AI-Assisted Contributions and Maintainer Load - PyCon US 2026

-- A talk I gave at PyCon US 2026

AI-assisted contributions are changing how open source work happens. This talk looks at real maintainer experiences from projects like Django, Python, GNOME, and OCaml, focusing on review load, responsibility, and the governance questions that appear when AI replaces understanding instead of supporting it.

https://www.paulox.net/2026/05/15/pycon-us-2026/

Transcript

1. AI-assisted contributions & maintainer load Paolo Melchiorre www.paulox.net
2. 2001: A Space Odyssey (1968) — Stanley Kubrick
3. “(its) use … cripples the mind, its teaching should be regarded as a criminal offense.” — Edsger
4. “You can’t trust code that you did not totally create yourself.” — Ken
5. Edsger Dijkstra Shortest-path algorithm creator Talking about COBOL 1975
6. Ken Thompson Co-creator of Unix Talking about code from its company 1984
7. © 2023 Paolo Melchiorre · CC BY-SA
8. © 2020 Randall Munroe · CC BY-NC
10. © 1946 William Gottlieb (Public Domain)
11. © 1946 William Gottlieb (Public Domain) · AI-modified
13. Amplifier More contributors More low-value contributions More code More untested code
14. AI does not change the nature of Open Source. It changes the scale.
15. Paolo Melchiorre 🦄 DSF board member 🐍 PSF fellow 🇮🇹 PyCon Italia organizer 🐬 Python Pescara founder 🚀 Djangonaut Space navigator 🧡 Django Girls+ organizers 🌐 www.paulox.net
16. © 2026 Paolo Melchiorre · CC BY-SA
17. AI can change the nature of Open Source by changing the scale.
19. OCaml Huge diff +13,323 -36 Wide scope No prior discussion Review impossible github.com/ocaml/ocaml/pull/14369
21. Godot AI-generated PRs Low understanding High review effort Maintainer fatigue bsky.app/akien
23. Jazzband Too many projects Limited maintainers Growing backlog Sustainability crisis jazzband.co/sunsetting-jazzband
25. MemPalace Famous identity reused Ambiguous authorship Synthetic credibility Trust becomes fragile pypi.org/project/mempalace
26. Blade Runner (1982) — Ridley Scott
27. Matplotlib AI-generated PR Maintainer rejection Escalation outside repo Automated conflict github.com/matplotlib/matplotlib/pull/31132
29. 2001: A Space Odyssey (1968) — Stanley Kubrick
30. “I'm sorry, Dave. I'm afraid I can't do that.” — HAL 9000
31. The problem is no longer just bad code. It is automated conflict.
32. The Matrix (1999) — The Wachowskis
33. “As with using any tool, the resulting contribution is the is the responsibility of the contributor.” — Python - Generative AI
34. “If the human effort put in a PR ... is less than the effort ... to review it, please don't submit the PR. — FastAPI - Automated Code and AI
35. “Unverified AI-generated contributions create ... maintenance burden ... Submissions that show no evidence of manual verification may be closed without review, ...” — Django - AI-Assisted Contributions
36. “To ... preserve limited core developer capacity, we will flag and reject low-value contributions that we believe are AI generated.” — Matplotlib - Use of Generative AI
37. “Pull requests that have an LLMproduct listed as co-author ... will be closed without further discussion.” — attrs - AI Policy
38. “No LLMs for issues. No LLMs for pull requests. No LLMs for comments ...” — Zig - Strict No LLM / No AI Policy
39. The Matrix (1999) — The Wachowskis
40. The Matrix (1999) — The Wachowskis
41. The Matrix (1999) — The Wachowskis · AI-modified
42. AI is a tool that amplifies problems and possibilities…
43. ... so responsibility and protection must scale.
44. We need more human in the loop!
45. © 2022 Paolo Melchiorre · CC BY-SA
46. © 2025 Paolo Melchiorre · CC BY-SA
48. djangomed.eu
49. License CC BY-SA 4.0 This work is licensed under a Creative Commons Attribution - ShareAlike 4.0 International License.
50. Paolo Melchiorre www.paulox.net