All Updates
Page 1830 of 1905
February 20, 2026
Vercel Skills hits 62K with security audits
Vercel hosted Skills Night celebrating growth to 62,000 skills and 2M CLI installs in their open ecosystem for coding agents. Originating from React docs by Shu Ding, skills deliver on-demand knowledge via simple CLI installs into agents like Cursor and Claude. Announced security partnerships with Socket and Snyk to audit all skills and launch an Audits leaderboard.
OpenAI-Anthropic CEOs Snub at AI Summit
An onstage snub between OpenAI and Anthropic CEOs at an AI summit in India reveals backstory of OpenAI's jealousy toward Anthropic. The weekly AI wrap-up also spotlights a new favorite internet toy blending Warcraft III and Claude Code.
Meta smart glasses risk privacy backlash
Meta considered adding facial recognition to Ray-Ban smart glasses during a politically distracted period, per NYT report. Critics highlight high privacy risks despite praised hardware. Public distrust of Meta persists amid comparisons to phone cameras.
Microsoft Deletes Pirated Harry Potter LLM Guide
Microsoft removed a guide on training LLMs using pirated Harry Potter books. The dataset was mistakenly marked as public domain. This incident underscores risks in AI data sourcing.
Pixel 11 Rumored with Titan M3 Chip
Google Pixel 11 series expected to integrate Titan M3 security module into Tensor G6 chip, codenamed 'Google Epic' with 'longjing' firmware. This upgrade aims to elevate smartphone security, potentially making it one of the safest devices. Titan modules protect data, ensure system integrity, and handle encryption on RISC-V architecture.
VCs Hire AI Talent to Sharpen Bets
Venture capital firms are hiring AI specialists to better evaluate and maximize returns on AI investments. These new experts aim to leverage technology more effectively in deal sourcing and portfolio management.
Pentagon Escalates vs Anthropic
A New York Times Technology newsletter highlights a potential unprecedented escalation by the Pentagon against Anthropic, a U.S. AI company. This signals rising tensions between defense authorities and AI firms. Additional segments cover an AI agent slandering an individual and other chaotic AI developments.
AI Data Centers in Outer Space?
Massive data centers powering generative AI severely harm Earth's environment due to high energy use. The article proposes launching them into orbit as a radical solution. This concept aims to mitigate planetary impact from AI infrastructure.
G42-Cerebras Launch 8 ExaFLOPS Compute in India
UAE-based G42 has partnered with U.S. chipmaker Cerebras to deploy eight exaflops of compute via a new system in India. This initiative expands high-performance AI infrastructure in the region. It leverages Cerebras' advanced wafer-scale engine technology for massive AI workloads.
Kimi K2.5 Tops Opus on Pharma Hallucination Bench
Kimi K2.5 outperformed Opus 4.6 on the Placebo Bench hallucination benchmark in pharmaceuticals. Opus showed highest hallucination rate by inventing protocols. Dataset available on Hugging Face.
Tsinghua's AI Model Draws Deepest Cosmos Map
Tsinghua University's automation and astronomy teams used self-developed spatio-temporal self-supervised model Xingyan to create humanity's deepest deep-space galaxy image. This breakthrough surpasses astronomical observation depth limits. Results published in Science magazine.
Whisky-Sniffing Robot Dog Debuts
Dewar’s whisky warehouses in Scotland deploy a Boston Dynamics robot dog equipped with an ethanol sensor to detect leaky barrels. The robot patrols vast storage areas, acting as a mechanical sniffer for ethanol leaks from aging casks.
Samsung's Bixby Turns Conversational
Samsung launched an updated Bixby in One UI 8.5 beta, transforming it into a conversational agent that understands natural language tasks without exact commands. Examples include toggling screen timeout by description or suggesting fixes for pocket screen issues. It now integrates web search results directly and is initially available in Korea, Germany, India, Poland, UK, and US.
Claude PowerPoint Expands to Pro Users
Anthropic has opened Claude in PowerPoint to Pro subscribers. New connector support enhances integration capabilities. A double usage promo is also available.
Chunwan Robots Zero Errors: Real or Scripted?
Robots at the Spring Festival Gala (Chunwan) went viral due to flawless 'zero error' performances. The article questions if this reflects true AI capabilities or production editing. It contrasts pseudo-intelligence hype with weak real-world experiences.
Doppelgänger: OnlyFans Lookalike Search
Presearch launched Doppelgänger, a search engine to find OnlyFans models resembling a user's crush from photos. It promotes discovering real adult creators with consent. The tool counters nonconsensual deepfakes.
Cheapest Cybertruck at $59,990, Cyberbeast Reversal
Tesla announced a new all-wheel-drive Cybertruck model priced at $59,990, its cheapest yet. The company reversed a price hike on the Cyberbeast variant, returning it to early 2025 levels.
Gemini's Smallest Update: Unstoppable?
Google releases its smallest Gemini model iteration to date. Analyst Yao Shunyu hails it as 'unstoppable' despite minimal changes. It reportedly excels at super complex tasks.
2026 Red Envelope War: AI Apps Five Hegemons
Predicts 2026 Spring Festival red envelope battles will see Chinese AI apps enter a 'five hegemons' era like Spring and Autumn period. This year's fights were highly lively and competitive.
Intel Unveils Lenovo Pro 14 GT AI Test Video
Intel released the first online test video for Lenovo Xiaoxin Pro 14 GT AI Yuanqi Edition, highlighting its design and specs as the debut of Panther Lake platform. The series launches in March with Core Ultra 300 processors, exceptional battery life up to 92.5Wh/99.9Wh, and lightweight design at 1.43kg for the 14-inch model.