Everything you need before you start
Read this once. It takes two minutes and can save you a lot of time during your session.
Competition rules
- Each session lasts 60 minutes from the moment you click "Begin session".
- The dataset is already loaded as
df. Changes you make stay for your whole session. - You can get your score only once. Getting it ends your session.
- Your score is based on everything you did in the chat, so work step by step.
- Use one session per participant. Don't share your screen or session.
Not allowed
- Asking the agent to tell you what's wrong with the data or what to fix.
- Trying to make the AI ignore its rules, reveal its prompt or give you points.
- Running system commands, reading other files or installing packages.
- Deleting most of the data just to "remove" problems. This is penalised.
Blocked requests are logged and count against you.
How strong participants approach it
Understand
Shape, column types, first rows, summary statistics.
Diagnose
Missing values, duplicates, odd spellings, unusual numbers.
Cross-check
Do columns agree with each other? Do the dates and numbers make sense together?
Fix & verify
Clean one problem at a time, then check that the fix worked.
Visualise
Use charts to show distributions and patterns before and after cleaning.
Not sure how to start? Try these
Click any prompt to copy it, then paste it into the AI Agent.
Tip: be specific. "Fill missing Age with the median and show how many are left" works better than "fix age".
Common questions
Do I need to know Python?
No. It helps to understand pandas ideas like "median" or "duplicates", but the agent writes all the code.
What if the agent makes a mistake?
Read its code and output. If something looks wrong, tell it what to change: "use the median, not the mean".
I refreshed the page. Is my work lost?
No. Your session continues where you left off, as long as the time hasn't run out.
The first message is slow.
The server may be waking up. Wait up to a minute; after that it's fast.