I think another is: Medium's technical staff don't really have much of a clue what they are doing, and the editorial staff don't really know what they want them to do.

They fiddled about with the Topic pages, 'Recommended' stories and 'Who to follow' recently, and 'hot mess' is an absurdly generous description of what they have done.

They simply don't have the means to matchmake articles and readers effectively, yet rely on suggesting that this is precisely how distribution works. It's simply not true: their analysis of what an article contains and what a reader might be interested in is as useless as Netflix saying 'If you liked Terminator, you'll also like Titanic'. Or worse.

That is why they devised the Boost system. Like a shop whose Internet has gone down fishing out the old slidey clunk-click credit card machine with the carbon paper.

Let's do it manually - our automated systems don't work.

It's true that (a) the workings of the algorithm will constantly shift and be tinkered with, and (b) even if they knew what it was doing and why (which, I would contend, they don't), they couldn't tell people too openly or else it would become more gameable.

But they rely on our assuming that this means that behind the curtain there is a system which has been intelligently and effectively coded, and produces logically satisfactory results.

I simply don't think that is the case, based on what gets pushed my way, and the way in which follow-up pieces I have written to similarly themed and tagged stories which had generated hundreds or thousands of reads have disappeared with barely half a dozen reads, even as the 'system' was sending me as a writer up to the top of their 'who to follow' rankings for that specific subject.

What Zulie said in her Medium Day talk was the theory of how it works (useful to anyone who has been on the platform for just a month or so, old hat for anyone else). The problem is that in practice, that simply does not happen.