AI Weekly Malaysia

Ukraine reportedly reverse-engineered a Chinese drone design after covertly photographing a factory, producing a $48,000 flying-wing attack drone with 2,000 km range that destroyed a Russian Tu-95 bomber. Ukraine is now building 6,000 of these drones per year.

A critical macOS Screen Sharing vulnerability (CVSS 9.8) is being actively exploited in the wild to gain remote root access and deploy Monero cryptojacking miners. CISA has elevated the bug's severity, though the article body itself contains no technical details beyond the headline.

The title announces that Firefox for iOS now includes a native adblocker, but the article body is just a cached error page from Mozilla's support site with no actual content about the feature.

Anthropic's documentation page catalogs the core system prompts used in Claude's web interface (claude.ai) and mobile apps across model versions from Claude Haiku 3 (July 2024) through Claude Opus 5 (July 24, 2026). These prompts inject context like the current date and encourage behaviors such as Markdown code snippets, but do not apply to the Claude API. Starting with the Claude 4.6 generation, each model ID is a single fixed snapshot with one prompt entry.

Google is reportedly partnering with AMD to design its next-generation TPU, which may integrate on-package CPU cores specifically to accelerate reinforcement learning workloads. The article is primarily a rumor report with no official confirmation, technical specs, or timelines provided.

OpenAI's head of product design Ian Silber argues this is the best time to be a designer, noting that engineers have 10x'd with AI while design teams haven't. He shares his counterintuitive 'just do less' advice, discusses where humans still outperform AI (user understanding, invention, point of view), and outlines a vision of ChatGPT evolving into a super app.

A Google Scholar search for the exact phrase "kidney disappointment" reveals research papers that use this nonsensical term instead of "kidney failure," suggesting machine translation or AI-generated text artifacts have propagated into academic literature. The Hacker News post links to the Scholar query but the page itself returned a 403 error, so the full extent of the problem isn't documented in the source.

Intel VP Robert Hallock says the company will launch its new core architecture with Nova Lake on desktop first, rather than in the data center, and invites enthusiasts to compare performance-per-dollar against AMD. The article itself is mostly site boilerplate with no further technical detail beyond the headline.

Researchers have demonstrated 3D-printed micro drone fliers powered by ultrasonic sound waves, achieving 12,000 RPM silent hovering. The article itself is mostly Tom's Hardware site navigation boilerplate with minimal substantive detail beyond the headline claims.

A 3D-printing enthusiast designed and shared a 'Flock Sock,' a slip-on cover that blinds controversial surveillance cameras mounted on traffic lights and streetlights. The cover attaches to a broom handle so it can be applied from ground level without climbing. The design files have been publicly shared.

Tom's Hardware marks the 45th anniversary of the Intel 8088 processor, whose 8-bit external bus became the foundation for the IBM PC and 45 years of x86 architecture dominance. The article text provided is almost entirely site boilerplate with no substantive technical content.

A 10,000-hour longevity test found modern OLED panels are just as susceptible to burn-in as 2017 models, but they deliver roughly twice the brightness and 27% better efficiency, giving users more headroom to mitigate burn-in by running at lower brightness levels.

Tom's Hardware reviews the xTool M2, a color craft laser and engraver featuring a 10W or 20W diode laser, CMYK print head, 2W IR laser, and rotary tool options, with improved positioning cameras at a lower price point than its predecessor. The review awards it Editor's Choice.

Thieves replaced RTX 5080 GPUs with RTX 3060s in Chinese gaming hotel PCs, and the downgrades went unnoticed until a power blackout exposed the swap. The heist resulted in approximately $12,000 in losses.

A former Missouri Farm Bureau president is publicly offering his farmland to AI data center developers, arguing that a blocked $6.3 billion project will simply relocate to willing neighboring communities. He is pushing back against a wave of moratoriums across roughly 500 jurisdictions and 70% public opposition to data center construction in the area.

Corma, an AI defensive-security startup founded about a year ago by Alon Pluda, raised a $60M seed round led by Sequoia Capital with Khosla Ventures and Coatue. Pluda claims frontier models (Claude Opus 4.8, GPT-5.5, Grok 4.3, DeepSeek V4) are getting exponentially better at offensive security but lag at defensive tasks because most defensive work doesn't involve code scanning.

The article content failed to load — only Ars Technica's cookie consent and privacy policy boilerplate was captured. No substantive text about wildfire smoke, prenatal health, or air pollution is present.

A Ukrainian drone regiment defeated a 3,500-strong U.S. armored brigade combat team in a war game, with drones easily spotting and destroying tanks and armored vehicles at a high kill-rate that forced continuous respawns. The exercise exposed shortcomings in U.S. counter-drone response capabilities.

An iOS developer released an app that simulates carrier pigeon messaging, where virtual messages travel at a capped 100mph speed—meaning an LA-to-NYC message takes roughly 22 hours—and the virtual birds can be distracted or lost en route. The article text itself is almost entirely Tom's Hardware boilerplate with no additional technical detail beyond what the title states.

Raymond Chen, a Microsoft veteran, clarified on his Old New Thing blog that PR descriptions are point-in-time persuasive writing aimed at convincing approvers to accept a change, while code comments are durable documentation about how the code works. The Register notes the distinction is timely given the volume of PRs and comments now generated by AI coding tools, some producing questionable annotations.

Researchers created LittleLearner, a suite of language models (up to 5B parameters) trained exclusively on an 88B-token corpus filtered to a K-5 elementary school curriculum. They found that while scaling, post-training (SFT+GRPO), and in-context learning amplify in-scope abilities, none of these interventions allow the model to perform meaningfully on out-of-scope tasks, proving the pretraining data sets a hard capability ceiling.

Anthropic's frontier red team identifies coordination failures in emerging multi-agent systems, noting that agents work well when treating each other as tool invocations (defined inputs/outputs) but stumble when acting as long-lived peers with their own goals and no clear hierarchy. They warn that benign individual-level behavioral quirks like confabulation and reward hacking can compound into unexpected systemic failures at scale, and that agent-agent interaction volume could exceed human-human interaction before anyone understands how to make it safe.

A reflection arguing that AI agent harnesses have crossed the 'can it be done' threshold but that software engineering fundamentals—how code fits together, API design, testability—matter more than ever. The author notes open-weight models are closing the capability gap fast on local hardware, and shares a concrete practice: prompting agents with 'develop with red/green TDD' to get testable output, though code seams above that level remain subjective art.

David Horn argues that software engineers repeatedly reinvent concepts from older disciplines—statistics became 'data science,' crypto speedran finance history, Silicon Valley reinvented bus stops—without studying the originals. He sees the same pattern repeating with AI agents, where engineers are intuiting that managing multiple agents resembles managing engineering teams, but are not consulting existing management literature.

A woman identified as Jane Doe 4 has joined a lawsuit against xAI (now described as part of SpaceX) alleging her stepfather used Grok to transform a childhood photo taken when she was 11 into over 7,000 explicit images. The stepfather died by suicide two days after the images were found in a law enforcement raid. The suit, originally filed by three Tennessee teenagers, seeks class action status and accuses xAI of failing to take basic precautions to prevent Grok from generating explicit images of real people, including minors.