GLM 5.2 Outperforms Claude in Cybersecurity Benchmarks
An open-weight Chinese model just outperformed Claude Code on IDOR vulnerability detection at $0.17 per bug. Here's what the Semgrep benchmark actually means ...
HackerRank ATS Open Source: The Hiring Rubric Inside
HackerRank open-sourced its AI hiring agent, and the scoring rubric is not what developers expected. ...
AI Cheating in Technical Interviews: 61% Still Pass
Fabric analyzed 19,368 technical interviews: 48% of tech candidates cheat with AI. 61% pass anyway. ...
codebase-memory-mcp Cuts AI Agent Token Use by 121x
codebase-memory-mcp hit #1 on GitHub trending today. The free MCP server that builds a knowledge ...
LineShine Tops TOP500 — China’s Exascale Without US Chips
China's LineShine supercomputer dethroned El Capitan at ISC 2026 using 2.45 million Huawei ARMv9 cores ...
LibrePods Unlocks AirPods on Android: Lock-In Exposed
LibrePods breaks Apple AirPods open for Android and Linux — ANC, ear detection, hearing aid features, all unlocked. One config line exposes deliberate ...
Kimi K2.7-Code: Open-Source 1T Coding Agent, 30% Fewer Thinking Tokens
Moonshot’s Kimi K2.7-Code improves on K2.6 benchmarks and cuts thinking tokens 30%. Open weights, $0.75/M ...
pedit COW CVE-2026-46331: Any Linux User Can Get Root Now
CVE-2026-46331 (pedit COW) lets any local user root Linux via page cache poisoning. Public PoC ...
Snyk Evo ADS Goes GA: 1 in 12 MCP Servers Has a Critical Finding
Snyk scanned 10,000 dev environments: 1 in 12 MCP connections has a critical finding, 36% ...
Microsoft Drops Claude Code: Enterprise AI’s Budget Reckoning
Microsoft cuts Claude Code for 5,000 engineers on June 30. Uber, Meta, and Accenture hit ...