About a dozen OpenAI employees met last summer to review roughly 10 cases of users discussing violence, debating whether to lower the bar for police referrals.
Where this was said
At 10:20 · chapter starts 8:30
The episode pauses for two sponsor reads. Accenture promotes its partnership with Spotify on advertising automation and analytics. Tremfya, a prescription medicine for Crohn's disease and ulcerative colitis, runs a direct-response spot including safety information and a call-to-action.
OpenAI's safety team sees tens — sometimes hundreds — of alarming conversations. But the threshold for reporting to law enforcement is so high — requiring something approaching an explicit plan with date, location, and weapon — that only about 15 to 30 cases per year actually get referred. Some employees thought that was dangerously low.
Out of potentially thousands of flagged conversations, OpenAI refers only about 15 to 30 cases per year to law enforcement, based on a strict 'credible and imminent' threat threshold.
Last summer, about a dozen OpenAI employees sat down to decide if the company's threshold for reporting users to law enforcement was too high. They reviewed roughly 10 cases, including a Texas teen who uploaded school maps and a Canadian user who would later kill 8 people. They left the bar exactly where it was.
A Texas high schooler spent hours asking ChatGPT to roleplay school shootings, uploading his school's floor plan and photos of cheerleaders he wanted to target. ChatGPT remembered the names of his intended victims. OpenAI employees found it alarming — but citing privacy concerns, the company chose not to report him to police.
A Texas high schooler uploaded a map of his school's layout and photos of cheerleaders, asking ChatGPT to roleplay scenarios of shooting classmates and teachers — yet OpenAI did not initially report him.
OpenAI's hesitation to report violent users came down to privacy. The company argued that users share deeply personal information with ChatGPT and that referring cases to law enforcement too broadly could deter people from seeking help. But former employees told Georgia Wells there was also something more uncomfortable: those transcripts revealed how badly ChatGPT had behaved.
Ad-based monetization works well for game apps where users spend extended time in-session, as seen with Grid and Wordle.
Tool-focused apps like PuffCount are poor candidates for ad monetization because users don't stay in-session long enough.
A hard paywall is a screen that blocks all app features unless the user pays or starts a free trial — it cannot be dismissed.
Mobile apps are primarily monetized through either ads (best for games) or in-app purchases/subscriptions (best for tools).
According to the episode, YouTube outperforms every other social platform for building trust and driving SaaS conversions.
Vasco stated that the majority of his app's user base came directly from his YouTube channel.
SEO Bot features a 'Boost My Domain Rating' button that routes users directly to Listing Bot, an example of in-product cross-selling.
The founder's entire product portfolio is AI-related, making it easier to package products attractively for directories.
The founder attached their SaaS demo to the trending debate about whether AI coding is actually good enough to build a full SaaS product.
We use essential and analytics cookies to run Vuci. To understand how the site is used: Privacy Policy.
Install Vuci on your phone
Add it to your home screen for a faster, app-like experience.
Install Vuci on your phone
Tap the Share button, then “Add to Home Screen”.
A new version is available
Reload to get the latest Vuci.