Tech Brew Ride Home

Episode: TikTok Sold (Again) (?)
Date: December 19, 2025
Host: Brian McCullough
Duration: ~15 minutes

Overview

This episode dives into the latest developments regarding TikTok’s US operations sale (again?), lays out a major privacy ruling in Pennsylvania, explores YouTube's crackdown on AI-generated fake trailers, discusses big AI model advances from Meta, previews League of Legends’ next-gen overhaul, and finishes with long-read recommendations on Meta’s AI ambitions, space-based data centers, and the booming market for AI training data.

Key Discussion Points & Insights

TikTok US Sale: It's (Finally?) Happening

[03:05 - 07:45]

• Internal Memo Leak: Axios obtained an internal memo confirming that TikTok has agreed to sell its US unit, expecting to close the deal on January 22, 2026.

  • Oracle, Silver Lake, and Abu Dhabi-based MGX to collectively own about 45% of TikTok’s US entity, with ByteDance retaining approximately 20%.
  • The new entity to be called TikTok USDS Joint Venture LLC.
  • One-third of the company will be held by existing ByteDance investors.

• Security & Independence Measures:

  • “The US joint venture will be responsible for US data protection, algorithm security, content moderation and software assurance.” (Brian quoting the memo, 04:20)
  • The recommendation algorithm will be re-trained on US user data to ensure the content feed is free from outside manipulation.
  • Oracle will act as the "trusted security partner," overseeing compliance with US security terms and moderating authority over content and data.

• Deal Size and Structure:

  • Valuation estimated at $14 billion.
  • Oracle and Silver Lake bring deep entertainment ties:
    • Larry Ellison’s active role in Hollywood buyouts.
    • Silver Lake’s involvement with WME, TKO, and acquisition of EA in partnership with Saudi Arabia’s PIF.

Notable Quote:

“Upon the closing, the US Joint Venture ‘will operate as an independent entity with authority over US data protection, algorithm security, content moderation and software assurance...’”
— Brian McCullough, reading internal memo (05:45)

Pennsylvania Supreme Court: Warrantless Search of Your Google Query History

[07:46 - 11:05]

• Major Privacy Precedent:
  The Pennsylvania Supreme Court rules that police do not need a warrant to access Google search histories.

  • The rationale: It’s “common knowledge” that ISPs and apps collect and may sell user data, thus no expectation of privacy.

• Case Example & Expert Reaction:

  • Police used a “reverse keyword search” to connect a defendant to a crime scene by querying Google for users who searched the victim’s address.
  • Law professor Andrew Ferguson warns:

    “A list of your questions to Google is a direct link to your thinking and we usually try to prevent the government from having access to that information.” (Andrew Ferguson, 10:25)

• Potential National Impact:

  • Though only binding in PA, the expert expects it will encourage similar policing practices nationwide.

Notable Quote:

“The danger of a reverse keyword search is that it allows the police to rummage through the digital questions and queries—and thus by inference, our minds.”
— Andrew Ferguson, quoted by Brian (10:12)

Meta’s Next AI Bets: “Mango” for Video, “Avocado” for Text

[11:06 - 12:39]

• AI Model Pipeline:

  • Meta’s Chief AI Officer Alexander Wang discussed new models at an internal Q&A:
    • Mango (image and video-focused, coming 2026).
    • Avocado (a next-gen language model aimed at excelling in coding and “world models”—AI that learns via vision and environment).

• Competitive AI Race:

  • Meta is vying with OpenAI and Google; recently, Meta released the video gen tool “Vibes” (with Midjourney), just before OpenAI dropped Sora.
  • Google’s “Nano Banana” image generator reportedly boosted Gemini’s user base from 450M to 650M in just months.

YouTube Terminates Big Fake Movie Trailer Channels Over AI Use

[12:40 - 14:45]

• Enforcement Against AI-Generated Content:

  • YouTube bans Screen Culture and KH Studio for making fake movie trailers with AI.
  • Channels had a history of quickly splicing official footage and generative images to create misleading content, sometimes ranking above real trailers.
  • Previously regained monetization by labeling videos “fan trailer” but dropped caveats, prompting the takedown.

• Behind the Scenes in Hollywood:

  • Some studios reportedly asked YouTube to route ad revenue from these AI-heavy trailers to themselves, rather than remove them.

Notable Quote:

“The monster was defeated.”
— Unnamed YouTuber, Deadline via Brian (14:15)

League of Legends “League Next”: Biggest Update in Title’s History

[15:15 - 17:50]

• Overhaul Planned for 2027:

  • Riot Games is reworking League of Legends’ visuals, UI, and gameplay to attract new players.
  • Aims for improved updates, easier onboarding, and a modern new player experience.

• Why Now?

  • Player base erosion, rise of other Riot titles like Valorant, and historical challenges with diversifying revenue.

Notable Quote:

“We’re overhauling the new player experience so that once we’re done, it should be the best time ever to get your friends into League.”
— Paul Bellezza, League of Legends executive producer (17:29)

Long Reads: Meta’s AI, Data Centers in Space, and the Value of AI Training Data

[17:51 - 20:10]

Mark Zuckerberg’s “All-In” Bet on AI at Meta

• Intense internal drama between Zuck and AI leadership:
  • Wang: “...finds Zuckerberg’s micromanagement of the company’s AI work, quote, suffocating.”
  • Debate over Meta’s ability to beat OpenAI, intense pressure to ship products (e.g., ‘Vibes’ rushed to beat OpenAI’s Sora).
  • “When you’re a friend of Zuck, you have more room for error.” (former insider, 19:10)

Data Centers in Space: Harsh Economic Realities

• Detailed look at why orbital data centers are still economically impractical.
  • Building in space costs 3x more per watt; thermal management is a major bottleneck.
  • “The… financial reality is savage.” (Andrew McCallop, quoted 19:40)
  • SpaceX is the only plausible near-term contender due to vertical integration.

AI Training Data: The New Gold Mine

• Companies like Merkore and Cursor are booming, supplying AI labs with custom training data.
  • Merkore hit $500 million annualized revenue at age 22.
  • Most AI companies lose money, but data suppliers are often profitable.
  • Labs now need specialized data, not just scale.

Notable Quotes & Timestamps

• “Did you ever notice the whole TikTok thing was never done?”
  Brian McCullough, 03:08

• “The danger of a reverse keyword search is that it allows the police to rummage through the digital questions and queries—and thus by inference, our minds.”
  Andrew Ferguson (as quoted by Brian), 10:12

• “Upon the closing, the US Joint Venture ‘will operate as an independent entity…’”
  Internal memo, read by Brian, 05:45

• “We’re overhauling the new player experience so that once we’re done, it should be the best time ever to get your friends into League.”
  Paul Bellezza, 17:29

• “Wang has told associates he finds Zuckerberg’s micromanagement of the company’s AI work suffocating.”
  FT report, summarized 18:40

Other Interesting Segments

• Meta’s Vibes vs. OpenAI Sora: Race to release AI video tools first (13:15)
• Hollywood’s Paradoxical Stance: Studios directing ad revenue from fake trailers (14:25)
• Data Center Futurism: Orbital compute as a possible stepping stone for a real space economy (19:50)
• AI Data Market: Surprising profitability trend—data providers are the ‘only AI companies turning a profit’ (20:05)

Episode Tone & Closing

Brian delivers his news hits with a dry, lightly cynical wit—quick to point out US policymaking absurdities, tech hype cycles, and the ironies of enforcement in tech and content. The show maintains a fast pace while balancing reporting with color commentary.

For Next Week:

• Top 10 Tech Stories of 2025 episode on Christmas Day
• History of Toys “R” Us episode Friday