Something like 90% of startups fail. Therefore, if articles were randomly sampled among people who start startups, we should see 90% of articles about failed startups and 10% about successful startups. But if anything it's closer to the reverse. So even though some people do write about their failures, it's definitely not a proportionate sample.