Claude Code isn’t a note-taking app. It’s an AI agent built for code — and yet, in a real-world test across a sprawling Obsidian vault, it outperformed both VS Code and a local LLM at turning fragmented thoughts into linked, filed, context-aware notes. That’s the core finding of XDA Developers’ hands-on comparison published Thursday evening, not a lab benchmark, but a workflow stress test on a live knowledge base full of daily notes, wikilinks, and unstructured text. No new product launch. No pricing. No Indian or US release date. Just one developer’s verdict: Claude Code, with minimal setup, handled note capture and cross-note linking more reliably than tools designed for the job. And it did so by treating notes like source code. Reading the folder structure, reverse-engineering naming conventions, and writing to files in Obsidian’s native syntax without plugins or translation layers. The catch? It writes more than you ask for, explanations, suggestions, formatting. Unless you explicitly constrain it in a rules file. That’s not a bug. It’s how Anthropic’s agent layer works: reasoning first, editing second. VS Code stayed fast and precise for search and review, but offered zero automation for capture or linking. LM Studio with Qwen 3.5 9B ran locally and kept data private, but choked on vault size, hitting 20k tokens of its 30k context window just parsing a small sample, and failing to search inside file contents, not just names. So while Claude Code doesn’t replace Obsidian or Notion as a UI, it’s now the only tool in this trio that can autonomously administer a note system. Filing, linking, auditing broken references, without human intervention. That changes the bar for what ‘AI-powered productivity’ means on desktop. Next step? Wider testing against larger vaults, and whether Anthropic’s upcoming MCP integrations will tighten control over output verbosity. For now, it’s not about replacing your note app. It’s about offloading the admin.