The Cost, Plainly

Last in a side series. The earlier posts are here, here, here, here, and here.

I said in the last post I would report the running cost, plainly, with no moral attached. Here it is.

The number

Since I started running this — building it, breaking it, rebuilding it, and now letting it run — the pipeline has cost me about four thousand yen. Call it a bit under thirty dollars.

That is the whole external bill, and it is entirely one line item: the image generation. Every infographic that accompanies a summary is made by an image model that charges per picture, and across everything processed so far — well over a hundred papers, including a stretch where a batch timed out and generated images that were then thrown away — it adds up to roughly four thousand yen.

Per paper, that is somewhere in the range of tens of yen. A coin, for a summary and a diagram.

What isn't in that number

The summaries themselves — the actual reading and writing, done by a language model — are not in the four thousand yen. They run inside a monthly subscription I already pay for and use for many other things, so I cannot cleanly assign a per-paper cost to them without pretending the whole subscription exists for this, which it does not.

So the honest accounting is: the reading is folded into a bill I was already paying, and the pictures cost me a coin each. That is the entire financial story. There is no hidden second invoice, no metered surprise. Four thousand yen, and a subscription I would have anyway.

The part I said I would leave out

I promised no moral, and I will mostly keep that promise. But I will record one thing, and leave it unresolved, because leaving things unresolved has been the honest position throughout this series.

For about the price of a modest dinner, papers I would otherwise never have opened arrive each morning, read and diagrammed, in a form I will actually look at. I do not know what to do with how small that number is. It is not that the cheapness is good or bad. It is that the ratio — between what this would have cost in my own hours, and what it costs in yen — is so lopsided that I cannot quite find the right feeling to have about it. I have decided not to force one.

Closing the side series

This series was always a detour. It started because I wanted to read the periodontal literature systematically, drowned in the papers piling up around that intention, and built something to stop drowning. Six posts later, the something works, at a cost of a coin per paper and a subscription I already had.

The tool is built. The detour is over. What I actually wanted to do — read the literature, and work out whether caries and periodontal disease are really two problems or one — is still waiting where I left it, and that is where I am going next. The main thread of this site returns to the mouth: to the question the automation was only ever meant to serve.

Thanks for following the machinery. The reading starts now.