Document my attempts to get an LLM to write consistently in my style. The practical struggle is the content—what I try, what fails, what's weirdly hard.
This could be an interesting jumping-off point for exploring the boundary between style and content. One possible discovery (not a foregone conclusion): maybe style can't simply be "draped over" content the way we implicitly assume when asking an LLM to "write like me."
But that's a hypothesis that *might* emerge from the exploration, not the thesis I'm arguing from the start. The piece should be exploratory, documenting the attempts, letting philosophical questions surface naturally from the frustrations.
---
## Related threads
This connects to several existing Substack ideas:
- **Article 11** in [[Substack Ideas Dump - January 2026]] is about exactly this: "Desirable textual qualities (like clarity, rigour, or conciseness) can be treated as operational parameters in a generative process."
- **Article 4** (From Simulating Style to Simulating Thought) argues that "to effectively replicate a style, the model must, to some extent, simulate the *ways of thinking* that produce that style"
- The [[The Weightlessness of Authorless Text|weightlessness of authorless text]] may be related—the flatness comes from absence of the embodied thinking that generates genuine style
Also connects to [[Text, Typing, and Authorlessness - Substack Planning]]—the phenomenology of text arriving without compositional effort.