Open source news — new project releases, community governance, licensing updates, and open-source AI models including Llama, Mistral, and others.
"Copy Fail," a dangerous Linux local-privilege escalation vulnerability that exploits a kernel memory corruption flaw to provide a controlled 4-byte write into the Linux page cache. This allows attackers to rewrite cache…
Forge is a reliability layer for self-hosted LLM tool-calling that uses guardrails and context management to dramatically improve performance on multi-step agentic tasks, boosting an 8B model from 53% to 99% accuracy. It…
The article announces the launch of a new, open-source forum software called "bsBB" that combines a nostalgic, old-school forum structure with modern features. It replaces traditional user authentication with Bluesky acc…
On May 11, 2024, TanStack suffered a sophisticated supply-chain attack that compromised 42 npm packages and published 84 malicious versions within six minutes. The attackers exploited GitHub Actions cache poisoning and u…
Security vulnerabilities found in Epsilon, a WebAssembly (WASM) runtime written in Go. AI agents discovered over 20 security flaws, including sandbox escapes that allowed malicious WASM modules to break isolation and acc…
Based solely on the provided text, the article's title and body do not contain a factual summary of SAP's AI strategy. The body consists entirely of a list of unrelated article headlines and event promotions. Therefore, …
User-space security agents are structurally flawed because they share the same privilege level as the processes they monitor, allowing attackers with root access to simply kill the agent or erase logs. It explains that e…
Armorer v0.1.19, an experimental local control plane designed to manage the operational challenges of running AI agents, such as configuration drift and failed runs. Unlike an agent framework, Armorer functions as a loca…
Using GitHub Copilot CLI as a triage tool to manage open-source contributions across over 40 repositories. The workflow relies on four key prompts: explaining a repo's architecture, locating specific files, explaining fu…
Tool-using research agent built for the Gemma 4 DEV Challenge, which runs locally on the Gemma 4 e2b model via Ollama using roughly 200 lines of Node.js code. The agent accepts a question, selects between two tools to re…
The author successfully ran Google's new Gemma 4 open models (2B and 4B parameters) on a laptop using Ollama during Google I/O 2026 week. To make the small 2B model function as a useful research agent, they built a five-…
To measure and monitor prompt cache hit ratios in Hermes Agent to reduce costs, as cache-breaking significantly increases expenses. It introduces a Python library called cachebench that wraps model client calls to track …
Fully local AI system using Llama 3 and LangChain to enable natural language querying of an internal SQLite database, eliminating any data being sent to third-party servers and avoiding legal compliance issues. The syste…
Hermes Agent is an open-source agent framework developed by Nous Research that can run on various infrastructures, from a low-cost VPS to a GPU cluster. Unlike most agent frameworks, it is self-improving, learning from e…
Relying on autonomous OSINT (Open Source Intelligence) AI agents, which are gaining popularity in Japanese security communities, leads to "skill atrophy" among human analysts. It describes a "Skeleton Implementation" pro…
The author, an AI professional, built an AI narrative platform called WYRD in about a month for roughly $467 using the OpenClaw agent framework, despite not being a developer. The platform allows users to interact with m…
The article investigates whether increasing an AI model's "thinking time" through additional recurrent loops during inference improves its performance, using a small-scale OpenMythos model trained on multi-digit addition…
PyTorch's `torch.load()` function uses the Python pickle format for serialization, which inherently executes arbitrary code when loading a file, making it trivially exploitable for remote code execution (RCE) attacks. Th…
Choosing between Make.com, n8n, and custom Python agents should be based on a team's automation maturity level rather than a feature comparison. It recommends Make.com for non-technical teams with simple SaaS workflows, …
Semble Cloud, a tool that adds semantic code search to Claude Code via a single command line, requiring no setup, API keys, or configuration. It wraps the open-source Semble semantic search engine into an MCP endpoint, a…