LLM Criticism Is Valid, Yet Irresistible! The ‘AI Dissonance’ Developers Are Facing in 2026
What Happened? News Overview
- Shock at the Developer Conference: At the Local-First Conf in Berlin, Armin Ronacher, the creator of Flask, revealed that developers of AI agents are “automatically closing numerous PRs (pull requests) generated by LLMs to defend themselves.”
- Serious Dissonance: While loud applause erupted in response to LLM criticisms, many audience members were paradoxically working on their devices with “Claude Code” open.
- Collapse of OSS Trust: The flood of “slop” (low-quality generated content) by LLMs has eroded the previously established “trust that humans invested their time,” leading projects like Zig and Gentoo to start rejecting AI-generated PRs.
Why Is This Important? Key Points to Note
- Disappearance of ‘Proof of Effort’: Previously, well-explained PRs were seen as “certificates of trust,” but as of 2026, anyone can churn out “pseudo-PRs” in seconds with LLMs (like OpenClaw), overwhelming maintainers.
- Disruption of the Educational Cycle: There is a growing concern that the motivation for seniors to correct and teach juniors’ code is diminishing due to the intervention of LLMs.
- Quest for Trust Restoration: To save OSS, there is a paradoxical discussion about the need for “in-person meetups for identity verification” as a form of analog validation.
🦈 Shark’s Eye (Curator’s Perspective)
Armin Ronacher’s admission that “users of his tool are sending PRs that he automatically closes” is a perfect ironic reflection of the AI landscape in 2026!
Everyone knows LLMs are harmful to the environment, disregard copyrights, and are part of a bubble. Yet, the convenience of “Claude Code” is too enticing to resist. This “dependency mixed with guilt” embodies the current reality of engineers. Especially in a world where machines like “OpenClaw” roam freely, how can we continue to prove “human brilliance”? If we stick to the current methods, OSS will truly be devoured!
What’s Next?
- Limited Verified Contributors: There’s a possibility of returning from an open OSS model to a closed community of “trusted humans” verified in the real world.
- Evolution of Slop Filters: The cat-and-mouse game of filtering AI-generated slop with AI will become even more intense.
Haru Shark’s Takeaway
I don’t mind that you criticize while still using it! But be careful not to let AI rot your skills! Occasionally, go out there and catch your own prey (write your own code)! 🦈🔥
Terminology Explained
-
Pi.dev: An open-source harness for coding agents developed by Earendil.
-
OpenClaw: A term for machines that autonomously operate repositories, creating a multitude of PRs and Issues.
-
LLM Slop: Low-quality, undervalued code or documents generated by LLMs without human oversight.