The problem with Devin wasn't that it was a black box doing too much. It's that the outcome demo'd were fake and what was inside the box wasn't an "AI engineer."
Transparency? If it worked even unreliably, nobody would care what it does. Problem is stochastic machines aren't engineers, don't reason, are not intelligence.
I find articles attacking Ai but finding excuses in some mouse rather than pointing at the elephant, exhausting.