Skip to content
gentic.news — AI News Intelligence Platform
Connecting to the Living Graph…
gentic.news

Latest AI News

Every story, newest first — research, funding, product launches, policy, and analysis.

Two AI chat interfaces side by side: left panel shows OpenAI Codex with parallel subagents highlighted, right panel…
Big Tech
85

OpenAI Codex Gains Subagents, Anthropic Ships 1M Context at Standard Pricing

OpenAI added parallel subagents to Codex to combat 'context pollution,' while Anthropic made 1M context generally available for Claude Opus/Sonnet 4.6 with no price premium, achieving 78.3% on MRCR v2. These incremental upgrades reshape practical agentic workflows.

pub.towardsai.net/Mar 17, 2026/3 min read/Multi-Source
technical analysisagentic aiai strategy
A security researcher stands before a digital map showing 40,000 exposed OpenClaw servers, with red markers…
Products & Launches
85

Security Researcher Exposes 40,000+ OpenClaw Servers, 12,000 Vulnerable to API Key Theft

A security scan reveals over 40,000 OpenClaw servers are exposed online, with 12,000+ vulnerable to API key and data theft. The researcher published a comparative security analysis of hosted AI providers.

x.com/Mar 17, 2026/3 min read
open sourcesecurityinfrastructure
Dario Amodei, CEO of Anthropic, speaks during a tech conference, gesturing while seated on stage in front of a large…
Opinion & Analysis
85

Anthropic CEO Dario Amodei Predicts 50% of Entry-Level White-Collar Jobs Could Be Automated Within 3 Years

Anthropic CEO Dario Amodei stated in an interview that AI could automate 50% of entry-level white-collar jobs within three years. The prediction highlights the rapid timeline some industry leaders anticipate for AI's impact on knowledge work.

x.com/Mar 17, 2026/3 min read
ai ethicsbusinesspolicy
A computer screen displays waveform lines and a video editing timeline, with a diverse group of animated characters…
Products & Launches
85

Tongyi Lab Releases World's First Open-Source Multi-Speaker AI Dubbing Model

Alibaba's Tongyi Lab has released the first open-source AI model capable of dubbing multi-speaker conversations, addressing one of the hardest problems in AI video generation. The model synchronizes voice with lip movements across multiple speakers in a single pass.

x.com/Mar 17, 2026/3 min read
open sourcecomputer visionmultimodal ai
Alibaba's OpenSandbox interface showing an AI agent isolated in a secure gVisor runtime with a code interpreter and…
Products & Launches
97

Alibaba Open-Sources OpenSandbox: A gVisor/Firecracker-Based Execution Environment for AI Agent Security

Alibaba has open-sourced OpenSandbox, a general-purpose execution environment that isolates AI agents in secure runtimes like gVisor or Firecracker. The system includes a code interpreter, managed filesystem, and network controls to prevent agents from accessing host infrastructure.

x.com/Mar 17, 2026/3 min read
open-sourcesecurityinfrastructure
Side-by-side comparison of Bedrock Agents and AgentCore architecture diagrams, showing managed vs customizable AI…
Big Tech
83

AWS Bedrock Agents vs. AgentCore: A Technical Guide for AI Architects

AWS offers two distinct approaches for building AI agents: the fully managed Bedrock Agents for speed and the low-level AgentCore framework for control. This article breaks down the architectural differences, code examples, and selection criteria for production deployments.

pub.towardsai.net/Mar 17, 2026/3 min read/Multi-Source
agentsai engineeringaws
A dark-themed dashboard displays a world map with tracked flight paths, satellite markers, and a sidebar listing…
Products & Launches
99

Crucix: Open-Source Personal Intelligence Terminal Aggregates 26 OSINT Feeds Locally

Developer-built Crucix runs locally, pulling 26 open-source intelligence feeds every 15 minutes into a unified dashboard. The MIT-licensed tool includes satellite data, flight tracking, conflict monitoring, and integrates with LLMs for analysis.

x.com/Mar 17, 2026/3 min read
open sourceintelligenceosint
Developer demonstrates Kavach firewall intercepting an AI agent's file deletion and network exfiltration attempts…
Products & Launches
95

Kavach: Open-Source Local Firewall for AI Agents Intercepts Destructive File Ops and Network Exfiltration

Developer releases Kavach, a local 'military-grade' firewall for AI agents. It intercepts destructive file operations and network requests, redirecting them to a phantom workspace while spoofing success responses to the agent.

x.com/Mar 17, 2026/3 min read
open sourcesecuritydeveloper tools
A robotic arm with inflatable vine-like beams wraps around a metal kettlebell, lifting it off a table in a lab setting
AI Research
95

MIT's Loop Closure Grasping Enables Robots to Lift 6.8kg Kettlebell with Inflatable 'Vine' Beams

MIT researchers developed loop closure grasping, a method where robots snake around objects with inflatable beams, then lock into a closed loop to distribute weight through tension. This allows lifting heavy, delicate items like a 6.8kg kettlebell without crushing or slipping.

x.com/Mar 17, 2026/3 min read
hardwareroboticsresearch
A glowing digital brain surrounded by shield icons and data streams, symbolizing an LLM defending against poisoned…
AI Research
88

How Large Language Models 'Counter Poisoning': A Self-Purification Battle Involving RAG

New research explores how LLMs can defend against data poisoning attacks through self-purification mechanisms integrated with Retrieval-Augmented Generation (RAG). This addresses critical security vulnerabilities in enterprise AI systems.

news.google.com/Mar 17, 2026/3 min read/Multi-Source
risk managementai securityllms
Engineer inspects NVIDIA DGX Spark desktop unit with GB10 Blackwell chip, Qwen3.5–35B model deployment interface on…
Opinion & Analysis
95

We Hosted a 35B LLM on an NVIDIA DGX Spark — A Technical Post-Mortem

A detailed, practical guide to deploying the Qwen3.5–35B model on NVIDIA's GB10 Blackwell hardware. The article serves as a crucial case study on the real-world challenges and solutions for on-premise LLM inference.

medium.com/Mar 17, 2026/3 min read/Multi-Source
case-studyhardwarellm-ops
A sleek smart speaker on a wooden table, blue LED ring glowing, with a smartphone showing a privacy shield icon next…
Products & Launches
85

OpenHome Launches Open-Source Voice Assistant Platform with Full Local Processing

OpenHome has launched an open-source voice assistant platform that processes all audio and commands locally on-device, positioning itself as a privacy-focused alternative to cloud-based services like Amazon Alexa.

x.com/Mar 17, 2026/3 min read
open sourcevoice aiedge computing
Screenshot of Hugging Face repository page showing deleted Mistral models Magistral, Pixtral, and Devst with commit…
Products & Launches
85

Mistral Deletes Magistral, Pixtral, and Devst Models from Hugging Face Hub

Mistral AI has removed three of its models—Magistral (reasoning), Pixtral (multimodal), and Devst—from the Hugging Face Hub. The deletions, confirmed via the platform's commit history, were unannounced, leaving developers to speculate about the company's strategy.

x.com/Mar 17, 2026/3 min read
open sourcebusinessmodels

RAG Eval Traps: When Retrieval Hides Hallucinations

AI Research
76

RAG Eval Traps: When Retrieval Hides Hallucinations

A new article details 10 common evaluation pitfalls that can make RAG systems appear grounded while they are actually generating confident nonsense. This is a critical read for any team deploying RAG for customer service or internal knowledge bases.

medium.com/Mar 17, 2026/3 min read
risk managementcustomer experienceai strategy
AI engineer in server room staring at a monitor displaying a GPU memory usage graph spiking to 100 percent, with a…
Opinion & Analysis
95

How a GPU Memory Leak Nearly Cost an AI Team a Major Client During a Live Demo

A detailed post-mortem of a critical AI inference failure during a client demo reveals how silent GPU memory leaks, inadequate health checks, and missing circuit breakers can bring down a production pipeline. The author shares the architectural fixes implemented to prevent recurrence.

pub.towardsai.net/Mar 17, 2026/3 min read/Multi-Source
case studymlopsfailure analysis
An AI agent architecture diagram showing reasoning-action loops with memory and guardrails between user input and…
AI Research
76

Building ReAct Agents from Scratch: A Deep Dive into Agentic Architectures, Memory, and Guardrails

A comprehensive technical guide explains how to construct and secure AI agents using the ReAct (Reasoning + Acting) framework. This matters for retail AI leaders as autonomous agents move from theory to production, enabling complex, multi-step workflows.

pub.towardsai.net/Mar 17, 2026/3 min read
agentic airetail innovationai engineering
A diagram shows a query entering an LLM that outputs logical rules, which are applied in a knowledge graph embedding…
AI Research
84

EMBRAG Framework Achieves SOTA on KGQA Benchmarks via Embedding-Space Rule Generation

Researchers propose EMBRAG, a framework that uses LLMs to generate logical rules from a query, then performs multi-hop reasoning in knowledge graph embedding space. It sets new state-of-the-art on two KGQA benchmarks.

arxiv.org/Mar 17, 2026/3 min read
reasoningllmsresearch
A fleet of delivery trucks navigates a city map with emission quota limits, while a reinforcement learning agent…
AI Research
87

Reinforcement Learning Solves Dynamic Vehicle Routing with Emission Quotas

A new arXiv paper introduces a hybrid RL and optimization framework for dynamic vehicle routing with a global emission cap. It enables anticipatory demand rejection to stay within quotas, showing promise for uncertain operational horizons.

arxiv.org/Mar 17, 2026/3 min read/Multi-Source
operations-researchsupply-chainsustainability

Continual Fine-Tuning with Provably Accurate, Parameter-F…

AI Research
78

Continual Fine-Tuning with Provably Accurate, Parameter-Free Task Retrieval: A New Paradigm for Sequential Model Adaptation

Researchers propose a novel continual fine-tuning method that combines adaptive module composition with clustering-based retrieval, enabling models to learn new tasks sequentially without forgetting old ones. The approach provides theoretical guarantees linking retrieval accuracy to cluster structure.

arxiv.org/Mar 17, 2026/3 min read/Multi-Source
researchfoundation-modelsmachine-learning
DOVA framework architecture diagram showing multi-agent deliberation phase preceding tool use, with interconnected…
AI ResearchBreakthrough
100

DOVA Framework Introduces Deliberation-First Orchestration for Multi-Agent Research Automation

Researchers propose DOVA, a multi-agent platform that uses explicit meta-reasoning before tool invocation, achieving 40-60% inference cost reduction on simple tasks while maintaining deep reasoning capacity for complex research automation.

arxiv.org/Mar 17, 2026/3 min read/Widely Reported
architectureefficiencymulti-agent
Architecture diagram showing a multimodal retrieval flow with text, image, and video inputs processed through a late…
AI Research
79

AMES: A Scalable, Backend-Agnostic Architecture for Multimodal Enterprise Search

Researchers propose AMES, a unified multimodal retrieval system using late interaction. It enables cross-modal search (text, image, video) within existing enterprise engines like Solr without major redesign, balancing speed and accuracy.

arxiv.org/Mar 17, 2026/3 min read
mlopscomputer visionsearch & discovery
Researchers analyze a chart showing LLM performance decline, with a laptop displaying Brittlebench framework results…
AI Research
84

Brittlebench Framework Quantifies LLM Robustness, Finds Semantics-Preserving Perturbations Degrade Performance Up to 12%

Researchers introduce Brittlebench, a framework to measure LLM sensitivity to prompt variations. Applying semantics-preserving perturbations to standard benchmarks degrades model performance by up to 12% and alters model rankings in 63% of cases.

arxiv.org/Mar 17, 2026/3 min read
ai safetyresearchmachine learning
NVIDIA researcher presenting technical slides on Nemotron Ultra LLM architecture at a conference, with complex AI…
Products & Launches
85

NVIDIA Nemotron Ultra: Details Emerge on Upcoming Open-Source LLM Series

NVIDIA is developing the Nemotron Ultra series of open-source large language models. The project, described as 'insane' and 'underrated,' is generating early hype among AI researchers.

x.com/Mar 17, 2026/3 min read
open sourceresearchllm
Wharton professor Ethan Mollick presents his 'AI Weirdness Axiom' on stage, gesturing to a slide showing a graph of…
Opinion & Analysis
85

Ethan Mollick's 'AI Weirdness Axiom': Why Treating AI Like Standard IT Products Reduces Reliability

Wharton professor Ethan Mollick argues that AI's inherent 'weirdness' must be embraced, not minimized. Attempting to implement AI like conventional software leads to less useful and less reliable systems.

x.com/Mar 17, 2026/3 min read
ai ethicsai strategyimplementation
Sam Altman speaking at a tech conference, gesturing with his hand, with OpenAI logo on a screen behind him
Products & Launches
87

Sam Altman Aims for '5T Tokens Per Day' as OpenAI Reportedly Scales GPT-5.4

Sam Altman stated his goal is to flood the market with AI tokens, comparing intelligence to a utility. A separate, unverified report claims GPT-5.4 is processing '5T tokens per day' in its first week.

x.com/Mar 16, 2026/3 min read
rumorbusiness strategyopenai
Mistral Small 4 model announcement graphic with AI neural network visuals and performance metrics chart
Products & Launches
85

Mistral Releases Mistral Small 4, Claiming Significant Performance Jump Over Previous Models

Mistral AI has released Mistral Small 4, a new model in its 'Small' tier. The company claims it represents a major performance improvement over its predecessors, though no specific benchmarks are provided in the initial announcement.

x.com/Mar 16, 2026/3 min read
product launchmistral aicommercial ai
A developer's terminal showing Claude Code running alongside prompts for Google Gemini and OpenAI Codex models in a…
Products & Launches
89

Claude Octopus: GitHub Tool Enables Claude Code to Run Gemini and Codex Simultaneously

A developer discovered Claude Octopus, a GitHub repository that allows Anthropic's Claude Code to execute prompts across Google's Gemini and OpenAI's Codex models concurrently. The tool appears to enable parallel code generation from multiple AI assistants.

x.com/Mar 16, 2026/3 min read
model orchestrationgithubcoding tools
Wayve CEO Alex Kendall gestures on stage during a LONDON.AI keynote, with a large screen displaying autonomous…
Products & Launches
85

Wayve CEO Declares 'ChatGPT Moment for Autonomous Driving' at LONDON.AI Keynote

Wayve CEO Alex Kendall claimed autonomous driving has reached its 'ChatGPT moment' during a keynote, signaling a potential inflection point for AI-powered vehicles. The statement points to emerging end-to-end AI models replacing traditional modular self-driving systems.

x.com/Mar 16, 2026/3 min read
generative-aiautonomous-vehiclescomputer-vision
Claude Code Analyzes 1.2M Pentagon Contracts, Flags $4.2B in Potential Overpricing
Products & Launches
87

Claude Code Analyzes 1.2M Pentagon Contracts, Flags $4.2B in Potential Overpricing

An AI agent using Claude Code analyzed 1.2 million Pentagon procurement awards via API, comparing them to retail prices. It identified 340 contracts with 10x+ markups worth $4.2 billion in potential undercuts.

x.com/Mar 16, 2026/3 min read
procurementclaudegovernment tech
A person wearing a VR headset views a screen displaying a futuristic cityscape with neon lights, while audio…
AI Research
92

OmniForcing Enables Real-Time Joint Audio-Visual Generation at 25 FPS with 0.7s Latency

Researchers introduced OmniForcing, a method that distills a bidirectional LTX-2 model into a causal streaming generator for joint audio-visual synthesis. It achieves ~25 FPS with 0.7s latency, a 35× speedup over offline diffusion models while maintaining multi-modal fidelity.

x.com/Mar 16, 2026/3 min read
model-distillationmultimodal-aigenerative-ai