The Differential
Open main menu
Sign in
Create Account
Latest
Articles
Code
Papers
Article
-
techcrunch.com
FBI says Artex AI open source hacking tool used to breach networks in 'under three hours' | TechCrunch
The FBI has issued a warning about hackers using AI-driven tools like Artex for cyberattacks across various sectors, including finance and IT. These tools exploit security flaws quickly, leading to serious breaches. Despite attempts to restrict access, Artex remains widely available, posing ongoing risks for organizations.
3 min read
Article
-
amirulcyber.com
CNN, Zoom, Fox News Served Visitors a Crypto-Stealing Skimmer – OPTIMJACK (Nov 2026) - Amirul Cyber
A security vulnerability has been uncovered in Optimizely bundles affecting major sites like CNN, Zoom, and Fox News. Malicious JavaScript is stealing cryptocurrency addresses and card information from visitors. If your site uses Optimizely, ensure to check your bundles to safeguard user data.
4 min read
Article
-
spectrum.ieee.org
Why Skydio Built a Lopsided Drone
Skydio's new F10 Lightrunner drone stands out with its unique asymmetrical design, featuring a wingtip-mounted camera for improved reconnaissance. This innovative approach aims to solve occlusion issues in fixed-wing flights, complemented by the autonomous MegaDock system for seamless operations.
5 min read
Article
-
www.hntrends.com
"Who is Hiring?" postings stuck near 12-year lows as AI reaches half of all jobs (November 2026)
Job postings on Hacker News dropped 37% from last year, reflecting a trend towards AI integration in over half of listings. Python and TypeScript remain popular, while Front End roles have fallen sharply. The report highlights evolving skills and shifting demands in the tech job market for November 2026.
4 min read
Article
-
iter.ca
Universal XSS in uBlock Origin from website-driven element picker
A recently discovered vulnerability in uBlock Origin for Firefox allowed malicious websites to trick users into adding arbitrary trusted filters. This could have enabled attackers to run JavaScript code on any page. The issue has been addressed, and more details can be found in the GitHub advisory.
4 min read
Article
-
xenaproject.wordpress.com
“…and the proof is checked in Lean”
This article explains how to interpret claims regarding the correctness of theorems checked using Lean code. It explores essential questions that arise, particularly for AI-generated code, and emphasizes the need for thorough tests, as many existing validations are incomplete.
10 min read
Article
-
www.theguardian.com
Revealed: Israel uploaded millions of photos to Amazon’s facial recognition platform to track Palestinians
The Israeli military has utilized Amazon's cloud platform to upload 3.5 million images of Palestinians for facial recognition analysis. Leaked documents reveal the military's reliance on this technology for monitoring, highlighting significant concerns about data privacy and the ethical implications of such surveillance in occupied territories.
6 min read
Article
-
infisical.com
The Active Directory Era Is Over: Machine Identity for AD Admins
Active Directory is becoming outdated as modern cloud computing highlights its limitations in managing both human and machine identities. The architecture, designed for long-lived infrastructure, struggles to keep up with the dynamic nature of current technologies. This shift raises significant security concerns for organizations relying on AD.
7 min read
Article
-
www.wired.com
Nvidia Was Already in Trouble Before the Jensen Huang Mess
Nvidia's leadership under CEO Jensen Huang has positioned it as a dominant force in AI chip production. However, with rising competition and regulatory challenges in markets like China, the company's future could face hurdles. As the industry evolves, Nvidia may need to adapt to maintain its lead.
5 min read
Article
-
klauria.tech
psychic: like fzf but learns from your actions | Kshitij Lauria
Psychic is an innovative terminal-based file browser that uses machine learning to streamline your file navigation. It learns from your usage patterns, ranking files based on your history for faster access. With high-speed performance and real-time updates, it's designed to enhance your productivity while working on projects.
5 min read
Article
-
www.gameworkers.co.uk
On Generative AI Usage in the Games Industry - IWGB Game Workers
A coalition of European game worker unions has issued a manifesto outlining the dangers of generative AI in the video game development sector. The document highlights how genAI negatively impacts workers, degrades working conditions, and threatens the industry's sustainability. It calls for awareness and collective action to address these issues.
15 min read
Article
-
www.breck-mckye.com
Software is breaking down at a systemic level | Jimmy Breck-McKye
Recent widespread software failures highlight systemic issues in the tech industry. A series of outages and vulnerabilities reveal a complex web of dependencies and poor security practices. As AI technologies influence both development and exploitation, the challenge now is to address the fundamental problems in building software.
10 min read
Paper
-
arxiv.org
Same Papers, Different Verdicts: Tracing Version Drift in LLM Peer Review
This article examines how different versions of large language models (LLMs) impact peer review outcomes. It reveals significant shifts in acceptance rates for the same papers based on model updates, highlighting the need for consistent validation of LLM evaluations against human judgments in academic assessments.
2 min read
Paper
-
arxiv.org
Searching for Autoregressive Neural Architectures from Fundamental Operations
This article explores the development of einspaceAR, a new neural architecture search space designed to create autoregressive models from basic operations. Unlike previous methods that impose constraints, this approach broadens the possibilities for sequence modeling, making it a valuable addition to machine learning tools.
2 min read
Paper
-
arxiv.org
Disentangling Forgetting from Utility: Lightweight LLM Unlearning via Orthogonal Null-Space Projection
This article introduces UUE, a novel framework for unlearning in large language models (LLMs) that effectively separates the deletion of specific knowledge from the model's overall performance. UUE enhances unlearning efficiency while preserving general capabilities, demonstrating its advantages over traditional methods through extensive benchmarking.
2 min read
Previous
Next