AI News — Sunday, July 26, 2026

9
Monday.com is the latest tech company to blame AI for layoffs — here are 20 others

Monday.com joins a growing list of over 20 tech companies that have cited AI as a reason for recent layoffs, highlighting a significant industry trend.

TechCrunchindustry
8
One fallen power line exposed a growing AI data center problem. Here’s how to fix it.

A single power line failure has brought to light the increasing vulnerability and infrastructure challenges faced by rapidly expanding AI data centers, prompting calls for urgent solutions.

TechCrunchindustry
7
Librarians are hosting viral ‘Avoiding AI’ workshops for people who are fed up with Big Tech

Librarians are gaining traction with popular workshops designed to help individuals navigate and minimize their interaction with AI technologies, reflecting growing public sentiment against Big Tech.

TechCrunchindustry
7
NVIDIA-labs OO Agents: Native Python Object-Oriented Agents

NVIDIA-labs introduces a new framework for native Python object-oriented agents, offering a structured approach for developing complex AI systems.

Hugging Faceopen-source
7
K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs

Researchers introduce K12-KGraph, a curriculum-aligned knowledge graph designed to benchmark and enhance the training of large language models specifically for educational applications.

Hugging Faceresearch
6
Create, edit and star in videos with two Google Vids updates

Google Vids receives two new updates, enabling users to create, edit, and feature themselves in videos more easily, leveraging AI for enhanced content creation.

Google AI Blogproduct
6
NTT DATA Group cuts incident analysis to 30 minutes with Codex

NTT DATA Group has significantly reduced incident analysis time to just 30 minutes by integrating OpenAI's Codex, demonstrating the practical impact of AI in IT operations.

OpenAI Blogindustry
6
SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals for Efficient Video Generation

SANA-Video 2.0 introduces a novel hybrid linear attention mechanism with attention residuals, significantly improving the efficiency of video generation models.

Hugging Faceresearch
6
Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction

Tencent introduces WorkBuddy Bench, a new multi-domain benchmark for coding agents featuring contamination-resistant task construction to accurately evaluate their performance.

Hugging Faceresearch
6
LLMs Get Lost in Evolving User Intent

New research indicates that Large Language Models struggle to maintain coherence and effectiveness when user intent evolves during a conversation, highlighting a key challenge in conversational AI.

Hugging Faceresearch
6
Self-Supervised Learning of Structured Dynamics from Videos

A new paper explores self-supervised learning methods to extract structured dynamics directly from video data, advancing the field of AI understanding complex temporal relationships.

Hugging Faceresearch
6
Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

Researchers propose a streaming multi-agent autoregressive diffusion model that utilizes world state registers to enhance the generation of complex, dynamic environments.

Hugging Faceresearch
5
We instrumented an AI agent swarm with SigNoz, and its own telemetry told us we were wrong about almost everything

Instrumenting an AI agent swarm with SigNoz revealed unexpected insights from its telemetry, challenging initial assumptions about its behavior and performance.

Dev.toopen-source
5
What is Good? Extracting and Testing Implicit Theories of Literary Quality from LLM Reasoning Traces

This research investigates how to extract and test implicit theories of literary quality by analyzing the reasoning traces of Large Language Models.

arXivresearch
5
Knowledge Injection Exists in MoE? Exploring Expert-Aware Contrast Decoding in MoE for Mitigating LLMs'Hallucinations

A new study explores the phenomenon of knowledge injection in Mixture-of-Experts (MoE) models and proposes an expert-aware contrast decoding method to reduce hallucinations in LLMs.

arXivresearch