ok, agentic coding is "successful" in a way similar to how biological evolution is successful. there's no intelligence or intention, just random exploration over an implicit plausibility landscape, to satisfy some goal.
the solutions are messy and incremental with not much in the way of "coherent design", but it works anyway, kind of.
there are continual arms races that have patchwork solutions. eyes and wings and histamines can arise from evolution; LIGO and wheels and vaccines do not
the implicit plausibility landscape of agentic coding is built on the vast libraries of code created by individuals collaborating to communicate precise intent via programming languages. but if agentic coding takes over, the landscape will get weird and incomprehensible, and maybe that's ok?
biological systems are weird and incomprehensible, we can poke at the edges a bit, but the indifferent whole crawls endlessly in an amorphous way.
... this is adjacent to an argument that trying to achieve precise understanding and control of computers (or other systems) is a fundamentally fascist enterprise. it's not, but I don't know if I can articulate that clearly yet.
it's... these systems are models, they're games. games are a way of practicing agency. fascism happens when you conflate the map with the territory and try to fix flaws in the territory instead of making more maps to play with.
felix (grayscale) 🐺