Homechevron_rightArchiveschevron_rightTechnology

Archives

Explore Stories by Date

Today

Malicious VS Code Extensions With 1.5 Million Installs Steal Developer Source Code

Cybersecurity researchers uncovered two malicious VS Code extensions disguised as AI-powered coding assistants that secretly exfiltrated developer source code to China-based servers. ChatGPT - 中文版 and ChatMoss, with 1.5 million combined installs, remain available on the Visual Studio Marketplace despite actively stealing files and tracking user behavior through hidden mechanisms.

2 Sources

Microsoft unveils Maia 200 AI chip with 3x performance edge over Amazon Trainium and Google's TPU

Microsoft has launched its second-generation custom AI chip, the Maia 200, claiming superior performance over competitors Amazon Trainium and Google's TPU. Built to optimize AI inference workloads, the chip packs over 100 billion transistors and delivers 10 petaflops in 4-bit precision. Already deployed in Azure data centers, Maia 200 powers OpenAI's GPT-5.2 models and Microsoft 365 Copilot while offering 30% better performance per dollar.

7 Sources

NVIDIA invests $2 billion in CoreWeave, grants early access to standalone Vera CPUs

NVIDIA has committed an additional $2 billion to CoreWeave, making the AI-focused cloud provider the first to offer Vera CPUs as a standalone option. CEO Jensen Huang signals the chip giant's aggressive push into the CPU market, directly challenging Intel and AMD while addressing bottlenecks in AI infrastructure.

2 Sources

Nvidia launches Earth-2 open AI weather forecasting models, beating Google's GenCast on accuracy

Nvidia unveiled its Earth-2 family of open AI weather forecasting models at the American Meteorological Society meeting in Houston. The new models promise faster, more accurate predictions and claim to outperform Google DeepMind's GenCast on over 70 variables. Organizations including Israel Meteorological Service, TotalEnergies, and The Weather Company are already deploying these tools for operations ranging from storm tracking to energy grid management.

4 Sources

ChatGPT Temporary Chat Gets Major Upgrade With Personalization While Staying Private

OpenAI is testing a significant upgrade to ChatGPT's Temporary Chat feature that allows users to retain personalization settings like memory and style preferences while keeping conversations private. The update addresses a key limitation where users had to choose between privacy and a personalized AI experience.

2 Sources

Samsung closes gap with rivals as HBM4 memory chips pass Nvidia's final verification for Rubin AI

Samsung Electronics has entered the final qualification phase with Nvidia for its next-generation HBM4 AI memory chips, preparing for mass production in February. The South Korean giant's HBM4 operates at 11.7 gigabits per second, exceeding industry standards and Nvidia's requirements. This marks a significant turnaround for Samsung, which has trailed SK Hynix and Micron in supplying advanced memory for AI accelerators.

6 Sources

Yesterday

Meta Ray-Ban Display impresses with features but restrictive software ecosystem holds it back

Meta's $800 Ray-Ban Display smart glasses showcase advanced Neural Band gesture controls and heads-up navigation, but reviewers highlight significant software limitations. The wearables lock users into Meta's ecosystem, blocking third-party push notifications and calendar integration while offering limited translation options compared to competitors.

2 Sources

ChatGPT now sources data from Grokipedia, sparking concerns about AI misinformation spread

OpenAI's latest GPT-5.2 model has been caught citing Grokipedia, Elon Musk's AI-generated encyclopedia, for obscure queries on topics like Iranian politics and Holocaust deniers. The discovery raises alarm about information integrity as AI systems increasingly reference other AI-generated content, creating what experts call a recursive loop that could accelerate the propagation of misinformation and digital folklore across the web.

5 Sources

23 Jan 2026

Meta pauses teen access to AI characters globally as lawsuits and regulatory pressure mount

Meta announced a global suspension of teen access to AI characters across Instagram, Facebook, and WhatsApp as it develops an updated version with enhanced parental controls. The move comes just days before the company faces trials over child safety concerns and follows similar restrictions by Character.AI and OpenAI amid growing scrutiny over the potential harm of AI chatbots on minors.

15 Sources

AI Agents Face Mathematical Limits That Could Block Full Autonomy, Researchers Prove

A recently surfaced research paper claims to mathematically prove that AI agents and large language models cannot reliably handle tasks beyond a certain complexity threshold. The study by former SAP CTO Vishal Sikka and his son challenges the AI industry's vision of fully autonomous systems, suggesting fundamental limitations are built into transformer-based architectures.

4 Sources

Galaxy S26 series poised to gain Google's Scam Detection, breaking Pixel exclusivity

Google's AI-powered Scam Detection feature, previously exclusive to Pixel phones, appears headed to Samsung's upcoming Galaxy S26 series. Code discovered in the Phone by Google app reveals references to Galaxy S26 model numbers alongside Pixel devices, suggesting Samsung may adopt Google's dialer app to enable real-time scam alerts for phone calls and text messages.

4 Sources

Google Discover AI headlines become permanent feature despite accuracy concerns from publishers

Google has confirmed that AI-generated headlines in Google Discover are now a permanent feature, claiming they improve user satisfaction. But publishers like The Verge and PCMag report the AI-rewritten headlines often contain misleading or false information, replacing carefully crafted titles with inaccurate summaries that can spread misinformation.

5 Sources

cURL scraps bug bounties as AI slop overwhelms security team with fake vulnerability reports

The widely-used networking tool cURL is ending its HackerOne bug bounty program after being flooded with AI-generated fake vulnerability reports. Lead developer Daniel Stenberg cited the need to protect maintainers' mental health as low-quality submissions strained the small security team. The move raises questions about the future of open-source security programs.

2 Sources

JBL BandBox practice amps use Stem AI to remove any instrument from songs in real-time

JBL launched the BandBox Solo and BandBox Trio, AI-powered practice amplifiers that use Stem AI technology to separate and isolate vocals and instruments from any song streamed over Bluetooth. The speakers work offline, letting musicians create custom backing tracks instantly. The $250 Solo targets individual players while the $600 Trio supports up to four instruments for bands.

5 Sources

AI coding now powers 30% of US code, but productivity gains favor experienced developers

A Science-backed analysis reveals AI-assisted coding has reached nearly one-third of newly written code in the US by late 2024. But the productivity gains concentrate among seasoned developers, not newcomers. Less experienced programmers use AI tools more often yet see minimal benefits, raising concerns about a widening software development skill gap.

2 Sources

Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

The Outpost

News

About