A blog post titled "AI is a tool and using it well takes skill," from a writer at yyhh.org, climbed the Hacker News front page this week, collecting 287 points and 238 comments. Its target is a claim attributed to Rich Hickey: that LLMs are automation rather than tools. The post argues the opposite. In its telling, AI can be wielded skillfully when a human supplies judgment, feedback, and oversight.
It's a short sentence with a big fight inside it, and the comment count suggests plenty of readers had opinions.
The Argument in Plain English
Per the post, the distinction between "tool" and "automation" is not just wordplay. The author frames LLMs as instruments that amplify the user's habits, standards, and expertise rather than replace them outright.
That's the angle worth chewing on: AI as a mirror of the user's judgment. If the post is right, the same chatbot in two different pairs of hands produces two very different results, because the output reflects what the person brings to it.
Why a Mirror Is a Fun (and Slightly Uncomfortable) Metaphor
Pop culture has trained us to picture AI as either a genie or a robot butler. You make a wish, the thing delivers. The post's framing is less magical and more like a guitar. Hand a Stratocaster to a beginner and a seasoned player, and you hear the difference immediately. Nobody blames the guitar for the beginner. (That comparison is mine, not the author's.)
The mirror idea cuts both ways, and that's where it gets interesting. If AI amplifies your standards, it also amplifies your lack of them. Sloppy habits in, sloppy output out, just faster and with more confident formatting. It's hard not to wonder how many of the internet's most cringeworthy AI moments are less about the machine and more about a person who stopped checking.
Why the Timing Matters
According to the coverage, the conversation is shifting. As AI becomes embedded in writing, coding, research, and education, the debate is moving from "should we use it?" to "what counts as responsible use?"
The post pushes back on two camps at once: the hype crowd who think AI replaces expertise, and the rejection crowd who think using it means giving up your own thinking. Its middle path centers on skill, agency, and accountability. You're still the one on the hook for what ships.
That last part is the quiet punchline. A tool doesn't take the blame. If you publish it, merge it, or cite it, it's yours.
Tool vs. Automation: Why Anyone Cares
The disagreement with Hickey is the spicy part. As the post presents it, Hickey's claim is that LLMs are automation, not tools. The author's rebuttal is that guided use makes them tools after all. I can only speak to the post's summary of the dispute, not to Hickey's full argument, so it's worth reading both sides before picking a team.
But the stakes are easy to see. Call something automation and the natural question is what it replaces. Call it a tool and the question becomes how well you use it. One framing invites suspicion, the other invites practice.
What "Using It Well" Looks Like
The post's emphasis, as summarized, is on workflows where AI supports learning, critique, and iteration instead of substituting for expertise. A few ways that translates in practice, as my own reading of the idea rather than the author's checklist:
- Ask it to critique, not just create. Use it as a sparring partner for your draft or your code rather than the author of it.
- Keep your hand on the wheel. Feedback and oversight are the human contributions the post puts at the center.
- Use it to learn. If you can explain the result afterward, you used a tool. If you can't, you may have just hit a button.
Programmers and researchers are the obvious audiences here, and the post's home on Hacker News fits. But the logic stretches to anyone who writes, studies, or makes things.
A Word on Treating It Like a Person
One of the themes flagged around the piece is the danger of anthropomorphizing AI. It's an easy trap. The chat window talks like a friendly colleague, and our brains are happy to play along. The tool framing is a useful corrective: you wouldn't thank a spreadsheet for its insight or trust a calculator because it sounded sure of itself.
The Takeaway
Whether or not you side with the blog post over Hickey, the post lands on a point that outlasts this week's thread. The most interesting thing about AI may not be what it can do, but what it reveals about the person holding it. The mirror doesn't flatter. It just shows you your standards, at speed.
