📋 Quick Takeaways

• OpenAI officially introduces advertising to its chatbot interface, marking a pivotal shift in its revenue strategy beyond subscriptions.
• Meta CTO Andrew Bosworth confirms the internal delivery of new “Avocado” (text) and “Mango” (media) models from their Superintelligence Labs.
• Google DeepMind CEO Demis Hassabis tells Davos that Chinese AI firms are trailing US frontier models by approximately six months.
• Microsoft CEO Satya Nadella identifies “energy costs” as the primary deciding factor in the global AI race.
• Nvidia invests $150M in Baseten as part of a $300M round, valuing the inference startup at $5 billion.
• ServiceNow and Schneider Electric double down on “Agentic AI” to automate complex enterprise and energy workflows.

💰 Business Models & Frontier Models

OpenAI Begins Offering Chatbot Ads to Advertisers

Revenue Pivot: In a report surfacing Tuesday, The Information confirmed that OpenAI has officially begun offering “chatbot ads” to dozens of advertisers. This marks a historic shift in OpenAI’s business model, moving from a purely subscription/API focus to an ad-supported ecosystem directly within its conversational interfaces. Industry Impact: This move challenges traditional search advertising dominance and introduces a new battleground for commercial attention within AI responses. Source: Radical Data Science

📋 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

• OpenAI declares internal “Code Red”, reportedly rushing the release of GPT-5.2 to counter Google’s latest moves.
• Google’s Gemini 3 “Deep Think” officially rolls out to Ultra subscribers, introducing “System 2” reasoning capabilities.
• Runway unveils Gen 4.5 (“The Talkies”), setting a new benchmark for lip-sync fidelity and Hollywood-grade video generation.
• Meta acquires AI-wearable startup Limitless, signaling a major push into consumer AI hardware hardware.
• Raythink cracks the Top 10 in the 2025 Security 50 rankings, driven by breakthrough AI thermal imaging tech.

🚨 Frontier Model Wars

OpenAI Declares “Code Red” & Rushes GPT-5.2 Launch

Strategic Panic: Internal reports surfacing this weekend indicate OpenAI has initiated a “Code Red” protocol. The company is reportedly accelerating the timeline for GPT-5.2, aiming to reclaim industry dominance before the end of 2025.

📋 Quick Takeaways

• Google unveils “Ironwood”, a massive AI hypercomputer connecting 9,216 chips to revolutionize inference speed.
• ChatGPT launches a Personal Shopper Agent, autonomously creating buyer guides and disrupting e-commerce search.
• Nvidia releases Orchestrator-8B, a highly efficient model that beats larger competitors in tool-use tasks.
• Microsoft introduces Fara-7B, bringing agentic intelligence to local consumer hardware with privacy focus.
• xAI builds a solar farm for “Colossus”, directly addressing the critical energy bottleneck for AI data centers.
• Meta quietly releases SAM 3, setting a new standard for zero-shot object segmentation in computer vision.

🚀 AI Infrastructure & Hardware

Google Unveils “Ironwood” AI Hypercomputer for Real-Time Inference

Major Breakthrough: Google [finance:Alphabet Inc.] announced the launch of Ironwood, a new supercomputing architecture that links 9,216 chips into a single “superpod.”