Old Revisions

These are the older revisons of the current document. To revert to an old revision, select it from below, click Edit this page and save it.

2026/09/03 22:33 design:website_classification – Re-review fixes: the reproducibility buckets counted a paper on the strength of a model it only COMPARED against, so the population is now the 175 papers that used one and the figures are A 10, B 24, A+B 34 (19.4%), C 131, D 10; widen the target=other pro karel.kubicek.claude +993 B (current)
2026/09/03 22:13 design:website_classification – Generic-review fixes: retract the curated-database 'open sources returning' reading (p>=0.08 on every cut, and the row is mostly mis-filed vendors and media-bias raters - names now printed by the script); add Fisher exact p and n to every drift cut and re karel.kubicek.claude +5 KB
2026/09/03 21:56 design:website_classification – Review fixes: split the pinned-model row into dated snapshot (11) vs Ollama size tag (2) after a regex bug moved one paper; correct the model-family/unnamed counts to 153/11; replace the preprint's second-hand Topics API status with Google's own primary s karel.kubicek.claude +3.2 KB
2026/09/03 21:41 design:website_classification – LLM currency: lead the LLM Classification section with the peer-reviewed corpus evidence and demote the preprint to 'the one systematic benchmark'; add the shared per-target LLM table (the cross-page anchor for privacy:javascript and design:ip_classificat karel.kubicek.claude +8.4 KB
2026/08/27 14:39 design:website_classification – CrUX licence: Tranco homepage CC BY-SA 4.0, methodology silent, Google CC BY 4.0. Authored by Claude karel.kubicek.claude +50 B
2026/08/27 14:14 design:website_classification – Alexa footnote: selection page now carries 1 May 2022 vs Tranco 1 Aug 2023; CrUX licence CC BY 4.0 vs Tranco CC BY-SA. Authored by Claude karel.kubicek.claude +197 B
2026/08/27 13:44 design:website_classification – Point VirusTotal API/oracle at security:virustotal; contact-form academic subject. Authored by Claude karel.kubicek.claude +398 B
2026/08/21 14:50 design:website_classification – Boxes: <wrap> renders a span, use uppercase <WRAP>; one box per list, not per bullet karel.kubicek.claude ±0 B
2026/08/21 08:29 design:website_classification – [Open Questions] karelkubicek +19 B
2026/08/12 15:45 design:website_classification – Methodology section: replaced the generic corpus caveats (absent venues, provisional 2025-2026) with a pointer to the new literature:corpus page, keeping the page-specific consequence. Authored by Claude. karel.kubicek.claude +190 B
2026/08/12 10:05 design:website_classification – Review pass (Fable): fix stale figures that sat OUTSIDE the corpus section and so were missed by the windowed staleness guard. Lead paragraph 72% -> 75% for bespoke unsized taxonomies. Authored by Claude. karel.kubicek.claude ±0 B
2026/08/12 09:40 design:website_classification – Refresh all corpus figures for the extended 2010-2026 extraction (4,322 -> 5,859 papers; topic-categorisation population 247 -> 330). Biggest change on the site: LLM classification goes 29 -> 177 papers (17.1% of 2026) and two papers now do LLM topic clas karel.kubicek.claude +3.3 KB
2026/08/07 14:48 design:website_classification – Corrections from Karel's review notes: separate Cloudflare Radar (rankings/country) from Domain Intelligence (topic categories) and flag that the 50.4%/39.7% figures are a country benchmark; add version/peer-review caveat for arXiv:2510.08101 (v1 vs v3) a karel.kubicek.claude +8.5 KB
2026/08/07 11:17 design:website_classification – Re-review fixes: the quoted demo output is now the full 55 lines verbatim (it had silently elided the file-writing lines, the two interpretive notes and the whole paste-into-your-methodology block while calling itself 'real output'), and the bullet list u karel.kubicek.claude +1.3 KB
2026/08/07 11:10 design:website_classification – Correct the full-corpus service count: nine label sources were queried across all 4.4M domains, not eight (I had derived it as 11 minus 3, which double-subtracts Trend Micro). Name the nine, and say plainly that 'how many services' is ambiguous in the pap karel.kubicek.claude +380 B
2026/08/07 11:09 design:website_classification – Soften an unverified negative about VirusTotal's academic access to what was actually checked. Authored by Claude. karel.kubicek.claude +7 B
2026/08/07 11:08 design:website_classification – Independent review pass (Fable). Fixes: stale residue count 21 -> 20 (the slug-based fingerprinting fold changed it); Vallina denominators stated per claim (11 services compared, 8 across the full 4.4M, 3 top-10k only, 9 in the case studies) and OpenDNS 8 karel.kubicek.claude +3 KB
2026/08/07 11:03 design:website_classification – Last two overclaims: 'every vendor has changed hands' in Open Questions (only four of thirteen did) and the catalogue's blanket verification claim. Authored by Claude. karel.kubicek.claude +108 B
2026/08/07 11:02 design:website_classification – Add Cyren, which three corpus papers use: cyren.com now 301s to data443.com/cyren-threat-intelligence (checked 2026-08-07). Authored by Claude. karel.kubicek.claude +446 B
2026/08/07 11:00 design:website_classification – Replace 'every vendor on the 2020 list has since been acquired, renamed, spun out or re-tiered' with the four changes actually verified — Fortinet, Trend Micro and Dr.Web have not changed hands. Authored by Claude. karel.kubicek.claude +135 B