On AI and Writing in 2026

Craig Mod in his most recent Roden newsletter on “Claude Voice”

I like knowing a literary text comes from a person because I sense, in some way, that I could know the person or understand the mind that produced the text.
[…]
AI-generated text doesn’t have this knowability. It has legibility (sometimes), of course, but the source is unknown, there is no “mind,”

I agree with the latter assertion that AI text output has legibility. We can read it and it (mostly) makes sense.

The knowability aspect is the purview of AI interpretability researchers, though it’d be interesting to know if interrogating the model for the reasons for, and source of, its outputs would yield much more beyond “because you asked for it”.

Well worth a read for the full concerns around writing across all domains and how downstream of that you can’t trust what you read on platforms as their incentives reward output velocity rather than human provenance.