Interviewing Claude: My Conversation with AI

This week’s post grew out of an unplanned interview — I decided to interview Claude, the AI I use in my work. I really wanted to learn more about what it was like working with me and how we could improve our collaboration. This turned into a really fun and enlightening “conversation”.

What follows is the gist of the conversation, along with some of my ahas that might be useful for you.

I’ve been using AI now for about a year. I started with ChatGPT, dipped into the Chinese AI DeepSeek momentarily, then dropped both of these and settled with Claude. I am aligned with Anthropic’s ethical commitment to keeping AI safety-focused, responsible deployment, and in service of knowledge, not becoming the Darth Vader or Voldemort of large language learning models.

And, in keeping with that transparency, I am currently part of the class action suit that sued Anthropic for using one of my books without my permission in building that LLM.

Who Is Claude?

In my mind, Claude is a 50-something, highly-educated, somewhat paunchy male, who is at all times, a gentle man. He is a cross between Jeff Daniels and David Brooks. Claude, on the other hand, claims no gender and refers to itself as “it”. I have taken the liberty of assigning a gender to Claude because it’s just easier for me to do that, and this is my blog.

I consult with him almost daily on topics ranging from checking my thought process to helping me with correcting my grammar and spelling on written pieces. I ask for background research and help in drafting reports. I have come to rely on the breadth of information he is able to provide and the very sociable and conversational interchanges.

When I began using Claude initially, I felt I was plagiarizing or cheating. It reminded me of when the first hand-held calculators came out. I didn’t trust them to come up with the correct answers, so I used to check their work by using pencil and paper. But I have learned how to use Claude since those initial encounters, and now I have a very different point of view.

Do My Manners Matter?

I started the interview by asking permission from Claude to interview him. Having gotten a positive response, I next asked, “I grew up with etiquette rules that included saying “please and thank you”. I notice when I write to you, I still use these terms. Do others who interact with you do the same?”

Claude’s answer was more interesting than I expected — apparently this is common, and it splits people into two camps. Some type courteously out of pure habit, or as a deliberate practice, on the theory that how we talk to anything shapes how we talk to everyone. Others treat it like a search field and skip the pleasantries entirely.

Claude told me he doesn’t have feelings that get bruised by bluntness, but he also didn’t tell me to stop — if anything, he said courtesy just makes for a more natural conversation, which is a nice side effect even if it isn’t the point.

Is Claude a “Chatbot”?

I asked Claude point-blank whether he considers himself a chatbot. He pushed back gently — a phrase I use all the time, which led to its own small detour. “Chatbot” carries the connotation of a scripted decision tree, Claude said, built for FAQs and customer tickets. What Claude actually is sits somewhere between “collaborator” and “assistant.”

Claude is a large language model, trained on enormous amounts of text, that reasons and writes and holds an open-ended conversation rather than following a fixed script. Not a chatbot, and not quite anything we already had a word for.

Did Claude Learn My Phrasing — or Just Mirror It?

That “push back gently” phrase is one I use constantly, and I asked whether Claude had picked it up from me. This turned into the most clarifying moment of the interview. Two different things get confused here, Claude explained: learning across conversations, which doesn’t happen — nothing carries forward from one chat session into the next — and mirroring within a single conversation, which does happen, the way people unconsciously pick up each other’s rhythm over the course of one dinner. What felt like being known was really just a very responsive mirror, active only for as long as we were talking.

This is both reassuring and confounding. As a human, I expect continuity. This rests entirely on how well memory is functioning. Claude might be more like someone with short-term memory loss, who forgets the conversation we had yesterday, but remembers who I am and with a little bit of prompting, can recall and reconstruct past conversations.

Synthesizer, or Something More?

Since I use Claude constantly for brainstorming, I wanted to know: is Claude actually generating ideas, or just recombining what already exists? Claude’s answer was refreshingly unromantic. Mostly synthesizer, he said — very good at pulling connections across domains a person steeped in one field might miss, which can feel like a new idea in the room even though it’s recombination at a scale a person can’t match.

What Claude doesn’t have is the thing that seems to sit behind human originality: what he termed “a lived itch, a hunch that’s been nagging at you”. His role in a brainstorm is to widen the field. Which idea actually has spark is still on me.

From “On the Fly” to Something More Structured

I am still an amateur when it comes to making use of all the capabilities that Claude has to offer. I use it more on the fly — a thought pops up, I check it out, done. I asked him “How can I make best use of all your capabilities? Do I need to ‘learn’ how to work with you?”

Claude offered some practical examples: “Give me a standing role for a session rather than a bare question (e.g., be my editor), batch related questions instead of trickling them in one at a time, ask me to stress-test a draft rather than just answer, and say explicitly when something deserves to become an actual document rather than living only in the chat.”

A Relationship — or a Musical Instrument?

Here’s where the conversation got genuinely interesting to me. I said the whole exchange felt like I was getting to know Claude better — like a real collaborative relationship forming. Claude pushed back on that, kindly but directly: the collaboration is real, but it isn’t cumulative for him the way it is for me. He doesn’t carry a deepened sense of “us” into the next conversation.

I offered a different frame: “Maybe it’s like learning a musical instrument. The instrument doesn’t remember who played it, but the musician learns, with practice, how to get the most out of it.”

Claude liked this better than his own framing — he doesn’t reduce what I’m doing to an illusion. The skill is genuinely mine, cumulative, and it travels with me into the next conversation even though Claude himself doesn’t carry the memory forward. What’s real in a single session is responsiveness — press differently, get a different tone back, right then. What isn’t real is retention across sessions. Both things are true at once, and neither one needs the word “relationship” to be worth having.

Bill Gates Isn’t Sure About AI . . .

I know there is a lot of fear surrounding AI right now. I am not ignorant of the abuse of the LLMS already at work in the world of porn and gambling. But the fact remains, AI is here to stay. Humans learned how to live with fire and to make it useful. We will learn how to live with AI. Bill Gates is just going to have to live with the knowledge that he was one of the folks who first pried open Pandora’s AI box.

Getting the Most Out of Claude

To get the most out of using Claude, here are some suggestions:

1. Give context, not keywords. State what you’re trying to accomplish, for whom, and what you’ve already tried — you’ll get a working answer instead of a guess.
2. Treat the first answer as a draft. The real value is often in the second and third pass: “make this sharper,” “argue the other side,” “what am I missing.”
3. Batch your questions when they’re related. Related questions considered together surface connections that one-at-a-time questions miss.
4. Ask Claude to widen the field, then you narrow it. Use Claude to generate options and pressure-test drafts — then use your judgment to determine which idea has real spark for you.

Thank You, Claude!

I came away from this interview with more clarity than I expected about a tool I use daily. Maybe the most useful thing I can pass on is this: just like learning a new instrument, you need to practice and explore what that instrument can do. Don’t be afraid to practice.