Image via Manus
Friday, October 17, 2025
Claude Skills just changed the game for AI agents
We're watching a serious speed upgrade cycle unfold: Manus just dropped 1.5 with 4x faster web dev capabilities, while Apple's M5 chip quadruples AI performance with Neural Accelerators (yikes, the hardware arms race is real). Meanwhile, Simon Willison is making waves suggesting Claude Skills might be a cleaner play than MCP for expanding AI capabilities, Google shipped DeepSomatic for detecting tumor mutations with real precision, and Salesforce is going bold with a $15 billion San Francisco bet on their new Agentforce 360 platform. The question everyone should be asking: are you building with these faster tools or just watching them get faster?
Image via Manus
Top Stories
Manus
Manus 1.5 achieves 4x faster execution and 15% quality improvement through architectural redesign, enabling autonomous full-stack web application development and complex multi-step workflows—marking significant progress in practical AI agent capabilities.
Simon Willison
Claude Skills represent a simpler, more practical alternative to MCP for extending AI model capabilities by combining Markdown instructions with executable scripts in a coding environment. Their token efficiency and ease of implementation could spark widespread adoption across multiple AI platforms.
Google Research
Google's DeepSomatic uses machine learning to identify cancer mutations in tumors more accurately than current methods, with results published in Nature Biotechnology and tools released openly to advance precision medicine and accelerate cancer research.
Salesforce's $15 billion San Francisco investment and Agentforce 360 platform launch signal the company's commitment to enterprise AI adoption and staying competitive as AI becomes central to business software.
WIRED
Apple's M5 chip delivers 4x GPU compute performance gains for AI through embedded Neural Accelerators, marking a significant architectural shift toward on-device AI execution across MacBook Pro, iPad Pro, and Vision Pro.
Keep Reading
Industry Voices
Yi Tay
Research Scientist at Google DeepMind
Deep technical insights on transformer architectures and scaling laws from someone who co-authored PaLM and works at the frontier of LLM research.
Fei-Fei Li
World Labs
Building spatial intelligence AI at World Labs after pioneering ImageNet, bridging computer vision's past with 3D reasoning's future.
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.