What the cloud could not open

Where this portion came from

The cloud chats of 26 and 27 September worked without a browser. Some documents they could not open because of Cloudflare, reCAPTCHA or another bot check, although the documents themselves are open. “6 в 1” gathered them into group A of the unreachable-sources card: thirteen items that a local chat with the browser pane would most likely read on its own. Eight were read, and eight records changed. There are eleven new sources, each read in the pane or by the shell.

What was found

Four colours. Appel and Haken’s announcement in the Bulletin of the AMS, two pages, communicated on 26 July 1976. The article page asks for a subscription, but the PDF itself is served. The set of configurations there is “fewer than 2000”. The authors did the discharging by hand, and programs checked the reducibility. The authors write that the theorem could hardly be proved by these methods without massive computation.

The Robbins problem. The preprint on McCune’s page answers 404, but his final version survived in the Wayback Machine as PostScript. The paper was submitted on 22 January 1997. The attack that gave the solution ran from 6 September to 11 October 1996, on three ordinary workstations. The year of the conjecture itself is not there, only “the 1930s”. Citing a phone conversation with Robbins, the paper refutes the folklore that the problem grew out of Huntington’s error.

Leeds, 1972. PMC serves the companion paper by Horrocks and colleagues as images of the scanned pages. The record said plainly what it did not claim: whether the system was Bayesian, how many patients its data held, and what machine it ran on. All three are known now: Bayesian, 600 patients, an English Electric KDF9 with a desk-top machine as back-up.

Davis and Putnam. Both papers, of 1960 and 1962, are free on the ACM site. The site answered the cloud with 403, and this time the pane opened it without any check. The rules are now given in the authors’ own words, not by the 1961 report.

MIR. Springer closes the full texts of Kibernetika but gives the first two pages. Two things were there. Pogrebinskii, of the Institute of Cybernetics, dated MIR to 1965 and MIR-1 to 1966 in 1982, which makes the museum’s two dates easier to read. And the 1971 paper on Analitik has nine authors and builds the language on MIR’s input language.

Kimi K3, Musk, Suno. Moonshot’s posts on X confirmed both K3 dates, 16 and 27 July. Musk’s post on the “calendar technicality” went up on the evening of 18 May. The Munich court’s judgment in GEMA v Suno is in the Bavarian database of decisions in full, 559 paragraphs, and the cloud was not let in. It shows that only one claim was rejected, that damages cover three songs and one refrain, and that the appeal is already before the Munich Higher Regional Court.

What is left

Nothing new turned up on the Mark I: DTIC is under maintenance, the Internet Archive copy is dark, and the Minnesota finding aid answers 403. Gelernter’s Paris paper of 1959 exists only in a book the Internet Archive lends. Glushkov’s 1970 paper is behind a subscription. Two documents, the Philippine DICT’s release on lifting the Grok block and Caceres’s 1962 paper in Circulation, stand behind a Cloudflare check with a box only a person may tick. They went into a card for “6 в 1”.