XNews

Open-source AI projects and curated resource rundowns

Open-source AI projects and curated resource rundowns

Curated open-source AI projects and resource roundups for 2025–2026. - Latent Space: 2025 must-read papers (123 papers, 800MB download) - Awesome-From-Video-Generation-to-World-Model maps video→world-model evolution - Awesome LLM Skills catalog standard LLM workflows and templates - Open Scouts provides continuous monitoring AI 'scouts' (Next.js + Supabase) - Fincept Terminal: open financial analysis terminal with 100+ data connectors - OpenCode claims easier onboarding for Anthropic users - NVIDIA opened GLM-4.7 and minimax-m2.1 APIs free for registered users - Dolphin: ByteDance document parser with two-stage parsing and v2 (3B) model - llm-madness: end-to-end LLM training UI and visualization - NanoBanana Studio: natural language image editing tool - Second Brain AI Assistant: course for building personal RAG agents

Key insights

Comprehensive reading lists and paper collections surfaced for 2025

Latent Space compiled a 2025 'must read' paper list and reportedly downloaded 123 papers (≈800MB) for community use. Other GitHub curations map domain research—e.g., video generation to world models—helping practitioners track the fast-moving literature.

Open-source tooling targets agentic and document workflows

Projects like Awesome LLM Skills, Open Scouts, Dolphin (ByteDance), and Second Brain AI Assistant provide reusable skills, continuous monitoring agents, robust multimodal document parsing and RAG course material, respectively. These repos include templates, connectors and deployment guidance to accelerate production use.

Industry players and vendors release accessible models and APIs

NVIDIA opened free APIs for GLM-4.7 and minimax-m2.1 to registered users. Community open-source terminals (Fincept), image editors (NanoBanana Studio), and training UIs (llm-madness) lower the barrier for domain-specific model experiments and deployment.

Source posts