Method signature hinting and other forms of autocomplete-like functionality are very actively harmful to productivity and prevent you from holding a mental model of the codebase. The larger the codebase, the worse this effect gets and the more critical it is to only approach writing code when you’ve got a sufficiently burned in mental model of the subsystem you’re dealing with, by searching and reading code with tools & interfaces wholly distinct from what you use for writing code.
It’s like the difference between someone who can navigate the highway system with interstate signs & atlases alone vs someone who is unable to navigate at all without a GPS system.
Even if the GPS system experiences no downtime, they still lack skills that make the driving lower quality, like anticipating changes or adjusting for weather.
Coders heavily reliant on IDE tools simply understand the ramifications of what they are writing less well, because they haven’t invested in the mental map of the code and mistakenly believe they can safely offload many parts of that cognitive requirement to an IDE.
Worst of all, they believe that code shipped this way is somehow reinforcing evidence that the IDE was a value additive tool, which is the wrong counterfactual for comparison. You have to consider the value lost from what better solution could otherwise have been shipped for lower cost using a different set of tools.