Gemini (Google DeepMind flagship multimodal AI, 2023)

Gemini is the flagship product-instance of Google DeepMind — Google's natively multimodal foundation model series and its defensive-flywheel response to ChatGPT/OpenAI. The instance arc: launched as "Bard" March 21 2023 (PaLM 2 chatbot; reactive defensive posture — ChatGPT had launched November 2022 and Google declared a "code red" internal response in late 2022); rebranded to Gemini February 8 2024. Gemini 1.0 (Ultra/Pro/Nano tiers) December 6 2023 — first natively multimodal foundation model (trained end-to-end on text, images, audio, video, code simultaneously rather than retrofitted multimodality). Gemini 1.5 February 2024: 1-million-token context window (experimental); landmark long-context capability. Gemini 2.0 December 11 2024: agentic posture (Project Astra = universal AI agent demo at I/O May 2024; Project Mariner = browser agent). Gemini 2.5 Pro March 2025 (state-of-the-art benchmark performance). Gemini 3.0 late 2025 (reported; [EXTRAP]). Institutional context: Google DeepMind formed April 20 2023 by merger of Google Brain + DeepMind under Demis Hassabis as CEO — Gemini is the first product series from the merged org. Flywheel mechanic (motor=flywheel): Google's ~2B+ Search users + 3B Android users + 3B+ Workspace users generate multimodal query data → Gemini trains on usage signal → model improves → integrated into all Google product surfaces → more multimodal usage → next training cycle. Defensive posture vs. ChatGPT: Google held the OPP for information retrieval (Search) and feared LLM-native search eroding the core advertising revenue engine. AI Overviews: launched May 2024 at Google I/O — AI-generated summaries at top of Search results page. Controversial "glue on pizza" / "eat rocks for minerals" hallucination episodes within first week of launch (May 2024); forced rapid quality-gating revisions. This is the canonical ontological-doubt signal: Gemini-powered AI Overviews at Google Search scale (~8.5B queries/day) creates mass-scale hallucination-as-infrastructure, creating need for LM ontological-doubt-infrastructure machines that help users distinguish trustworthy from AI-confabulated information. Gemini Advanced: $20/month subscription (launched Feb 8 2024 alongside rebrand). Workspace integration: Gemini for Workspace in Gmail, Docs, Sheets, Meet ($20-30/seat/month add-on). Android integration: on-device Nano models; Pixel 8+ hardware AI features. Substrate note: Substrate.institutional NOT in v0.1 enum; closest available is `social` (for Google's institutional position and user-base social substrate); flagged as [STUB-substrate-enum-gap] pending v0.2 enum extension. Sibling competitive context: machine:chatgpt-instance-2022 (OpenAI; primary competitor; GPT-4 launched March 2023, two weeks before Bard); machine:claude-instance-2024 (Anthropic; competitor for enterprise/API market). Both are intra-DM couplings (C1 verified). V0.2 GAP: (1) Substrate.institutional missing from enum — Google's institutional gravity (search monopoly, Android OEM relationships, Workspace enterprise lock-in) is a load-bearing substrate for Gemini that the current [semiotic, cognitive, social] bundle undersells; (2) commodity enum gap for AI-overview-impressions, tokens-processed — forces commodity: null + [STUB] on key throughputs; (3) motor=flywheel correctly captures the Google defensive-flywheel but the asymmetry between Gemini-the-product and DeepMind-the-lab (motor=push) creates a class/instance motor divergence that the schema handles but warrants note.