> This requires the ability to write well (or well enough to get what you want from the agent) and clearly communicate intent (in your language of choice, not code; very different IMO).
I do not see why you can't write your spec in pseudocode if you really want to - communicating your intent to the LLM, for how the code should be developed is far closer to programming than writing skillwise.
I was referencing https://www.neobrutalism.dev/ and https://www.retroui.dev/ and slopped my way through it. A lot of it was just asking Claude Code "is this a proper design system?", then I kept doing that until it didn't have anything useful to add. Now I'm using my that as the template for understanding such things in more detail.
I think it's not just puzzle solving - for me it's the idea of creating something from raw materials where that something is itself a standard building block. it appeals to the same part of me that programming does.
we used ninja as a parallel task runner in pytype - had our whole-project analyser generate a ninja file with a task graph, and then just evoke ninja to run it, taking care of dependencies and parallel execution. it worked very nicely indeed.
I really hope that one effect of ai code generators making code cheaper to write is that the calculus around "accept vendor lock in return for getting up and running faster" changes dramatically
reply