/* Attention / Shared Following View — dedicated page for cohort information diet + cohesion.
Visualizes the following lists data (globalRanked from agg + per-member attention).
Complements the Pedigree Map (past trajectory) with current "who we collectively follow".
*/
const { useState, useEffect, useMemo } = React;
const CAT_COLOR = {
'in-cohort': 'var(--accent)',
'ai-lab': '#f59e0b',
'researcher': '#10b981',
'tool': '#3b82f6',
'media': '#8b5cf6',
'vc': '#ec4899',
'musk-orbit': '#dc2626',
'founder': '#14b8a6',
'other': 'var(--text-tertiary)'
};
const CAT_LABEL = {
'in-cohort': 'In-cohort',
'ai-lab': 'AI Labs',
'researcher': 'Researchers',
'tool': 'Tools / Infra',
'media': 'Media / Papers',
'vc': 'VC / Angels',
'musk-orbit': 'Musk Orbit',
'founder': 'Founders / Operators',
'other': 'Other'
};
function xUrl(handle) {
return 'https://x.com/' + String(handle || '').replace(/^@/, '');
}
function AttentionView({ onOpenFollows }) {
const [agg, setAgg] = useState(null);
const [loaded, setLoaded] = useState(false);
const [openCat, setOpenCat] = useState(null);
useEffect(function () {
var tried = false;
function done(d) {
if (d) setAgg(d);
setLoaded(true);
}
fetch('/data/xai-following-agg.json')
.then(function (r) { return r.ok ? r.json() : null; })
.then(done)
.catch(function () {
// Fallback: compute a minimal global view from per-member attention in the loaded cohort
if (!tried && window.COHORT && window.COHORT.members) {
tried = true;
var members = window.COHORT.members.filter(function (m) { return m.attention && m.attention.topFollowed && m.attention.topFollowed.length; });
var by = {};
members.forEach(function (m) {
(m.attention.topFollowed || []).forEach(function (f) {
var h = f.handle;
if (!by[h]) by[h] = { handle: h, name: f.name || h, category: f.category || 'other', followedByCount: 0 };
by[h].followedByCount++;
});
});
var ranked = Object.values(by).sort(function (a, b) { return b.followedByCount - a.followedByCount; }).slice(0, 50);
done({ globalRanked: ranked, internalVsExternalRatio: { internal: 0, external: 0, ratio: 0 }, categoryStats: {} });
} else {
done(null);
}
});
}, []);
const cohortMembers = (window.COHORT && window.COHORT.members) || [];
const members = cohortMembers.filter(function (m) {
return m.attention && m.attention.topFollowed && m.attention.topFollowed.length > 0;
});
const summary = useMemo(function () {
if (!agg) return null;
var internal = agg.internalVsExternalRatio ? agg.internalVsExternalRatio.internal : 0;
var external = agg.internalVsExternalRatio ? agg.internalVsExternalRatio.external : 0;
var ratio = external > 0 ? Math.round((internal / (internal + external)) * 100) : 0;
return {
members: members.length,
totalRanked: (agg.globalRanked || []).length,
internalRatio: ratio
};
}, [agg, members.length]);
// Live top from members (in case agg not loaded or for freshness)
const liveTop = useMemo(function () {
var count = {};
members.forEach(function (m) {
(m.attention.topFollowed || []).slice(0, 30).forEach(function (f) {
var k = f.handle;
if (!count[k]) count[k] = { handle: k, name: f.name, category: f.category, c: 0 };
count[k].c++;
});
});
return Object.values(count).sort(function (a, b) { return b.c - a.c; }).slice(0, 30);
}, [members]);
var top = (agg && agg.globalRanked && agg.globalRanked.length > 0) ? agg.globalRanked : liveTop;
// Category stats for bars (prefer agg, fallback live)
var catStats = useMemo(function () {
if (agg && agg.categoryStats && Object.keys(agg.categoryStats).length > 0) return agg.categoryStats;
var s = { 'in-cohort': 0, 'ai-lab': 0, 'researcher': 0, 'tool': 0, 'media': 0, 'vc': 0, 'other': 0 };
members.forEach(function (m) {
var cs = m.attention.categoryCounts || {};
Object.keys(s).forEach(function (k) { s[k] += (cs[k] || 0); });
});
return s;
}, [agg, members]);
// Scale bars to the largest *signal* category (exclude the catch-all 'other', which
// dwarfs everything) so meaningful categories stay readable. 'other' clamps at 100%.
var maxCat = Math.max.apply(null, Object.keys(catStats)
.filter(function (k) { return k !== 'other'; })
.map(function (k) { return catStats[k] || 0; })) || 1;
// Per-category account lists for the expandable rows. Prefer the precomputed full
// lists from the agg; otherwise group whatever ranked accounts we have by category.
var catRanked = useMemo(function () {
if (agg && agg.categoryRanked && Object.keys(agg.categoryRanked).length > 0) return agg.categoryRanked;
var grouped = {};
(top || []).forEach(function (t) {
var cat = t.category || 'other';
if (!grouped[cat]) grouped[cat] = [];
grouped[cat].push({ handle: t.handle, name: t.name, followedByCount: t.followedByCount || t.c || 1 });
});
Object.keys(grouped).forEach(function (k) {
grouped[k].sort(function (a, b) { return b.followedByCount - a.followedByCount; });
});
return grouped;
}, [agg, top]);
// High-overlap members (from current members)
var highOverlap = useMemo(function () {
return members
.filter(function (m) { return m.attention.cohortOverlapScore > 0; })
.sort(function (a, b) { return b.attention.cohortOverlapScore - a.attention.cohortOverlapScore; })
.slice(0, 8);
}, [members]);
if (!loaded) {
return
Loading attention data…
;
}
// Render whenever we have ANY attention signal: either the precomputed global
// aggregate (xai-following-agg.json) OR per-member attention on the cohort.
// `top` falls back to member-derived liveTop, so a non-empty `top` means we
// have something useful to show even when per-member .attention is absent
// (e.g. the deployed snapshot only ships the agg file).
const hasData = (top && top.length > 0) || members.length > 0;
if (!hasData) {
return (
Shared Attention
No following data yet in this data snapshot.
To populate for the prototype: run e.g. COHORT=xai-t1 pnpm fetch:followings (or a T2 subset), then pnpm export:data (this attaches .attention to xai.json and builds the agg), commit the updated public/data/*.json files (or refresh the Railway /data volume), push, and redeploy.
Per-member attention (when present) is also visible in the Directory drawer.
);
}
return (
{/* Hero */}
Shared Attention
Current information diet of the cohort — who they follow, and where their collective attention overlaps.
Complements the Pedigree Map (past) with live signals (present).
{members.length > 0 ? members.length + ' members with following data · ' : ''}{top.length} unique high-signal accounts surfaced
{agg && agg.internalVsExternalRatio ? ' · ' + agg.internalVsExternalRatio.ratio + ' internal ratio' : ''}
Accounts the cohort follows most
(size = follows from active insiders)
Each bubble is an account that multiple xAI insiders independently follow — not who follows them, but who they pay attention to. What makes this high-signal isn't the raw count: following lists are only pulled for the members our scoring ranks as the most active, high-influence accounts on the team. The theory — someone who lives on X (high frequency, backed by the xAI halo) has the sharpest read on who's actually worth following, and heavy users refresh their follows often, so it's a fresh signal. Any one person carries social-media noise, but aggregating across many independent active insiders cancels it out. The names that keep recurring are who the team genuinely watches — and a quiet recruiting radar.
{top.slice(0, 20).map(function (t, idx) {
var count = t.followedByCount || t.c || 1;
var size = Math.max(38, Math.min(92, Math.round(38 + Math.sqrt(count) * 7)));
var col = CAT_COLOR[t.category] || 'var(--text-tertiary)';
return (
Concentration on researchers + ai-labs is the core cohesion signal.
{/* High-overlap members from T1 / current data */}
{highOverlap.length > 0 && (
Most representative members
Cohort members whose own follow list overlaps most with the group's collective attention —
i.e. the people whose information diet best mirrors everyone else's.
Higher overlap score = more of who they follow is also followed by ≥3 other members.
{highOverlap.map(function (m, i) {
var a = m.attention;
return (
Data from prioritized T1 fetch (high-activity + high-influence members) + previous runs. Total {members.length} members with following lists.
Classification uses expanded seeds (researcher / vc / ai-lab etc.). Re-run fetch + export to refresh.
);
}
/* ── Full ranked-list sub-page (#attention/follows) ──
A tabular complement to the bubble cloud: every external account the cohort
follows, ranked, with its follow count (= "weight"), category, and a relative
magnitude bar. Color encodes category (same legend as the bubbles). */
function AttentionFollowsView({ onBack }) {
const [agg, setAgg] = useState(null);
const [loaded, setLoaded] = useState(false);
const [cat, setCat] = useState('all');
const [sortKey, setSortKey] = useState('weight');
useEffect(function () {
fetch('/data/xai-following-agg.json')
.then(function (r) { return r.ok ? r.json() : null; })
.then(function (d) { if (d) setAgg(d); setLoaded(true); })
.catch(function () { setLoaded(true); });
}, []);
var rows = useMemo(function () {
if (agg && agg.globalRanked && agg.globalRanked.length) {
return agg.globalRanked.map(function (t) {
var count = t.followedByCount || t.c || 0;
return { handle: t.handle, name: t.name || t.handle, category: t.category || 'other', count: count, weight: t.weightedScore != null ? t.weightedScore : count };
});
}
var members = ((window.COHORT && window.COHORT.members) || []).filter(function (m) {
return m.attention && m.attention.topFollowed && m.attention.topFollowed.length;
});
var by = {};
members.forEach(function (m) {
(m.attention.topFollowed || []).forEach(function (f) {
if (!by[f.handle]) by[f.handle] = { handle: f.handle, name: f.name || f.handle, category: f.category || 'other', count: 0 };
by[f.handle].count++;
});
});
return Object.values(by).map(function (r) { r.weight = r.count; return r; });
}, [agg]);
var maxWeight = useMemo(function () {
return rows.reduce(function (m, r) { return Math.max(m, r.weight); }, 1);
}, [rows]);
var catCounts = useMemo(function () {
var c = {};
rows.forEach(function (r) { c[r.category] = (c[r.category] || 0) + 1; });
return c;
}, [rows]);
var filtered = useMemo(function () {
var list = cat === 'all' ? rows.slice() : rows.filter(function (r) { return r.category === cat; });
list.sort(function (a, b) {
if (sortKey === 'name') return (a.name || '').localeCompare(b.name || '');
if (sortKey === 'count') return b.count - a.count;
return b.weight - a.weight;
});
return list;
}, [rows, cat, sortKey]);
if (!loaded) {
return
Loading follows…
;
}
var cats = ['all'].concat(Object.keys(CAT_LABEL).filter(function (k) { return catCounts[k]; }));
function Chip(props) {
var active = props.active;
return (
);
}
return (
Accounts the cohort follows — full ranked list
Every external account followed by our scoring-prioritized active insiders, ranked by
follow count (the bar = weight relative to the top account).
Color = category. {rows.length} accounts total.
{/* Category filter chips */}
{cats.map(function (k) {
return ;
})}
{/* Sort control (segmented tab) + weight explainer */}
Sort by
{[['weight', 'Weight'], ['count', 'Follows'], ['name', 'Name']].map(function (o) {
var active = sortKey === o[0];
return (
);
})}
{/* How weight is calculated */}
How "Weight" is calculated.{' '}
Every follow is one vote — but votes aren't equal. Each member's vote is scaled by their own
priority score (our activity + influence ranking),
mean-normalized to 1, so{' '}
weight = Σ (memberScore ÷ mean){' '}
across everyone who follows that account. A uniform cohort → weight ≈ follows; but a follow from a more active,
higher-reach insider counts for more. Follows stays the honest head-count.
{/* Table header */}
#
Account
Category
Follows{sortKey === 'count' ? ' ▾' : ''}
Weight{sortKey === 'weight' ? ' ▾' : ''}
{/* Rows */}
{filtered.map(function (r, i) {
var col = CAT_COLOR[r.category] || 'var(--text-tertiary)';
return (