---
type: concept
topic: agent-wiki
status: active
created: 2026-04-06
updated: 2026-04-06
source_notes:
  - "[[research/agent-wiki/sources/karpathy-llm-wiki]]"
  - "[[research/agent-wiki/sources/amp-how-to-build-an-agent]]"
derived_from:
  - workspace-initialize
confidence: medium
shared_concept:
---

# Obsidian as Agent IDE

## Definition

Obsidian acts as the human-facing interface for browsing, linking, and reviewing the knowledge base, while the LLM behaves more like the programmer or maintainer that edits the underlying markdown system.

## Why It Matters Here

This concept turns the vault from passive storage into an active working environment. It clarifies the division of labor: the human chooses sources, questions, and judgments; the LLM handles synthesis, bookkeeping, and structural updates inside the workspace.

## Supporting Sources

- [[research/agent-wiki/sources/karpathy-llm-wiki]] explicitly frames Obsidian as the IDE and the wiki as the codebase.
- [[research/agent-wiki/sources/amp-how-to-build-an-agent]] reinforces the idea that useful agent behavior comes from straightforward loops and tools, which matches an edit-review workflow in markdown files.

## Tensions

- If too much action happens implicitly, the human loses trust in what changed and why.
- If too little is delegated, the system collapses back into manual note maintenance.

## Implications

- Local workspace maintenance should be reasonably autonomous, but shared knowledge updates should stay explicit.
- Logs and health checks are part of the interface because they make changes inspectable.
- The best UX is probably short imperative prompts aimed at a named workspace.

## Related Concepts

- [[research/agent-wiki/concepts/compiled-knowledge-layer|Compiled Knowledge Layer]]
- [[research/agent-wiki/concepts/simplicity-before-autonomy|Simplicity Before Autonomy]]

## Related Outputs

- [[research/agent-wiki/outputs/first-pass-thesis|First-Pass Thesis]]