A coin toss will give the correct answer 50% of the time. The Gods of academia will give you a correct answer 30% of the time.
More straw man while showing your backside and that you don't know anything about statistics.
Did you read the Nature article about where the 70% came from?
Did you understand and note the key reasons for the lack of repeatability? Such as pressure to publish often tied to degree or grant objective (sometimes combined) which leads to incomplete methodology documentation to copy by other researchers? Simple lack of publication room to go through every methodology nuance and process. Lack of incentive to truly copy anyhow because there's little fame or name recognition (or degree granting) for not coming up with something novel? Lack of deeper blind peer review of statistical methods to make sure they are suitable? And those that have done this for a living also know that failure is sometimes as simple as having 94.9% correlation about some factoid (which in itself is somewhat arbitrary) is considered failure even though it should more correctly be considered strong support.
None of these processes is even a factor or even attempted in most other "sources" of information-- And honestly most of the big ideas of interest to the public aren't about the vast majority of research--it's about big things.