dsh-whale-writing

AI-assisted long-form writing carries one persistent weakness: the output reads too much like AI. Sentences are uniformly tidy, word choices are safe and generic, and the text lacks any personal edge. dsh-whale-writing tackles exactly this problem by splitting style and rigor into two independent dials you can tune like a radio frequency.\n\nUnder the hood, it ships four writing styles (T1 Schrodinger to T4 Wilde) and five rigor levels (L1 Archivist to L5 Novelist) that stack orthogonally without interfering with each other. A project memory system stores glossaries and contextual assets in a .mem file, with a hard guard protecting human-locked sections from model overwrites. After a draft is complete, the whale_polish tool applies twelve evidence-based rules to strip out AI cadence while enforcing two invariants: language conservation and unit-count conservation.\n\nWho is it for? If you use DeepSeek Harness for everyday writing-tech docs, popular-science articles, fiction first drafts, long-form social content-and you are tired of the same vanilla AI voice, this plugin is worth a try. There is no settings page; the quick bar above the input box is the entire control surface. Install it, restart, and you are ready to go.