While tokens are an LLM's energy source, structured markdown files in a system like Obsidian act as its perfect, persistent memory. This organized, interlinked data is the true "oxygen" that allows an AI to develop a deep, evolving understanding of your context beyond single-session interactions.

Instead of creating a bespoke memory or messaging protocol for agent-to-agent communication, Notion leverages its core primitives. Agents compose by writing to and reading from shared Notion pages and databases, creating a decoupled, human-editable, and transparent system for coordination.

The simple, text-based structure of Markdown (.md) files is uniquely suited for both AI processing and human readability. This dual compatibility is establishing it as the default file format for the AI era, ideal for creating knowledge bases and training documents that both humans and agents can easily use.

Instead of a complex database, store content for personal AI tools as simple Markdown files within the code repository. This makes information, like research notes, easily renderable in a web UI and directly accessible by AI agents for queries, simplifying development and data management for N-of-1 applications.

The underlying infrastructure for AI agents ('harnesses') becomes obsolete roughly every six months due to rapid advances in AI models. At Notion, this means completely rewriting the harness multiple times a year, demanding a culture comfortable with constantly rebuilding core systems and discarding previous assumptions.

Standard APIs for human developers are often too verbose for AI agents. Notion created agent-centric APIs, like a special markdown dialect and a SQLite interface, by treating the AI as a new type of user. This involved empirical testing to understand what formats agents are naturally good at using.

To fully leverage rapidly improving AI models, companies cannot just plug in new APIs. Notion's co-founder reveals they completely rebuild their AI system architecture every six months, designing it around the specific capabilities of the latest models to avoid being stuck with suboptimal implementations.

Early AI agents like OpenClaw use simple markdown files for memory. This 'janky' approach is effective because it mirrors a code repository, providing a rich mix of context and random access that agents, trained on code, can efficiently navigate using familiar tools like GREP.

Instead of relying on platform-specific, cloud-based memory, the most robust approach is to structure an agent's knowledge in local markdown files. This creates a portable and compounding 'AI Operating System' that ensures your custom context and skills are never locked into a single vendor.