speaking from personal experience only, i think it pretty much boils down to owning the code, i.e always having a deep and up to date understanding of the abstr

speaking from personal experience only, i think it pretty much boils down to owning the code, i.e always having a deep and up to date understanding of the abstractions and the underneath assumptions. ironically this isn't new, it's tempting to just hand-over ownership to the AI so llms kind of just made the trap easier to fall into