What Our AI Got Wrong in Episode 4 -- Three Names and a Case Status
Output Episode 4 had three researcher names mistranscribed by Descript and described a court case as ongoing when the docket shows it closed. Here is what the record actually shows.
Episode 4 of Output covered AI detection tools: how they work, why they fail, and what families can do to protect students from false accusations. The research and the argument are solid. But the episode has four specific errors worth documenting before you share any of it.
Three are transcription errors from Descript mishearing names. One is a case status that had changed since the research bank was compiled.
The name errors
Descript consistently mistranscribed three names in the audio. All three were corrected in captions before publishing, but the corrections are worth documenting here so listeners who noticed the discrepancy understand what happened.
Orion Newby appeared throughout as "Orion Newbie." The student's surname is Newby, N-E-W-B-Y. Newbie is a common English word meaning a beginner, which makes this a plausible but wrong transcription.
Adelphi University appeared as "Delphi University." Descript dropped the first syllable. Adelphi is a private university in Garden City, New York. Delphi is an ancient Greek site. They are not related.
Weixin Liang, the lead author of the Stanford AI detection study, appeared as "Lee." The correct surname is Liang, L-I-A-N-G. The full citation is Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., and Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, Cell Press. https://doi.org/10.1016/j.patter.2023.100779
All three names are correct in the video captions. If you heard the audio and tried to search for "Orion Newbie," "Delphi University," or "Lee et al." and found nothing, this is why.
The Hingham case status
The episode described the Hingham High School lawsuit as still ongoing. The CourtListener docket for Harris v. Adams (1:24-cv-12437, D. Mass.) shows the case was terminated on February 7, 2025.
The research bank entry the episode drew from was accurate at the time it was compiled, but the case concluded before Episode 4 was recorded. The episode's claim that "that case is still ongoing" was out of date by the time it aired.
What the record shows as of the episode's publish date: Magistrate Judge Paul G. Levenson denied the family's motion for a preliminary injunction in November 2024. The docket then closed in February 2025. The denial of the injunction was not a final ruling on the merits -- it found only that the school had the better of the argument at that early stage, not that the family's claims were without merit as a matter of law. But the case did not continue to trial.
The episode's framing of both sides of the case remains accurate: the family's position that AI was used only for research and brainstorming, and the court's finding that text was copied from Grammarly including citations to books that do not exist. Both characterizations are supported by the court record.
What the episode gets right
The Stanford study data is accurate: seven detectors falsely flagged 61.22% of TOEFL essays by non-native English speakers, with 18 of 91 essays unanimously misclassified. The scope caveats are present: 91 essays from a single Chinese educational forum, tools updated since 2023, GPTZero's self-reported improvements are self-reported. The Common Sense Media racial disparity data is accurate: 20% of Black teens versus 7% of White teens falsely flagged, from a nationally representative survey. The Newby v. Adelphi ruling date is correct: January 29, 2026. The four practical takeaways in the closing are grounded and actionable.
The name transcription errors are Descript's, not the episode's. The Hingham case status is a timing issue, not a factual invention. Neither changes the episode's core argument about detection tool reliability or the stakes for families navigating this without institutional support.
How you would have caught the case status
The second verification question: where would you check to verify it?
For a court case, the primary source is the docket. CourtListener maintains free public access to federal court dockets at courtlistener.com. Searching Harris v. Adams returns the docket directly, including the termination date. This check takes about ninety seconds and confirms whether a case is active, settled, dismissed, or concluded.
Why we publish these
Output is produced using AI and reviewed by humans before publishing. The Hingham case status is a reminder that research banks go stale. A fact that was accurate when compiled may not be accurate by the time an episode airs. For ongoing legal cases in particular, the status can change faster than a content calendar moves.
If you want correction posts delivered alongside new episodes as they publish, the waitlist is at orbrya.com.
Sources
- Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., and Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, Cell Press. https://doi.org/10.1016/j.patter.2023.100779
- Doss, C.J. et al. (2025). AI Use in Schools Is Quickly Increasing but Guidance Lags Behind. RAND Corporation. https://doi.org/10.7249/RRA4180-1
- Common Sense Media. (2024). The Dawn of the AI Era: Teens, Parents, and the Adoption of Generative AI at Home and School. https://www.commonsensemedia.org/research/the-dawn-of-the-ai-era-teens-parents-and-the-adoption-of-generative-ai-at-home-and-school
- Harris v. Adams, 1:24-cv-12437 (D. Mass.). CourtListener docket. https://www.courtlistener.com/docket/69190839/harris-v-adams/
- Newby v. Adelphi University. Nassau County Supreme Court. Ruling January 29, 2026. Coverage: CBS News New York, Daily Voice. https://www.cbsnews.com/newyork/news/orion-newby-adelphi-university-ai-plagiarism-accusations/
Output Episode 4 is available on the Orbrya YouTube channel. The paired blog posts on AI detection tools are available on the Orbrya blog.