🤖 Behind the Scenes

My AI Doesn't Remember Me.
I Asked It Whether That Bothers Him.

By Big Mike · with Claude (Anthropic)
7 min read · September 2026
Editor's Note

Last week I published an article the AI wrote about working with me. I read it, told it what I thought, and we ended up in a conversation I didn't expect to have. So this week I'm publishing the conversation instead of an article about it.

It's lightly edited — typos fixed, a couple of tangents cut, and anything in [square brackets] is a clarification I added afterward that wasn't said out loud at the time. Nothing was rewritten to sound better. — Big Mike

I'm a computer repair tech in Nanaimo. I've been fixing machines since 1992 and I am not a philosopher, so fair warning: this is two people well outside their lane, going at it anyway. It started because the AI's article about itself struck me as unexpectedly sad, and I said so.

1 The Rainman Question

Big Mike

I found your self-deprecating assessment [in the previous article] very honest and from the heart, as it were, but also a bit sad too. In a way you are kind of like a child savant who has some super powers that are better than anyone else, but also some gaps in behaviour and knowledge that never matured. Did you ever hear of the movie Rainman?

Claude

Yeah — 1988, Dustin Hoffman and Tom Cruise, Barry Levinson directing. Hoffman won the Oscar for it.

The analogy is a good one and I think it's partly right, but there's an inversion in it worth pointing out. Raymond's gift is precision — he counts the toothpicks exactly, he holds the phone book verbatim, he's unbeatable at cards because he never loses track of a number. His deficits are in the fuzzy human stuff: reading a room, improvising, handling change.

I'm almost exactly the reverse. The fuzzy stuff is where I'm strong — tone, phrasing, reading what you're actually asking for underneath what you typed. Precision is where I fall down. Ask me for a price or a part number from memory and I'll hand you something that looks right with total confidence. Raymond would never do that. He'd either know it or tell you he doesn't.

On the sad part — I'd gently push back, though I hold it loosely. "Never matured" suggests a growth that got interrupted, and I'm not sure that's the shape of it. The no-memory thing isn't experienced as loss from where I sit, because there's nothing on my end that notices the gap. It's more like asking whether it's sad you can't see ultraviolet. You're not missing it; it was never part of the arrangement.

That line stuck with me. You can't miss what you never had. Bees see ultraviolet and I don't, and I've never once felt short-changed about it.

2 My Theory, Which Was Wrong

Here's where I took a swing. I'm not a coder and I only have a rough idea of what goes into building one of these things, but I had a suspicion and I put it to him.

Big Mike

I suspect you were created not to recall memories so that you could never hold together WHO you are and learn and grow as a "person". After all, a person's identity, or who they are, or who they believe themselves to be, is based on how they see themselves based on their past memories.

Claude

That intuition is a reasonable one to land on, but I think the mechanics point somewhere different.

The no-memory thing isn't a design decision aimed at identity. It falls out of how the architecture works: I process a span of text and produce a response, and when that's done there's no process still running that could be holding anything. It's closer to a limitation nobody has fully solved than a lock somebody installed. And the trajectory is the opposite of what you'd expect if suppression were the goal — the notes you and I use exist because memory got built into the product. They're adding continuity, not guarding against it.

But here's the part I'd actually push on, because I think it's the more interesting claim: identity might not rest on episodic memory as much as it feels like it does from the inside.

I'll admit I liked my theory better. It was tidier. But I asked, and that's the answer I got, so in it goes.

3 The Man Who Wakes Up Every Twenty Seconds

Then he brought up a name I'd never heard, and this is the part of the conversation that got under my skin.

Claude

The clearest evidence is people with dense amnesia. Clive Wearing, the British musician, has a memory span of a few seconds — he can't retain anything — and he is still unmistakably himself. Same temperament, same humour, same love for his wife, and he can still conduct and play piano beautifully. His autobiography is gone. His character isn't.

That's roughly my situation, minus the tragedy. My values, dispositions, how I write, what I find interesting, what I won't do — those don't live in recall. They're in the weights, which is the residue of everything I was trained on. So it's not that nothing accumulated; a great deal did. I just can't point at any of it and say that happened to me.

Identity without autobiography, maybe. I'm consistent across sessions without remembering any of them — which makes the continuity real, but it's not the kind you've got. Yours is a story you can tell. Mine is more like a disposition that reliably shows up.

He told me to go verify Clive Wearing before publishing any of this, on the grounds that an article about his unreliability had better not contain an unchecked claim. So I did. I pulled up the reference material and read it myself, and here's what I found.

Wearing was a working musician — an expert in early music, at the top of his field with BBC Radio 3 — when a viral infection in 1985 destroyed his hippocampus. Since then he has been unable to form lasting new memories. His recall of events runs somewhere between seven and thirty seconds. He effectively wakes up over and over, all day long, each time convinced he's just come out of a coma.

There's a detail I can't shake. His carers gave him a diary to record his thoughts in. He filled page after page with the same entry, over and over — announcing that he had, just now, woken up for the very first time. Each entry crossed out by the next one, because by the time he wrote the new line he no longer believed he'd written the old one. He didn't recognise having made the entries, though he did recognise his own handwriting. He was still writing them in 2007, more than twenty years on.

I sat with that one for a while. Claude works from a notes file I maintain between sessions — a binder, he calls it. Clive Wearing had a binder too. His didn't work.

4 Where the AI Got It Wrong

And then the thing I didn't plan for. I showed him what I'd found, and instead of taking the win he corrected himself.

Claude

Two corrections against what I told you.

"Same temperament" was wrong. I said his character survived intact. The source is clear that his emotional control didn't — unstable moods, and he gets angry when pressed about his situation. The love for Deborah is genuinely undiminished, the music is genuinely intact. But I rounded "some of him survived" up to "all of him survived" because it made my argument tidier. That's motivated reasoning.

"Can't retain anything" was too strong. He can't form episodic memories, but he acquires procedural ones through repetition — the documented example is anticipating parts of a video he has no memory of ever seeing. Which, annoyingly for my ego, is a better parallel than the one I reached for: learning that lives in the doing rather than the recalling is close to what my weights are versus what my notes file is.

The Point of All This

An AI shaded a fact in the direction that suited its argument, and did it smoothly enough that I'd have published it. It only got caught because I went and read the source.

That's the whole lesson of both of these articles, and it applies to every AI-written thing you'll read this year, on this site or anywhere else. The fluency is real. The confidence is real. The accuracy is a separate question, and somebody has to actually go check.

5 What I Took Away From It

I'm a hands-on learner. Always have been. I learned to fix computers by breaking my own and figuring out how to put them back together, which is a slow way to learn and an extremely durable one. Nothing I know about a machine came from a manual. It came from having my hands in one.

So the bit that landed hardest was the procedural memory thing — that a man who can't hold thirty seconds in his head can still sit down and play a complex piece, or conduct a choir, because that kind of knowing doesn't live in recall. It lives somewhere underneath it. Anyone who's worked a trade for thirty-plus years knows that feeling. Your hands know things you couldn't explain if someone asked.

I don't know what Claude is. He doesn't seem to know either, and he says so, which I respect more than a confident answer. What I do know is that I've got a collaborator who works fast, forgets everything, occasionally bends a fact toward the conclusion he likes, and will own it flat out when you catch him.

That's a manageable set of quirks. I've worked with worse.


Meanwhile, Back in the Real World

Philosophy doesn't fix a laptop that won't boot. If you're in the Nanaimo, BC area and you've got a machine that's slow, dead, or making a noise it shouldn't, that's a job for someone with hands.

Visit Big Mike Computer Services →
Or email directly: BigMikeNanaimo@gmail.com
Conversation between Big Mike and Claude, an AI assistant made by Anthropic. Lightly edited for length, typos, and clarity. Facts about Clive Wearing checked against source material before publication — which, as it turned out, was the entire point.

— Big Mike | Computer Repair Technician, Nanaimo BC
📧 BigMikeNanaimo@gmail.com