Nous Research's Hermes Agent Adds Multi-Agent Kanban Workflow

Nous Research's Hermes Agent Adds Multi-Agent Kanban Workflow

Nous Research dropped Hermes Agent v0.12.0 "The Curator Release" this week, introducing multi-agent collaboration through Kanban-style task management [4]. The open-source AI agent now supports parallel work where multiple agents can claim tasks, hand off responsibilities, and share workspaces with live dashboards and SQLite storage for persistence.

What sets Hermes apart is its self-improving architecture with a built-in learning loop that creates new skills from experience [5]. The demo video gained serious traction with 3,800+ likes, highlighting growing interest in autonomous agent workflows that can actually coordinate and learn rather than just execute single tasks [6].

LLM Wiki Turns Documents into Evolving Knowledge Graphs

A new open-source desktop app called LLM Wiki automatically transforms PDFs, documents, and web content into interconnected Markdown knowledge bases with AI-driven gap-filling research [7]. Released just three days ago, it's inspired by Andrej Karpathy's LLM Wiki concept and integrates seamlessly with Obsidian for visual knowledge exploration.

The tool represents a shift toward AI as knowledge engineer rather than just search assistant [8]. Instead of static document storage, it actively builds connections, identifies knowledge gaps, and suggests research directions — essentially creating a living, breathing knowledge base that evolves with your content [9].

GraphRAG Shows 30% Performance Jump Over Vector Search

New analysis confirms GraphRAG's superiority for complex queries, achieving 80% accuracy on multi-hop questions compared to 50% for traditional vector RAG — a substantial 30% improvement [10]. The approach excels at relational and global questions where understanding connections between entities matters more than semantic similarity alone.

The trade-off remains significant: GraphRAG indexing costs can run 100x higher than vector approaches [11]. However, practitioners in 2026 increasingly view it as essential augmentation rather than replacement for vector search, particularly in knowledge management systems where relationship understanding drives value [12].

What This Means For Your Meetings

These developments point toward a fundamental shift in how we capture and leverage professional knowledge. xAI's voice cloning and Hermes Agent's multi-agent workflows suggest we're moving beyond simple transcription toward systems that can synthesize, coordinate, and act on meeting insights. Imagine AI agents that not only transcribe your discussions but can clone participant voices for follow-up clarifications or spawn specialized agents to research action items mentioned during calls.

The GraphRAG performance gains are particularly relevant for meeting intelligence platforms. Traditional keyword search falls short when you need to understand how decisions from different meetings connect, or when tracking how a project discussion evolved across multiple stakeholders over time. Knowledge graphs excel at these relationship-heavy queries that define real professional work.

LLM Wiki's approach to evolving knowledge bases mirrors what meeting platforms should become — not static archives but living systems that identify knowledge gaps, suggest connections between past discussions, and proactively surface relevant context. The goal isn't just better search; it's transforming your meeting history into an active knowledge partner that grows smarter with every conversation. Key takeaway: Meeting intelligence is evolving from passive recording to active knowledge synthesis, with AI systems that understand relationships, coordinate follow-ups, and continuously improve their understanding of your work context.

Sources

1. https://x.com/xai/status/2050355373052223585
2. https://www.basenor.com/blogs/news/xai-launches-voice-cloning-via-api-80-voices-28-languages?srsltid=AfmBOoomki9EP-bx-A_ukypI9iguAB68v4mkyIA1JNdZtZD5_CANUzdT
3. https://venturebeat.com/technology/xai-launches-grok-4-3-at-an-aggressively-low-price-and-a-new-fast-powerful-voice-cloning-suite
4. https://github.com/nousresearch/hermes-agent
5. https://hermes-agent.nousresearch.com/
6. https://www.linkedin.com/posts/nousresearch_hermes-agent-v0120-the-curator-release-activity-7456036098146844672-nM1y
7. https://github.com/nashsu/llm_wiki
8. https://www.reddit.com/r/ObsidianMD/comments/1shntdn/new_plugin_llm_wiki_turn_your_vault_into_a
9. https://aimaker.substack.com/p/llm-wiki-obsidian-knowledge-base-andrej-karphaty
10. https://medium.com/ai-engineering-simplified/graphrag-vs-vectorless-rag-vs-vector-rag-a-2026-guide-to-advanced-context-engineering-e8e9264cab38
11. https://flur.ee/fluree-blog/graphrag-vs-vector-rag-when-knowledge-graphs-outperform-semantic-search
12. https://medium.com/graph-praxis/graph-rag-in-2026-a-practitioners-guide-to-what-actually-works-dca4962e7517