Clojurians Log v2
ai-skeptics 2026-07-01
There are probably twice the number of public git clojure repos than there was before AI. Newer models are thus trained on many AI-produced clojure repos... And the same can be said about any proglang ecosystem. Where is this leading?
@henrik "Prompt Injection" by Peter Watts is a very pertinent essay. Not sure if there are any other sources of the text, but I found it in this book: https://www.hatjecantz.com/products/66041-worldbuilding
github overall commit volume is up something like 30x y/y due to AI > There were 1 billion commits in 2025. Now, it's 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler: it won't.) https://x.com/kdaigle/status/2040164759836778878 (April, these are Q1 numbers) https://github.blog/news-insights/company-news/an-update-on-github-availability/
Model collapse?
The long-term loop as well. Humans start writing prose and code like models do, even when it’s actually hand-written.
a temporal search qualifier for github search would be useful, i barely don't google without "before:2022" any more
Imagine how different The Terminator would be if made today. No big nuclear war, just a bunch of tired people drowning in slop, and then explaining the situation with wildly inaccurate information once they travel back in time: “it all started when the 5G towers beamed nanobots into the water supply”
What may be funny with the original question is that it's hard to answer with confidence. I'd say AI adds a lot of non-determinism not only to its own responses, but also to its own impact on the world. If it makes sense...
I bet distinguishing human from AI agents is going to be well funded endeavor. I fear this could lead to further justifications for biometrics.