Protecting Your Second Brain: Why AI Should Stay Out of Your Notes

Simon Späti argues that integrating AI into personal note-taking apps like Obsidian undermines the essential human thinking required to build a meaningful knowledge base. He warns that AI-generated content creates noise and dilutes the personal insights that make a 'Second Brain' valuable over time. Ultimately, he recommends using AI only for technical search tasks while keeping the core creative process strictly human-driven.

Key Points

• The value of a Second Brain lies in the process of thinking and making deliberate connections, not just in the retrieval of information.
• Mixing AI-generated text with personal notes leads to 'AI Slop,' which dilutes original insights and creates noise that hinders long-term learning.
• AI summaries often lack the personal conviction and unique 'signal' that make human-written notes valuable for future reference.
• Automating organization and tagging prevents the intellectual compounding that occurs when a user manually manages their own knowledge graph.
• AI should be restricted to technical assistance like advanced search or kept in isolated environments to protect the user's creative process.

Sentiment

The overall sentiment is cautiously supportive of the article's main point. The community largely agrees that generative AI should not be allowed to write or organize the core of a personal knowledge base without strict provenance boundaries, and many commenters see the same risk in code comments and project documentation. The disagreement is mostly about scope: commenters are more receptive to AI as a search, transcription, review, prototyping, or scratchpad tool than as a source of durable personal thought.

In Agreement

• The value of personal notes comes from the act of thinking and writing, so outsourcing summaries, tags, or ideas can remove the main benefit of maintaining a knowledge base.
• Mixing generated text with human notes creates a provenance problem, making it harder later to tell whether an insight was personally discovered or machine-produced.
• AI-written comments and explanations can pollute codebases in the same way, because future agents and maintainers may treat plausible hallucinated rationale as canonical context.
• Separate vaults, directories, labels, or scratchpads are useful boundaries for keeping AI output available without contaminating the user's own thinking archive.
• AI can still be valuable for semantic search, transcription, research assistance, review, and on-demand summarization when it does not become the authoritative source of notes.

Opposed

• Some commenters argue that AI-generated notes are manageable if they are clearly labeled or stored in dedicated folders, so the provenance issue is practical rather than fundamental.
• Several people find AI useful for completing execution, developing prompts into seed ideas, or surfacing angles that humans can then refine, so they reject a blanket prohibition.
• Others say summaries and transcriptions are useful even when hand-written notes would produce deeper learning, especially for lectures, meetings, or reference material.
• Some participants suggest that agentic coding or AI-assisted production can work if guarded by deterministic tests, linters, type checks, and other mechanical constraints.
• A minority challenges whether AI output is really hard to distinguish from human writing, implying that the article's concern may be overstated for current models or careful users.