AMD Releases Fully Open MoE with Complete Training Artifacts
AMD's Instella-MoE-16B-A3B stands out for publishing weights from every training stage, data mixtures, configs, and inference code alongside a...

Created by Sahil Panhotra
Fast‑rising GitHub projects across all languages, sourced from official Trending and tech news
Explore the latest content tracked by GitHub Trending Pulse
AMD's Instella-MoE-16B-A3B stands out for publishing weights from every training stage, data mixtures, configs, and inference code alongside a...
npm restricted granular access tokens configured to bypass 2FA from performing account, organization, or package management operations effective July...
Community repos tackle Claude Code's session memory gaps and project quirks, with awesome-claude-code serving as the central index for commands,...
A broad audit of popular open-source AI projects uncovers widespread security gaps.
MemoryCustodian solves context retention for agents like Codex and Claude by storing decisions and constraints as plain Markdown inside the repo, with...
KotlinLLM introduces smart macros that turn simple function calls into LLM-generated Kotlin source, compiled and hot-reloaded at runtime via JDI with...
OpenSpec keeps AI coding agents aligned across sessions by storing intent in version-controlled proposals, delta specs, designs, and tasks rather than...
YourAIProject lets you browse 135 weekly snapshots of GitHub Trending, compare historical ranks, and spot long-term open-source patterns. Snapshots...
GitHub's June 2026 updates directly counter npm and Actions attack patterns seen in 2025-2026 incidents like Shai-Hulud and tj-actions. Read-only...
GitHub counted 4.3 million AI repos in 2025 with a 178% YoY jump in LLM projects, driven by demand for multi-agent coordination beyond single-model...
Keychron is extending its open-source push into gaming mice via ZGM firmware on Zephyr, launching Q1 2027, explicitly calling out industry firmware...
Trending Dockerfiles highlight wide-ranging interests in autonomous systems and infrastructure:
Run Kimi K3 delivers a 2.8T-parameter MoE LLM on a single Apple Silicon Mac, streaming MXFP4 experts on demand into local cache with fused NEON/Metal...
Fonderie stands out as a trending open-source backend skill for AI coding assistants, stopping LLMs from reinventing auth, billing, and permissions every session.
Buzz stands out on GitHub Trending as a self-hostable Nostr-based workspace where humans and AI agents share rooms, keys, and full audit...