I’m trying to figure out how to get code completions to favor solutions that can be inferred from code introspection (i.e. old-school pre-AI style) instead of an AI prediction. (I’m using Python and iPython notebooks, but this applies to other languages as well).
To make up an example, if a library module has a ‘write_to_db()’ function, the AI might suggest calling ‘save()’ even though no such method exists on that module. The situation gets worse for named function arguments. In this circumstance, the AI is actively anti-productive and needs to be turned off.
I imagine that a careful adjustment of the settings would help, however terms like ‘suggestion’ and ‘completion’ seem to be used interchangeably or are not rigorously defined. This ambiguity is unclarifyed by modifier words, leading to “quick suggest” or “snippet suggestion” or “inline suggest” which sound very similar but are referenced in settings in such a way that implies they are distinctly different features.
(I don’t mean to be harsh… written in the spirit of constructive feedback. 