The rise of AI agents making consumer choices (e.g., shopping) won't upend business models as feared. It will likely follow an evolutionary path similar to how Search Engine Optimization (SEO) adapted to become Generative Engine Optimization (GEO), with value capture mechanisms compounding on existing principles.
Google is in the 'crawl' phase of implementing agentic AI in its main products. This is a deliberate strategic choice, not a technical limitation. The company feels a stewardship responsibility to its massive user base, prioritizing gradual adoption and user comfort over deploying its most advanced capabilities.
What we call an AI 'model' is no longer just a set of weights but an entire system with scaffolding for tool calling, search, and code execution. This external 'harness' indicates future native capabilities, as the model eventually 'eats' the scaffolding and incorporates these functions directly, pushing the innovation frontier outward.
The cultures of major AI labs are extensions of their leaders' personas. DeepMind embodies the scientific rigor of Demis Hassabis ('Nobel Prize scientist'), OpenAI reflects the business acumen of Sam Altman ('world's best businessman'), and Anthropic mirrors the more esoteric nature of Dario Amodei, shaping their respective strategies.
The path to AGI won't be uniform. Instead, we'll see 'jagged superintelligence,' where models achieve superhuman capabilities in specific verticals with high verifiability, such as coding, finance, and scientific research. These specialized peaks of excellence will appear long before a generalized intelligence is achieved.
Large AI labs must serve a vast portfolio of products, preventing them from focusing intensely on any single vertical. This creates a significant opportunity for startups. By concentrating all resources on a specific domain, startups can 'run laps around' even the best-resourced labs, leveraging focus as their primary competitive advantage.
Google is shifting its product integration strategy. Initially, the Gemini model API was the common thread. Now, its 'Antigravity' agent harness is the new standard, embedding action-taking capabilities natively across all products, from Search to the Gemini app.
Google realized it's difficult to build a top-tier coding model without a dedicated product for complex, agentic tasks. Developing Antigravity created the essential internal engine and feedback loop, driving massive token consumption and accelerating model progress. The product was necessary to spin the model improvement flywheel.
The concept of a 'world model' is evolving from action-conditioned video predictors to single, multimodal models like Google's Omni. Omni demonstrates a deep, scalable understanding of the world, shown through nuanced video editing, representing a more practical approach than traditional, computationally expensive architectures.