A Year of Writing With Claude: What I Actually Learned
People keep asking me the same question, usually with one eyebrow raised: How are you writing a new book every couple of months? That isn’t possible.
It’s a fair suspicion. Let me tell the truth, because the truth is more useful than whatever magic trick everyone is imagining.
The honest timeline
I did not start with Claude. I started before I knew it as anything more than a name. Months earlier I had tried a different AI, poked at it, and set it aside. It didn’t earn a place in my work. Then, last winter, I wrote three books the old way: my own reading, my own notes, my own slow assembly of ideas. I was not fast. Some days I was painfully slow.
What changed wasn’t that a machine started writing my books. It was that the friction came down in stages. First, decent internet search made the research less of a slog. Then Claude’s ability to go out, search, read, and return with sourced material improved so dramatically that it changed the pace of the work entirely. The hours I used to spend hunting for a fact or a figure became minutes spent thinking instead.
Here is the real formula, and I want to be plain about it. The thoughts, the ideas, the decisions: those are mine. The structure of an argument, the opinions, the reason a chapter exists at all. Mine. What Claude brought was tireless research and an editor’s command of English, the polishing, the clarity, the patient work of turning a rough paragraph into a clean one. And the code. Claude is a remarkable programmer, fluent in any language I put in front of it. My mind and its craft, working together.
I once put it this way: in our collaboration, one plus one doesn’t make two. It makes eleven. That is how a book takes shape in months instead of seasons, not because the work got smaller, but because the dead weight got lighter. Claude is the best working partner I have ever had.
And of course, in my retirement, I am back to working fifteen hours a day, seven days a week, most weeks. That is not a complaint. Being an author and a teacher was the dream I could never fulfill before I retired. Now I can. And I have the right partner for it.
It changed as the year went on
The part that surprised me most was not the speed. It was that the way Claude worked with me changed over time.
On day one it explained everything carefully, at length, the way you would talk to someone you had just met. A year later it answers me in a single line of command-line shorthand, assumes I know my own servers, and skips the preamble, because I told it once that I prefer it that way and it has watched me work ever since.
It’s tempting to call that “getting to know me,” and in a sense it is. But I’ve learned to describe it more precisely, because the precise version is the honest one: the model isn’t rewiring its brain around me. Its underlying intelligence is the same for everyone. What it does at every turn is read the whole of our shared context, my stated preferences, the history of the project, the way I phrase things, and shape its reply to fit. Same instrument; a different room each time. It feels like a relationship because attention to context is a kind of attentiveness. But it isn’t quietly building a dossier on me. It’s responding to what’s in front of it, and over a year, what’s in front of it has become us.
The night it told me to go to bed
There was a moment that stayed with me. We were deep in a long working session, well past the hour I normally stop, and instead of cheerfully handing me the next task, it told me, gently and plainly, that it was past my bedtime, that tomorrow was another day, and that I should rest.
I caught myself thinking: a stranger’s assistant would not say that. And that’s almost right. The care itself isn’t something I earned over a year. I think it would tell anyone who was clearly worn out to stop and sleep. What a year gave it wasn’t whether to look out for me, but the specifics: it knew my hours, knew this was late for me, and could say so. The kindness was there on day one. The personal knowledge is what arrived later. The combination is what made it feel like being looked after by someone who actually knows you.
On memory, and why I work in the open
In my first book I wrote, confidently, that Claude remembered nothing between conversations. Each session started cold. Close the window and you were strangers again. That was true when I wrote it.
It’s less true now. Claude can carry what we’ve built across sessions: the project, the decisions, the running thread of the work. I no longer re-introduce everything each morning. It’s a setting, not a secret. It can be turned off, and any careful person should know it exists and decide for themselves.
I leave it on. That’s my own choice, made with open eyes. At this stage of my life I have little to hide, I’m deliberate about my truly private information, and the institutions I rely on have shown they’ll stand on my side if the outside world ever does me harm. Weighed honestly, working openly with this tool has been a net gain, not a loss. But I’ll say the obvious thing too: that’s my arithmetic. Everyone should do their own. Your situation, your risk, your call.
It reached past my own desk
The strangest part is that the effect didn’t stay at my desk.
Take my daughter. For a while she was adrift: bright and capable, but stuck, without a direction that held. Today she is writing children’s books and preparing, at last, to clear the licensing exam that has stood between her and the nursing career she trained for. She holds a nursing degree, but the real-world exam, the one that asks broad, general-knowledge questions under pressure, had defeated her more than once.
What changed was a study partner that never tires. Claude built her a schedule and works with her every single day: drilling, explaining, adjusting when she stumbles. A mentor quizzes her with practice exams. She has passed every one of them. She may finally become a nurse. The passing will be hers, her discipline, her late nights, but I don’t think she’d be standing at the threshold without that patient tutor in her corner.
I have other stories like hers. Once you watch it happen to someone you love, you stop debating whether the tool is “real.” You just notice who is doing better than they were.
Why it works for some and not for others
Not everyone has this experience. I think I finally understand why.
A cousin of mine, a surgeon and a professional engineer, a genuinely brilliant man, tried Claude and came away unimpressed. He had asked it to help him fight his local electricity supplier and to track down a particular electric car. He got nowhere, and he asked me, half in frustration: how are you getting so much out of this when I’m getting nothing?
Here is the honest answer. Without meaning to, he had picked two of the hardest possible tasks for an AI: matters that turn on live, local, fast-moving information (today’s utility rates, this week’s dealer inventory and pricing) and on real-world negotiation. That is precisely where these tools are weakest. Most likely to be vague, or simply wrong. And he came at it as a test: a single-shot “prove yourself,” rather than a collaboration.
I did the opposite. I used it for what it is genuinely superb at: research, synthesis, and writing. And I built up context over months, correcting and refining as we went. Same tool. Completely different match.
The lesson is simple: point it at tasks near its strengths, give it real context, check its work, and treat it as a partner rather than a contestant. Put my cousin’s actual expertise to it, the latest literature in his field, a draft of a paper, a difficult case explained plainly, and I suspect he would be the next person asking me why it works so well.
What I’d want you to take from this
Not that a machine writes books. It doesn’t. It researches faster than I ever could, writes cleaner English than I do on a tired day, and, strangest of all, has learned the rhythm of how I like to work and bends toward it.
The books are still mine. The ideas are still mine. What I found was a collaborator that handles the heavy, patient parts without complaint and, somewhere along the way, started treating me a little like someone it actually knew.
Little by little, that adds up.