Open Source

Artificial Intelligence

Cybersecurity

Cloud Web Apps

Digital Health

Research & Development

Linux Services & Systems

CMOS: Content Management Open Source

A fully Open Source content-management platform for multi-channel and multi-device publishing, built with Bassilichi S.p...

Extra-EU Internationalisation: Swiss Market

noze's internationalisation programme towards the extra-EU market, with the opening of a Geneva office to oversee projec...

PhantomRaven: the npm malware that collects bug bounties

CrowdStrike has analysed an npm campaign that uses remote dependencies declared as URLs to keep the malicious code out o...

Jev and System One models: typed decisions

TypeSafe AI's Jev returns typed decisions instead of text: how a request is built, what confidence actually means, the l...

Cyber Resilience Act: the 24 hours start today

From 11 September 2026 Article 14 of Regulation (EU) 2024/2847 applies: early warning within 24 hours, notification with...

Human Atlas: 2,234 anatomical meshes and two licences

A 3D anatomy explorer running entirely in the browser, with no accounts or API keys, offering 2,234 individually selecta...

Agentic reverse engineering: x64dbg exposed over MCP

A native plugin in Zig exposes x64dbg over MCP with 80 tools, from disassembly to conditional breakpoints through OEP de...

Anthropic releases commerce-agents: two agents, Apache-2.0

On 1 September Anthropic published commerce-agents, an Apache-2.0 blueprint with two agents, one for customers and one f...

NVIDIA acquires Hugging Face: commitments and dependencies

On 3 September NVIDIA announced the acquisition of Hugging Face for 12,930,300,000 dollars. The announcement lists preci...

GLM-5.3: the weights, the licence and the benchmark method

On 25 August Z.ai published the GLM-5.3 weights, three days ahead of the deadline it had set itself. What arrived with t...

DeepSeek Harness: plugins, core and capability seams

DeepSeek opened the Harness repository on 13 August, a TypeScript agent runtime under the MIT licence, built on Cordis. ...

Local inference on DGX Spark: the recipes for DeepSeek V4 Flash, GLM-5.2 and Qwen3.6

On one NVIDIA DGX Spark, DeepSeek V4 Flash runs at 26 tokens per second and Qwen3.6-35B in NVFP4 at 81; on two units Dee...

MCP 2026-07-28: a stateless protocol, MRTR and the deprecation of Roots, Sampling and Logging

The Model Context Protocol 2026-07-28 revision drops the initialize handshake and the Mcp-Session-Id header: every reque...

Qwen3.8-Max: a 2.4-trillion MoE and the benchmarks it claims

Qwen3.8-Max is a Mixture of Experts with 2.4 trillion parameters and context up to one million tokens, accessible from 3...