⏱️ AI capability accelerating beyond expectations ⏱️
Doubling every 7 months

A study from research group METR reveals that frontier AI models are improving at a pace far exceeding previous estimates. The complexity of tasks LLMs can handle is doubling approximately every seven months - significantly outpacing Moore's Law. For perspective, Claude 3.7 Sonnet can now manage tasks that take humans about an hour, compared to GPT-4 (2023) handling five-minute tasks and GPT-3.5 (2022) limited to one-minute tasks. If this trend continues, AI capable of tackling four-hour human tasks could arrive within two years, with models potentially handling month-long projects by 2029.

🎬 Pika’s revolutionary video object manipulation 🎬
Edit without disrupting scenes

Pika has unveiled a preview of its next-generation AI video editing tool that enables users to manipulate individual objects and characters within footage while keeping the surrounding scene perfectly intact. This precision technology represents a significant leap beyond current capabilities, allowing for creative modifications without introducing artifacts or distortions that typically plague such edits. The tool opens new creative possibilities for content creators, offering unprecedented control over video elements and potentially transforming post-production workflows.

🔍 Google’s AI-powered Gmail search 🔍
Finding what you actually meant

Google has implemented advanced AI capabilities in Gmail search, dramatically improving the system's ability to find relevant emails even with vague or incomplete search terms. The new functionality understands intent rather than just matching keywords, making email retrieval significantly more intuitive. This enhancement uses LLM technology to interpret search context and user history, enabling Gmail to surface meaningful results despite imprecise queries—effectively bridging the gap between how humans naturally think about their emails and how traditional search systems operate.

INTERESTING NEWS

🎨 Adobe’s agentic AI showcase 🎨

At the recent Adobe Summit, the company demonstrated seven experimental AI agents that could revolutionize creative workflows. Standout projects included Project Slide Wow for instantly converting datasets into polished PowerPoint presentations, Project Get Savvy which coordinates multiple specialized agents to create entire ad campaigns from scratch, and Project Panorama which generates interactive diagrams visualizing user journeys through websites. These experiments highlight Adobe's vision for how agentic AI could fundamentally transform content creation across its product ecosystem.

🧠 Microsoft + InAIT’s brain simulation breakthrough 🧠 

Microsoft has announced a partnership with InAIT to develop revolutionary AI systems through advanced brain simulation technology. The collaboration, led by Human Brain Project founder Henry Markram, aims to create digital brain models that more accurately reflect human neural processing. This approach differs significantly from current deep learning methods and could potentially lead to AI systems with more human-like reasoning capabilities and better alignment with human values. The partnership represents a significant investment in alternative approaches to artificial intelligence development.