There is a lot going on in my industry right now, and LLM-augmented development is changing what is possible (not so much at a technical level, but in terms of how quickly it is possible to do certain things).

The challenge is that this can be a bit tempting: the phrase "move fast and break things" has been around for a while, and as someone who doesn't really like to "break things" when possible it's wild to have the value equation so dramatically shift (i.e. you can now move, say, 30x faster while breaking 25% less).

Another example of a "math shift": third party dependencies. There will always be some argument for just building a library / tool yourself rather than have to worry about upstream communities and their decisions. Before LLMs there were a few reasons to avoid this: community wisdom, separation of concerns, maintaining something locally means you have to spend some % of your energy on it. With an LLM the cost of those concerns is less (particularly the last one).

That said, the value of community, the value of having more human eyes trying to find and fix bugs within a given tool... there is a lot to be said.

I don't have answers on these things, and even if I had them they would need to be revisited in a month!