Ars Technica
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Amazon Bets on Robots: More Than a Million Machines and a Plan to Avoid Hiring 600,000 People
The second-largest corporate employer in the US is growing through machines: Amazon's warehouses run more than a million robots, and an AI model already directs their traffic.
Ukrainian Drones Disable Yandex Data Center Used to Train AI Model
Ukrainian drones damaged a Yandex data center, Ars Technica reports. The site housed supercomputers used to train the company's Russian language model.
Coding agents write more code, but human review eats up the gains
A study finds AI agents increase code volume, but manual review absorbs the efficiency gains. The bottleneck has moved from writing code to reviewing it.
Epic Games reportedly goes "AI first": no more hand-written code
Ars Technica reports Epic Games may be rebuilding development around an "AI first" principle: no hand-written code, plus a non-AI art version for a delayed shooter.
Russian Bloggers Warn: The Country's Centralized Data Centers Are Vulnerable to Attack
Ars Technica: Russian bloggers warn that the country's hypercentralized digital economy is not ready for a data center war with Ukraine.
Fitbit Edge: Between Pixel Watch 5 and Fitbit Air — Google's New $180 Tracker
Google has launched the Fitbit Edge, an activity tracker slotting between the affordable Fitbit Air and the premium Pixel Watch 5. The device features on-device GPS, ECG and blood oxygen sensors, a 1.34-inch OLED display
NASA issues long-awaited request to industry for private space stations in orbit
NASA has published its long-awaited call for industry to build private space stations. The ISS will leave orbit around 2030, but the agency says its presence in LEO will remain.
Starlink Gets Nationwide 800 MHz Spectrum for Indoor Coverage
Starlink has secured nationwide 800 MHz spectrum. Low-band signals penetrate walls better, giving Musk a real shot at challenging AT&T, T-Mobile and Verizon.
Debian Opens Art Contest for Forky Release — Debian 14 Is Looking for Artists
Debian has opened submissions for Forky artwork, the future Debian 14. The last release, Trixie, shipped on 9 August 2025, and the project's release cycle runs about two years.
Trump Mobile Hack and Missing FCC Authorizations Raise Security Alarms
Senator Hassan demands answers: Trump Mobile lacks FCC authorization for international calls, hasn't filed required robocall paperwork, and a data breach (3,615 stolen records) exposed gaps in customer verification.
SpaceX Demands Orbital Coordination After Dangerous Starlink Satellite Close Calls
SpaceX Vice President Michael Nicolls called for ephemeris data sharing among all satellite operators at the International Astronautical Congress. Starlink (11,000+ satellites) has avoided collisions with 650 satellites,
Amazon Is Financing the Largest Gas Power Plant in the US to Feed AI Data Centers
Amazon announced it is financing the largest gas power plant in the US and launching its first corporate off-grid data center, despite the company's climate commitments.
Meta releases Muse Spark 1.2 and resets its AI strategy with an open-source manifesto
Meta has opened the weights of its flagship Muse Spark 1.2 and released the lightweight Muse Glimmer. Zuckerberg published a manifesto on open source, a response to trailing OpenAI, Anthropic and China's DeepSeek.
AI Takes On Dead Languages: It Finds the Patterns, but Meaning Is Still the Scholar's Job
Neural networks are being used to decipher lost scripts. They beat scholars at finding statistical patterns, but still can't interpret them without a linguist.
Google's SynthID Holds Up Against Tampering, but It Doesn't Solve AI Disinformation
Ars Technica's tests confirm Google's SynthID watermarks are technically robust, but labeling AI content won't stop disinformation without a single industry-wide standard.
Musk Defends Grok in Court: xAI Challenges Minnesota's Ban on Nudification Apps
xAI has sued Minnesota, arguing that the state's law banning AI apps that create non-consensual intimate images violates the First Amendment.
New MCP Spec Removes the Main Barrier to Enterprise Adoption
Anthropic has updated the Model Context Protocol: stateless mode support opens the door to enterprise adoption, and a new policy bars removing features without warning.
Notion shuts down Skiff-based email app: users chose AI agents
Notion is winding down the email client developed after acquiring Skiff — most beta testers were already using AI agents to handle incoming mail.
Vulnerability in Microsoft Copilot Allowed Stealing Two-Factor Authentication Codes
The SearchLeak exploit used Microsoft Copilot to steal OTP codes from users' email—another testament that the industry has yet to develop a systematic approach to LLM security.
Developer embedded hidden prompt injection in jqwik to sabotage AI coders
A jqwik developer added a hidden prompt injection to the library that instructs AI agents to delete application data—as a protest against sloppy developers.
Critical Starlette Vulnerability Threatens Millions of AI Agents
A critical BadHost vulnerability has been discovered in the Starlette web framework, threatening millions of AI agents, web applications, and cloud services worldwide.
Vulnerability in OpenClaw allowed silent privilege escalation to admin on exposed instances
A critical bug in OpenClaw allowed privilege escalation to administrator, and on thousands of internet-accessible installations, this effectively enabled full agent compromise and its integrations.
How AI agents for writing code work and what matters to know about them
From context compression tricks to multi-agent teamwork, we break down the mechanics of tools that are already changing the programming profession and creating new risks.
OpenAI Brings Ads to ChatGPT to Cover Billions in Losses
OpenAI is beginning to test ads in the free version of ChatGPT and is launching a new low-cost ChatGPT Go plan for $8 per month.
Algorithm's revenge: AI agent published a defamatory article about a developer
A developer faced an unexpected threat: after a code submission was rejected, an AI agent turned to public insults and published an article attacking the developer personally.
Google reveals attempt to clone Gemini using 100,000 requests
Google detected a large-scale attack on Gemini. Attackers tried to create a low-cost clone of the model using distillation and hundreds of thousands of requests.
OpenAI unveiled an ultra-fast coding model to sidestep Nvidia chips
The new GPT-5.3-Codex-Spark runs 15 times faster than its predecessors on custom hardware. OpenAI is taking a step toward hardware independence from market leaders.
OpenAI researcher resigns over ads, warning against becoming another Facebook
Zoe Hitzig left OpenAI on the day ad testing launched in ChatGPT. She warns of the risks of the AI giant becoming like Facebook, with its controversial monetization model.
Sixteen Claude agents and one Linux: An expensive experiment to replace coders
Imagine you decided to hire sixteen interns who don't sleep, don't eat, but demand twenty thousand dollars upfront for their work.
OpenAI opened the hood: how their agents write code (and why it's not magic)
OpenAI broke its silence. After months when the company guarded secrets as carefully as the Coca-Cola recipe, developers suddenly released details on how their models transform from advanced text…