On Newton-like methods