Google Seizes the Consumer AI Crown — A New Era with Gemini Spark and Antigravity

Google Is Taking the Consumer AI Throne

The Economist reported that "Google is dethroning OpenAI as the king of consumer AI," drawing attention across the industry. The string of announcements at Google I/O 2026 — especially the arrival of the always-on AI agent Gemini Spark and the new Antigravity platform — makes it clear that the axis of AI competition has shifted decisively from "who is the smartest" to "who is the most usable."

Gemini Spark: An AI Agent That Works Around the Clock

Announced at Google I/O 2026, Gemini Spark is an "always-on personal AI agent" that keeps running in the cloud even when your PC or phone is powered off. Unlike conventional chatbots, once you give it instructions, it continues processing tasks autonomously in the background. It leverages Gemini 3.5 Flash and the new Antigravity platform, and notably operates without consuming device resources.

In addition, Antigravity's lightweight terminal UI, the Antigravity CLI (AGY CLI), has been released, letting developers tap into AI capabilities directly from the command line. Reviews comparing its code generation capabilities against the existing Gemini CLI are already underway.

A New Phase in the AI Model Race

In spring 2026, the AI development race has entered a new stage. OpenAI is rolling out GPT-5.5 for enterprises through a partnership with Databricks, optimized for enterprise workflow agents. Meanwhile, Anthropic has been steadily gaining support, with some observing that "the industry's star is shifting to Anthropic." Every player has now entered a phase of differentiating not just on model intelligence, but on ecosystem usability and practicality.

Notably, there are also reports that Google quietly downgraded the contents of its AI Pro plan — a hint of the twists and turns in its product strategy.

Cohere Releases 218B Open-Source Model "Command A+"

Cohere and Cohere Labs have released Command A+ under the Apache 2.0 license. It uses a Mixture of Experts (MoE) architecture with 25B active parameters and 218B total parameters, and supports a 128K context length. It is optimized for agentic tasks, multilingual processing, and reasoning-heavy workloads, and also supports image input (vision).

The W4A4 quantized version is recommended, with benchmark quality degradation described as nearly negligible. The weights are available on Hugging Face, and the model can also be tried out in a Space.

Fujitsu Cuts Personnel Transfer Planning Work by 98% with AI — Deployed at TRUSCO NAKAYAMA

Fujitsu has deployed an AI-powered system for drafting personnel transfer plans at TRUSCO NAKAYAMA. By consolidating scattered HR data and automatically computing the optimal plan from an enormous number of assignment patterns, the company reportedly reduced the effort of drafting transfer plans by approximately 98%. The case is drawing attention as a sign that practical AI adoption in the enterprise is steadily advancing.

WordPress 7.0 Ships with a Built-in AI Client

It has been announced that WordPress 7.0 will include an AI Client feature. Introduced on the official development blog as "Introducing the AI Client in WordPress 7.0," the direct integration of AI capabilities into the CMS powering roughly 40% of the world's websites is significant. Notably, it is built in as a core feature rather than a plugin.