AI Techniques
Under Technique, AI Techniques covers how you apply language models and agents in software and knowledge work—patterns, frameworks, and conventions—not the ML training stack or the agent products themselves.
In this subcategory: portable skill formats (Agent Skills, Agent Skills - Sources), vault/brain conventions (gbrain two-layer pages, MECE resolvers), prompt and context patterns, and integration techniques (e.g. wiring agents into IDEs via Agent Client Protocol). Sibling subcategory: Artificial Intelligence & Machine Learning — classical ML/DL methods, datasets, and evaluation (see hub AI & ML).
Not here: runnable agent tools (cursor-agent, Claude Code, Codex → AI Agent under Tool); editor platforms (Cursor → IDE); model hosts (Ollama → Platform).
Garden stance: trial gbrain patterns (two-layer pages, light MECE resolvers)—adapt to your vault, not the full opinionated stack; adopt Agent Skills where your agent supports SKILL.md; assess new skill sources before importing. Pair techniques with adopt cursor-agent / Cursor for execution, not omnichannel bots (hold under AI Agent).
Tag an item here when the note documents a repeatable pattern for using AI in systems, not a vendor SKU or raw ML theory.