I've seen two separate analyses recently that suggest that the use of LLM to create software reduces collaboration during development.

https://lucumr.pocoo.org/2026/7/13/the-tower-keeps-rising/

https://blog.mempko.com/a-million-kakapos/

I think that this is plausible. I think certain domains will be left where it's still worthwhile to collaborate: software that is high risk if fails, software that is deeper down the stack (OSes, languages, and some libraries and frameworks etc), very complex sets of rules, and software pushing the boundaries of what's possible.

Outside of software, areas of concern where truth and sources are valued strongly, like Wikipedia.

It also really sucks on a human level. Here we have this wonderful semi-anarchist open source commons...

And on a personal level: we're going to be collaborating less. It was already bad enough where often the only collaboration happens on PR review...

I saw a shaky metaphor. I created another shaky metaphor and then explicitly said "see this is shaky too" to invite people to examine the original one.

I did not expect a deconstruction of my shaky metaphor, but people held me to be unkind when I pointed that out.

I thought I'd be clever and not try to drearily deconstruct the original one but instead make people think but I failed at that entirely.

Doing so in a thread where everyone was having fun with the original one was a bad idea too.