AI Briefing
Hacker News news
The latest AI and developer news from Hacker News, with short summaries.
Visit website →5 Frontier LLMs Give Different Answers on 67% of 1,000 Real-World Fact-Check Items
When 5 top-tier LLMs analyzed 1,000 real-world fact-check items, the models disagreed with each other on 67% of the items.
Sticker Shock Spreads Amid Surging Enterprise AI Adoption Costs
As companies grapple with low ROI relative to AI adoption costs, they are focusing on cutting reckless AI usage and finding efficiency.
Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction
This research introduces a study that uses a Multi-Agent LLM System to automatically discover and reproduce software vulnerabilities.
DuckDuckGo Search Visits Jump 28% After Google Says People Love AI Mode
Following Google's rollout of AI search features, users seeking AI-free search are flocking to DuckDuckGo, driving a surge in visits and app installs.
The Verdict on VibeSec
It warns of the security vulnerability risks in code generated by AI coding tools and emphasizes the need for technical guardrails.
Italy's Lombardy Region: 200% Tax on Data Centers Built on Green Spaces and Farmland
Italy's Lombardy region is imposing an additional tax of up to 200% on data centers built on green spaces and farmland to curb their unchecked expansion.
XLIDE: Editing VBA Without Excel
Introducing XLIDE, which lets you edit VBA in VS Code without installing Excel and integrates with GitHub Copilot.
Nvidia Vera CPU Benchmarks: Olympus Cores Show Outstanding Performance
NVIDIA's next-generation CPU for AI data centers, Vera, has signaled strong performance with its self-designed Olympus cores.
Language Models Need Sleep Too
To overcome the context-scaling limits of Transformers, a 'sleep' mechanism is proposed that compresses recent information into fast weights.
Launch HN: Minicor (YC P26) – Large-Scale Windows Desktop Automation Solution
A self-healing AI agent platform for legacy Windows desktop automation
The combination of outsourcing and local AI will soon become more economical than frontier models
As frontier model costs rise, combining engineers from low-wage countries with local AI is emerging as a more economical alternative.
Eagle 3.1: A Collaboration Between the EAGLE Team, vLLM Team, and TorchSpec Team
Eagle 3.1 was released through a collaboration between the EAGLE, vLLM, and TorchSpec teams.
Uber President Says It's Becoming Increasingly Hard to 'Justify' AI Spending
Uber is struggling to justify its AI investment because there's a missing link between rising AI token consumption and actual feature improvements.
BadHost – CVE-2026-48710: Starlette Host-Header Authentication Bypass Vulnerability
A security vulnerability has been discovered in which insufficient Host header handling in Starlette can allow authentication to be bypassed in major AI infrastructure such as vLLM and LiteLLM.
Apple macOS 26.5 Kernel Vulnerability Discovered by Claude: CVE-2026-28952
A security vulnerability in the Apple macOS kernel, CVE-2026-28952, was discovered using Claude.
Uber COO: 'It's Getting Harder to Justify Spending on Tokenmaxxing'
Uber's COO pointed out the cost-efficiency problem arising from maximizing LLM token usage.
Memory's Share of AI Chip Component Costs Rises to About Two-Thirds
HBM's share of AI chip component costs is projected to surge to 63% by the end of 2025.
DeepSeek Applies Permanent 75% Discount to Its Flagship AI Model
DeepSeek is permanently cutting the price of its flagship AI model by 75%.
Constraint Decay: Vulnerability of LLM Agents in Backend Code Generation
It has been confirmed that a phenomenon called 'Constraint Decay' occurs, in which LLM agents' performance degrades sharply as structural constraints increase during backend code generation.
Perceptual Image Codec: The Key to Practical Learned Image Compression
Introducing PICO, which is optimized for the human visual system and dramatically improves compression ratio and speed compared to existing codecs.