PLUS: OpenAI's agentic AI hire, India's $100B investment, and Japan's AI toilet maker

Happy reading

Anthropic has just upended its own model hierarchy with a new release. The company's new mid-tier Sonnet 4.6 is not only more affordable but is now outperforming its previous flagship model, Opus 4.5, on a range of complex tasks.

The update makes near-flagship AI capabilities significantly more accessible for developers. But what does it signal for the industry when a company's best model is made obsolete by a cheaper alternative in just a few months?

In today’s Next in AI:

  • Anthropic's new Sonnet model outperforms Opus

  • OpenAI's strategic agentic AI hire

  • India's $100B investment in AI infrastructure

  • The toilet maker in the AI hardware supply chain

Opus Performance, Sonnet Price

Next in AI: Anthropic just released Claude Sonnet 4.6, a mid-tier model that delivers a clean upgrade in performance. Early testing shows users often prefer it over the previous flagship model, Opus 4.5, for complex tasks.

Explained:

  • In head-to-head coding tests, developers preferred Sonnet 4.6 over the former top model Opus 4.5 59% of the time, citing better instruction-following and fewer hallucinations. You can review its full capabilities and safety in the official system card.

  • The model sets a new record on the OSWorld benchmark thanks to major computer use upgrades. This allows it to better handle tasks like navigating complex spreadsheets and automating actions across multiple browser tabs.

  • Sonnet 4.6 provides this leap in performance while maintaining the same price as its predecessor—40% cheaper than the current top-tier Opus 4.6 model—and is now the default for all free and pro users.

Why It Matters: This release makes near-flagship AI performance significantly more affordable and accessible for developers and businesses. The rapid pace of improvement signals that today's top-tier capabilities are quickly becoming tomorrow's industry standard.

The Agentic AI Tipping Point

Next in AI: OpenAI has hired Peter Steinberger, the creator of the viral open-source agent OpenClaw. The move signals a major strategic shift across the industry, moving beyond conversational chatbots and toward autonomous AI that can take action on your behalf after he is joining OpenAI to work on bringing agents to everyone.

Explained:

  • OpenClaw exploded in popularity because it moves past text generation to execution, allowing it to browse the web, execute code, and complete multi-step tasks autonomously across a user's entire computer.

  • The agent's power comes with significant security risks, prompting companies like Meta to ban the tool from corporate devices due to its unpredictable nature and the potential for data breaches.

  • This is a major win for OpenAI and a missed opportunity for Anthropic, which originally inspired the project but sent a cease-and-desist over its name, pushing the viral project into the arms of its chief rival.

Why It Matters: This hire confirms the industry's center of gravity is moving from AI that can talk to AI that can do. The key challenge now is bridging the gap between what's possible in open source and what's safe enough to deploy within the enterprise.

India's $100B AI Bet

Next in AI: India’s Adani Group announced a massive $100 billion investment over the next decade to build a network of AI-ready data centers. The initiative aims to establish India as a global leader in AI infrastructure, shifting its role from a consumer to a creator of intelligence.

Explained:

  • The project is projected to create a $250 billion AI ecosystem, stimulating growth in server manufacturing, sovereign cloud platforms, and other supporting industries.

  • Adani plans to expand its data center capacity from 2 GW toward a 5 GW target, supported by strategic partnerships with major tech players like Google and Microsoft.

  • A core part of the strategy involves powering the entire data center network with renewable energy, tackling the immense power demands of AI with a sustainable approach.

Why It Matters: This colossal investment signals India's intent to become a top-tier player in the global AI race, directly challenging the infrastructure dominance of the U.S. and China. By building its AI future on a green foundation, India is also creating a potential blueprint for how to sustainably power the next generation of technology.

The AI Toilet Maker

Next in AI: An activist investor is pushing Japan's famous toilet maker, Toto, to embrace its role as a key player in the AI hardware boom. The company's hidden gem is a highly profitable division that produces critical components for AI memory chips.

Explained:

  • Toto leverages its long history in ceramics to produce electrostatic chucks, advanced ceramic components that hold silicon wafers steady during the manufacturing of NAND memory chips.

  • Investor Palliser Capital believes the division has a five-year competitive 'moat' and is urging Toto to invest more heavily, projecting over 30% revenue growth in the next two years.

  • This reflects a growing trend of non-tech Japanese companies becoming critical to AI supply chains, with the stock surge seen as an unlikely boon for the legacy manufacturer.

Why It Matters: The AI hardware supply chain is revealing surprising and essential players well outside of traditional tech. As demand for specialized components continues to soar, expect more legacy industrial companies to become pivotal to AI's infrastructure.

AI Pulse

A report revealed the U.S. military used Anthropic's Claude AI model during a raid in Venezuela, marking the first known use of a major AI developer's tool in a classified Department of Defense operation.

Alibaba released its Qwen3.5 model series, a new open-weight AI with enhanced agentic capabilities designed to compete with the latest tools from OpenAI and Anthropic.

GitHub added a feature allowing repositories to disable pull requests entirely, a direct response to open-source maintainers being overwhelmed by low-quality, AI-generated code submissions.

SpaceX joined a secret Pentagon competition with its subsidiary xAI to develop voice-command software for coordinating drone swarms across air and sea.

Keep Reading