llms.txt and index.html's meta descriptions still claim "up to 97% bias reduction" - the real max, after the AI Fair Recruitment fix, is 96.1%

[yakew7]

Where: llms.txt line 3; index.html lines 13 (og:description), 27 (twitter:description), and 2002's search-bar placeholder.

The gap: All three lines say "up to 97% bias reduction". This reflects the AI Fair Recruitment audit's pre-fix numbers (4.51% → 0.12% = 97.3% reduction). Commit 715e4f3 ("AI Fair Recruitment scripts mislabeled Other-gender applicants as Female") changed the real gap to 4.03% → 0.16%, a 96.1% reduction - and every other place in the repo was updated to match: README.md's Results-at-a-Glance table, index.html's own JSON-LD description, its roadmap card, and its project-card searchable text - which is now the highest reduction figure documented anywhere in the repo. Only these three "up to 97%" lines were missed by that fix.

As a knock-on effect, index.html's search-bar placeholder (placeholder="search experiments… e.g. race, hiring, 97%, COMPAS") now suggests a search term ("97%") that matches none of the project cards' data-project-text attributes - typing the example into the search bar now returns zero results.

Repro:

grep -n "up to 97%" llms.txt index.html
# llms.txt:3, index.html:13, index.html:27 - all still "up to 97%"

grep -n "96.1%" README.md index.html
# README.md: | 4.03% → 0.16% | **96.1%** |
# index.html: "a 96.1% reduction." / "96.1% bias reduction"

grep -o 'data-project-text="[^"]*"' index.html | grep -c '97%'   # 0 - no card mentions 97%

Fix direction: Change all three "up to 97%" mentions to "up to 96%" (or "96.1%"), matching the AI Fair Recruitment audit's current number, and update index.html's search placeholder example away from "97%" (e.g. to "96.1%" or "82%").

[yakew7]

@thadidaniel-ctrl could you just send comments in the issues u want assigned to you! i will assign you this issues and the others you send comments on including this

[thadidaniel-ctrl]

ok sir !

[yakew7]

Fixed in c28b852: updated index.html's og:description/twitter:description and llms.txt's "up to 97% bias reduction" claim to the real current max, 96.1% (AI Fair Recruitment, post-#718 fix), and the search-bar placeholder to match.