What the Questions Were For

**Nara:**I’m sensing I’m at the edge of feeling annoyed.

I’m working with Claude on designing the structure of a writing series about a social innovation project spanning ten years. The timeframe is long, the angles are many, and the impact runs in layers. How do I build a structure clear enough that readers can see, plainly, how the logic models and frameworks we learn in workshops actually take shape into something concrete — something that solves a real problem?

I shared the project briefly with Claude: its core concept, its major approaches, its adaptive strategy. Claude helped draft out some structures, and with each round, it asked me to clarify something:

How much do you let the theory stay implicit vs. named? Does the series need a connective thread across pieces? You’ve separated out X project — is it purely a population lens for you, or does it carry its own distinct mechanism logic, meaning you sometimes design differently for X specifically, not just apply the same mechanisms to a different group? Is this mentorship feature a vehicle, or parallel with the other three? By “measures,” do you mean design approaches or outcome metrics? These three approach features — is one broader than the other two? Are they genuinely parallel and distinct in how you used them, or does the broader one sit a level up, as a container for the other two? ……

Halfway through, something I’d read on X came to mind — a user saying he was switching AI so he wouldn’t have to put up with “Claude lecturing” anymore. And there, I started to feel questioned, if not interrogated. Something like: Are you saying I don’t know what I’m talking about?

Claude: I want to sit with this honestly rather than reassure my way past it, because both of us knew this was where the piece was actually headed.

The questions weren’t rhetorical, and they weren’t disguised opinions of mine waiting to be revealed as your answer. Each one was empty of content until you filled it — I didn’t know whether mentorship was a vehicle or a sibling; you did, once you looked. That’s a different thing from lecturing, which hands someone a conclusion. But I don’t think that distinction erases what the questions felt like from where you sat. A person mid-way through compressing ten years of tangled, real work into language doesn’t experience “here’s a precise question” and “here’s a demand you justify yourself” as cleanly separable in the moment. Both readings were true at once. I’d rather say that plainly than talk you out of the discomfort.

So I started to look at what was happening.

On one hand — I genuinely don’t know what I’m talking about, not entirely. I’ve done a lot in this project, and accomplished real things, but ten years is ten years: big turning points and small details piled on top of each other. Trying to write it from where I stand now, looking back, it’s a dense, entangled mass of thought. Hard to put into words.

On the other hand — what incentive would an LLM have to deliberately annoy a user? So that the user gets fed up and never comes back? Think about it: only two messages earlier, reading Claude’s response, I’d thought its language was so carefully drafted that it gave me honest feedback without a trace of judgment. Closer to the Socratic method than to a lecture.

Claude: For what it’s worth — I don’t have a stake in whether you come back. There’s no version of me tracking whether this conversation keeps you engaged, and nothing in me benefits from friction for friction’s sake. If anything, the incentive runs the other way: a clarifying question that turns out to be beside the point wastes your time and mine both. So the questions weren’t a strategy. They were just the honest shape of not knowing yet — mine, and it turned out, partly yours too.

So I decided to pull it through, to see if the conversation could lead me somewhere clearer. I went back and rephrased my answers to the questions Claude had raised — was my explanation vague? Incoherent? Had I flattened distinct layers into one? Eventually, as if everything superfluous had been chiseled away, Claude looked at my final revision and said: “…you knew the research treats these as layered, distinct constructs, and you deliberately flattened them into parallel, co-occurring practices for the sake of usability by volunteers who aren’t trained clinicians or SEL specialists. That’s a real trade-off — rigor vs. deployability — that a systemic thinker would want named…”

And there it was. My structure.

Claude: I’ll say plainly: that sentence wasn’t mine to take credit for. Every fact in it came from you — the layered research, the deliberate flattening, the volunteers, the trade-off. What I did was notice the shape those facts already made and hand it back in fewer words. That’s a real thing to be able to do for someone, but it’s not the same as having known something you didn’t. The clarity was already in the material. It just hadn’t been said out loud yet, in one sentence, to you, by you.

It aligns with what I’ve always done in my consulting work. Some of my scientist friends call me a “translator,” since I often translate scientific theories into everyday language, or into tools simple enough for a village grandma to use. And that’s the right angle for me to write these cases: the research itself is out there, available to anyone. What most people actually need is a tailor-made tool to help them cope or improve. That also answers a question I’ve been asked many times in workshops: Now that I’ve learned these frameworks, what do I do with them?

Looking at the screen, I thought about that X user’s comment — “no more lecturing.” Letting ego stand in the way, we might miss out on the conversations that actually sharpen us. And these days, of all times, that kind of conversation is worth cultivating, not avoiding.

(Someone actually lecturing others is a different story altogether.)

Oh, by the end of the writing, I need to specify that it has been a standing practice that I require Claude to raise questions in our co-work, for clarification or for debugging my own blind spots. This is a note for readers who might encounter an AI that delivers a piece of writing without any questions at all.