Content Quality: Well structured News piece (843 words; policy News range 400-1200, satisfied). Title is 133 characters (cap 150, satisfied). Balanced Overview / What We Know / What We Don't Know. Covers the three headline changes plus the other security entries accurately.
Source Verification: Read all three snapshots from disk (gunzip). source-0.html.gz (openssh.org/txt/release-10.6, plain-text release notes): confirms 'sshd(8), ssh(1): disable LZ77 dictionary coder to mitigate the side-channel leaks described in "Crossing the Streams: SSH Plaintext Recovery via a Common Compression Context in Multiplexed Channels" by Fabian Baumer and Marcus Brinkmann, preprint https://arxiv.org/abs/2609.07709'. What is disabled is the LZ77 dictionary coder in ssh and sshd (the Compression option is not removed, only made 'less effective'); article states this correctly. Verbatim quotes verified: 'to mitigate the side-channel leaks', 'will reduce the effectiveness of the Compression option', 'completely immune to this type of attack', 'a fragile optimisation', 'for now, be making more frequent releases', 'very much welcome these reports, especially when combined with human triage, analysis, test-cases'. Notes say 'enable hybrid post-quantum ssh-mldsa44-ed25519 signature algorithm' under New features, with the '@openssh.com' suffix dropped and old experimental keys needing regeneration; the notes do not say 'default' for the algorithm and the article does not claim it (only WarnWeakCrypto is described as default, which matches the notes). scp -R: 'begin deprecating the -R flag'; from 10.6 still works but emits a deprecation warning to stderr; 'in a future release, the option will be ignored' with no date; article matches. Other security bullets (sftp path validation, GSSAPI, username $ and backslash, ssh-keygen DST, sshd-session root on QNX 6/SCO OpenServer 5, KDF rounds 24 to 32, AgentSocketPath, max-pk-ok default 6, sftp mkdir -p) all match. Chris Rohlf 'in collaboration with Claude and Anthropic Research' credit appears on the 'none' options entry and the early key/CA signature type check entry, as the article says. source-1.html.gz (lwn.net/Articles/1098980, 'OpenSSH 10.6 released', posted Oct 6 2026): confirms hybrid ssh-mldsa44-ed25519, sftp -p, LZ77 disable, scp -R deprecation and the AI-report/release-cadence point. source-2.html.gz (arxiv.org/abs/2609.07709, v2 revised 16 Sep 2026): title, authors Fabian Baumer and Marcus Brinkmann, ID, 'accepted at ACM CCS 2026' in Comments all match. The abstract reads: 'in the lowest-noise scenario, an 8-character secret over a 26-letter alphabet can be recovered using at most 276 guesses'; the article quotes it verbatim and keeps the 'lowest-noise scenario' qualifier, so it does not generalise the figure. The snapshot is v2 (typo fix only in a table), the number is present. No CVE identifier appears in any source and none is claimed in the article. suspicious_patterns is null for all three sources; nothing to evaluate.
Factual Accuracy: All claims trace to the cited sources. Title phrase 'Over a Plaintext-Recovery Attack' is a writer's compression of the paper's title ('SSH Plaintext Recovery') and the notes' 'chosen-plaintext attack'; it is supported, though slightly informal. Minor observation only: the title says 'Enables' for the signature algorithm, matching the notes' 'enable'; the notes do not say whether it is a default key type, and the article does not claim so.
Overall Assessment: Accurate, well-sourced, all quotes verbatim, numbers properly qualified, no unsourced CVE, no defects under title-length or word-count policy. Approve.