<?xml version="1.0" encoding="UTF-8"?>
<oembed>
  <version>1.0</version>
  <type>link</type>
  <title>EQ-05: Direct Step-Level Routing vs. Confidence Cascades (Paul &amp; Nandy, 2026; Chen et al., 2023) (Hong Kong COO Practice) | ANDYGROVE.SYSTEMS</title>
  <author_name>ANDYGROVE.SYSTEMS Continuing Professional Development &amp; Operations Research Collective</author_name>
  <author_url>https://andygrove.systems/education</author_url>
  <provider_name>ANDYGROVE.SYSTEMS — High Output Operations &amp; Fresh Preprints Upskilling Hub</provider_name>
  <provider_url>https://andygrove.systems</provider_url>
  <cache_age>86400</cache_age>
  <thumbnail_url>https://andygrove.systems/og-card.png</thumbnail_url>
  <thumbnail_width>1200</thumbnail_width>
  <thumbnail_height>630</thumbnail_height>
  <url>https://andygrove.systems/hong-kong/eq-05</url>
  <description>EQ-05 (Direct Step-Level Routing vs. Confidence Cascades (Paul &amp; Nandy, 2026; Chen et al., 2023)) applied by IE/OR specialists in Hong Kong: C_route(k) = (1 − p_F(θ_k)) c_S T_k + p_F(θ_k) c_F T_k   vs.   C_casc(k) = c_S T_k + p_def(θ_k) c_F T_k. A &lt;5ms step classifier (AgentRouter) dispatches routine tool/parsing steps directly to c_S and hard synthesis directly to c_F without paying c_S twice (~72% cost cut at 97.3% frontier accuracy). | Fresh Preprints: https://andygrove.systems/preprints | 42-Hr CPD Hub: https://andygrove.systems/education (Independent platform; not affiliated with Intel Corporation or the Grove family)</description>
</oembed>