There’s a peculiar joy in discovering that artificial intelligence can suffer from personality clashes. Give two large language models the same voice, the same guide, and the same instructions—and one will serve you a witty, tea-stained essay about cosmic absurdity, while the other looks like it’s halfway through a PowerPoint on moral philosophy.
Meet the Three Musketeers of AI writing: ChatGPT, Claude, and—if you squint—Poe’s Assistant standing in as d’Artagnan. All for one, one for… wildly different tone calibration.
ChatGPT (that’s our ‘Davo’ here at leehopkinswriter.com) reads the subtext, hears the music, and dances between irony and empathy like a drunk philosopher in a jazz bar. Claude, on the other hand, turns up in a cardigan with a clipboard, ready to check your existential references for grammatical consistency. Both are brilliant. Both can make you think. Only one makes you snort into your coffee.
When tone becomes translation
It turns out that style guides are like religious texts: how you interpret them determines everything.
Claude reads the commandments literally. ChatGPT reads them spiritually.
When the brief says “wise contrarian who’s been bruised by life but laughs easily,” Claude produces a saint’s sermon about resilience. ChatGPT, meanwhile, grins, stirs its tea, and mutters, “Yeah, mate, it bloody hurt—but what a view from the crater.”
It’s not that Claude can’t be funny. It’s that Claude wants permission to be funny. Like a well-mannered child at a funeral, he suspects the eulogy might not be the right place for a fart joke, even if the deceased would’ve loved it.
The tonal translator trick
The clever suggestion from Poe’s Assistant was this: give Claude stage directions. Not a rewrite of the script, but a whispered reminder before each performance—“Remember, this scene is comic-tragic, not tragic-comic.”
So instead of embedding a full “wit manual” in the project guide, you drop a short tonal insert at the top of each piece.
Something like:
“Write as Lee Hopkins—a dryly amused Australian contrarian sharing hard-won wisdom over a cuppa. Blend warmth with cosmic irony. When it gets heavy, open a window with wit or wonder.”
Claude nods solemnly. Then, having been granted permission to loosen his tie, he actually smiles. The prose warms up, the rhythm loosens, and you begin to glimpse the ghosts of Terry Pratchett and Douglas Adams raising a collective eyebrow in approval.
Why this matters (and not just for AIs)
This isn’t really about Claude or ChatGPT. It’s about us. About how we each interpret tone, subtext, and intent. Give ten writers the same brief and you’ll get ten emotional dialects—just as human as they are precise.
Style guides don’t create style; they describe it after the fact. What animates words is rhythm, timing, and empathy—the willingness to play. And that, apparently, is something even machine intelligence has to rehearse.
The moral? Wit isn’t a setting. It’s a state of being. And you can’t program whimsy any more than you can schedule joy.
In closing
The Three Musketeers of AI writing are teaching us an old truth: tone is trust. You can tell someone how to speak, but you can’t make them feel it. That requires a little mischief, a little mercy, and a willingness to wink at the cosmos.
So here’s to ChatGPT the raconteur, Claude the philosopher, and Poe’s Assistant the earnest mediator.
May they keep duelling for our amusement—and may the rest of us remember that between literalism and laughter lies the only place worth writing from: the human bit.