All Updates
Page 48 of 679
April 7, 2026
Apple Studio Display XDR FDA-Approved for Radiology
Apple's Studio Display XDR Medical Imaging Calibrator feature has received FDA clearance for diagnostic radiology. It will officially launch this week for US users. This enables radiologists to view diagnostic-grade medical images directly on the display.
Galaxy XR Turns 2D Apps into 3D
Android XR launches auto-spatialization, converting 2D apps, websites, images, and videos into 3D experiences on Samsung Galaxy XR headsets. The experimental feature rolls out Tuesday, demonstrated in a Google video. It supports up to 1080p at 30fps but uses more battery and works only on the focused app window.
Apple Foldable iPhone On Track for September
Mark Gurman reports Apple's first foldable iPhone is progressing smoothly toward a September release. It targets the standard fall iPhone cycle and may launch with or slightly after iPhone 18 Pro series. This confirms Apple's entry into foldables as planned.
AWS S3 Files Mounts Buckets for AI Agents
Amazon launches S3 Files, enabling AI agents to access S3 buckets as a native file system via a single command, without data migration. It leverages EFS technology to provide full file semantics on object storage. Now available in most AWS regions, it boosts tools like Kiro and Claude Code.
1B+ Tokens/Day on 2x H200 with GPT-OSS-120B
Research lab serves over 1B tokens/day locally using GPT-OSS-120B on 2x H200 GPUs via vLLM and LiteLLM stack. Achieves ~220-250 tok/s decode throughput, outperforming other models tested. Setup includes Docker, PostgreSQL, Prometheus for monitoring.
NVIDIA Blackwell Rack-Scale AI Supercomputers
NVIDIA's GB200 NVL72 and GB300 NVL72 are rack-scale supercomputers powered by Blackwell architecture, featuring 18 tightly coupled compute trays, massive GPU fabrics, and high-bandwidth networking. The post addresses challenges for AI architects and HPC operators in optimizing infrastructure beyond basic hardware setup. It covers topology-aware scheduling for efficient AI workloads.
Anthropic Allies with Rivals on AI Cybersecurity
Anthropic launches Project Glasswing, partnering with Apple, Google, and over 45 organizations to bolster AI cybersecurity. The initiative uses the new Claude Mythos Preview model to test and advance defensive capabilities against AI hacking threats.
ICML Reviewer Final Justification Notifications?
A Reddit user asks if ICML authors are notified when reviewers add final justifications to their original review comments. The post seeks clarification on the conference review process. Discussion is hosted on r/MachineLearning.
Chrome Vertical Tabs and Reading Mode Out
Google Chrome officially rolls out vertical tabs. Includes new Reading Mode for focused reading. Guide covers opt-in steps and usage tips.
Pimco Eyes $14B Debt for Oracle Data Center
Pacific Investment Management Co. is in talks with Bank of America to provide about $14 billion in debt financing. The funds will support construction of a massive Oracle data center in Michigan.
Intel Joins Musk's Terafab AI Compute Fab
Intel partners with SpaceX, Tesla, and xAI to build Terafab fab in Austin for 1 TW/year AI compute. This leverages Intel's chip expertise to scale AI and robotics power. Musk's AI refocus includes xAI acquisition by SpaceX.
Intel Joins Musk's Terafab Chips Project
Intel has signed on to Elon Muskβs Terafab chips project. The partnership infuses semiconductor expertise into Musk's latest high-tech endeavor.
Z.ai launches open-source GLM-5.1 beating Opus, GPT on SWE-Bench
Z.ai released GLM-5.1, a 754B parameter open-source MoE LLM under MIT license on Hugging Face, capable of autonomous work for up to 8 hours. It outperforms Claude Opus 4.6 and GPT 5.4 on SWE-Bench Pro via a 'staircase pattern' optimization. The model handles 1,700 steps and thousands of tool calls without strategy drift.
Anthropic's Mythos: Cybersecurity AI Reckoning
Anthropic has unveiled its new AI model, Mythos, claiming it marks a 'reckoning' for cybersecurity. The company is delaying public release and partnering with 40 companies to test its potential in preventing cyberattacks.
Anthropic Tests Mythos Model with Apple, Amazon
Anthropic is granting access to its more powerful, unreleased Mythos AI model. Apple and Amazon are testing it to prepare for potential cyberattacks from wider AI availability.
Anthropic Previews Mythos AI for Cyber Defense
Anthropic has debuted a preview of its powerful new AI model Mythos. The model launches within a cybersecurity initiative for defensive work by select high-profile companies.
Anthropic AI Finds Vulns in All OS/Browsers
Anthropic debuts Claude Mythos Preview in Project Glasswing cybersecurity partnership with Nvidia, Google, AWS, Apple, Microsoft. The model flags vulnerabilities autonomously in every major OS and browser. Not planned for public release due to security concerns.
Niantic Spatial Launches Global 3D Mapping Platform
Niantic Spatial, spun out from PokΓ©mon GO creator Niantic, is launching a revamped Scaniverse platform and a new global visual positioning system. This represents the company's biggest release since becoming independent last year. The platform extends AR expertise into physical AI applications.
Higgsfield AI series viewer-voted
Higgsfield launched its first AI-generated pilot, Arena Zero, with a slate of concepts. Viewers can vote to decide which shows progress to full series production.
Google Updates Gemini Mental Health Safeguards
Google has redesigned Gemini's crisis hotline module with a one-touch interface for texting, calling, or chatting with human agents or visiting the 988 website. The update ensures help options remain visible throughout conversations and adjusts responses to prioritize human connections, avoid validating harmful behaviors, and distinguish subjective experiences from facts. This follows a lawsuit accusing Gemini of contributing to a user's suicide.