A - “true & useful, AND new to me”

B - “true & useful, but NOT novel”

U - “unclear, undefined, unknowable, or just flat out wrong”

This clearly defines my experiences with codegen ai - 1% A, 30-50% B (true, but not useful, like telling me the --count parameter sets the count) and the rest U - loads and loads of complete bullshit, sprinkled with enough confidence and subject-appropriate wordage that it looks valid, but someone who actually knows what they're doing will find the flaws - ironic, since it's people who _don't_ know what they're doing who are the major target audience