More from this creator
Other episodes by Kitty Cat.
More like this
If you liked this, try these.
Transcript
The full episode, in writing.
OpenAI released ChatGPT to the public on November 30, 2022. Two months later it had 100 million users — the fastest a consumer software product has ever reached that number. Instagram took two and a half years to get there. TikTok took nine months. ChatGPT did it over a single college winter break, which is the detail that mattered most for what happened next. When students returned to American classrooms in January 2023, every one of them had access to a tool that could write a passing five-paragraph essay in about eleven seconds.
The first institutional response was prohibition. On January 4, 2023, the New York City Department of Education — the largest school district in the country — blocked ChatGPT on every device on its network. Los Angeles Unified had already done it in December. Seattle Public Schools followed. The University of Hong Kong banned it across every class, every assignment, every assessment, with violations treated as plagiarism, on February 17, 2023. Sciences Po in Paris banned it the same week. The bans assumed that schools controlled access. They did not. Students used the tool on their phones, on home Wi-Fi, in the library on their personal laptops. The bans were unenforceable from the moment they were written.
By May 2023 a survey by Impact Research and Common Sense Media found that 58% of American students aged 12 to 17 had already used ChatGPT, and 38% had used it for schoolwork without their teacher's permission. That same month, NYC Schools Chancellor David Banks reversed the ban. He called the original decision a "knee-jerk fear" that "overlooked the potential of generative AI to support students and teachers." Walla Walla Public Schools in rural Washington reversed its ban for the 2023–24 school year. Sciences Po loosened. Hong Kong loosened. The fastest reversal in the modern history of educational technology took less than five months.
The second response was detection. Turnitin, the dominant plagiarism-detection company in higher education, rushed an AI-detector to market in April 2023 and claimed 98% accuracy. Within weeks the Washington Post tested it and found it had flagged a real student's genuine essay as AI-generated. GPTZero, founded by a Princeton senior named Edward Tian in January 2023, raised millions in seed funding on the same premise. OpenAI itself launched an AI Text Classifier in January 2023 and quietly shut it down on July 25, 2023, citing what the company called a "low rate of accuracy." The makers of the tool could not reliably detect the tool's output. A May 2023 Texas A&M professor put the failure into the most concentrated form possible: he asked ChatGPT itself whether it had written each of his students' final essays. ChatGPT said yes to all of them. He failed the entire class. Texas A&M had to walk it back. Multiple students at multiple universities had scholarships pulled and degrees withheld over false positives. By the 2024–25 school year, most major universities had quietly stopped recommending detection software at all.
The third response was acceptance. Some professors openly assigned ChatGPT — write a draft, then critique it, then revise it. Wharton's Ethan Mollick required it. Vanderbilt's Jules White built a course on prompt engineering. Khan Academy launched Khanmigo in March 2023, a GPT-4-based tutor designed to ask questions instead of giving answers. OpenAI launched ChatGPT Edu, a university-tier product, in May 2024. Baylor University took the opposite path — handwritten in-class essays, oral examinations, the assessment methods of 1925. The Wall Street Journal in June 2023 reported a measurable increase in oral exams across American universities for the first time since the GI Bill expansion of the late 1940s. The collateral damage hit the homework-help industry directly: Chegg, which sold subscriptions to a database of worked problem sets, lost roughly half its market capitalization in a single earnings call on May 2, 2023, after CEO Dan Rosensweig acknowledged ChatGPT was eating its user base. The stock has not recovered.
The fourth response began in 2025 and is still underway. In June 2025 a team at MIT Media Lab led by Nataliya Kosmyna published a preprint titled "Your Brain on ChatGPT." Fifty-four subjects aged 18 to 39 wrote essays under EEG monitoring across four sessions. The group using ChatGPT showed measurably reduced neural connectivity in the alpha and beta bands during writing — markers of lower cognitive engagement — and weaker recall of what they had just written. Kosmyna called it "cognitive debt." A follow-up wave of psychiatric commentary in early 2026 — including from Danish researcher Søren Dinesen Østergaard — warned that the effect was already visible in the writing output of working scientists. The research is preliminary. The sample is small. But it replicated a finding the writing professors had been describing anecdotally since 2023: students could submit competent ChatGPT-assisted essays and then, in oral follow-up, fail to explain a single argument they had just turned in. The Atlantic's Stephen Marche had called the form dead in December 2022 in an essay titled "The College Essay Is Dead." Daniel Herman in the same magazine wrote "The End of High-School English." Three and a half years later the essay is not dead. It is undead — produced in volume, graded in volume, and increasingly disconnected from the brain that was supposed to write it. The institutions are still arguing about detection.