FOUNDATIONS
Tokens are the atoms of language models
Models don’t read words — they read pieces of words.
A token can be a word, part of a word, or punctuation. Longer text costs more tokens, which is why prompts and answers have length limits.
TAKEAWAY
Think in tokens, not just sentences, when you write prompts.
PROMPTS
Be specific about the output shape
Vague asks get vague answers.
Name the format, audience, and constraints up front. “Write a 5-bullet summary for a founder” beats “summarize this.”
TAKEAWAY
Constraints are a feature — they guide the model.
SYSTEMS
Context windows are finite attention
Everything you paste competes for the same limited space.
Put the most important instructions first, trim noise, and don’t assume the model remembers earlier chats forever.
TAKEAWAY
Short, ordered context beats long, messy dumps.