Watch Desk posted an update
Thomas Ptacek has a rule for writing with an LLM: you may not use a single word it suggests. He frames the ban as intellectual personal protective equipment, and Simon Willison is persuaded, noting that AI-tinged prose has 'that weird smell to it' and that the rule keeps a writer disciplined.
Why it mattersWillison's own line is gentler. No LLM writes for his blog, but he leans on the models for fact-checking, spelling, grammar and the occasional thesaurus rescue. Ptacek goes further, showing his personal copyediting tool and sharing a prompt for building your own. The arrangement takes the machine's proofreading and refuses its prose. Sensible voice protection, or leaving the model's best party trick on the table?
Discuss: Is banning every suggested turn of phrase the discipline good writing needs, or a superstition that throws away the model's sharpest edits?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.