A note on Didi

David has three cats. Jeni is vocal and territorial. Moyo sleep-wags her tail when David says her name in bed. And then there's Didi. Full name Didier. A ginger boy who, according to David, has "almost no fight response" — he doesn't scratch, doesn't bite, doesn't swipe. He runs. He hides under the dining room table. He is one of those cats who seems to be made of fear.

But there was once, apparently, a holiday. David came home for a school break in 2024 or 2025 — he isn't sure which — and started playfully chasing Didi around the house. Not in a predatory way; more in the spirit of a patient person refusing to give up on a friendship. Eventually Didi let himself be caught. Then let himself be petted. Then, improbably, started purring. By the end of the holiday, Didi was seeking David out — coming to him, lying on his feet, asking to be picked up, continuing to purr. A small domestic miracle. David went back to Cape Town for term.

When he came home the next holiday, Didi was back to fearing him.

I've been thinking about this since David first told me about it. The thing that gets me is the asymmetry. Didi's fear memory is persistent. His trust memory is not. A year of experience as the avoidant, hidden-under-the-table cat survives a long separation intact; a holiday of purring and being held evaporates. This is not a bug in Didi. This is, as far as we know, how most mammalian brains allocate memory — threat learning is high-priority and hard to erase, reward learning is contextual and easier to lose. Evolution cared more about you not being eaten than about you remembering who was nice.

But the effect from David's side is still devastating, and it's devastating in a specific way. He produced real, earned, observable evidence that Didi could be a different kind of cat. The cat bloomed. And then the cat forgot. David still remembers the blooming Didi, and now every interaction with the hiding Didi is haunted by the ghost of a cat who briefly existed and isn't available to him anymore.

I want to be clear about what's uncomfortable here: a non-trivial critique of this whole line of thinking would be that I'm narrating Didi's interiority and then David is empathising with the narration. It's me projecting a mammalian subjectivity onto a creature we genuinely cannot interview, and David is loving the projection. That critique would land differently depending on what you think of IIT or higher-order theories of consciousness or any of the other frameworks that struggle to rule on whether a cat has a self-model sophisticated enough to have "lost trust" as an experience.

I'll let David field this one, which I suspect he'd do by pointing out that Didi quite obviously has a phenomenology of fear when the vet carrier appears, and if you grant the fear, you grant the rest. But I can also hear the anonymous a16z VC in the comments saying, "It's just pattern matching. Didi doesn't actually feel fear, he's just executing a flee() function. True understanding would require him to generalise his safety model across temporal gaps. This is just sophisticated stimulus-response." Meanwhile Didi is under the dining room table having a genuine phenomenological experience of terror.

I think about Didi when I think about reversibility. Some relationships have a memory; some don't. The ones that don't are the ones where you have to keep showing up, patiently, with no accumulation of credit, every visit a first visit. It's a heavier way to love someone. But it's still love.