I'm building a personal Second Brain in Obsidian, using AI coding assistants to help design and build the tools around it. It started as a place to keep notes. It has become the system I run my day from. Ideas land there, tasks move forward, and what I learn gets turned into something I'll remember.
One page for the day. Today's note brings together tasks, quick thoughts, links and voice memos. Completed and discarded items move into their own sections at the bottom instead of being deleted, so the day stays readable and nothing is lost. A day can be formally closed out, and navigation skips straight to the days that are still open.
From capture to action. AI helps turn loose capture into structured tasks that link back to where they were written. A single keyboard shortcut opens a picker that moves a task between active, paused, blocked, refined, handed to an AI agent, and done. A short "frog" list keeps the day's most important tasks at the top. Due dates grow more visible as they approach. A floating "Up next" list holds short routines without cluttering the day.
Working with Claude Code. Claude Code works alongside the Markdown vault, helping me turn rough ideas into scoped briefs, build the small tools that support my routines, and review the results. A task can be expanded into a written brief and handed to a coding agent. It stays marked as "away with an agent" until I review what came back. The note is the shared context, rather than an idea disappearing into a chat history.
What that looks like in practice. I can drop a half-formed project idea into my daily note, keep a link to its source, and later turn it into a next action or a focused brief. If it needs more thought, it stays paused rather than being forced into today's priorities. A learning question can become a note first, then a reviewed flashcard candidate. Capturing an idea and deciding what to do with it are deliberately separate steps.
Illustrative workflow: capture an idea → clarify the next action → write a brief with Claude Code → review the result → keep the useful learning.
Learning built in. Any line can become a flashcard candidate through a small form. Candidates collect in a learning home base, where I review them before they reach Anki. Voice recordings, saved YouTube videos and LinkedIn posts arrive in the vault automatically as notes I can process later.
Knowing what to trust. Notes carry a confidence level, from rough draft to confirmed, so the AI tools I work with know what to rely on and what to treat as unverified.
Under my control. It is all plain Markdown files with full version history. AI suggests and builds, and I review every change before it becomes part of the system.