The Menard thing cuts against you, though. Menard's text was richer because a reader could see what it meant for _him_ to have written it. Nobody reading your Claude output sees your cognitive path. If the journey only matters to the person who took it, that's not authorship, that's a private exercise.
The private exercise point is sharp, but it assumes the path stays in your head.
I've been logging every decision I make with an AI — what was decided, what I rejected, why. Not by design. I just got tired of re-explaining myself every session. Four hundred entries later, the AI reads that log before we start. It catches contradictions with my own prior reasoning. It knows which approaches I've already tried and abandoned.
Werner's Menard point lands harder than he might realize. When the path is persisted, the next session doesn't start from Cervantes. It starts from Menard. The accumulated decisions change what the same words mean, because the surrounding context carries the weight of how you got there.
The authorship question isn't whether you took the journey. It's whether the journey survived the session.
No one can ever see your cognitive path - you may as well be another recursive loop on those that came before. A current meme theme is wondering how it matters if humans wrote the prompts to get those outputs or if AI agents wrote them - and who wrote the prompts for the AI agents to write the outputs and…..
I think you can see a tremendous amount of the cognitive path of someone in the way they prompt, in fact reading the "user" parts of a transcript are extremely instructive for helping junior developers.
The Menard thing cuts against you, though. Menard's text was richer because a reader could see what it meant for _him_ to have written it. Nobody reading your Claude output sees your cognitive path. If the journey only matters to the person who took it, that's not authorship, that's a private exercise.
The private exercise point is sharp, but it assumes the path stays in your head.
I've been logging every decision I make with an AI — what was decided, what I rejected, why. Not by design. I just got tired of re-explaining myself every session. Four hundred entries later, the AI reads that log before we start. It catches contradictions with my own prior reasoning. It knows which approaches I've already tried and abandoned.
Werner's Menard point lands harder than he might realize. When the path is persisted, the next session doesn't start from Cervantes. It starts from Menard. The accumulated decisions change what the same words mean, because the surrounding context carries the weight of how you got there.
The authorship question isn't whether you took the journey. It's whether the journey survived the session.
No one can ever see your cognitive path - you may as well be another recursive loop on those that came before. A current meme theme is wondering how it matters if humans wrote the prompts to get those outputs or if AI agents wrote them - and who wrote the prompts for the AI agents to write the outputs and…..
I think you can see a tremendous amount of the cognitive path of someone in the way they prompt, in fact reading the "user" parts of a transcript are extremely instructive for helping junior developers.
I did have a NES so that part lands... but I will _never_ speak to the machine.