All articles

Five universities with active AI research, as measured by the AI Index. We examine how to interpret indicators when comparing universities, drawing on the number of notable AI models and research structures at Stanford, CMU, Tsinghua University, UC Berkeley and MIT.

OpenAI’s Rosalind Academic Plan is for research labs at U.S. universities. This article outlines who is eligible and the terms for getting up to 30 seats free for up to 12 months, the role of GPT-Rosalind, a model specialized for the life sciences, and the program’s review and safety measures.

NVIDIA, IIT Madras, and AI4Bharat have announced plans to release around 2.8 trillion tokens of training data for 22 Indian languages, along with the generation pipeline. The release stands out not only for its data volume, but also for opening up the process from translation and quality evaluation to correction.

Microsoft researchers tracked tens of thousands of in-house engineers for four months and analyzed their use of Claude Code and Copilot CLI. Adoption was linked to colleagues’ and managers’ use of the tools, while sustained use was associated with developers’ day-to-day coding volume. Those who used the tools during their first week had an average of 24% more merged pull requests.

Google has revamped how links appear in AI Mode and AI Overviews. It is increasing the chances that links will appear at the top of responses and is also testing prominent placement for official websites and links to creator profiles. Here’s a breakdown of what’s changing—and what isn’t.

Meta has announced Muse Spark Open, an open-weight model with 816 billion parameters and a 1-million-token context window. It is designed to run on a Mac, though it reportedly requires at least 1TB of memory. We look at what the gap between the model’s capabilities and its hardware requirements means for the use of open-weight AI.

Anthropic has shown investors revenue of $16.8 billion for the third quarter of 2026. That is up about 45% from the previous quarter, though growth has slowed. Ahead of its planned November IPO, we examine the outlook for its valuation through the lenses of growth, profits, distillation countermeasures and the competitive landscape.

Google is updating AI Mode and AI Overviews. The changes include showing links in answers higher more often, and adding guidance to official pages, news carousels, and pathways to creators’ profiles. Here’s a breakdown of confirmed changes and measures still being tested.

Bun, the JavaScript runtime developed by Anthropic, has released version 1.4.3, whose built-in type checker is said to be more than three times faster than TypeScript 7. Here’s a look at its 100% conformance test pass rate, lower memory use, and other updates.

Does giving an AI assistant its entire conversation history make it smarter? In a small-scale test using a fictional character, retrieving and supplying only relevant details achieved both a higher success rate and lower cost than providing the full history.