Free preview: first minute
Transcript
Click on the transcript to go to that point in the video. Please note that transcripts are auto generated and may contain minor inaccuracies.
Loading transcript...
I recently did a livestream where I used Claude as a thinking partner, a kind of rubber duck, while analyzing fake data on food pantry volunteers. I ran into several wrong numbers, and instead of having Claude rewrite my code, I read its responses and made the fixes myself so I could understand what was going wrong.
You'll learn:
- How to use an AI agent as a debugging partner that explains a problem instead of writing the fix for you
- How to spot suspicious output, like a volunteer count of 2,045 and total hours of 6,000, that points to a problem in the code
- A way to check a fix by re-running the code, watching for the many-to-many join message to disappear, and asking Claude whether the problem is fully solved
- Why a partial fix can leave other bugs behind, such as counting no-show or canceled shifts in the volunteer totals
Tools mentioned:
Key moments:
- 00:23 Rubber duck debugging explained
- 00:39 Livestream setup: using Claude as a thinking partner
- 00:45 Food pantry volunteer data and the analysis goals
- 01:02 Suspicious totals for hours and volunteer count
- 01:20 Making the changes myself instead of letting Claude do it
- 01:40 Joining shifts and volunteers without the many-to-many message
- 01:48 Partial fix and asking Claude if it is solved
- 02:02 Shift status (no-shows, cancellations) still affecting the count
- 02:22 Running the adjusted code and confirming it works