📋 Quick Takeaways

• DeepSeek publishes “striking breakthrough” mHC paper introducing Manifold-Constrained Hyper-Connections for more efficient AI model training
• OpenAI developing new audio model for Q1 2026 release with natural speech and real-time conversation capabilities
• Baidu’s AI chip unit Kunlunxin files Hong Kong IPO valued at $3 billion amid China’s semiconductor self-sufficiency push
• MCP becomes industry standard as OpenAI, Microsoft, and Google adopt Anthropic’s agentic AI protocol
• Tesla reports Q4 miss, loses #1 EV maker title to BYD after second consecutive year of declining deliveries
• Google tests Nano Banana 2 Flash image generation model as part of Gemini’s fast model lineup
• Microsoft CEO Nadella doubles down on AI strategy across all products and services for 2026

🔬 AI Research Breakthroughs

DeepSeek Publishes “Striking Breakthrough” mHC Architecture Paper

Major Research Development: Chinese AI startup DeepSeek kicked off 2026 with a groundbreaking research paper introducing Manifold-Constrained Hyper-Connections (mHC), a new approach to training AI models that could significantly reduce computational costs while improving scalability.

📋 Quick Takeaways

• Meta acquires AI agent startup Manus for over $2 billion to boost advanced AI capabilities.
• Nvidia finalizes $5 billion investment in Intel shares, stabilizing the US chip ecosystem.
• SoftBank agrees to buy DigitalBridge for $4 billion, aggressively expanding AI data center holdings.
• Octopus Energy spins off Kraken, its AI energy platform, at an $8.65 billion valuation.
• MIT & Recursion release Boltz-2, an AI model predicting protein structures in seconds.
• Google rolls out Gemini 3 Flash globally, integrating native audio and complex workflow handling.

💰 Major Acquisitions & Corporate Moves

Meta Acquires AI Agent Startup Manus for $2 Billion+

Strategic Expansion: On December 30, Meta agreed to acquire the AI agent startup Manus for a reported $2 billion+. This acquisition targets the enhancement of Meta’s advanced AI agent capabilities and “reasoning” features within its ecosystem.

Quick takeaways (ranked by impact)

• 1) Synthetic media enters high-stakes geopolitics: An AI-generated hostage video directly addresses President Donald Trump, highlighting how deepfakes can shape diplomatic pressure and public opinion. Source: CNN [web:26]
• 2) New hyperscale model milestone: SK Telecom unveils A.X K1, described as South Korea’s first 500B-scale hyperscale AI model, signaling intensified competition in frontier-scale language models. Source: Macau Business [web:36]
• 3) AI moves deeper into family life: Reporting describes Chinese parents using AI tools to monitor homework and children’s online behavior, raising privacy and autonomy concerns. Source: ABC News (Australia) [web:35]
• 4) AI accelerates bioacoustics research: Scientists apply AI techniques to decode wolf howls, pointing to broader real-world gains from machine learning in ecology and conservation. Source: PBS [web:20]

1) AI + Geopolitics: Deepfake pressure tactics

CNN reports an AI-generated video of a Gaza hostage addressing President Donald Trump, illustrating how synthetic media can be used as a political persuasion tool in real-time crises. [web:26]
The case underscores a growing trust-and-verification gap: audiences may struggle to distinguish authentic footage from AI-generated messaging during fast-moving events. [web:26]

📋 Quick Takeaways

• Startup Poetiq scores 54% on ARC-AGI-2, beating Google’s Gemini 3 Deep Think model at half the computational cost.
• Apple prepares to open underlying AI technology to third-party developers, signaling a major ecosystem expansion.
• Seagate announces 55TB+ HAMR hard drives for 2026, addressing the critical storage shortage for AI training data.
• Cloudflare reports AI bot traffic now rivals human traffic in volume, marking a fundamental shift in internet usage patterns.
• OpenAI’s “GPT Image 1.5” clashes with Google’s “Nano Banana Pro”, with the former claiming 4x speed advantages.
• ChatGPT users get new “Warmth” and “Enthusiasm” sliders to fine-tune the model’s personality.

🧠 Frontier AI Models & Research

Poetiq Cracks Major Reasoning Benchmark, Undercuts Google

Efficiency Breakthrough: AI startup Poetiq has achieved a score of 54% on the ARC-AGI-2 benchmark, a rigorous test for abstract reasoning capabilities.

📋 Quick Takeaways

• President Trump greenlights Nvidia H200 chip exports to China, conditional on a 25% government cut on sales
• Unconventional AI secures massive $475M seed round at $4.5B valuation to build new AI hardware
• NeurIPS 2025 highlights include breakthroughs in “Artificial Hiveminds” and stable LLM attention mechanisms
• Google Cloud CEO unveils 3-part energy strategy to tackle data center electricity bottlenecks
• Instacart launches “Agentic Commerce” pilot inside ChatGPT for autonomous meal planning and shopping
• Intuit AI Chief warns CEOs that 2026 must be the year of AI-driven revenue growth, not just cost-cutting

🌏 Global Policy & Trade

Trump Approves Nvidia H200 Sales to China with 25% “US Cut”

Major Policy Shift: In a significant reversal of strict export controls, President Donald Trump announced on December 8 that Nvidia will be permitted to sell its advanced H200 AI chips to China.

📋 Quick Takeaways

• Jeff Bezos returns as co-CEO of Project Prometheus, $6.2B AI manufacturing startup targeting computers, automotive, and aerospace
• Tether invests $1.16 billion in Neura Robotics, aiming for “iPhone moment” in humanoid robotics with 5M units by 2030
• Fei-Fei Li’s World Labs launches Marble, groundbreaking 3D world model creating immersive environments from text, images, or video
• Microsoft unveils Fairwater 2GW AI superfactory, world’s most powerful AI datacenter with hundreds of thousands of NVIDIA GB200/GB300 chips
• Japanese AI ecosystem surges with Sakana AI ($2.6B valuation) and Turing ($388M) securing major funding rounds
• Mind captioning breakthrough enables AI to translate brain activity into descriptive text using fMRI and language models
• GMI Cloud announces $500M Taiwan datacenter with NVIDIA chips, leveraging semiconductor supply chain advantages
• RIKEN researchers create 100-billion-star Milky Way simulation using AI and supercomputing, 100x faster than previous methods
• Nebura launches N Box platform, accelerating enterprise access to customized AI computing infrastructure

🚀 Major AI Startup Launches & Investments

Jeff Bezos Returns as Co-CEO of $6.2 Billion AI Manufacturing Startup

Historic Move: Amazon founder Jeff Bezos announced on November 17 his return to active CEO duties for the first time since leaving Amazon in 2021, co-leading Project Prometheus, an AI startup focused on revolutionizing manufacturing across computing, automotive, and aerospace industries with $6.2 billion in funding—one of the largest early-stage rounds in startup history.

📋 Quick Takeaways

• Google Quantum AI releases five-stage framework to guide quantum computing from theory to practical real-world applications, calling for industry-wide collaboration
• Microsoft launches world’s first AI superfactory connecting Wisconsin and Atlanta datacenters with dedicated network, reducing AI training time from months to weeks
• AlphaProof achieves silver medal level at International Mathematical Olympiad using reinforcement learning and formal proof verification in Lean
• Human-aligned vision AI breakthrough published in Nature shows models aligned with human perception are more robust and reliable
• AlphaEvolve discovers breakthrough algorithms for matrix multiplication and chip design, achieving 23% speedup in Gemini training

🔬 Quantum Computing & AI Convergence

Google Quantum AI Unveils Five-Stage Roadmap for Practical Applications

Paradigm Shift in Quantum Computing: Google’s Quantum AI team released on November 13 a comprehensive five-stage framework to guide the entire quantum computing industry from theoretical algorithms toward verified, real-world utility.