Recently, I listened to the last episode of core.py podcast. In general, I understand less than fifty percent of the content :P.
I like to listen to them for three or four reasons:
- I have the idea that maybe, by some kind of radio-wave osmosis, I can learn from them only by listening.
- It is a good way to practice my eternally deficient English.
- They sound very honest to me.
- And of course, they are very seasoned Python developers.
But the most important thing is this: although I don't understand completely, I can always take one or two words, ideas, or concepts that keep resonating in my mind. Because of that, for me all the effort is justified.
The Idea That Stayed With Me
In this particular episode, Pablo Galindo talked about the impact of AI-generated PRs on the maintainers of Python core (in this case, but it could be any open source project).
To cut a long story short: it is an asymmetric relationship.
AI can write thousands of lines of code in seconds. On the other hand, we—as human beings—have much more limited attention, time, and capacity.
Nothing new under the sun.
However, what caught my attention was the idea of lost opportunities that this new scenario creates.
If maintainers spend their time interacting with code generated by an AI, then answer a review request from an AI, and after that their comments are taken again by an AI and fixed / applied / whatever...
...the whole thing becomes a Python developer talking with an external AI while burning tokens.
Ok. That is good because it avoids Łukasz and Pablo having to pay for them.
But the whole interaction destroys the opportunity (and the project's necessity!) for mentorship.
A Broken Social Contract?
It seems that the implicit social contract they expected has been broken:
"I review your code because it could help my project, and I can help you learn about the technology."
Nowadays, nobody but the AI learns something.
And intentionally I wrote "The AI"* as a separate entity, in the same way Karl Marx titled "The Capital" and not "The Bourgeoisie".
Qué tul!?
* Nothing to do with AGI and all of that bullshit.
Questions
A lot of questions came to my mind:
- How is the interaction between humans and non-humans going to be organized in open source ecosystems?
- Is there a best approach to regulating those interactions?
- Are we losing opportunities to be mentored or to mentor people?
- How are we going to connect with people as a foundation of open source development?
- Are communities a real foundation of open source projects?
- How can we participate in something bigger than ourselves or our LLMs?