AI Weekly Malaysia

llama.cpp launched an official website at llama.app with a one-line install script, package manager support (Brew, Winget), and a local coding agent workflow via the pi-llama plugin that auto-discovers your served model with no API keys or config. The site lists supported hardware ranging from Apple Silicon to H100 clusters and highlights models like Qwen 3.6, Gemma 4, GPT-OSS, and Gemma 3.

Why: If you want a fully local coding agent with zero API costs and no data leaving your machine, the `llama serve` + `pi install git:github.com/huggingface/pi-llama` + `pi` flow is now a three-command path worth trying on your existing hardware before paying for an API-based agent.

The trycua/cua project published a blog post on using GPU passthrough in macOS VMs to accelerate llama.cpp-based LLM inference on Apple Silicon. The post details the technical approach for passing Apple GPU resources through to a virtualized macOS environment.

Why: If you run local LLM inference on Apple Silicon but need VM isolation for CI, agent sandboxes, or multi-tenant setups, this approach could let you keep near-native Metal GPU performance inside a VM rather than falling back to CPU-only inference. Builders evaluating local agent infrastructure should test whether the passthrough overhead is acceptable for their workload before committing to a bare-metal-only deployment.

South Korean security firm Genians reports that North Korean threat group Kimsuky is running local LLMs via Ollama, GPT4All, and Msty, and experimenting with Cursor and RAG for local document search, keeping operations off cloud services to avoid detection. The group uses AI to craft polished phishing lures about virtual assets and finance, delivered via ZIP archives containing malicious LNK files that run PowerShell loaders, with GitHub repositories serving as command-and-control infrastructure.

Why: Threat actors are now using the same local LLM toolchain (Ollama, GPT4All, Cursor) that many developers and AI tinkerers run, meaning AI-generated phishing lures will be increasingly convincing and harder to spot. Builders should scrutinize unfamiliar GitHub repositories more carefully since Kimsuky uses public GitHub repos as C2 infrastructure, and treat unsolicited ZIP/LNK files—even those referencing research or events—as high-risk.

South Korean security firm Genians reports that North Korean hacking unit Kimsuky has set up an offline AI stack on its own infrastructure using Ollama, GPT4All, and Msty, with GPT4All's LocalDocs RAG database (localdocs_v3.db) configured to query a private document collection. The group has not trained a custom model but is assembling existing tools to automate malware development and improve phishing lures, which will increasingly lack the traditional tells of stilted translation and formatting errors.

Why: If you build or operate systems that rely on users spotting phishing by its clumsy language or formatting, that defense layer is eroding—AI-generated lures will read naturally. Shift detection toward correlating technical indicators like LNK execution, PowerShell activity, hidden scheduled tasks, and unusual GitHub traffic rather than judging emails by polish. For builders running offline LLM stacks (Ollama, GPT4All), note that these same tools are now documented in nation-state offensive operations.

Mark Zuckerberg publicly criticized 'closed' AI competitors while positioning Meta as returning to open model releases. The article is behind a paywall, so concrete details on which models, timelines, or policy shifts are involved are unavailable from the extracted text.

Why: Without the full article text, there is no actionable detail to guide a decision on whether to bet on Meta's open models versus closed alternatives. The headline alone signals Meta's strategic positioning but does not specify what builders should change.