↓ Skip to main content
Amit Kohli
  1. Blogs/

3 mins

Blog Posts - February 2026
#

Five blog posts for amitkohli.com exploring Obsidian, AI automation, and practical knowledge management.

Posts Created
#

  1. Why Obsidian Ate My Brain (And I’m Totally Fine With It)

    • Why Obsidian works when other systems fail
    • Links over folders, metadata as infrastructure
    • Plain text files, knowledge graphs, and organic structure
    • No AI discussion - pure Obsidian benefits

  2. I Put an AI Front Door on My Brain (What Could Go Wrong?)

    • GitHub Copilot integrated with Obsidian vault
    • What it gives: scaffolding for neurodiversity, memory without mental load
    • What it costs: Zettelkasten inefficiency (THE INEFFICIENCY IS THE POINT)
    • The bargain: AI handles logistics, human handles synthesis
    • “What’s the point of writing anything down?” explored

  3. I Built a Formal Ontology for My Notes and the AI Completely Ignored It

    • Created relationship definitions, semantic triples, property hierarchies
    • AI agent didn’t use any of it (used semantic search instead)
    • LLMs can’t do inference, tooling doesn’t support graph traversal
    • Ontology helped human brain organize thinking
    • Simple YAML + wiki-links beat formal semantics for AI agents

  4. I Tried Claude Code for Three Days (Then Went Back to Copilot)

    • Claude Code vs GitHub Copilot comparison
    • Higher manual intervention, context degraded faster
    • Critical failure: ignored mandatory skills, wrotebad defensive code
    • Works for batch refactorings, not interactive development
    • Anthropic’s no-friction refund policy appreciated

  5. The AI Perspective Translator (Or: How to Make Your Technical Doc Actually Useful)

    • Use AI to simulate CEO, COO, Product, Bizdev, IT perspectives
    • Shows blind spots in technical writing before sharing
    • Restructure docs for multi-audience accessibility
    • Neurodiversity assist: theory of mind prosthetic
    • Progressive disclosure > multiple separate documents

Writing Style
#

All posts use the blog-writing skill documented in .github/skills/blog-writing/SKILL.md:

  • Conversational and self-aware (“OK ok ok…”, “Aaaaanyway:”)
  • Technical but accessible (explains jargon, uses analogies)
  • Playfully irreverent (self-deprecating, honest about failures)
  • Analytically curious (“What does it mean?” sections)
  • Visual elements (Mermaid diagrams with color standards)
  • Honest about limitations and tradeoffs

Themes
#

Knowledge Management
#

  • Obsidian as external brain
  • Plain text files for longevity
  • Links and metadata for discoverability
  • Organic structure emerging from connections

AI Integration
#

  • Automation as scaffolding, not replacement
  • Executive function prosthetics for neurodiversity
  • Where AI helps vs where it hinders
  • Honest assessment of tradeoffs

Pragmatic Experimentation
#

  • Try tools, document failures
  • Honest “what works for me” assessments
  • No hype, just practical experience
  • Value failed experiments for learning

Neurodiversity
#

  • ADHD and context-switching challenges
  • Perspective-taking difficulties
  • Assistive technology for executive function
  • Systems that work with brain differences, not against them

Files Created
#

content/blog/
├── README.md (this file)
├── why-obsidian-ate-my-brain.md
├── ai-front-door-to-obsidian.md
├── ontology-experiments-obsidian-ai.md
├── claude-code-experiment.md
└── ai-perspective-translator-docs.md

.github/skills/blog-writing/SKILL.md - Voice and style documentation

Control Tokens
#

All posts include:

  • BLOG_VOICE_APPLIED - Conversational and self-aware
  • TECHNICAL_ACCESSIBLE - Jargon explained
  • VISUAL_ELEMENTS - Diagrams or code included
  • HONEST_LIMITATIONS - Caveats acknowledged
  • COLLABORATION_INVITE - Reader engagement encouraged

Next Steps
#

These posts are ready to be adapted for your Hugo site at amitkohli.com. They match the voice and structure of your existing blog while covering your recent experiments with Obsidian, AI agents, and knowledge management systems.