You need the same (and probably more) robust controls in place as you would use for a more junior developer.
A human will be influenced by the potential outcome of any proposed solution. For example they might worry about the impact to their employment status if they are wrong, while an LLM will just spit out an answer that is "most likely correct".
> It can't use European ones because of different wheelbase.
Just a quick point of clarification for anyone trying to look this up. The correct term is "gauge" (distance between the rails) instead of "wheelbase" (distance between front and rear wheels). Most European countries use tracks with a gauge of 1,435mm but Ukraine uses 1,520mm.
Is there any evidence he's ever driven (or even ridden in) a production unit?