docs(blog): bind context comparisons to their measured baselines

apps/presentation/site/public/blog/context-delivery/index.html

@@ -6,18 +6,21 @@
 <meta property="og:type" content="article"><meta property="og:site_name" content="LoopX"><meta property="og:locale" content="en_US"><meta property="og:title" content="An Agent needs a way to look things up, too"><meta property="og:description" content="One voice task, from choosing facts to fitting the first read, preserving exact values, and recovering without replaying work."><meta property="og:url" content="https://loopx-project.github.io/loopx/blog/context-delivery/"><meta name="twitter:card" content="summary"><link rel="icon" href="data:,"><link rel="stylesheet" href="../blog.css">
 </head><body><a class="skip-link" href="#main">Skip to content</a>
 <header class="site-header"><a class="brand" href="../../"><span class="product-mark" aria-hidden="true"><i></i><i></i></span>LoopX</a><nav aria-label="Main navigation"><a href="../">Blog</a><a href="../../docs/book/">Docs</a></nav><div class="header-actions"><a class="language" href="../zh/context-delivery/" lang="zh-CN" hreflang="zh-CN">中文</a><a class="github" href="https://github.com/loopx-project/loopx">GitHub ↗</a></div></header>
-<main id="main"><div class="article-heading"><a class="back" href="../">← All articles</a><p class="eyebrow">Context delivery</p><h1>An Agent needs a way to look things up, too</h1><p class="lead">One voice task, from choosing facts to fitting the first read, preserving exact values, and recovering without replaying work.</p><div class="byline"><a href="https://github.com/loopx-project/loopx">LoopX</a><span>·</span><time datetime="2026-11-01">November 1, 2026</time><span>·</span><span>Updated November 10</span></div></div>
-<div class="article-layout"><aside class="toc"><details><summary>On this page</summary><nav aria-label="On this page"><a href="#multitask-context-byte-ratios">8-task context byte ratios</a><a href="#ablation-evidence">Multi-task and layered ablation</a><a href="#example">Follow one task: give an assistant voice input</a><a href="#known-paths">When you know the locations, read them together</a><a href="#discover">When the structure is unfamiliar, index the relevant subtree</a><a href="#requirements">See requirements first, then open their evidence</a><a href="#stage-scopes">Read scope before choosing</a><a href="#sharing">Share exact records within one delivery</a><a href="#cycle">On the next continuation, deliver what changed</a><a href="#native-path">When enabled, the default output is a reading entrypoint</a><a href="#first-read-budget">The first response has to arrive complete</a><a href="#transport-cost">Count the trips and the copies</a><a href="#recovery">Recover missing context while preserving the original work</a><a href="#evidence">What the measurements establish</a><a href="#reproduce">Read the same material yourself</a></nav></details></aside><article class="prose"><p>Long-running work accumulates material: the original goal, later constraints, attempted routes, failed experiments and current progress. Sending everything again on each continuation repeats more and more content; keeping only a summary can make the original basis of a decision difficult to recover.</p>
-<p>LoopX's context delivery separates these jobs: <strong>preserve complete sources, read the facts needed for the current decision, expand supporting material around questions, and reuse unchanged structures exactly within a valid compaction cycle.</strong></p>
-<h2 id="results">Results first: progressive reading, deltas and task ablation</h2>
-<h3 id="multitask-context-byte-ratios">Eight tasks: progressive reading and in-cycle deltas</h3>
-<p>This section recalculates the published eight-task Four-Arm Context ON 18h aggregates to distinguish Progressive and in-cycle Delta transport costs. The source reports 213 model-facing wakes: 101 new-session or post-compaction baseline reads and 112 in-cycle continuations. Per-task cells retain the published values; totals are recomputed from those cells. Raw per-wake logs are not included here.</p>
+<main id="main"><div class="article-heading"><a class="back" href="../">← All articles</a><p class="eyebrow">Context delivery</p><h1>An Agent needs a way to look things up, too</h1><p class="lead">Read necessary facts first, reuse verified structures within a compaction cycle, and recover the complete baseline when continuity is lost.</p><div class="byline"><a href="https://github.com/loopx-project/loopx">LoopX</a><span>·</span><time datetime="2026-11-01">November 1, 2026</time><span>·</span><span>Updated November 12</span></div></div>
+<div class="article-layout"><aside class="toc"><details><summary>On this page</summary><nav aria-label="On this page"><a href="#four-stage-cascade">Four mechanisms and their baselines</a><a href="#multitask-context-byte-ratios">8-task single-layer vs stacked</a><a href="#ablation-evidence">Multi-task and layered ablation</a><a href="#example">Follow one task: give an assistant voice input</a><a href="#known-paths">When you know the locations, read them together</a><a href="#discover">When the structure is unfamiliar, index the relevant subtree</a><a href="#requirements">See requirements first, then open their evidence</a><a href="#stage-scopes">Read scope before choosing</a><a href="#sharing">Share exact records within one delivery</a><a href="#cycle">On the next continuation, deliver what changed</a><a href="#native-path">When enabled, the default output is a reading entrypoint</a><a href="#first-read-budget">The first response has to arrive complete</a><a href="#transport-cost">Count the trips and the copies</a><a href="#recovery">Recover missing context while preserving the original work</a><a href="#evidence">What the measurements establish</a><a href="#reproduce">Read the same material yourself</a></nav></details></aside><article class="prose"><p>Long-running work accumulates material: the original goal, later constraints, attempted routes, failed experiments and current progress. Sending everything again on each continuation repeats more and more content; keeping only a summary can make the original basis of a decision difficult to recover.</p>
+<p>Selective JSON Pointer lookup by itself is straightforward—it folds long-tail history not needed for the immediate decision. The core mechanism in LoopX's context delivery is <strong>compaction-cycle-bounded structured delta context</strong>: complete sources are saved first, unchanged typed subtrees are reused exactly against a verified baseline within the active <code>context_epoch</code>, and any host compaction or digest mismatch atomically resets to a fresh complete baseline rather than chaining dangling deltas.</p>
+<h2 id="results">Evidence: separate context comparisons, cycle deltas and task ablation</h2>
+<h3 id="four-stage-cascade">Four mechanisms, with separate measurement baselines</h3>
+<p>Complete capture, Progressive views, compaction-cycle structured Delta and a bounded <code>--initial-read</code> are composable mechanisms, but their bytes do not decrease monotonically for every input. Delta requires a verified same-cycle baseline; a new epoch, mismatch or smaller full representation uses complete delivery. The archived tutorial below measures <strong>three independent comparisons, not a same-source four-stage cascade</strong>.</p>
+<figure class="diagram" data-comparison-progressive="5107/57014" data-comparison-cycle-delta="2164/5543" data-comparison-manual-select="674/51547"><p class="diagram-label">Separate tutorial comparisons and 18h compaction-cycle Delta observations</p><a href="../images/context-delivery/context-cascade-en.svg" target="_blank" rel="noopener" aria-label="Open the independent context comparisons at full size"><img src="../images/context-delivery/context-cascade-en.svg" alt="Independent tutorial comparisons: Progressive view 5,107 / 57,014 B (9.0%); in-cycle Delta 2,164 / 5,543 B (39.0%); manual selection 674 / 51,547 B (1.3% body). ANN and eight-task archived cohorts retain separate baselines." width="1440" height="690" loading="lazy" style="width:100%;height:auto;border-radius:10px;margin:8px 0 12px 0"></a><figcaption>A: Three archived tutorial comparisons from <code>measurements.json</code>: the necessary view is 5,107 / 57,014 B (9.0%); the next capture's Delta file is 2,164 / 5,543 B of its full Progressive delivery file (39.0%); manually selecting three paths from the separate 51,547 B <code>capture.json</code> returns 674 B (1.3% body), or 1,362 B including CLI metadata (2.6%). The 674 B is <strong>not</strong> a measured <code>--initial-read</code> after Progressive + Delta. These saved observations belong to the revision recorded in <code>measurements.json</code>; output size may change with source revision and reader paths. B1: ANN's archived 18h pairs compare Delta against selected packets, with 38.54% reduction across 46 events and 74.25% across 24 in-cycle Delta events. B2: Eight-task aggregates retain the separate 213-wake and 112-continuation denominators; their percentages cannot be chained into this tutorial.</figcaption></figure>
+<h3 id="multitask-context-byte-ratios">Eight tasks: single-layer payloads (Progressive / lossless Delta file) vs stacked first-read</h3>
+<p>When reading the published eight-task Four-Arm Context ON 18h table below, separate <strong>single-layer file payloads</strong> from the <strong>stacked first-read entrypoint</strong>. Across 213 model-facing wakes (101 new-session or post-compaction baseline reads and 112 in-cycle continuations), the published “In-cycle Delta payload” column (4.49 MB / ≈ 12.53 MB = <strong>35.8%</strong>) measures the <strong>unprojected lossless <code>delivery.json</code> file on disk (All → Delta)</strong>—a parallel single-layer cut alongside “Progressive projection” (7.29 / 22.41 MB = <strong>32.5%</strong>, All → Progressive), not stacked <code>Progressive + Delta</code>. Both mechanisms stack in the final column, <strong>Prog+Delta first-screen (3.0% body / 5.0% CLI)</strong>. Per-task cells retain the published values; totals are recomputed from those cells without raw per-wake logs.</p>
 <ul>
-<li><strong>Progressive reading:</strong> across all 213 wakes, full packets fall from <strong>22.41 MB to 7.29 MB (−67.5%)</strong>. The projection retains 32.5% of original bytes, with a per-task range of 26.5%–39.0%; unselected candidates and historical detail remain available on demand.</li>
-<li><strong>In-cycle deltas:</strong> across 112 continuations, corresponding full packets fall from <strong>approximately 12.53 MB to 4.49 MB (−64.2%)</strong>. Deltas retain 35.8% of same-wake full bytes, with a per-task range of 20.7%–47.1%.</li>
-<li><strong>Progressive + delta first read:</strong> for those same 112 continuations, the <code>--initial-read</code> body carries <strong>3.0%</strong> of original bytes, or <strong>5.0%</strong> including the CLI JSON envelope (−97.0% / −95.0%). First-read size is not whole-episode model input.</li>
+<li><strong>Single-layer progressive view (All → Progressive):</strong> across all 213 wakes, full packets fall from <strong>22.41 MB to 7.29 MB (−67.5%)</strong>, retaining 32.5% of original bytes (26.5%–39.0% per task) while keeping unselected candidates and historical detail expandable on demand.</li>
+<li><strong>Single-layer lossless delta file (All → Delta):</strong> across 112 continuations, unprojected lossless delta files fall from <strong>approximately 12.53 MB to 4.49 MB (−64.2%)</strong>, retaining 35.8% of same-wake full bytes (20.7%–47.1% per task).</li>
+<li><strong>Stacked entrypoint (Progressive + Delta First Read):</strong> for those same 112 continuations, stacking progressive selection with compaction-cycle deltas brings the <code>--initial-read</code> body to <strong>3.0%</strong> of original bytes, or <strong>5.0%</strong> including the CLI JSON envelope (−97.0% / −95.0%). First-read size is not whole-episode model input.</li>
 </ul>
-<figure class="diagram"><p class="diagram-label">8-task Four-Arm Context ON byte ratios (N = 213 wakes, 112 in-cycle Deltas)</p><a href="../images/context-delivery/multitask-byte-ratios-en.svg" target="_blank" rel="noopener" aria-label="Open the 8-task Progressive vs. Delta control-plane byte ratio chart at full size"><img src="../images/context-delivery/multitask-byte-ratios-en.svg" alt="8-Task Four-Arm Context ON 18h Control-Plane Byte Share: Progressive 32.5% (-67.5%), In-Cycle Delta 35.8% (-64.2%), Prog+Delta First-Screen 3.0% / 5.0%" width="1380" height="680" loading="lazy" style="width:100%;height:auto;border-radius:10px;margin:8px 0 12px 0"></a><figcaption>A: Per-task shares, with Progressive over all wakes and Delta/first-read over in-cycle wakes only. B: Separate cohort totals: Progressive 7.29 / 22.41 MB over 213 wakes; Delta 4.49 / approximately 12.53 MB over 112 wakes. First-read body / CLI totals are 0.374 / 0.631 MB for the same 112 wakes. The two denominators are not interchangeable. Open image for full size.</figcaption></figure>
+<figure class="diagram"><p class="diagram-label">8-task Four-Arm Context ON byte audit (N = 213 wakes, 112 in-cycle continuations)</p><a href="../images/context-delivery/multitask-byte-ratios-en.svg" target="_blank" rel="noopener" aria-label="Open the 8-task single-layer vs stacked context byte ratio chart at full size"><img src="../images/context-delivery/multitask-byte-ratios-en.svg" alt="8-Task Four-Arm Context ON 18h Control-Plane Byte Audit: All 100% vs Lossless Delta File 35.8%, Progressive View 32.5%, and Stacked Prog+Delta First-Read 3.0% / 5.0%" width="1400" height="680" loading="lazy" style="width:100%;height:auto;border-radius:10px;margin:8px 0 12px 0"></a><figcaption>A: Per-task shares comparing the two single-layer payloads (Progressive over all 213 wakes; lossless Delta file over 112 continuations) against the stacked first-read body. B: Cohort totals anchored at 100% Full: lossless Delta file ≈ 35.8% (4.49 / ≈ 12.53 MB), Progressive view 32.5% (7.29 / 22.41 MB), and stacked Prog+Delta first-read ≈ 5.0% (CLI) / ≈ 3.0% (body). The two denominators are not interchangeable.</figcaption></figure>
 <details><summary>Expand the eight-task data table</summary>
 <div class="table-scroll" role="region" tabindex="0" aria-label="8-task Four-Arm Context ON control-plane byte ratio table"><table style="min-width:880px">
 <thead>