Other TechCrunch

Elon Musk says X will send DMs when posts you’ve engaged with are corrected

X plans to notify users via DM when posts they've interacted with are corrected by Community Notes.

Tech Business/VC The Verge

Microsoft’s Xbox reset is pivoting Obsidian to make Fallout instead of Avowed

Microsoft is pivoting Obsidian Entertainment to focus on a new Fallout title, canceling multiple other projects as part of an Xbox reset.

AI/ML VentureBeat

OpenAI launches GPT-Live, a full-duplex voice upgrade that lets ChatGPT talk more like a person

OpenAI has launched GPT-Live, introducing a full-duplex voice architecture that allows for simultaneous listening and speaking, decoupled from the reasoning layer.

AI/ML arXiv cs.AI

Depression Symptoms and Relational Patterns in 187k ChatGPT Histories

A research study analyzes how individuals with depressive symptoms use ChatGPT as an informal support infrastructure based on 187k conversation histories.

AI/ML arXiv cs.AI

IMR: Iterative Mode-World Weighted Regression for Multi-Agent Trajectory Prediction

The IMR method is proposed for multi-agent trajectory prediction in automated vehicles, utilizing a mode-world weighted regression loss to improve accuracy and safety.

Software Engineering arXiv cs.AI

Plainbook: Data Science, in Plain Language

Plainbook introduces a natural-language centered notebook system for data science, aimed at making reproducible analysis accessible to non-coders.

AI/ML arXiv cs.AI

SCOReD: Student-Aware CoT Optimization for Recommendation Distillation

SCOReD is a new framework for optimizing Chain-of-Thought distillation in recommendation systems to reduce verbosity and improve student model performance.

Cybersecurity arXiv cs.AI

The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities

A systematization paper identifies critical gaps in execution security for AI coding agents, focusing on isolation, access control, and TOCTOU vulnerabilities.

Cybersecurity arXiv cs.AI

Unicode TAG-Block Concealment of Tool-Metadata Payloads in the Model Context Protocol: An Approval-View Fidelity Gap Across Three Independent Server Implementations

Researchers discover a vulnerability in the Model Context Protocol (MCP) where Unicode TAG-block concealment allows metadata payloads to be invisible to humans but visible to AI models.

Other Hacker News

What Do We Know About the Microplastics Inside Us?

A discussion on the prevalence and health impacts of microplastics within the human body.

Software Engineering Hacker News

Understanding B-Tree Indexes in PostgreSQL: A Comprehensive Guide– Part 1

A comprehensive technical guide explaining the inner workings of B-Tree indexes in PostgreSQL.

Other Hacker News

PlayStation can delete all your digital games after 3 years of inactivity (EU)

Reports that PlayStation may delete digital game libraries of EU users after three years of account inactivity.

AI/ML TechCrunch

Google Photos adds a new AI ‘Video Remix’ tool

Google Photos introduces 'Video Remix', an AI-powered tool for cinematic relighting, background swapping, and artistic styling.

Hardware/Chips Ars Technica

Aussie gov't tells volunteers to throw out thousands of functioning test routers

The Australian government has instructed volunteers to discard functioning test routers, despite their potential for reflashing.

Other Ars Technica

TikTok users don't have as much agency over their FYPs as they think

Analysis suggests TikTok users have less control over their For You Page algorithm than they believe, requiring active curation.

Hardware/Chips Ars Technica

US seeks cheaper hunter-killer drones after Iran destroys $1B worth of Reapers

The US military is seeking affordable hunter-killer drone replacements after significant losses of expensive Reaper drones.

Hardware/Chips Ars Technica

Miami-based City Labs achieves a first for commercial nuclear power in space

City Labs has achieved a milestone in commercial nuclear power for space travel via the BOHR mission.

AI/ML Ars Technica

Google updates Android Bench with new LLMs, but Gemini still lags behind

Google updates Android Bench with new LLMs, though Gemini continues to underperform relative to competitors.

AI/ML arXiv cs.AI

Beyond Accuracy: How Humans Evaluate Legally Correct but Socially Controversial Legal Advice from Machines

A research study examines how people perceive AI-provided legal advice, finding that while perceived as objective, it is often seen as lacking contextual sensitivity.

AI/ML Hacker News

Grok 4.5

Discussion regarding the release or capabilities of Grok 4.5.