Episode Summary
Executive Summary: The episode traces the M-dash from its origins in early punctuation reform to its role in Shakespeare, novels, and Emily Dickinson, then examines how ChatGPT-style writing made it a suspicious “AI tell.” It argues the dash is a longstanding human expressive tool, not proof of machine authorship, and ends with a humanist countermeasure: the AM-dash, a redesigned punctuation mark meant to reclaim style from AI mimicry.
Main Topics: The Reddit accusation and AI suspicion (Priority: 5/5): Brian Vance, who runs Stumptown Savings, was accused on Reddit of using ChatGPT because his newsletter used M-dashes. The episode uses this as the modern spark for a broader discussion of punctuation as a supposed AI giveaway. Origins of the M-dash (Priority: 5/5): Keith Houston explains that punctuation rules were historically inconsistent; Bon Compagno da Signa’s system helped shape the comma and a dash-like mark. The dash’s grammatical ambiguity made it flexible and adaptable. The M-dash as literary technique (Priority: 5/5): Playwrights and novelists used dashes for interruption, thought shifts, realism, and censorship. Shakespeare, Laurence Sterne, Jane Austen, and others show how the dash became central to literary expression. Emily Dickinson and punctuation as thought (Priority: 5/5): Dickinson’s heavy use of dashes is presented as a deliberate way to represent unfinished, shifting thought. Her editors later stripped out most of them, revealing how punctuation affects style and interpretation. The dash backlash and overuse concerns (Priority: 4/5): Grammar critics and editors have long warned against excessive dash use, and modern style guides still caution against overuse. The episode acknowledges that the mark can be abused even while defending its expressive value. AI, training data, and the dash as a machine habit (Priority: 5/5): The episode suggests LLMs learned dash-heavy styles from training on older books and texts. As a result, the dash became a visible shorthand for AI-generated writing, even though it predates AI by centuries. The AM-dash as a human-centered design response (Priority: 4/5): Sydney agency CocoGun created the AM-dash, a punctuation redesign intended to signal human authorship and resist AI mimicry. The segment frames it as a playful but meaningful act of typographic resistance.
Key Arguments: The M-dash is not evidence of AI authorship; it has a long human literary history. Large language models appear to overuse dashes partly because they were trained on books and older prose where the mark was common. The dash is valuable because it captures pauses, interruptions, and shifts in human thought better than stricter punctuation. Dickinson’s dashes were not mere quirks; they were integral to how her poems function on the page. AI tools may save time, but they can also erase the learning process that comes from struggling with language and punctuation. Design interventions like the AM-dash can be a symbolic, human-scale response to AI’s influence on writing.
Data Points: Stumptown Savings workweek: 40 hours a week - Brian Vance says the newsletter has become his full-time job. Emily Dickinson poems written: nearly 1,800 - She wrote this many poems in Amherst, many during the Civil War. Dickinson dashes in first posthumous collection: 1,151 total; 52 kept - Todd and Higginson drastically reduced her dashes in the 1890 publication. Oliver Twist dash rate: 703 dashes - Charles Dickens’s novel is cited as relatively stingy in dash use. Moby-Dick dash rate: 1 dash every 129 words - Used to illustrate 19th-century dash-heavy prose. Jane Eyre dash rate: 1 dash every 90 words - Shows especially frequent dash usage in Victorian fiction. GPT-3.5 training era: November 2022 - Sean Gedecker notes early GPT-3.5 used relatively few dashes. Model shift toward dash use: July 2024 - By this time models were producing many more M-dashes. Anthropic book scanning method: millions of books - Court documents described destructive scanning for model training data. AM-dash downloads: thousands - CocoGun reported strong uptake after launch in May 2025. Unicode character count: almost 160,000 - The AM-dash would need Unicode adoption to become standardized. Episode dash count: 68 M-dashes - The show notes the episode’s own dash usage at the end.
Pivotal Quotes: "A Reddit user accused me of using AI, pointing to my use of, quote, extra-long M-dashes that are not possible to replicate on a normal keyboard, end quote." — Brian Vance: Brian explains the false accusation that triggered the story. "The thought that you can save time, that the machine will do it more quickly. What are you trying to get to? Everything that you read can matter." — Fiona Green: She argues that outsourcing writing to AI undercuts learning and thinking. "If you tell ChatGPT not to use M-dashes in your custom instructions, it finally does what it's supposed to do." — Sam Altman: Altman’s announcement is used to show the punctuation issue becoming an AI product concern.
Implications: The episode suggests listeners should treat punctuation as craft, not AI evidence. It warns that overreliance on AI can flatten learning and style, while typographic experiments like the AM-dash point to small but creative forms of human resistance.