Biennale Pulse
0–39
0–39 · Watch — Archived, don't call yet
40–54
40–54 · Anecdotal — Tracked, not surfaced
55–69
55–69 · Early read — In the feed, monitored
70–84
70–84 · Developing — Featured, Board candidate
85–100
85–100 · Strong signal — Lead story, spotlight, top of feed
0→100

Methodology

How Biennale Pulse measures the art world — transparently

Last updated: July 24, 2026·Living document — updated as sources and capabilities expand

Living Artist & Contemporary Art Weighting

Biennale Pulse applies an editorial weighting bias toward living artists and contemporary art practice. This reflects our mission to surface what's happening in the art world right now, rather than reinforcing historical canon. The weighting is applied multiplicatively to the momentum sub-score only, so it boosts visibility without inflating credibility or novelty metrics.

Weighting Multipliers

Each boost is applied multiplicatively to the momentum sub-score only.

00.350.71.051.4Base SignalLiving ArtistContemporaryEmerging TagMax Combined1.00×1.15×1.10×1.05×1.33×

Max combined boost: +33% over base · Cap at 99

Worked Example

Select a scenario to see how weighting changes the final pulse score.

Living Artist+15%
Contemporary+10%
Emerging+5%
MomentumCredibilityNoveltyComposite025507510096657883
Without weighting: 71
With weighting: 83
Net boost: +33%

Composite Formula

Momentumweighted = Momentumraw × Multipliers
Composite = (Momentumweighted × 0.5) + (Credibility × 0.3) + (Novelty × 0.2)

Weighting is applied to the momentum sub-score only. Credibility and novelty are never inflated. This ensures boosts increase visibility without compromising source trust or freshness integrity.

Why This Matters

  • • Historical figures (B100 canon) retain authority weight, but it decays — living, active artists can outrank them on The Board.
  • • Contemporary signals are surfaced more prominently in feeds, digests, and trend timelines.
  • • The bias is documented and visible — every score page links here so readers understand what they're looking at.
  • • We do not penalize historical or non-contemporary work; we simply weight momentum toward the living and the now.

Pulse Score Composition

Every signal, trend, person, and city carries a composite Pulse Score (0–100) built from three sub-scores:

Momentum
50%

Growth velocity, source diversity, biennale connection, editorial boost. Living artist + contemporary weighting applied here.

Credibility
30%

Source tier, primary vs. secondary, independent confirmation, verifiable URLs.

Novelty
20%

First signal in 90+ days, new city on heatmap, trend graduation, dedup penalty.

85–99Strong signalLead story, spotlight, top of feed
70–84DevelopingFeatured, Board candidate
55–69Early readIn the feed, monitored
40–54AnecdotalTracked, not surfaced
<40WatchArchived, don't call yet

Source Universe

Biennale Pulse scans a curated universe of art world sources via Biennale Radar — RSS feeds, podcasts, video channels, Substack newsletters, and a monitored roster of social media accounts spanning curators, critics, fairs & biennales, galleries, and artists. Every source is scored for reliability (0–100) and tagged by region, type, and ecosystem category.

119

Active Sources

5

Source Types

83

Avg. Reliability

7

Regions Covered

The Foundation — Every Source, One Cell

119 sources

Each cell is one vetted source in the monitored universe, grouped by region and shaded by reliability score. Darker means more reliable — no source below 72 is in the roster.

Europe39
North America41
Global16
Asia Pacific10
Latin America6
Middle East4
Africa3

Reliability Distribution

median 84 / 100
3
70–74
22
75–79
40
80–84
40
85–89
14
90+
70–79 80–89 90+

By Editorial Category

Social Channels63 · 29%Blogs & Independent37 · 17%Fair Biennale17 · 8%Critic15 · 7%Galleries & Dea…13 · 6%Curator11 · 5%Artist7 · 3%Criticism7 · 3%

Area proportional to source count · hover for detail

By Source Type

social
66
rss
40
podcast
5
video
5
substack
3

Source-to-Signal Pipeline

How each source category flows into specific signal types — the direct line from inbox to instrument panel.

Source Categories

RSS Feeds

Art magazines, journals, blogs

Newsletters & Press

Gallery emails, museum press, Substack via email

Podcasts

Critics' shows, market talks

Video

Studio visits, fair coverage, interviews

Substack

Independent art writers

Social Signals

Cross-platform visibility scans

Signal Types Generated
Artist Rising
Curator Move
Market Shift
City Heating
Theme Emerging
Controversy
Breakout Work
Institutional Shift
Appointment
Art Travel
RSS Feeds
Newsletters & Press
Podcasts
Video
Substack
Social Signals

Each dot on a signal row indicates which source categories can generate that signal type. Some signals require corroboration across multiple categories — see Anti-Echo-Chamber Protocol below.

Browse the full source directory View feed endpoints

Automated Publishing Standards

Biennale Pulse does not use manual review. Every signal is evaluated automatically against codified quality rules and published or archived without human intervention. This ensures consistency, transparency, and speed — signals move from ingestion to publication in under 15 minutes.

Publishing Rules

Publish Threshold≥ 55 composite

Signals scoring 55+ on the composite Pulse Score are auto-published.

Archive Threshold< 55 composite

Signals below the publish threshold are archived with their computed scores.

Minimum Source Reliability≥ 50

Signals from sources below 50 reliability are archived regardless of score.

Title Quality≥ 15 characters

Signals with truncated or empty titles are archived.

Summary Quality≥ 20 characters

Signals without a meaningful summary are archived.

Dedup Window30 days

Duplicate signals (normalized title match) within 30 days receive reduced novelty scores.

How Scores Are Computed Automatically

The autoPublishSignals function computes three sub-scores for every candidate signal, then derives the composite:

  • • Momentum (50%): Base weight by signal type + source reliability bonus + biennale connection bonus. Living artist signals receive ×1.15 multiplier; contemporary art signals receive ×1.10 multiplier — both applied to momentum only.
  • • Credibility (30%): Source reliability score + bonus for verifiable URL + bonus for direct quote attribution.
  • • Novelty (20%): 65 for first-of-topic signals, 25 for duplicates. Breakout works and emerging themes receive novelty bonuses.
  • • Composite: momentum × 0.5 + credibility × 0.3 + novelty × 0.2
Runs every 15 minutes via scheduled automation.autoPublishSignals

Data Refresh & Access

Signal Ingestion

Every 15 minutes

RSS feeds, Substack, and social signals synced via Biennale Radar. New signals enter as 'draft' with initial credibility scores.

Auto-Publish Evaluation

Every 15 minutes

Draft signals evaluated against quality rules. Meeting thresholds are published with computed pulse scores; failing signals are archived.

Weekly Digest Generation

Mondays at 9:00 AM ET

Automated Pulse Check compiled from the prior week's top signals, trends, and movers.

Image & Metadata Sync

On demand

Wikipedia/Commons images and exhibition data refreshed as records are updated.

Access timestamps: Each signal, trend, and digest displays its last-updated date. Outgoing feed endpoints are stamped with generation time. Incoming sources show their last sync status on the Feeds page.

Citing Biennale Pulse

We encourage researchers, journalists, and students to cite Biennale Pulse data in papers and articles. Every trend, person, and digest page includes a "Share & Cite" panel with ready-to-copy citations in APA, MLA, and Chicago formats.

APA

Biennale Pulse. (2026). Posthuman Figuration. Retrieved July 4, 2026, from https://biennalepulse.com/trend/posthuman-figuration

MLA

"Posthuman Figuration." Biennale Pulse, July 1, 2026, https://biennalepulse.com/trend/posthuman-figuration. Accessed 4 July 2026.

Chicago

Biennale Pulse, "Posthuman Figuration," accessed July 4, 2026, https://biennalepulse.com/trend/posthuman-figuration.

Anti-Echo-Chamber Protocol

  • Trend Graduation Rule: No trend graduates from "watching" to "emerging" unless supported by ≥3 independent source types.
  • Counter-Read Rule: Any signal or trend hitting 85+ triggers a search for a dissenting view. Found → linked. Not found → flagged "uncontested."
  • BR Provenance Rule: Biennale Radar feed items inform signals but never satisfy the corroboration requirement.
  • Authority Decay: B100 authority weight decays from 100% in year 1 to 60% by year 5, ensuring canon is a diminishing prior, not a permanent boost.

Content Provenance & Editorial Review

The following entity types carry provenance tags: PulseSignal, Trend, PulseQuote, Person, and PulseDigest. Each tagged record documents how it was produced (content_origin) and its editorial clearance state (review_status). This allows readers to distinguish source-extracted factual content from AI-assisted editorial analysis on those entity types.

Source Extracted

Factual content drawn directly from a published source. Ingestion paths for this content class do not invoke generative AI for text creation.

Rule Generated

Computed by deterministic rules (scores, rankings) without LLM inference.

AI Generated

Editorial analysis or trend dossiers produced by an LLM. Labeled 'Automated analysis' until reviewed. AI-generated quotations are not stored as PulseQuote records.

Human Written

Editorial content authored or approved by a human editor.

Review Workflow

  • • Neutral factual signals (title + summary from source, no AI editorial take) may auto-publish if they meet quality thresholds.
  • • AI-generated editorial takes on signals are held for review and excluded from auto-publishing until a human editor approves, rejects, or requests revision via the Pulse Ops dashboard.
  • • AI-generated quotations are not stored as PulseQuote records. Only quotations verified against retained source evidence (exact text, source URL, publisher, speaker, evidence excerpt, extraction method) may be stored as PulseQuote. Quote rotation displays only approved quotations.
  • • Correction tracking: PulseSignal and Trend carry a correction_status field, and a Correction entity schema exists for audit-trail records. However, the public correction notice workflow, immutable history, and public rendering of corrections are not yet implemented. This system is classified as incomplete.
  • • Entity resolution: Signal cross-references in the SignalCrossRefs component use token-boundary matching, reducing classes of substring false positives (e.g. "Paris" no longer matching "Paris Hilton"). This reduces but does not eliminate all false matches.

Automation Health Monitoring

An hourly checkAutomationHealth monitor scans all automation run logs for consecutive failures. After 2+ failures, it sends a deduplicated alert email. When a previously-failing automation recovers, a recovery notification is sent. This ensures operational issues are caught quickly without alert fatigue.

Data Pipelines & Capabilities

Biennale Pulse operates an automated ingestion and enrichment pipeline. Each capability below is a backend function that runs on a schedule or on demand. As we expand sources and categories, new pipelines are added here.

Radar RSS Sync

syncRadarFeedsOn demand

Fetches and parses RSS feeds from curated art publications. Decodes HTML entities, extracts thumbnails/media images, and maps MYB tag taxonomy to signals.

Newsletter & Press Inbox Ingestion

syncGmailInboxOn demand

Ingests content delivered via email from art world newsletters, magazines, institutions, galleries, and press offices. Email is the delivery vehicle, not the source — each item is attributed to its originating publisher. Incoming content is graded against the Source reputation layer; only content from trusted/high-interest publishers is ingested as PulseSignals. MYB tag taxonomy applied during tagging.

Artist Image Sync

syncArtistImagesOn demand

Fetches portrait images from Wikipedia/Wikimedia Commons for Person records. Falls back to AI-generated conceptual imagery when no public-domain or open-access source photo is available.

City Image Sync

syncCityImagesOn demand

Fetches cityscape and venue imagery from Wikimedia Commons for CityPulse records.

Exhibition Feed Ingestion

syncExhibitionsOn demand

Ingests exhibition and event data from Biennale Radar (BR) and Meet You at the Biennale (MYB) feeds. Maps event types, calculates urgency levels, and links to external BR/MYB pages.

Social Visibility Scan

syncSocialVisibilityDaily

LLM-based scan of the monitored social source roster (Instagram, X, YouTube handles) — no direct platform API polling. Detected topics are weighted by each matched source's reliability score (0–100); standalone retail/merchandise posts are excluded.

Weekly Report Generation

generateWeeklyReportMondays 9:00 AM ET

Compiles top signals, trends, movers, and cultural shifts into a PulseDigest. Applies living artist and contemporary art weighting. Produces editorial intro with transparency about applied multipliers.

PDF Digest Export

exportDigestPdfOn demand

Generates a downloadable PDF of any PulseDigest record using jsPDF. Triggered from the Pulse Check page.

Image Sourcing & Rights Policy

Entity images carry an image-rights status documenting their provenance and usage rights. The canonical rights statuses are:

Public Domain

Expired copyright, verified public-domain source

Creative Commons (Verified)

CC-licensed imagery with verified terms

Verified Press Use

Press-use authorization with documentary evidence on file

Licensed

Licensed under a documented agreement

Permission on File

Explicit permission documented from rights holder

Review Required

Rights status not yet verified — pending review

Restricted

Usage restricted; image should not be displayed publicly

Legacy Unknown

Historical records with unverified rights status

Images marked review required, restricted, rejected, or legacy unknown are not displayed on public pages unless a separately documented permissible-use rule applies. Legacy status values (fair_use_review, press_image, open_access, internal_generated, do_not_use) are deprecated and no longer assigned by new code; existing records are being migrated to canonical statuses. fair_use_reviewis not treated as a legal or licensing status.

Methodology Changelog

This is a living document. As we add sources, categories, capabilities, or adjust weighting, changes are logged here so readers can trace how the instrument has evolved.

v1.7July 25, 2026
  • • Added a rule-generated sentiment pass to the ingestion pipeline: every signal is classified (positive / neutral / negative / mixed) with a -100 to +100 intensity score at publish time using a deterministic lexicon — no LLM inference. All previously published signals were backfilled.
  • • Automated degree-show tracking: a daily scan (scanEmergingArtists) monitors MFA degree shows, graduate exhibitions, and early-career coverage, adding only source-verified artists to the Early Artists radar.
  • • Added scrollytelling trend narratives ('The Story So Far') to trend pages: a four-chapter scroll-driven account of detection, signal accumulation, geographic spread, and the current read.
  • • Added the Source Foundation visualization to this page: every active source rendered as one reliability-shaded cell, with a reliability distribution histogram.
v1.6July 24, 2026
  • • Expanded the source universe to 104 active sources with a monitored social roster of 63 handles across five ecosystem categories: curators, critics, fairs & biennales, galleries, and artists.
  • • Phased social expansion: 35 targeted accounts (Phase 1), 12 regional biennales, independent commentary, and fair directors (Phase 2), and 16 global museum directors, kunsthalles, regional biennales, and art advisors (Phase 3).
  • • Social visibility scanning uses scheduled LLM-based content analysis rather than direct platform API polling; detected topics are weighted by each source's reliability score (0–100).
  • • Retail/merchandise-only social posts are excluded from scanning; ingestion focuses on exhibition- and artist-relevant content.
  • • Feeds page now supports secondary filtering of social sources by ecosystem category, with handle, platform, and category shown on each source card.
v1.5July 21, 2026
  • • Added content provenance tracking to PulseSignal, Trend, PulseQuote, Person, and PulseDigest: content_origin and review_status fields document how each record was produced.
  • • Implemented editorial review queue in Pulse Ops: AI-generated editorial takes on signals are held for human review before publication.
  • • Removed AI-generated quotations from public rotation: QuoteRotation displays only approved_for_rotation quotes. LLM-generated editorial language is no longer stored as PulseQuote records.
  • • Added Correction entity schema for audit-trail tracking of factual errors, attribution errors, and retractions. Public correction rendering and immutable history are not yet implemented — this system is classified as incomplete.
  • • Applied entity-resolution safety patch in SignalCrossRefs: signal cross-references now use token-boundary matching, reducing classes of substring false positives (e.g. 'Paris' matching 'Paris Hilton').
  • • Added automation health monitor (checkAutomationHealth) with deduplicated alert emails after 2+ consecutive failures and recovery notifications.
  • • Expanded image_rights_status taxonomy. Canonical statuses: review_required, verified_press_use, licensed, public_domain, creative_commons_verified, permission_on_file, restricted, rejected, legacy_unknown. Legacy values (fair_use_review, press_image, open_access, internal_generated, do_not_use) are deprecated and no longer assigned by new code.
  • • Fixed discoverAndEnrichTrends timeout: reduced LLM prompt input from 80 to 40 signals, shortened summary text, switched to platform-selected model (model_selection_mode: automatic).
  • • Fixed generateWeeklyReport: removed undeclared variable reference; reduced json_output to a summary contract to avoid field-size overflow.
v1.4July 5, 2026
  • • Replaced manual review with automated publishing pipeline. Signals are now evaluated against codified quality rules and auto-published or archived every 15 minutes.
  • • Added autoPublishSignals backend function with transparent scoring computation: momentum (50%), credibility (30%), novelty (20%).
  • • Living artist (×1.15) and contemporary art (×1.10) multipliers now applied automatically during score computation, not just documented.
  • • RSS ingestion updated to create signals as 'draft' with initial credibility scores instead of 'pending_review'.
  • • Documented all publishing thresholds: composite ≥ 55 to publish, source reliability ≥ 50, title ≥ 15 chars, summary ≥ 20 chars, 30-day dedup window.
v1.3July 4, 2026
  • • Added capabilities & pipeline disclosure section documenting all backend sync functions.
  • • Added image sourcing & rights policy section.
  • • Expanded source type tracking to include email-delivered newsletters, press releases, and internal signal sources.
  • • Added this changelog to make methodology evolution traceable.
v1.2July 2026
  • • Implemented advanced filtering (signal type, region, date range, sort) for the Signal Feed.
  • • Added PDF export capability for PulseDigest records.
  • • Created dedicated Methodology page and Share & Cite component for citation transparency.
v1.1June 2026
  • • Implemented Gmail inbox ingestion with reputation-based filtering.
  • • Added MYB tag taxonomy to Gmail ingestion and signal tagging.
  • • Automated weekly PulseDigest generation via scheduled task.
  • • Standardized on AI-generated conceptual imagery for entities lacking source photos.
v1.0May 2026
  • • Initial methodology: pulse score composition (momentum 50%, credibility 30%, novelty 20%).
  • • Living artist (×1.15) and contemporary practice (×1.10) weighting applied to momentum sub-score.
  • • Anti-echo-chamber protocol: trend graduation rule, counter-read rule, BR provenance rule, authority decay.
  • • Source universe: RSS, podcast, video, Substack, social signal tracking with reliability scoring.
base44
Edit with Base44