Tag
RESOURCE2SKILL is a framework that distills executable agent skills from multimodal resources like tutorial videos, code repositories, articles, and artifacts into a hierarchical SkillWiki, improving agent performance by 11.9 percentage points over no-skill agents.
Microsoft is reportedly training its sales team to negatively compare rival AI products from OpenAI, Anthropic, and Google against its own, emphasizing the full end-to-end system. The move comes after Microsoft dropped exclusivity with OpenAI and is swapping external models for its own in flagship apps.
Hundreds of Bethesda and Zenimax employees rallied at Bethesda HQ to protest sweeping Xbox layoffs, demanding Microsoft return to contract bargaining.
A new 4B parameter AI model trained with Social Reinforcement Learning (SocialRL) outperforms GPT-5 models in negotiation tasks, suggesting that traits making an AI pleasant assistant (agreeableness, transparency) hinder negotiation effectiveness.
A Windows 0-day vulnerability called LegacyHive allows non-admin users to escalate privileges by abusing how Windows loads user class hives, with Microsoft investigating and detection scripts available.
Introducing Microsoft's open-source tool markitdown, which can convert Word, Excel, PDF, PPT and other files into Markdown format, suitable as a side hustle.
Microsoft issued a record 570 security patches for Windows, Office, and other products, attributing the increase to AI-assisted vulnerability discovery.
Microsoft confirmed the existence of a Windows GDID device identifier that cannot be disabled, as documented in an FBI case filing, raising privacy concerns.
Microsoft has released a new embedding model called bitnet-embedding-0.6b on HuggingFace, but the model card is missing.
This article discusses the reverse information paradox in the AI era, pointing out that users are forced to leak proprietary knowledge when using AI services, while service providers gain more information. It calls on companies to protect core intellectual property and take control of the learning loop to address information asymmetry.
Satya Nadella criticizes the double standard in AI model cloning, highlighting ethical concerns in the industry.
Satya Nadella warns that companies using external AI services like OpenAI and Anthropic risk exposing proprietary knowledge, which may be used to compete against them, arguing for self-hosting AI to protect intellectual property.
Beginner-friendly virtual sessions in multiple languages to learn the GitHub Copilot app, an all-in-one control center for AI-driven software development.
A tweet argues that even Microsoft is discussing Sovereign AI and running model weights on local hardware, making a case for local and open-source AI.
Researchers at ESET discovered that Microsoft's Secure Boot has been vulnerable to trivial bypass for 13 years due to unrevoked signed shims, allowing attackers to install persistent firmware malware on both Windows and Linux devices.
Windows 11's July 2026 Patch Tuesday update allows users to pause updates for up to 35 days, with the option to extend the deadline repeatedly, along with security fixes and other improvements.
Microsoft released a record 570 security patches for Windows and other software, including 60 critical flaws and three zero-days, attributing the increase to AI-assisted vulnerability discovery.
Microsoft shares insights from shipping thousands of production AI agents at enterprise scale, covering the engineering challenges of moving from prototype to production, including the agent harness, retrieval-as-a-subagent, agent identity, and rubric-based evaluation loops.
Microsoft is testing a cleaner version of Windows 11's search menu that removes ads and recommended content, focusing on recent searches and local file results.
A field study analyzing the adoption and impact of agentic command-line coding tools (Claude Code and GitHub Copilot CLI) at Microsoft, finding that adoption spreads through social networks, retention correlates with coding activity, and adopters merge 24% more pull requests.