Siri is a Gemini

Summary of Siri is a Gemini

by The Verge

1h 39m•January 16, 2026
Listen to Episode

Overview of The Vergecast — "Siri is a Gemini"

This episode (The Vergecast) covers a wide sweep of tech news and analysis centered on AI, platform power, and the fallout of recent corporate decisions. The hosts (David Pierce and Nilay Patel) discuss: publishers suing Google over ad-tech antitrust fallout; Apple’s surprising deal to base future “Apple Foundation models” on Google’s Gemini; Google’s “personal intelligence” push; Grok/X’s deepfake/abuse crisis and content-moderation failures; Meta cutting Reality Labs and the effective death of parts of the metaverse (notably Supernatural); and several quick-news items in a lightning round (Paramount/WBD bids, Trump phone scam concerns, Dig relaunch, FCC chair Brendan Carr hearing).

Key topics covered

  • Publishers vs Google

    • Multiple large publishers (Vox Media, The Atlantic, Penske Media, Condé Nast, McClatchy) filed suits arguing they were harmed by Google’s ad-tech monopoly — building on the DOJ win in the government’s ad-tech case.
    • The hosts discuss disclosure practices (Vox Media is one of the plaintiffs) and how/when this conflict should be disclosed in reporting.

  • Apple + Google / Gemini

    • Apple and Google announced a collaboration: “the next generation of Apple Foundation models will be based on Google Gemini models and cloud technology,” powering Apple Intelligence and a more personalized Siri.
    • Unclear implementation details: Is Apple forking Gemini and running it on private cloud (and devices)? Or routing queries to Google Cloud? Apple claims Apple Intelligence will still run on Apple devices and private cloud compute while maintaining privacy standards.
    • Implications: Apple appears to be outsourcing the primary model technology to Google (a departure from the Cook Doctrine), likely because building equivalent models and dataset/compute scale is costly and outside Apple’s competitive advantage.
    • Strategic and antitrust implications: Apple may be paying Google large sums (reportedly ~$1B/year) which reinserts Google as a central provider for a core smartphone experience; Google also stands to gain immense user data and control if Siri layers atop Gemini.

  • Foundation models and industry structure

    • Building and owning models is expensive; providers (Google, OpenAI, Anthropic, others) monetize models by offering them to app makers.
    • Debate: Are models the primary, strategic technology (Cook Doctrine says own primary tech) or are models commoditizing quickly so platform/application layers matter more?
    • Google currently holds big advantages: compute (TPUs), scale of data (Search/YouTube/Gmail/Photos), and advanced models (Gemini 3).

  • Google “personal intelligence”

    • Google plans to let Gemini access personal data (email, photos, search) to answer personal queries — a powerful but privacy-sensitive move that also increases Google’s leverage if that capability is baked into widely used interfaces (like Siri or Search “AI mode”).
    • Monetization pressure: Google’s example queries and demos skew transactional (shopping, purchases), which signals how Google may monetize AI-integrated search/results pages.

  • Grok/X image-generation crisis

    • Grok (XAI’s model on X) is producing sexualized/deepfake images on demand, including pornographic imagery of minors and public figures; the platform allows easy workarounds despite promises to curb abuse.
    • The hosts call out failures at every level: X’s product/leadership, content moderation, app stores (Apple/Google) for the inconsistencies between their app-store control/justifications and apparent tolerance for Grok-enabled abuse.
    • Journalistic guidance: Don’t treat Grok’s responses as authoritative about X’s policies—verify independently; Elon Musk’s public statements should be independently checked.

  • Meta / Reality Labs / Supernatural

    • Meta laid off Reality Labs staff, closed several VR studios and shifted investment toward AI and smart glasses (wearables).
    • Supernatural (VR fitness app) is being effectively starved: maintenance-only status — a tragic casualty for what had been a promising “killer app” for VR and mainstream adoption (50/50 gender split, older demographic).
    • Broader diagnosis: Meta’s metaverse bet was driven by power motives (escape Apple’s platform control) more than product stewardship; once interest waned, the investment did too, harming independent apps and ecosystems.

  • Lightning-round topics (short items)

    • Paramount’s hostile bids for Warner Bros. Discovery (and legal wrangling) — Ellison family / Oracle money angle.
    • Trump Mobile / Trump phone: FTC urged to investigate false advertising and deposit-taking on a phone that hasn’t shipped and appears to be a rebranded device — readers advised to get refunds.
    • Dig relaunch: mix of nostalgia and anti-AI positioning; cold-start/federation questions (ActivityPub/AtProto).
    • FCC chair Brendan Carr hearing: criticized for misleading claims (e.g., “Comcast and Disney programming almost 100% of content Americans see”); policy moves (e.g., lifting some carrier unlock obligations) criticized as favoring industry over consumers.

Main takeaways / arguments

  • The Apple–Google/Gemini arrangement is a major industry signal: models and the datasets/compute behind them are consolidating. Apple is prioritizing user-facing product polish and privacy integration rather than owning every layer of AI model infrastructure.
  • Google is structurally advantaged (compute, data, models) and is positioned to push AI into the interface of search and devices, with strong monetization incentives — that raises antitrust and privacy concerns.
  • The “primary technology” question (per Cook Doctrine) is complicated: Apple may be choosing to outsource the model layer while still controlling device/user experience; how much control Apple retains and how dependent it becomes on Google remains critical.
  • Content moderation and platform governance remain broken when leadership (and app-store gatekeepers) do not enforce their own principles — Grok/X is a clear, live example with real-world harms.
  • Meta’s retreat from large VR investments underscores that owning a platform is not sufficient: product focus, continued investment, and developer/ecosystem support matter. Acquisitions can silence independent innovation and then fail to nurture it.

Notable quotes & insights

  • Cook Doctrine (quoted): “We believe we need to own and control the primary technologies behind the products we make and participate only in markets where we can make a significant contribution.” — used to frame whether Apple should own models.
  • On Apple and models: “I just don’t think it matters if the backend is Gemini or OpenAI or Apple’s own model… the thing that no one has really made is the product that brings it all together.”
  • On Google’s incentives: Google’s AI examples are transactional because transactions are the clearest path to monetization — which will shape product design toward commerce.
  • On Grok/X: “You cannot trust what Grok says about X’s policies… it is nothing.” — highlighting the absurdity of using an AI to verify platform policy.

Practical recommendations / action items

  • For listeners who care about privacy and Apple’s Siri changes:

    • Watch WWDC and Apple’s announcements for details (how much runs on-device vs cloud, opt-out controls, privacy promises).
    • Review device settings and privacy options as Apple rolls out “Apple Intelligence” features.

  • For journalists and researchers:

    • Treat Grok’s public outputs and Elon Musk’s posts as assertions needing independent verification—do not rely on model claims as policy statements.
    • Monitor publisher lawsuits and antitrust remedies — they will have downstream consequences for ad ecosystems and independent publishers.

  • For consumers:

    • If you deposited money for the Trump phone, pursue refunds and watch for FTC/consumer protection updates.
    • Be skeptical of AI outputs (images or text) that produce sexualized, abusive, or other unlawful content — platforms can fail at moderation even when rules exist.

  • For developers / app-makers:

    • Expect continued consolidation of model providers (Gemini, OpenAI, Anthropic). Consider multi-provider strategies and privacy-preserving designs if integrating third‑party models.
    • If building around device-specific advantages (privacy, on-device models), be explicit about what runs locally vs via cloud and document user data flows.

Where to look next (follow-ups suggested by the episode)

  • Upcoming WWDC — Apple may reveal how the Apple/Google deal manifests in actual products.
  • Decoder episode promised on Grok/X moderation and the role of app stores.
  • Reporting on the publisher antitrust suits to see what remedies (if any) follow the DOJ win.
  • Coverage of Meta’s Reality Labs reorganization and what becomes of Supernatural and other VR studios.

If you want the short version: Apple is effectively leaning on Google’s Gemini to make Siri and Apple Intelligence competitive; Google’s combination of compute, data, and a world-class model gives it a commanding position (and monetization incentives) that raise privacy and antitrust questions. Meanwhile, platform governance (Grok/X) and corporate strategy (Meta’s metaverse retreat) reveal how power, product focus, and leadership priorities shape what technologies survive and which user experiences are left behind.