Generating Code Faster Than You Can Debug It

Generating code faster than you can debug it?
You’re not the only one.
Velocity is up.
Clarity hasn’t caught up.
PyData is where we compare what actually holds up in practice.
No sales pitches—just people building real systems and sharing what works (and what doesn’t).
If you’re trying to write code you can reason about, maintain, and trust—you’ll fit right in.

18:30 – 18:55
Welcome & Community Update — Charles Fourie and Luis de Sousa
A quick read on what’s moving in the local data ecosystem before we get into the main session.

19:00 – 19:55
Talk: Sheena O'Connell
Powering up Django development with Claude Code
Description
Kernighan’s law still holds:
debugging is harder than writing the code.
LLMs don’t remove that problem—they amplify it.
We’re now generating code that:
works, but is hard to reason about
introduces hidden design trade-offs
accumulates technical debt faster than we notice
The question isn’t whether you can use these tools.
It’s whether you stay in control when you do.
Sheena walks through how she uses Claude Code in real Django projects:
structuring prompts to guide output
applying guardrails to avoid drift
knowing when to trust vs verify
spotting failure patterns early
Practical patterns that hold up in real projects.

Terms & Notes
Free, community-run event
Limited space—please cancel if you can’t attend

Call for Speakers
Built something that actually held up under pressure?
Submit here: https://sessionize.com/pydata-johannesburg/