2026-03 people starting to care much more about design principles and correctness
coding agents work better with well-designed code. Also coding agents are so high leverage that it makes sense to really lean into what makes them even marginally more effective
Also, people have more time to think about architecture because the agents are doing the actual coding. Hence people working with these systems seem to be placing an increased emphasis on design principles. And because the agents drive down the cost of refactoring, people can afford to actually move their systems towards their more ideal architectural patterns. In the past, it would have been too costly, so the projects were frequently deferred or cancelled.
Even with that, it's a really good idea to assume AI code is incorrect and to place a heavy emphasis on correctness. So now people are building the verification processes they always wish they had, because it's both needed and because it is high leverage.