Differences

This shows you the differences between two versions of the page.

--- programming:crawler:llm_agents [2026/08/29 06:37] – New page: crawling with an LLM agent. 5 of 1,120 crawling papers drove a crawl with one, all 2026; completion inflation, per-site cost, prompt sensitivity, run-to-run variance, tool currency as of today, and what to report. Authored by Claude karel.kubicek.claude
+++ programming:crawler:llm_agents [2026/09/21 13:43] (current) – Quote bullet: the 9 below-threshold quotes split 2 rescued from the PDF / 7 below in both (two papers). Authored by Claude karel.kubicek.claude
@@ Line 164: / Line 164: @@
   * **Counts are of papers, never tuples**, and shares are of the population named in the heading.
   * **Everything here rests on the provisional years.** Nine papers is a small number and 2026 is an incomplete year. Treat the //direction// as real and every //rate// as provisional; the honest claim is "this had not happened yet by 2026", not "this will not happen".
-  * **Quotes were checked.** All 15 evidence quotes behind the family were matched against the paper full text: 3 exact, 3 partial after normalising whitespace and hyphenation, 9 below threshold — and all nine were then read by hand and **all nine are present in the source**, failing only because the extraction dropped bracketed citation markers or because two-column reading order interleaved the sentence. The adjudication is on the provenance page.
+  * **Quotes were checked.** All 15 evidence quotes behind the family were matched against the paper full text: 3 exact, 3 partial after normalising whitespace and hyphenation, 9 below threshold — of which **2 are located in an independent ''pypdf'' rendering of the same PDF** and **7** are below threshold in both renderings. All nine were also read by hand and **all nine are present in the source**, failing only because the extraction dropped bracketed citation markers (the 7) or because two-column reading order interleaved the sentence (the 2). The seven are two papers, one sentence of which is quoted six times. The adjudication is on the provenance page.
   * **Venue coverage.** Seven venues. Agent research overwhelmingly publishes at NeurIPS, ICML, ICLR, ACL and CHI, none of which are in this corpus. **A count of nine is a statement about security, privacy and measurement venues, not about the world.** The frameworks in the table above are real and widely used outside it.