AI agent Codex migrates 1000 notes into Obsidian without data loss

Switching between note-taking applications is notoriously difficult due to the lack of direct bulk transfer options, especially with Apple Notes. Facing the challenge of consolidating eight years of notes from Apple Notes, Evernote, and Google Keep into a single Obsidian vault, an XDA writer decided to leverage artificial intelligence rather than handling the tedious task manually.

• Approximately 1,000 notes were processed and imported from three different apps in about ten minutes.
• The AI agent automatically sorted notes into subject-specific folders and applied proper tags.
• A detailed report was generated at the end, flagging files that could not be reliably categorized.

Leveraging Codex over standard import plugins

While Obsidian offers an excellent Importer plugin, it lacks the ability to automatically sort, categorize, and tag massive amounts of legacy content. To bridge this gap, the user turned to Codex, ChatGPT's agentic model, which imported the notes directly into the vault without dropping a single file.

The manual effort was limited to exporting the raw files from their respective sources—using the Exporter app for Apple Notes, Google Takeaway for Keep, and manual steps for Evernote. Once the data was prepped, Codex handled the heavy lifting of interpreting the text, organizing folders, and tagging documents, saving hours of manual cleanup.

Sorting accuracy and handling ambiguous notes

Over years of note-taking, context is easily lost, but Codex successfully interpreted the vast majority of entries and organized them accordingly. While it wasn't completely infallible—such as occasionally grouping a forgotten story idea with computer settings—its overall accuracy was remarkably spot-on.

Out of 1,000 notes, Codex was unable to categorize 83 files, which upon review turned out to be total gibberish, old passwords, or random snippets like a Skyrim puzzle. The user appreciated the AI's honesty rather than having it force incorrect categories into the vault.

Implications for users and agentic AI workflows

This experiment highlights the immense practical value of agentic AI in handling administrative and personal data management tasks. Tools like Codex, Claude Code, or even privacy-focused local LLMs demonstrate how artificial intelligence can streamline personal knowledge bases efficiently.