Why Google Isn't Chasing Claude Code

What It Covers

Google IO 2025 reveals a fragmented AI strategy across Gemini 3.5 Flash, AntiGravity 2.0, Gemini Spark, and Omni, while Gemini's monthly active users surged from 400 million to 900 million. The episode examines whether Google's product sprawl and Demis Hassabis's AGI-first priorities cost them ground against Anthropic and OpenAI's coding-agent momentum.

Key Questions Answered

• •Google's product sprawl problem: Google now fields at least eight distinct AI surfaces — Spark, AntiGravity, AI Studio, Flow, NotebookLM, Gemini CLI, Jules, and Omni — with no clear public guidance on which tool serves which user. Developers and consumers alike report confusion about when to use each product, creating friction that competitors like Anthropic avoid by consolidating around Claude Code and Claude Cowork.
• •Gemini 3.5 Flash token inefficiency: Despite being marketed on speed, Gemini 3.5 Flash uses roughly 3.5 times more output tokens than GPT-5.5 Medium on standardized benchmark tasks, and costs approximately twice as much as Gemini 3.1 Pro on equivalent workloads. Developers evaluating agentic coding tools should run token-cost comparisons on their specific task types before committing to Flash as a cost-saving option.
• •AntiGravity 2.0 architectural shift: AntiGravity 2.0 moves away from a full IDE environment toward a standalone agent-layer product with multi-agent teams, scheduled background tasks, native voice, and MCP integrations — mirroring the structural direction of Codex. Developers building on Google's stack should evaluate whether the new SDK and sub-agent scheduling capabilities now meet minimum parity requirements for production agentic workflows.
• •Omni's editing advantage over generation: Gemini Omni's primary value is not base video generation quality but granular video-to-video editing — changing scene settings, time of day, character outfits, and backgrounds while preserving shot structure. Teams producing video content should test Omni specifically for post-production editing workflows rather than benchmarking it against cinematic generation models like Sora or Seedance.
• •Hassabis world-model strategy vs. RSI path: Demis Hassabis is pursuing AGI through continual learning and world models rather than the recursive self-improvement path OpenAI and Anthropic are accelerating via coding agents. A reported internal faction led by Sergey Brin is pushing Google toward the RSI approach. Organizations building long-term AI infrastructure partnerships should monitor which internal Google strategy prevails, as it will determine model capability trajectories.