AI News — Thursday, November 19, 2026

9
Amidst fears of AI cyberattacks, Satya Nadella says lots of critical infrastructure is ‘unpatchable’

Microsoft CEO Satya Nadella highlights the alarming vulnerability of critical infrastructure to AI-powered cyberattacks, noting that much of it cannot be patched.

TechCrunchindustry
8
Is Coding with AI Actually Faster and Cheaper?

This article explores the practical implications and effectiveness of using AI tools in software development, questioning if they truly deliver on promises of speed and cost reduction.

Dev.toindustry
8
Gemini Intelligence: New features, on more devices

Google announces new features and broader availability for its Gemini AI, including support for lower-memory Android devices, expanding its reach and capabilities.

Google AI Blogproduct
7
Elon Musk says ousting Jensen Huang is like ‘kicking Michael Jordan off your team’

Elon Musk publicly expresses strong support for NVIDIA CEO Jensen Huang, emphasizing his irreplaceable value to the AI industry amidst hypothetical scenarios.

TechCrunchindustry
7
Beyond Answering: Learning to Speak First and Act in Streaming Video

New research introduces a model capable of proactively speaking and acting in real-time streaming video environments, moving beyond simple question-answering to more interactive AI agents.

Hugging Faceresearch
6
PROLA: Principal-Orthogonal Low-rank Adaptation for Predictive Spatiotemporal Weather Downscaling

Researchers propose PROLA, a novel low-rank adaptation method, to improve the efficiency and accuracy of predictive spatiotemporal weather downscaling using AI models.

Hugging Faceresearch
6
VideoNexus: Video Perception Beyond the Reasoning Context

This paper introduces VideoNexus, a new approach to video perception that allows AI models to understand and interpret video content beyond immediate reasoning contexts.

Hugging Faceresearch
6
Jeveration: System One Models Now Generate

A new paper introduces Jeveration, a system that enables 'System One' type AI models to generate content, potentially bridging intuitive and deliberative AI capabilities.

Hugging Faceresearch
6
The Resolution Hypothesis: Agent Evolution's Key Is Identifying Improvements

This paper proposes the Resolution Hypothesis, suggesting that the evolution of AI agents is fundamentally driven by their ability to identify and implement improvements.

Hugging Faceresearch
6
From Repositories to Training Environments: Scaling Long-Horizon Coding Agents with Environment Evolution Loop

Researchers present a method for scaling long-horizon coding agents by evolving training environments, enabling more complex and sustained AI-driven software development.

Hugging Faceresearch
6
ALIVE: Making Streaming Video Models Interactive

ALIVE presents a new approach to make streaming video models interactive, allowing for real-time engagement and manipulation of video content by AI.

Hugging Faceresearch
5
Game2Player: Scaling Gameplay Experience for Visual Agents through Environment-Player Co-Development

This research explores a co-development paradigm where AI agents and game environments evolve together to scale gameplay experiences for visual AI agents.

Hugging Faceresearch
5
Retrieval Ran First. Was Its Result Actually Ready?

This article discusses challenges in AI systems where retrieval components might execute prematurely, leading to suboptimal or incomplete results before the system is fully prepared.

Dev.toindustry
5
WorldRuler: Interactive World Modeling under Configurable Rules

WorldRuler introduces an interactive world modeling framework that allows AI systems to simulate and interact with environments based on user-configurable rules.

Hugging Faceresearch
5
FMT: Lossless Image Compression in a Fixed Number of Parallel Steps

This research introduces FMT, a novel method for lossless image compression that operates efficiently in a fixed number of parallel steps, promising faster processing.

Hugging Faceresearch