Student Guide

How To Learn Effectively In This Course

This course works best if you treat every week like a build lab, not a school worksheet. You are not trying to memorize definitions. You are trying to make the code feel less mysterious.

Use this learning loop:

1. Predict what the code will do.
2. Run it.
3. Compare the result to your prediction.
4. Change one thing.
5. Explain why the result changed.

If you follow that loop every week, you will learn faster than someone who watches videos passively.

Debugging Mindset

When something breaks, do not jump straight to “I am bad at this.”

Instead ask:

• What exactly is the error?
• Which line caused it?
• What value did I expect there?
• What value is actually there?
• What is the smallest print statement that would help me?

Good debuggers are not magical. They are patient and systematic.

Avoid Passive Consumption

Watching AI videos can feel productive even when you are not really learning. To avoid that trap:

• Keep a notebook open while watching.
• Pause every 5 to 10 minutes and write one sentence in your own words.
• Recreate examples yourself instead of trusting that you “get it.”
• Change the code after you copy it.

If you finish a video and cannot explain one concrete idea from it, rewatch a small section instead of moving on.

What Counts As Success

Success in this course is not “I never got stuck.”

Success looks like:

• I ran the code.
• I changed the code.
• I noticed what changed.
• I wrote down one honest thing I understand better now.

That is real progress.

Weekly Minimum Standard

Every week, make sure you:

• run the main file or notebook
• make one meaningful modification
• answer the reflection prompts
• explain one idea out loud to a parent, mentor, or even to yourself

Co-Author Mode

You are not only a student in this course. You are also helping shape a better version of it.

When you send feedback in OnPoint or Telegram, try to use this structure:

• What I was trying to do
• What felt confusing, slow, or annoying
• What change would help
• How important it feels: low, medium, or high

Example:

"I was doing Week 4 text generation. I understood the code, but I wanted one harder challenge after the first run. It would help if there was an optional trigram extension. Importance: medium."

That kind of feedback is gold because it helps improve both the course and the product.

When You Feel Lost

Shrink the problem.

Examples:

• Print one tensor instead of the whole model.
• Train for 10 steps instead of 1,000.
• Use 10 lines of text instead of a huge dataset.
• Test one function in isolation.

Small problems teach faster than giant confusing ones.