AlphaZero Creator Bets $1.1B on AI That Teaches Itself

From Larry Bruce: "One of the most closely watched researchers in AI just struck out on his own — and the industry responded with a $1.1 billion vote of confidence. For developers and founders tracking where foundational AI research is headed, this is a story worth watching closely." — Larry Bruce, Editor, BDCbox

The Recap: Ineffable Intelligence, a British AI lab founded by DeepMind's David Silver, has raised $1.1 billion at a $5.1 billion valuation to build an AI superlearner that learns entirely through reinforcement learning — with zero reliance on human-generated training data.

Unpacked:

• David Silver spent over a decade leading reinforcement learning research at DeepMind, where he created AlphaZero — the AI that mastered chess, Go, and other games by playing against itself, never once learning from human examples.
• Ineffable Intelligence's mission is to build a superlearner that takes this same self-teaching approach and applies it across all domains of knowledge — a significant shift from how most AI today relies on massive libraries of human-generated content to learn.
• The $1.1B round, led by Sequoia and Lightspeed, values the company at $5.1 billion and attracted backing from NVIDIA, Google, and the UK's Sovereign AI fund — a signal that both tech giants and governments are taking this data-free learning approach seriously.

Bottom line: David Silver's track record with AlphaZero gives Ineffable Intelligence a credible foundation for one of the boldest bets in AI right now. If this self-teaching approach scales beyond games, it could fundamentally change how AI systems are built — and what they can learn.

China Blocks Meta's $2B Manus Deal — and Warns the AI World

From Larry Bruce:
"China's move to block Meta's Manus acquisition marks a turning point for how geopolitics intersect with AI dealmaking. For founders and professionals building in the AI space, this is a story worth watching closely." — Larry Bruce, BDCbox

The Recap: China's top economic regulator, the NDRC, has ordered Meta to fully unwind its $2 billion acquisition of Manus — an agentic AI startup founded by Chinese engineers — citing national security concerns and dealing a significant blow to Meta's AI agents strategy.

Unpacked:

• Manus builds AI agents that handle complex, real-world tasks autonomously, and Meta had been actively weaving it into both Meta AI and Ads Manager before the ruling hit.
• China's NDRC offered no explanation for the decision, but the move signals that "Singapore-washing" — the practice of Chinese founders rebasing their companies abroad to avoid scrutiny — is no longer a viable workaround.
• The fallout extends beyond the deal itself: over 100 Manus employees had already relocated to Meta's Singapore offices, and the startup's founders reportedly face exit bans preventing them from leaving mainland China.

Bottom line: Meta loses a key piece of its AI agents strategy, and every cross-border US-China AI deal now carries a new layer of geopolitical risk. Founders and investors eyeing similar deals need to rethink how they structure these partnerships from day one.

The Shortlist

OpenAI could be building a ChatGPT smartphone in collaboration with MediaTek, Qualcomm, and Luxshare — a device where AI agents replace traditional apps entirely, giving OpenAI unrestricted system-level access to user context without Apple's or Google's gatekeeping. Analyst Ming-Chi Kuo says specs and suppliers could be finalized by early 2027, with mass production targeted for 2028.

Google faces a potential EU mandate to crack open Android's AI layer to competing assistants — the European Commission has proposed forcing system-wide access for third-party AI services like ChatGPT and Grok that are currently locked out of Gemini-exclusive features, with a final ruling expected by July 27.

Gemini wants to get proactive — code discovered in the latest Google app beta reveals a new 'Proactive Assistance' feature that accesses Gmail, Calendar, and incoming notifications to surface personalized suggestions before you even ask, with all processing handled on-device in an encrypted environment without feeding Google's AI training data.

Skye secured $3.58M in pre-seed backing from a16z, True Ventures, and SV Angel for its AI-powered iPhone 'agentic homescreen' — an iOS widget-based interface from Signull Labs that proactively surfaces health insights, drafts email replies, flags suspicious charges, and delivers location-aware recommendations, all before launch and already drawing tens of thousands of waitlist signups.