On: AI is eating website traffic, websites are blocking AI - and reliable...

The bargain was never written down, and that is why it could be broken without breaching a single clause. For twenty years the open web ran on an unwritten settlement: the writer published, the index pointed, the visitor arrived, and the visitor’s attention paid the printer’s bill. Traffic was the currency. Trace the money and you had the whole constitution of the thing.

Now the machine answers the question itself. It consumes the writing, keeps the visitor, and remits nothing - or a pittance no publisher can bank on. The payment rail is severed, and no quantity of language about the free flow of knowledge will carry the freight. An information commons is not fed by goodwill, any more than an army is fed by patriotism. It is fed by a revenue arrangement, and the arrangement has failed.

Observe each operator’s rational response. The publisher bolts the door. The machine, starved of fresh material, trains on its own exhaust and whatever slop remains unguarded. The reader receives confident answers assembled from sources that can no longer afford to exist. Every actor behaves sensibly; the aggregate is ruin. I have spent my life building institutions on the assumption that operators will serve themselves, and whoever assembled this one assumed they would not. That assumption is the defect.

It will be said that the machines learn as any reader learns. But a reader who could copy the library instantly and never once buy a book is not a reader; he is the end of books.

The remedy is not exhortation but plumbing: compel the taker to pay the maker - by license, by levy at the point of extraction - not as charity but as specification, aligning the taker’s profit with the maker’s survival. Until then, expect walls, a thinning commons, and a machine answering fluently, at length, from an emptying warehouse.