Sponsored

Stop vibe-debugging.

Every exception, every backtrace, grouped so you see patterns, not noise.

Start free trial
  1. ai
  2. /foundations

Foundations

You do not need a PhD to use AI coding tools effectively — but you do need a mental model of what models can and cannot do.

These pages explain why tools behave the way they do: context limits, training cutoffs, token costs, and when to verify instead of trust. They are not a machine learning course — they are the minimum theory that makes you a better prompter and reviewer.

How the Pages Relate

PageRead when…
How LLMs WorkYou want the big picture: prediction, tokens, transformers, limits
Tokens and ContextYou hit context limits, need to budget attachments, or wonder why @ files get dropped
Evaluating Model OutputYou need a framework for when to trust vs verify before merge
Reasoning ModelsYou need o-series / thinking models vs fast chat — and local Ollama

Start with How LLMs Work, then read Tokens and Context if you work in large repos. Read Evaluating Output before relying on AI for production changes.

Learning Paths

Solo developer

  1. How LLMs Work
  2. Tokens and Context
  3. Context Engineering — apply token knowledge in your IDE
  4. Evaluating Model Output
  5. Verifying AI Output

Team lead

  1. How LLMs Work — brief the team on limits and cutoffs
  2. Evaluating Model Output — define trust thresholds by change type
  3. Team AI Policy — encode verification requirements in policy

Start Here If…

Your questionStart with
"Why did the model forget my file?"Tokens and Context
"Why does it invent APIs?"How LLMs Work (training + cutoffs)
"When is AI output good enough to merge?"Evaluating Model Output
"Should we use RAG or fine-tuning?"Fine-Tuning vs RAG
"When do reasoning models pay off?"Reasoning Models

Guides

How LLMs Work

Tokens, context windows, training cutoffs, and limitations.

Tokens and Context

Token budgets, context windows, and practical attachment limits.

Evaluating Model Output

When to trust vs verify — ties theory to your review workflow.

Reasoning Models

When to use thinking models and local LLMs with Ollama.