A year ago, I concluded that we had failed in our quest to make scholarship open access (OA): the race had been won by pirates like SciHub (Green, 2017). Twelve months on, how do things look? Key points: We’re still failing to deliver open access (OA): around a fifth of new articles will be born free in 2018, roughly the same as in 2017. Librarians, funders and negotiators are getting tougher with publishers but offsetting, ‘Publish and Read’, deals based on APCs won’t deliver OA for all or solve the serials crisis. The authors of Budapest, Bethesda and Berlin OA declarations foresaw three changes with the coming of the internet. Flipping to a barrier to publish (APCs) from a barrier to read (subscriptions) wasn’t one of them. By itself, OA won’t reduce costs to solve the serials crisis: a digital transformation of scholarly communications based on internet-era principles is needed. Following the internet-era principle of ‘fail-fast’, what if papers are first posted as preprints and only if they succeed in gaining attention will editors invite submission to their journal In clinging onto traditional journals to advance the careers of the few (authors), OA is delayed for the many (readers): rebuilding the reputation economy to accept preprints could be the catalyst to deliver OA, solve the serials crisis and drive out predatory journals

Author's Note In keeping with the proposition in this paper, and following the advice of Pippa Smart, Editor of Learned Publishing who saw an early draft, I'm releasing it first as a preprint to test if it 'fails fast' or not. I will do my best to promote it so that it gains an audience and I invite readers to comment, propose improvements and point out where I've gone wrong. I also invite journal editors to consider whether it has 'succeeded' and if in their opinion it has I look forward to being asked to submit it for peer review and formal publication in due course.