llm-lore

graph TD
    A[What is an LLM?] --> B[Pattern matcher]
    A --> C[Autocomplete]
    A --> D[Eager to please]
    E[What isn't an LLM?] --> F[Not truthful by default]
    E --> G[Not a database]
    E --> H[Not conscious]
    B --> I[How to work with it]
    C --> I
    D --> I
    F --> I
    G --> I
    H --> I
    I --> J[Direct them]
    I --> K[Decompose tasks]
    I --> L[Create boundaries]

Overarching knowledge ABOUT LLMs as a technology. Not lessons from using them, but fundamental understanding of what they are and aren’t.

Contents

Root level - Conceptual explanations

• llm-is.md - What LLMs fundamentally are
• llm-is-not.md - What LLMs are not (and why that matters)

examples/ - Real-world illustrations of concepts

• forgetting.md - LLMs forget context, context management is paramount

What goes here

• Meta-knowledge about what LLMs fundamentally are
• What they are NOT (and why that matters)
• How they actually work at a conceptual level
• Core principles that explain their behavior
• The mental models you need to work with them effectively

Examples

• What is an LLM? (pattern matcher, autocomplete, eager to please)
• What isn’t an LLM? (not truthful by default, not a database, not conscious)
• How to work with these realities (direct them, decompose tasks, create boundaries)

Format

Clear explanations of overarching concepts. This isn’t “I tried X and it failed” - it’s “Here’s what LLMs fundamentally are and how that shapes everything.”