AI News — Friday, November 13, 2026

10
Democrats Launch Nvidia Investigation Over Alleged 'Scheme' with Trump

Democrats have initiated an investigation into Nvidia, sending a scathing letter to CEO Jensen Huang regarding an alleged scheme involving the company and former President Trump.

TechCrunchindustry
10
Google Search Gets Major Upgrade with Gemini 4 Hydrogen Integration

Google Search has received a significant enhancement through the integration of Gemini 4 Hydrogen, promising advanced AI capabilities for search queries.

Google AI Blogproduct
9
The Unfolding Jensen Huang Scandal: A Detailed Timeline

A comprehensive timeline details the unfolding scandal surrounding Nvidia CEO Jensen Huang, highlighting its potential implications for the tech industry.

TechCrunchindustry
8
Scaling Mathematical Proof Supervision from Structured Textbooks

New research explores methods for scaling the supervision of mathematical proofs by leveraging structured textbook data, aiming to improve AI's reasoning capabilities.

Hugging Faceresearch
8
Incentivizing Self-Learning in Large Language Models via Reinforcement Learning

A new paper proposes using reinforcement learning to incentivize self-learning behaviors in large language models, potentially leading to more autonomous and capable AI.

Hugging Faceresearch
8
Your MCP Servers Are Silently Draining Your Context Window: The Hidden Cost of Tool-Heavy AI Agents

An article highlights how tool-heavy AI agents can inadvertently consume significant context window resources, leading to performance issues and hidden costs for MCP servers.

Dev.toindustry
7
Muse Now Offers Paid Calling to US Businesses

The AI assistant Muse has expanded its capabilities to include placing calls to U.S. businesses, a feature now available to paying subscribers.

TechCrunchproduct
7
One Table, Many Queries: Benchmarking Generalist Tabular Models Across Generation, Imputation, and Prediction

Researchers introduce a new benchmark to evaluate generalist tabular models across diverse tasks including data generation, imputation, and prediction, providing a comprehensive assessment of their capabilities.

Hugging Faceresearch
7
SteerDistill: Quantizing LLM Agents via Closed-Loop Teacher Steering

A new technique called SteerDistill is proposed for quantizing LLM agents, utilizing closed-loop teacher steering to maintain performance while reducing model size.

Hugging Faceresearch
7
MACaQ: Fast, Model Agnostic Calibrated Quantization for Low-Precision Weights and Activations

MACaQ presents a fast and model-agnostic calibrated quantization method designed for low-precision weights and activations, enhancing efficiency in AI models.

Hugging Faceresearch
7
BirdDreams: Joint Perspective and BEV Video Generation for a Plausible, Reactive Driving World Model

BirdDreams introduces a novel approach for generating videos from both bird's-eye view (BEV) and first-person perspectives, creating a plausible and reactive world model for autonomous driving simulations.

Hugging Faceresearch
7
VS Code Extension Transforms Any Free Chatbot Website into an AI Agent

A new VS Code extension allows users to convert virtually any free chatbot website into a functional AI agent, offering a novel way to leverage existing conversational AI tools.

Dev.toopen-source
6
The Surprising Effectiveness of Deleting Weights in LLM Post-Training

Research reveals the unexpected effectiveness of strategically deleting weights in large language models after initial training, leading to potential efficiency gains without significant performance loss.

Hugging Faceresearch
6
Duo: A Reciprocal World Model for Full-Duplex Interaction

Duo introduces a reciprocal world model designed to enable full-duplex interaction in AI systems, allowing for more natural and continuous communication.

Hugging Faceresearch
6
Read Less, Solve More: Token-Efficient Sparse Reading of External Content for AI Agents

A new paper proposes a token-efficient sparse reading mechanism for AI agents to process external content, enabling them to solve tasks more effectively with fewer computational resources.

Hugging Faceresearch