microsoft

Microsoft's NextLat paper proposes a self-supervised training method where transformers predict their next hidden state instead of just the next token, leading to more compact world models, better planning and reasoning, and up to 3.3x faster generation.

Microsoft has open-sourced Fast Context, a tool designed to accelerate context retrieval in AI applications.

A rumor suggests that major gaming publishers like Sony and Microsoft are planning mass layoffs and studio closures, with Bethesda, BioWare, and others named as potentially affected.

Microsoft open-sourced bitnet.cpp, a 1-bit LLM inference framework that enables running 100B parameter models on local CPUs without GPUs, achieving 6.17x faster inference and 82.2% less energy consumption.

Chevron announces a 20-year power purchase agreement with Microsoft to develop a co-located natural gas power facility in West Texas, delivering approximately 2.67 GW to a Microsoft data center to support AI growth.

The article covers the upcoming expiration of a Microsoft Secure Boot certificate that Linux distributions rely on for booting via shim, and the complexities involved in updating system firmware to accommodate the replacement key.

Microsoft CEO Satya Nadella warns against concentration of AI power in the hands of a few companies, advocating for broader access and social benefit.

A Microsoft study using 43 weeks of data from 16,223 engineers found that GitHub Copilot increases pull request completion by 40.5% when holding development effort constant.

A Microsoft AI researcher built a simple neural network using goats in Age of Empires II to argue that if such a system can be considered conscious, then claims of AI consciousness in chatbots are equally absurd.

A 20-year-old woman taught herself computer science for two years and became an AI engineer at Microsoft, sharing her learning notes and resource repository X-Plore.

Windows 11's new Media Player consumes 3.5 times more RAM and now charges for popular video codecs, marking a controversial update to the built-in app.

This article explains that the Windows loader sets the bottom bit of an HMODULE handle to indicate the DLL was loaded as a data file (LOAD_LIBRARY_AS_DATAFILE), affecting resource lookup and freeing behavior without adding it to the module list.

Microsoft Activation Scripts (MAS) is an open-source tool hosted on GitHub that provides free and permanent activation for Windows and Office products, bypassing Microsoft's paid licensing. Despite Microsoft owning GitHub, the repository has remained untouched for six years with over 178,000 stars.

Microsoft is expanding its AI presence in China by reselling OpenAI models through Azure, navigating local regulations and competition.

A performance comparison shows Microsoft's new Outlook takes 10 seconds to load an email from a notification, while Outlook Classic does it instantly, highlighting ongoing issues with the web-based replacement.

An upcoming deadline requires Windows and Linux users to update Secure Boot keys to protect against UEFI-based bootkits. The expiration of Microsoft-signed certificates on June 24 could leave systems vulnerable if not updated.

The tweet discusses Microsoft's SkillOpt paper, which improved GPT-5.5 accuracy from 41% to 80% without retraining by using a small skills file to guide the agent.

A Microsoft and SJTU research paper introduces FastContext, a dedicated exploration subagent for coding agents that separates repository navigation from task solving, reducing orchestrator token usage by up to 60% and improving resolution rates by 5.5% on SWE-bench benchmarks.

Microsoft introduces SkillOpt, a method that trains an agent's skill documentation like a neural network, using epochs, batches, learning rates, and validation sets for optimization, without modifying model weights. It achieves top results across multiple benchmarks and can be transferred across models and tools.

Microsoft is testing Phi Silica support on Nvidia GPUs, allowing developers to run the small language model locally on Windows devices with RTX 30-series or newer GPUs, though it lacks NPU-only features like prompt compression.