https://arxiv.org/abs/2606.28344. The end of web parsing. The beginning of scalable pixel-native search. link: https://pixelrag.ai/
feat(web): ask-mode examples ChatGPT verifiably gets wrong (#20)
Zhifei Lidaeeb167c4b8
diff --git a/web/app/chat/page.tsx b/web/app/chat/page.tsxindex 65fc1c1..5a61957 100644--- a/web/app/chat/page.tsx+++ b/web/app/chat/page.tsx@@ -56,9 +56,9 @@ interface ChatMessage {}const EXAMPLES = [- { q: "Explain Van Gogh's The Starry Night", icon: "01" },- { q: "How do you read the periodic table?", icon: "02" },- { q: "Tell me about the Taj Mahal", icon: "03" },+ { q: "How many shots on target did Inter have in the 2010 Champions League final?", icon: "01" },+ { q: "Which district in Nagaland has the RTO code NL-03?", icon: "02" },+ { q: "Explain Van Gogh's The Starry Night", icon: "03" },{ q: "介绍一下兵马俑", icon: "04" },]diff --git a/web/components/SearchBar.tsx b/web/components/SearchBar.tsxindex e7857d1..6e20ceb 100644--- a/web/components/SearchBar.tsx+++ b/web/components/SearchBar.tsx@@ -19,10 +19,13 @@ const EXAMPLE_QUERIES = ["兵马俑",]+// The first two are verified "ChatGPT gets these wrong" questions: the answers+// live in table/infobox cells (match-stats table, RTO code box) that parametric+// memory and text scraping both fumble — but reading the rendered page nails.const ASK_EXAMPLES = [+ "How many shots on target did Inter have in the 2010 Champions League final?",+ "Which district in Nagaland has the RTO code NL-03?","Explain Van Gogh's The Starry Night",- "How do you read the periodic table?",- "Tell me about the Taj Mahal","介绍一下兵马俑",]