From 1abba23df347ba8e8a00dab7ab76cf4ae45137ac Mon Sep 17 00:00:00 2001
From: thePR0M3TH3AN <53631862+PR0M3TH3AN@users.noreply.github.com>
Date: Tue, 11 Aug 2026 20:43:35 -0400
Subject: [PATCH] docs: worked example of the ranking dimension as a feed
 multiplier
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The `ranking` dimension is the soft one — "still show this, but lower" — and
integrators reinvented turning it into a feed score. Show it instead.

Add examples/feed-ranking.mjs (runnable): it reads decision.ranking
({ effect: "normal" | "downrank", weight }) and turns it into a multiplier
on the app's own feed score (base ** weight for a downrank). Two authors
that led on base score get downranked by trusted spam reports and drop below
a clean lower-base author -- without being hidden, since visibility is a
separate dimension (capped at "restrict" on a feed). The multiplier lives in
the example, not the engine: core ships no such numbers on purpose.

Pinned by tests/feed-ranking.test.js (downrank never hides on a feed; a
high-base downranked author reorders below a clean one; a clean author is
untouched at x1). README headless section points to it.

Addresses the "BitGate feeding a feed ranking" item of the Tessera feedback
(docs/TODO.md) -- the last open BitGate item.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BXQCvJPA6sVSQ5j8uBd9bB
---
 README.md                  |  6 +++
 docs/TODO.md               | 11 ++++-
 examples/feed-ranking.mjs  | 97 ++++++++++++++++++++++++++++++++++++++
 tests/feed-ranking.test.js | 61 ++++++++++++++++++++++++
 4 files changed, 173 insertions(+), 2 deletions(-)
 create mode 100644 examples/feed-ranking.mjs
 create mode 100644 tests/feed-ranking.test.js

diff --git a/README.md b/README.md
index 31b11c5..2cd701f 100644
--- a/README.md
+++ b/README.md
@@ -161,6 +161,12 @@ No bundler? `npm run build:standalone` produces `packages/standalone/dist/bitgat
 ESM file. Self-host it and `import { evaluateTarget, decodeLabels } from "./bitgate.js"`;
 everything in this section works from that single file.
 
+The soft dimension, `ranking`, feeds an ordering rather than a gate:
+`decision.ranking` is `{ effect: "normal" | "downrank", weight }`, which an app
+turns into a multiplier on its own feed score — a reported author drops in the
+feed without being hidden. [`examples/feed-ranking.mjs`](examples/feed-ranking.mjs)
+is a runnable walkthrough.
+
 ## Extending to a new application
 
 Write an adapter saying what your object is in governance terms, and a policy
diff --git a/docs/TODO.md b/docs/TODO.md
index 88b2031..4ba0c99 100644
--- a/docs/TODO.md
+++ b/docs/TODO.md
@@ -282,5 +282,12 @@ adoption the obvious choice.
   whole headless path. Verified dependency-free (no bare imports in the output)
   and end-to-end (a smoke test composes a decision from the entry). Documented
   in the README Packages table and the headless section; wired into `npm run build`.
-- A worked example of BitGate feeding a feed ranking (the `ranking` dimension →
-  a score multiplier) would have saved us reinventing it in the feed recipe.
+- **[DONE 2026-08-11]** A worked example of BitGate feeding a feed ranking.
+  **Shipped** `examples/feed-ranking.mjs` (runnable): it reads `decision.ranking`
+  (`{ effect: "normal" | "downrank", weight }`) and turns it into a score
+  multiplier (`base ** weight` for a downrank), so trusted spam reports push two
+  high-base authors below a clean lower-base one **without hiding them** —
+  visibility stays a separate dimension, capped at `restrict` on a feed. The
+  multiplier lives in the example, not the engine (core ships no such numbers).
+  Pinned by `tests/feed-ranking.test.js`; the README headless section points to
+  it.
diff --git a/examples/feed-ranking.mjs b/examples/feed-ranking.mjs
new file mode 100644
index 0000000..bb729b3
--- /dev/null
+++ b/examples/feed-ranking.mjs
@@ -0,0 +1,97 @@
+/**
+ * The `ranking` dimension as a feed score multiplier.
+ *
+ * Moderation is not only hide/allow. The fourth dimension, `ranking`, is the
+ * soft one: it says "still show this, but lower." A feed reads
+ * `decision.ranking` — `{ effect: "normal" | "downrank", weight }`, where
+ * `weight` grows with each downrank signal — and turns it into a multiplier on
+ * whatever score the feed already computed. Nothing is hidden; the order just
+ * reflects the governance signal.
+ *
+ * The multiplier is an application policy, so it lives here, not in the engine
+ * (the core ships no such numbers on purpose). Tune `base` to say how hard a
+ * downrank bites.
+ *
+ *   node examples/feed-ranking.mjs
+ */
+import { SOCIAL_POLICY, createAuthorityState, createSnapshot, evaluateTarget } from "@bitgate/core";
+
+const ROOT = "a1".repeat(32);
+const NOW = 1_750_000_000;
+const trusted = (i) => i.toString(16).padStart(2, "0").repeat(32);
+const TRUSTED = Array.from({ length: 16 }, (_, i) => trusted(i + 1));
+
+// A base feed score the app already computed (recency, affinity, whatever).
+// Bob and Carol START ABOVE Alice — the point is that the ranking dimension
+// reorders them below her without hiding anything.
+// author -> { score, reports: [category, trustedReporterCount] }
+const authors = {
+  alice: { pubkey: "d1".repeat(32), score: 0.70, reports: [] },
+  bob: { pubkey: "d2".repeat(32), score: 0.95, reports: [["spam", 4]] },
+  carol: { pubkey: "d3".repeat(32), score: 0.90, reports: [["spam", 8]] },
+};
+
+const authority = createAuthorityState({ root: ROOT });
+
+/** Build a snapshot carrying `count` trusted reports of `category` against pubkey. */
+function snapshotFor(pubkey, reports) {
+  return createSnapshot({
+    authority,
+    trust: { contacts: new Set(TRUSTED) },
+    reports: new Map(
+      reports.map(([category, count]) => [
+        `user:${pubkey}`,
+        TRUSTED.slice(0, count).map((reporter) => ({ reporter, category, createdAt: NOW - 100 })),
+      ]),
+    ),
+  });
+}
+
+/**
+ * Application policy: turn a ranking decision into a score multiplier. A
+ * `normal` ranking is neutral (×1); a `downrank` bites harder the more signals
+ * back it (`weight`). Nothing here hides anything — visibility is a separate
+ * dimension.
+ */
+function rankingMultiplier(ranking, { base = 0.5 } = {}) {
+  if (!ranking || ranking.effect !== "downrank") return 1;
+  return base ** Math.max(1, ranking.weight);
+}
+
+const ranked = Object.entries(authors)
+  .map(([name, author]) => {
+    const decision = evaluateTarget({ type: "user", pubkey: author.pubkey }, snapshotFor(author.pubkey, author.reports), {
+      surface: "feed",
+      policyProfile: "feed",
+      policy: SOCIAL_POLICY,
+      now: NOW,
+    });
+    const multiplier = rankingMultiplier(decision.ranking);
+    return {
+      name,
+      base: author.score,
+      ranking: decision.ranking.effect,
+      weight: decision.ranking.weight,
+      multiplier,
+      final: author.score * multiplier,
+      visible: decision.visibility.effect, // still shown — ranking never hides
+    };
+  })
+  .sort((a, b) => b.final - a.final);
+
+console.log("Feed after applying the ranking dimension as a score multiplier:\n");
+console.log("  name    base   ranking    weight  ×mult   final   visible");
+for (const r of ranked) {
+  console.log(
+    `  ${r.name.padEnd(6)}  ${r.base.toFixed(2)}   ${r.ranking.padEnd(9)}  ${String(r.weight).padStart(4)}   ${r.multiplier
+      .toFixed(2)
+      .padStart(5)}   ${r.final.toFixed(2)}   ${r.visible}`,
+  );
+}
+console.log(`
+Bob and Carol led on base score, but their trusted spam reports downranked
+them (×0.5) and both dropped below clean Alice — nothing was hidden (visibility
+is a separate dimension, capped at "restrict" on a feed). weight is 1 here
+because each crossed the downrank threshold once; a target downranked by several
+independent signals accrues more weight and bites harder. That reordering is the
+ranking dimension doing its job.`);
diff --git a/tests/feed-ranking.test.js b/tests/feed-ranking.test.js
new file mode 100644
index 0000000..24aa468
--- /dev/null
+++ b/tests/feed-ranking.test.js
@@ -0,0 +1,61 @@
+// Pins examples/feed-ranking.mjs: the `ranking` dimension, applied as a score
+// multiplier, reorders a downranked author below a clean one without hiding it.
+import { describe, expect, it } from "vitest";
+
+import { SOCIAL_POLICY, createAuthorityState, createSnapshot, evaluateTarget } from "@bitgate/core";
+
+const ROOT = "a1".repeat(32);
+const NOW = 1_750_000_000;
+const TRUSTED = Array.from({ length: 16 }, (_, i) => (i + 1).toString(16).padStart(2, "0").repeat(32));
+const authority = createAuthorityState({ root: ROOT });
+
+const snapshotFor = (pubkey, category, count) =>
+  createSnapshot({
+    authority,
+    trust: { contacts: new Set(TRUSTED) },
+    reports: count
+      ? new Map([
+          [
+            `user:${pubkey}`,
+            TRUSTED.slice(0, count).map((reporter) => ({ reporter, category, createdAt: NOW - 100 })),
+          ],
+        ])
+      : new Map(),
+  });
+
+const rankingOf = (pubkey, category, count) =>
+  evaluateTarget({ type: "user", pubkey }, snapshotFor(pubkey, category, count), {
+    surface: "feed",
+    policyProfile: "feed",
+    policy: SOCIAL_POLICY,
+    now: NOW,
+  });
+
+// The same multiplier the example documents.
+const multiplier = (ranking, base = 0.5) =>
+  ranking?.effect === "downrank" ? base ** Math.max(1, ranking.weight) : 1;
+
+describe("ranking dimension as a feed multiplier", () => {
+  it("downranks a reported author but never hides them on a feed", () => {
+    const decision = rankingOf("d2".repeat(32), "spam", 4);
+    expect(decision.ranking.effect).toBe("downrank");
+    expect(decision.ranking.weight).toBeGreaterThanOrEqual(1);
+    // Ranking is the soft dimension: visibility stays out of hide/deny on a feed.
+    expect(["allow", "warn", "restrict"]).toContain(decision.visibility.effect);
+  });
+
+  it("reorders a high-base downranked author below a clean lower-base one", () => {
+    const clean = { base: 0.7, ranking: rankingOf("d1".repeat(32), "spam", 0).ranking };
+    const reported = { base: 0.95, ranking: rankingOf("d2".repeat(32), "spam", 4).ranking };
+
+    const cleanFinal = clean.base * multiplier(clean.ranking);
+    const reportedFinal = reported.base * multiplier(reported.ranking);
+
+    expect(reported.base).toBeGreaterThan(clean.base); // led on base score
+    expect(reportedFinal).toBeLessThan(cleanFinal); // ...but sinks after ranking
+  });
+
+  it("leaves a clean author's score untouched (×1)", () => {
+    expect(multiplier(rankingOf("d1".repeat(32), "spam", 0).ranking)).toBe(1);
+  });
+});