Global AI Weekly
Issue number: 129 | Tuesday, December 23, 2025
Highlights
Graphite is joining Cursor
Graphite has announced that it will be joining Cursor through a definitive acquisition agreement. This partnership will combine their strengths to deliver enhanced solutions and drive innovation, marking an exciting new chapter for both companies. Together, they aim to better serve their users and expand their capabilities.
cursor.com
AI Designs 843-Part Linux Computer, which Booted on the First Try
Quilter’s AI completed the design of an 843-component Linux computer in about one week, dramatically compressing a task that normally takes months of engineering work. The dual-board system achieved ~98 % routing and zero design violations, powering up and booting Debian on its first attempt with only ~38 hours of human cleanup time. Dubbed Project Speedrun, this physics-driven AI approach could transform hardware design workflows and reduce traditional PCB bottlenecks.
venturebeat.com
Cognizant, Infosys, TCS and Wipro emerge as Frontier Firms with Microsoft
Cognizant, Infosys, TCS, and Wipro have teamed up with Microsoft to lead the way in using Copilot and advanced AI technology across various enterprises. These collaborations focus on integrating agentic AI to revolutionize operations and drive innovation. By adopting these cutting-edge solutions, the companies aim to enhance productivity and create transformative business opportunities.
news.microsoft.comResearch
ReFusion: A Diffusion Large Language Model with Parallel Autoregressive Decoding
This paper introduces ReFusion, a diffusion-based large language model that incorporates parallel autoregressive decoding. The model aims to improve the efficiency and quality of natural language generation by combining the strengths of diffusion models and autoregressive techniques. It addresses key challenges in large language model decoding, offering innovative methods to enhance performance.
huggingface.co
Enabling small language models to solve complex reasoning tasks
MIT CSAIL's "DisCIPL" system uses a large language model (LLM) to guide smaller language models in tackling complex reasoning tasks together. This collaborative approach allows the smaller models to achieve accuracy comparable to top reasoning models while being more efficient and outperforming leading LLMs in precision.
news.mit.eduVideo
Silicon Minds, Human Hearts.- Jen Fox
In this episode of *Silicon Minds, Human Hearts*, Jen Fox from Microsoft discusses the evolving relationship between humans and AI, focusing on voice interfaces, accessibility, and multimodal interactions. She highlights practical AI tools like Copilot that help eliminate tedious tasks, freeing up time for more meaningful work. The conversation covers the potential of vision and gesture technologies, inspiration from pop culture like *Star Trek*, and the importance of creating AI that centers on human needs and experiences.
youtube.comArticles
Introducing Polly: Your AI Agent Engineer
Polly is an AI-powered assistant integrated into LangSmith, designed to help you debug, analyze, and enhance your agents effortlessly. Built with insights from working with thousands of developers, Polly simplifies the process of improving agent performance. It’s a smart addition to streamline and optimize creating production-grade agents.
blog.langchain.com
Google Cloud Launches Managed MCP Support
Google Cloud has unveiled fully-managed Model Context Protocol (MCP) servers, making it easier for developers to access and integrate APIs. This enterprise-ready solution simplifies API infrastructure while improving efficiency and enhancing AI capabilities for businesses. It offers a streamlined approach to building and managing applications.
infoq.com
Inside Kaggle's AI Agents intensive course with Google
Kaggle's AI Agents Intensive, in collaboration with Google, offered a free course designed to help participants build and deploy cutting-edge AI applications. This program provided hands-on learning opportunities, empowering learners to explore the potential of AI agents. The course aimed to equip participants with the skills needed to shape the future of AI innovation.
blog.google
When (Not) to Use Vector DB
This article explores the challenges of using vector databases for certain Retrieval-Augmented Generation (RAG) use cases and explains why a key-value store proved to be a better solution. It highlights the limitations of vector databases for scenarios requiring strict one-to-one lookups and faster retrieval times. By sharing this experience, it aims to help others choose the right database approach for their specific needs.
towardsdatascience.comUpcoming Events
AgentCon 2025
After an exciting set of AgentCon events before the summer holidays, AgentCon is happening in more locations during the second half of the year. The AI Agents World Tour is a series of one-day conferences for developers focused on autonomous AI agents. Spanning cities like San Francisco and Singapore, these events gather top engineers, researchers, and creators shaping the future of AI technologies. It’s an opportunity to connect, learn, and innovate in the fast-evolving world of autonomous AI. Check all events and register for an event near you!
globalai.communityCode
Build and deploy scalable AI agents with NVIDIA NeMo, Amazon Bedrock AgentCore, and Strands Agents
This post shows how to create and deploy scalable AI agents using Strands Agents, Amazon Bedrock AgentCore, and NVIDIA NeMo Agent Toolkit on AWS. It covers the full process from development to production, including building, evaluating, and optimizing the agents for performance. The guide highlights the integration of these tools to streamline AI deployment on Amazon Web Services.
aws.amazon.com
GitHub - langchain-ai/lca-lc-foundations
This GitHub repository is dedicated to the development of "lca-lc-foundations" by LangChain AI. It provides resources and tools foundational to LangChain's AI initiatives, encouraging collaboration and contributions from developers. Users can create an account to participate in the project and explore its codebase.
github.com
GitHub - m0n0x41d/quint-code: Structured reasoning framework for Claude Code, Gemini, Cursor, and Codex
This GitHub project introduces "quint-code," a structured reasoning framework designed for tools like Claude Code, Gemini, Cursor, and Codex. It emphasizes hypothesis-driven decision-making while providing an auditable trail of evidence. The framework is aimed at enhancing clarity and accountability in coding and decision-making processes.
github.com
Shannon - AI Pentesting Tool that Autonomously Checks for Code Vulnerabilities and Executes Real Exploits
Shannon is an advanced AI-driven tool designed to perform autonomous penetration testing for web applications. It analyzes code to identify potential attack vectors and validates these vulnerabilities by executing real exploits in a live browser environment. Shannon offers a powerful solution for ensuring web application security without manual intervention.
cybersecuritynews.comPodcast
The Artificial Intelligence Show
The Artificial Intelligence Show is a podcast designed to make AI understandable and useful for growing your business. Hosts Paul Roetzer and Mike Kaput explore important AI news, offering practical insights and strategies to help enhance your company and career. Brought to you by the team behind the Marketing AI Institute and MAICON, the show aims to boost AI literacy and make this transformative technology accessible to all.
open.spotify.com