5. HARKing: "Hypothesis after results known" is also a problem to produce even "legitimate" papers by discarding the original hypothesis after negative findings and come up with a new one that fits the data. Its similar to p-hacking and also skews science into "promising" dead-ends.

> but seeing Nobel-prize-winning labs publish photoshopped blots

And they were caught because they edited poorly which could be detected automatically or by human eyes. Think about well versed image manipulators or generative AI...

I think this erosion of trust is the worst, AI can produce and it will propably come.