1. Fundamentals of AI-Assisted Coding
Learn how LLMs work and write better prompts.
| Duration: 30 min | Tools: Cursor, R |
What You’ll Learn
- Why tokens matter in AI prompts
- How to write prompts that get better results
- The “prompt formula” for asking AI to help with data
Key Idea: Tokens & Context
Models don’t see words — they see tokens (roughly 0.75 words each). More importantly, every time you send a message, the model receives your entire conversation history, not just your latest message.
This means: Keep prompts short and specific. Rambling wastes tokens and confuses the AI.
Try It: Visualize Tokens with Tiktokenizer
Test how many tokens your prompts use:
🔗 Open Tiktokenizer in new tab
Example to try:
- Go to Tiktokenizer
- Paste the “Bad (vague)” prompt below and count tokens
- Then paste the “Good (structured)” prompt and compare!
Bad (vague):
Tell me about my data.
Good (structured):
I have data/penguins.csv with 344 penguins and 8 columns.
Load it with readr and group by species to show average bill length and body mass.
Use dplyr. Show as a tibble rounded to 1 decimal.
🎨 Artwork: Palmer Penguins Illustrations by @allison_horst
Result: The structured prompt uses more tokens but is far more useful to the AI!
The Prompt Formula
Good prompts have 4 parts:
"I have [data]. I want to [task]. Use [constraints]. Show [format]."
Example:
“I have
data/penguins.csvwith 344 penguins and 7 columns. Load it with readr and group by species to show average bill length and body mass. Use dplyr. Show as a tibble rounded to 1 decimal.”
Bad (vague):
“Tell me about my data.”
Why the good version works: It tells the AI what data, what analysis, what tools, and what output format.
Quick Try-Out
Open Cursor Chat (Cmd+L) and try one of these prompts:
"Load `data/penguins.csv` with readr and show me the first 5 rows and column names."
OR
"I have `data/penguins.csv` with penguin measurements. Group by species and count how many in each. Show as a table."
Key Takeaways
- Be specific — the more detail, the better
- Be concise — don’t ramble
- State your format — say exactly what output you want (table, plot, list, etc.)
- Tools matter — mention which R packages to use (dplyr, ggplot2, etc.)
Resources
Loading last updated date...