This feels like another software engineering truism that gets forgotten in the LLM age - if the ticket requires more than a two week sprint's worth of work (ie 8+ points), it should be broken down. Otherwise you bury so much context and tech debt while also making your reviewers deeply unhappy and cognitively depleted.
replyHave your LLMs treat you like you want to be treated as a code reviewer!
Agile/Scrum was a thing before AI. What's the point now
replyThis is the right question. The science on multitasking and frenetic context switching is pretty clear. New software doesn’t change that.
replyBecause LLMs and tools are not instant.
replyEven pre-AI I would work in parallel while I waited for tests to finish, or deploy to go out or whatever.
They're not, but the way we're filling our time while waiting for LLMs involves much more context switching than how it was with tools and human bottlenecks like manual coding. I think it's fine to work on two or three things at once, provided we don't let the LLMs decide when we should tend to which task.
replyI.e. a “manager” schedule not a “maker” schedule. [1]
replyThe agents are the makers; you switch among them. Either that or watch grass grow.