mean when he says AI is not a tool?
Hickey uses “tool” to mean an instrument a person wields, not simply any technology that helps accomplish a task. The crucial feature is ongoing control: the user acts with intention, directs what happens, and gets feedback while the work is in progress.
He captures that idea in the phrase, “A tool is something you wield.” A tool is “held in your hand” as “a source of both control and feedback,” and wielding involves “intention and continuous directing of the course of events.” In this account, a tool does not merely produce an outcome after receiving an instruction. It remains responsive to a person who is guiding the action as it unfolds.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Hickey’s conclusion follows from that definition: “As a lifelong tool builder and user, I know what tools are. LLMs are not tools.” He considers them automation because a user makes a request and the model then acts on its own, rather than being continuously steered. That is a classification of how the interaction works, not a claim that an LLM cannot be useful.
Popular on Amazon
How is an LLM different from a conventional software tool?
The contrast is not simply that one technology is new and the other is old. Hickey’s distinction is about the relationship between a person and a system during a task: how much the person directs, what feedback the system provides, and whether its behavior is bounded and dependable.
| Dimension | Tool in Hickey’s sense | LLM automation in Hickey’s account |
|---|---|---|
| Control | The person continuously directs the course of events. | The person requests an outcome; the model carries out work without continuous direction. |
| Feedback | The tool provides feedback as the person uses it, helping guide the next action. | The user is not continuously directing the model’s course of action while it completes the request. |
| Predictability | Hickey describes good tools as predictable. | He argues that LLMs do not have the predictability he expects of good tools. |
| Reliability | Hickey describes good tools as reliable. | He argues that LLMs do not have the reliability he expects of good tools. |
| Scope | Good tools are relatively simple and bounded in function. | LLMs do not fit his model of a simple, bounded instrument. |
| Skill and responsibility | The user remains engaged in doing the underlying work and can develop skill through that engagement. | When the model takes on the work, the user may also miss the learning that doing it would have provided. |
The table describes Hickey’s criteria and argument; it is not a controlled comparison or a measurement of every tool or LLM. His point is that the word “tool” can conceal a meaningful difference between actively operating an instrument and delegating a task to an automated system.
Why does continuous control matter?
For Hickey, control and feedback are linked. A person has an intention, takes an action, observes what happens, and uses that information to decide what to do next. Continuous direction keeps the person inside the process rather than limiting their role to stating a desired result.
Free tools Windows power users keep installed
One-click scans. No signup required.
That distinction helps explain why a system can be interactive without being a tool by Hickey’s definition. A user may type a request, read the response, and ask for a revision. But the model still performs each requested stretch of work without the user directing every step of its execution. The user can guide later requests; Hickey’s test asks whether the person is wielding and continuously steering the system while the work itself proceeds.
Also popular on Amazon
This is a definition-based argument, not a universal rule about how everyone must use the word “tool.” Someone using “tool” broadly to mean any technology that helps achieve a goal may reasonably include LLMs. Hickey chooses a narrower meaning that emphasizes agency during the task. Understanding that choice makes his conclusion clearer: he is objecting to treating delegated automation as though it were an instrument under continuous human control.
What does Hickey mean by calling LLMs automation?
Automation performs work after a person specifies a request or outcome. In Hickey’s framing, the user initiates the task, but the model then acts without ongoing direction. That makes automation a better category for LLMs than “tool,” if “tool” is reserved for an instrument the person actively wields.
The distinction concerns the mode of use, not whether a human is involved at all. A person chooses the request, evaluates what comes back, and decides what to do next. But those points of involvement do not necessarily amount to continuous control over the model’s intermediate work. Hickey’s claim is that the user delegates a task rather than directly operating an instrument through each part of it.
It also does not follow that every use of an LLM involves the same amount of delegation. A person might ask a model to complete a task, or use it for a limited response and then do the rest themselves. Hickey’s category draws attention to where the work is being performed and how much of its course the user directs; it does not, by itself, settle how much delegation is appropriate in a particular situation.
Also popular on Amazon
Does using AI outsource learning as well as work?
Hickey’s concern is that reading and writing are not only ways to produce finished work. Doing them also gives a person opportunities to learn how to write, form arguments, observe, judge authors, and think. If an LLM takes over those activities, a user may receive an output while missing some of the learning that would have come from producing or examining it themselves.
He makes the point sharply: “What do you miss by abdicating doing the reading? Pretty much everything.” The word “abdicating” matters. His concern is not merely that reading can be shortened; it is that a person may hand over the act of reading and, with it, the judgments and discoveries made along the way.
The same reasoning applies to writing. Drafting an argument involves deciding what matters, how claims relate, and how to express them. If the model supplies the writing, it can also take over some of the decisions through which the writer would have practiced those skills. Hickey’s argument therefore asks users to consider two outputs: the immediate result and the understanding they would gain by doing the work themselves.
This is Hickey’s philosophical argument about learning, not an empirical finding that AI use always reduces learning or productivity. The essay does not establish how large any effect is, which users or tasks would be affected, or whether a particular use of an LLM improves or harms learning. It identifies a possible cost of delegation and a question for users to consider.
Also popular on Amazon
How can you apply Hickey’s distinction to an AI task?
Use his criteria to examine the interaction rather than deciding based on whether a product is called a tool. The practical question is whether you are wielding the system through the work or asking it to perform work for you.
- Name the work. Identify what you want done, such as reading, writing, or another task. Be clear about which parts you intend to do yourself and which parts you are handing over.
- Locate your control. Ask whether you are continuously directing the task as it unfolds or mainly providing a request and waiting for the model to act.
- Identify the feedback. Consider what information you receive while the task is in progress and whether it lets you steer the model’s course of action, rather than merely evaluate a completed response.
- Check the expected behavior. Hickey values tools that are predictable, reliable, relatively simple, and bounded in function. Ask whether those qualities describe the system and task you are considering, rather than assuming they do.
- Account for learning. If you delegate reading or writing, consider which opportunities to practice observation, judgment, argument, or expression you will give up. That cost may matter even when delegation gets you an output.
This checklist does not prove that using an LLM is a mistake. It makes the trade-off visible: delegation can change not only who performs the task, but also who takes part in the process that can build skill.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What does Hickey’s essay establish—and what does it leave open?
Hickey offers a philosophical and software-design argument: define tools by wielding, continuous control, and feedback; then classify LLMs as automation because users request work that the models carry out without continuous direction. He also argues that doing reading and writing can be a source of learning, so delegating those activities may mean delegating some of that learning.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Those claims clarify his position, but the essay does not settle the wider debate about whether LLMs improve productivity, how AI affects education, or what happens to learning across different tasks and users. Nor does a disagreement about terminology have to be resolved before someone can consider the practical issue Hickey raises: whether handing over a task also means handing over valuable engagement with it.
Also popular on Amazon
Frequently Asked Questions
Is software with AI always automation rather than a tool?
Not necessarily under every definition of “tool.” Hickey’s distinction applies to the interaction he describes: a user makes a request and the model performs work without continuous direction. Whether a particular use fits that description depends on how the person and system share the task.
Can an LLM still be useful if it is not a tool?
Yes. Hickey’s classification is about what kind of system an LLM is, not a verdict that it has no value. A person can find automation useful while still asking what work—and what learning—they are choosing to delegate.
Does Hickey’s essay prove that AI makes people worse at writing?
No. His essay raises the possibility that delegating writing can remove practice in forming arguments and expressing ideas; it does not measure changes in writing ability. It should be read as a concern about what a person may miss, not as proof of a universal effect.
Why do people disagree about whether AI is a tool?
They may be using “tool” differently. A broad definition can include technology that helps accomplish a goal, while Hickey’s narrower definition requires wielding, feedback, and ongoing control. Clarifying the definition often reveals that people are debating the relationship between user and system, not just the label.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
