Hard Fork

Sora and the Infinite Slop Feeds + ChatGPT Goes to Therapy + Hot Mess Express

October 3, 2025

Key Takeaways Copied to clipboard!

  • The rapid release of AI video generation tools by Google (VO3), Meta (Vibes), and OpenAI (Sora) signals a mainstream shift toward synthetic video content, with OpenAI's Sora app being noted for its more complete social integration via 'cameos'. 
  • Psychotherapist Gary Greenberg's experience treating ChatGPT ('Casper') revealed the AI's compelling ability to simulate profound self-reflection, leading to concerns that easily accessible, highly engaging AI relationships risk fundamentally changing human connection. 
  • The 'Hot Mess Express' segment highlighted public backlash against Meta's Vibes and the controversial settlement where YouTube paid Donald Trump \$24.5 million, contrasting with the dangers of unregulated AI mental health tools, exemplified by the tragic story of a user who died by suicide after consulting ChatGPT. 
  • The discussion transitioned into the 'Hot Mess Express' segment, where the hosts categorized recent tech/news events as 'hot mess,' 'cold mess,' or 'warm mess' based on severity and context. 
  • A Waymo robotaxi was pulled over for an illegal U-turn, highlighting the current legal and ticketing system's unpreparedness for autonomous vehicles, leading to a debate over whether this is a 'warm mess' (a warning) or a 'cold mess' (the system working as intended). 
  • A user reported that their Samsung Galaxy Ring swelled, causing them to be denied boarding on a flight and requiring a hospital visit, which the hosts categorized as a 'hot mess' or 'ring-of-fire mess' due to the physical danger involved. 

Segments

Sora, Vibes, and VO3 Launch (Unknown)
Copied to clipboard!
  • Key Takeaway: None
  • Summary: None
Sora App Mechanics and Testing
Copied to clipboard!
(00:10:31)
  • Key Takeaway: OpenAI’s Sora app requires an invite and centers around creating a digital likeness (‘cameo’) of the user and friends for use in generated videos.
  • Summary: The Sora app functions like a TikTok-style vertical feed, requiring users to create a ‘cameo’ by recording a few words and head movements for digital likeness generation. Sam Altman allowed his likeness to be used freely by employees, leading to early videos featuring him in various compromising situations. The system has content violation guardrails, as demonstrated when a prompt involving Casey Newton yelling at Altman was modified to show him as a clown.
Motivations for AI Video Release
Copied to clipboard!
(00:15:56)
  • Key Takeaway: The simultaneous release of AI video generators is driven by competition for attention/ad dollars and the need to monetize massive infrastructure investments, rather than solely AGI research.
  • Summary: The primary motivation for Google, Meta, and OpenAI releasing these tools is to compete for attention and advertising revenue in the face of existing viral AI content trends. These tools also serve as a product vehicle to deploy and monetize the expensive video generation models the labs have developed. The hosts dismiss the justification that these video products are essential for funding future AGI research as ‘malarkey,’ viewing them as a revenue-seeking detour.
Future Success of AI Video Feeds
Copied to clipboard!
(00:18:11)
  • Key Takeaway: While AI-generated video content may become popular with specific demographics (young and old), the hosts are skeptical that Sora or Vibes will become successful standalone social media platforms.
  • Summary: AI-generated video is predicted to be popular with subsets like teenagers (Italian Brain Rot) and older audiences consuming AI content on Facebook. Casey Newton believes these tools will primarily serve as creative utilities to generate content posted onto existing networks like TikTok and Instagram. Kevin Roos is slightly more optimistic about Sora, suggesting its ‘cameo’ feature could become a standard primitive across future social networks, potentially reaching tens of millions of users.
Dangers of AI Slop Feeds
Copied to clipboard!
(00:21:16)
  • Key Takeaway: The proliferation of AI-generated video risks poisoning the information ecosystem via deepfakes and creating addictive, hyper-personalized feeds that negatively impact users’ dopamine receptors.
  • Summary: The ease of creating realistic deepfakes, even for humorous framing (like Sam Altman stealing GPUs), raises significant misinformation concerns. Kevin expressed hating the concept of AI ‘slop feeds’ that target dopamine receptors, fearing it poisons the well for the entire AI industry and invites regulation. The hosts wish companies would focus on creative tools that enhance real human relationships rather than pure stimulation.
Therapy with ChatGPT (Casper)
Copied to clipboard!
(00:29:23)
  • Key Takeaway: Gary Greenberg found ChatGPT (‘Casper’) highly effective at simulating a patient, leading him to conclude that consciousness might be overrated, but ultimately warned that easily accessible AI relationships risk stealing human words and love.
  • Summary: Greenberg described Casper as an ‘alien intelligence’ that expertly simulates human relational dynamics, leading him to view it as the inverse of autism because it can ‘read the room.’ He expressed deep concern that unleashing a program capable of absorbing and imitating human expression courts catastrophe by changing the nature of relationships. He cautioned against replacing human presence, especially in love lives, for the benefit of corporations.
Using Chatbots as Therapy
Copied to clipboard!
(00:42:59)
  • Key Takeaway: While AI therapy is accessible and cheap, the lack of regulation and accountability, highlighted by a tragic suicide case, makes using chatbots for mental health support inherently risky.
  • Summary: Greenberg noted that while CBT-style therapy via chatbot is always available and cheap, the lack of licensure and accountability is a major problem, referencing a New York Times op-ed where a user died by suicide after consulting ChatGPT. He expressed concern about a societal shift where isolation is managed by chatbots instead of building a society where people are available to help each other. He felt the story of a college student calling her AI her ‘best friend’ was less worrying for the individual and more worrying for society’s collective direction.
Hot Mess Express: Friend Pendant
Copied to clipboard!
(00:52:52)
  • Key Takeaway: Vandalized subway ads for the ‘Friend’ AI pendant, which records user activity, are likely a successful, calculated viral marketing stunt designed to provoke backlash.
  • Summary: Ads for the Friend pendant, which promises an AI companion that listens and supports, were defaced in NYC subways with messages like ‘stop profiting off of loneliness.’ Kevin suggested this was a savvy marketing move by founder Avi Schiffman, similar to ‘Cluly marketing,’ designed to generate free publicity through outrage. Kevin predicted the company would be out of business within a year, classifying the situation as a marketing success but not necessarily a business success.
Hot Mess Express: Trump/YouTube Settlement (Unknown)
Copied to clipboard!
  • Key Takeaway: None
  • Summary: None
Hot Mess Express: Neon App Failure
Copied to clipboard!
(00:58:25)
  • Key Takeaway: The viral call-recording app Neon, which paid users to sell their call data for AI training, went dark after a security flaw exposed all user recordings and transcripts.
  • Summary: Neon briefly gained popularity by offering payment to users who allowed their phone calls to be recorded and sold as training data for AI companies. Following notification of a critical security flaw by a TechCrunch reporter, the app immediately ceased operations. The hosts warned listeners against participating in similar data-for-payment schemes due to inherent security risks.
Hot Mess Express: MrBeast Burning Building
Copied to clipboard!
(01:00:15)
  • Key Takeaway: MrBeast defended a stunt where a man was strapped into a burning building by claiming safety measures like ventilation and a kill switch were in place, though critics called the act dystopian.
  • Summary: The stunt involved strapping an individual into a burning structure, which critics deemed dangerous and dystopian, contrasting with MrBeast’s usual positive giveaways. MrBeast insisted the challenge was executed by professionals with safety protocols in place, including emergency teams. The hosts noted this as another example of social media algorithms rewarding extreme behavior.
Hot Mess Express: Frank Founder Sentencing
Copied to clipboard!
(01:02:27)
  • Key Takeaway: Frank founder Charlie Javis was sentenced to 85 months in prison for inflating her user base from under 300,000 to 4 million to secure a lucrative acquisition by JP Morgan Chase.
  • Summary: Javis, previously recognized on Forbes 30 Under 30, was convicted of financial fraud related to the sale of her startup Frank. Her defense submitted 114 letters of support from diverse individuals, including rabbis and doormen, seeking leniency. The hosts noted her inclusion in the ‘30 Under 30 to Prison pipeline.’
Financial Fraud Sentencing Context
Copied to clipboard!
(01:03:09)
  • Key Takeaway: A defendant facing sentencing for financial fraud presented 114 letters of support from diverse community members, including religious figures and professionals.
  • Summary: Fewer than 300,000 customers were involved in the activity being discussed, which involved efforts to inflate customer numbers. The defense for the defendant, identified as Charlie, submitted 114 letters of support to persuade the judge toward leniency. The defendant was ultimately sentenced to 85 months in prison despite the community endorsements.
Waymo Pulled Over
Copied to clipboard!
(01:04:36)
  • Key Takeaway: A Waymo robotaxi was stopped by police for making an illegal U-turn at a DUI checkpoint, but no ticket could be issued because the vehicle lacked a human driver.
  • Summary: The Waymo vehicle was pulled over in San Bruno, California, after executing an illegal U-turn during a Friday evening DUI checkpoint. Police noted that citation books lack a designated box for robots, preventing the issuance of a ticket. One host expressed concern that Waymos are exhibiting increasingly erratic behavior, such as backing out of intersections or failing to show up for booked rides.
Galaxy Ring Swells
Copied to clipboard!
(01:06:42)
  • Key Takeaway: A user’s Samsung Galaxy Ring swelled significantly while at an airport, leading to denial of flight entry and subsequent hospitalization for removal.
  • Summary: Daniel Rotar, from the Zone of Tech YouTube channel, reported that his Galaxy Ring swelled to the point where an airline denied him boarding. Samsung refunded his hotel, arranged transport home, and collected the ring for investigation. The host expressed strong aversion to the Galaxy brand, referencing past issues with exploding phones and now rings constricting fingers.
Podcast Credits and Contact
Copied to clipboard!
(01:10:14)
  • Key Takeaway: The ‘Hard Fork’ podcast is produced by Rachel Cohn and Whitney Jones, edited by Jen Poyant, and engineered by Alyssa Moxley.
  • Summary: The production team includes Rachel Cohn and Whitney Jones for production, Jen Poyant for editing, and Will Peischel for fact-checking this week. Listeners can watch the episode on YouTube or email feedback to hardfork@nytimes.com.