01-02 Counting What the Model Sees

Why this matters

Token-frequency counting is the first measurable signal a language model can extract from data.

Intuition first (no jargon)

A frequency table is the model's first memory: it records what appears often.

Code walkthrough

js
export function countChars(text) {
  const counts = {};
  // fill counts
  return counts;
}

Your task

Implement countChars(text).

• Return an object mapping each character to count.
• Include spaces as characters.
• Return {} for empty input.

Hints

• Iterate with for (const ch of text).
• Initialize with counts[ch] = (counts[ch] || 0) + 1.
• Keep the function pure.

Check your thinking

1. Why might spaces matter for language?
2. What does a high count mean?
3. Does counting alone preserve order?

Stretch (optional)

Add countWords(text) and compare character vs word counts.

Likely test focus

• Exact counts for toy strings.
• Handles empty strings.
• Includes whitespace tokens.

What should improve

You can now compute reproducible corpus statistics that drive simple probabilistic baselines.

Bridge to next lesson

Next lesson: add weighted randomness for generation.