AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.
Rust agent-memoryai-agentsbm25claude-codeclicode-searchcursordatabaseelasticsearchfull-text-searchhybrid-searchjsonllocal-firstmcpmcp-servermodel-context-protocolragrustsearch-enginevector-search
polish(site): rc.80 gates post lede names rc.80 again (#1551)
xerj-org0225cdcd21f9
diff --git a/landing/blog/the-rc80-gates.html b/landing/blog/the-rc80-gates.htmlindex 41a6437b2..8ac2d72b1 100644--- a/landing/blog/the-rc80-gates.html+++ b/landing/blog/the-rc80-gates.html@@ -142,7 +142,7 @@<section class="verdict-panel" id="verdict"><div class="vh"><span class="wn">THE VERDICT</span><span class="wm">split, measured, with the losses left in</span></div>- <p>The program behind rc.81 shipped ten features, and every item that could carry a number was filed with a measured gate that binds at ship time. This post is the packaging half of that rule, and the scoreboard is a split. <strong>Three gates passed outright</strong>: <code class="inline">/_ask</code> turned 230 prompt/gold pairs into validated DSL at macro-F1 <strong>0.9975</strong> with zero invalid queries and wall-clock p50 <strong>1.129 ms</strong> against a 300 ms budget; <code class="inline">xerj_map</code> took unknown-field 400s on MCP-shaped DSL from <strong>20/30 guessed to 0/30 map-informed</strong> — a 100% reduction against a ≥80% bar; and the calibration layer took FiQA rerank ECE from <strong>0.3109 raw to 0.0088 held-out</strong> against a ≤0.10 bar. <strong>Three failed with the numbers</strong>: the local judge loses to its own first stage everywhere it was measured (SciFact <strong>0.7045 → 0.5967</strong>), the decision flywheel answers <strong>19.72%</strong> of later traffic at ≥0.8 confidence against an 80% bar, and the tier-2 decide head misses its accuracy bars by 3 to 59 points. <strong>One split</strong>: stemming gained +0.0160/+0.0179 BM25 nDCG@10 (bar +0.01) and still left 15 zero-hit queries against a bar of ≤10 — and no stemmer can fix the 15. The failures filed four issues, <strong>#1091–#1094</strong>. No number below was tuned after its gate was read.</p>+ <p>The program behind rc.80 shipped ten features, and every item that could carry a number was filed with a measured gate that binds at ship time. This post is the packaging half of that rule, and the scoreboard is a split. <strong>Three gates passed outright</strong>: <code class="inline">/_ask</code> turned 230 prompt/gold pairs into validated DSL at macro-F1 <strong>0.9975</strong> with zero invalid queries and wall-clock p50 <strong>1.129 ms</strong> against a 300 ms budget; <code class="inline">xerj_map</code> took unknown-field 400s on MCP-shaped DSL from <strong>20/30 guessed to 0/30 map-informed</strong> — a 100% reduction against a ≥80% bar; and the calibration layer took FiQA rerank ECE from <strong>0.3109 raw to 0.0088 held-out</strong> against a ≤0.10 bar. <strong>Three failed with the numbers</strong>: the local judge loses to its own first stage everywhere it was measured (SciFact <strong>0.7045 → 0.5967</strong>), the decision flywheel answers <strong>19.72%</strong> of later traffic at ≥0.8 confidence against an 80% bar, and the tier-2 decide head misses its accuracy bars by 3 to 59 points. <strong>One split</strong>: stemming gained +0.0160/+0.0179 BM25 nDCG@10 (bar +0.01) and still left 15 zero-hit queries against a bar of ≤10 — and no stemmer can fix the 15. The failures filed four issues, <strong>#1091–#1094</strong>. No number below was tuned after its gate was read.</p></section><p class="uc-lead">One window, ten features, eight gates, all measured on 2026-09-30 against bars that were written down before the work started. The wins and the losses sit in the same tables, every figure links to the run that produced it, and the four defects the gates exposed are filed in the open tracker with the measurements that found them. That is the whole post.</p>