Threads Chronology

threadsdatainformation-theory

Threads Chronology & Linguistics

Page: /glyphary/threads Source file: src/pages/glyphary/threads.astro (~3,650 lines) Data sources: index.json, post-tags.json, sleep.json, threads/posts.json (raw) Corpus: 37,912 Threads posts from @maybe_foucault

Overview

The main Threads page is the largest single page in the project — a seven-section Foucauldian analysis of the entire Threads posting archive. Every visualization is pure SVG generated at build time in the Astro frontmatter. The page is organized by Roman-numeral sections following the structure of an archaeological excavation: from raw inventory (I) through epistemic strata (II), confession (III), temporal apparatus (IV), cross-domain correlations (V), reception metrics (VI), and information theory (VII).

I. The Archive

“A quantitative inventory of utterances”

Stats Card Row

Eight metric cards providing a census of the corpus:

  • Total utterances (post count)
  • Words written (total word count)
  • Unique vocabulary (distinct words)
  • Average words per post
  • Average unique words per active day
  • Quote ratio (percentage of quotes vs. originals)
  • Active days (days with at least one post)
  • Profanity percentage

The Accumulation of Statements

Visualization: Stacked bar chart (monthly) Shows monthly output broken down into three post types: original posts, quote-posts, and replies. Each month is a vertical bar with stacked segments colored by type.

The Calendar of Presence

Visualization: Calendar heatmap Daily utterance density displayed as a year-long calendar grid. Each cell represents a single day, colored by posting volume (lighter = fewer posts, darker = more). Reveals weekly and seasonal rhythms.

The Instruments of Form

Visualization: Horizontal bar chart Media format distribution across the corpus: text, image, carousel, video, audio. Shows how posting style has shifted across media types.

II. The Archaeology

“Epistemic strata and discursive formations”

Discursive Strata

Visualization: Stacked area chart (monthly) Topic proportions over time, rendered as geological strata. Each of the 20 LLM-classified tags is a colored band. The area chart reveals how discourse topics shift, grow, and recede month over month.

The Twenty Discourses

Visualization: Multi-line chart Top 10 of the 20 discourse tags plotted individually over time. Uses gap-aware rendering — months with zero posts for a tag break the line rather than drawing to zero. Shows which topics are persistent vs. episodic.

I Contain Multitudes

Visualization: Line chart Shannon entropy of the topic distribution measured monthly. Higher entropy = more diverse topic spread; lower entropy = concentrated discourse. Named after Whitman’s Song of Myself. Tracks whether posting becomes more or less topically diverse over time.

The Drift Between Voices

Visualization: Line chart Monthly ratio of original posts to quote-posts. Measures the balance between monologic (original) and dialogic (quoting others) discourse. Tracks whether the account shifts toward commentary or original expression.

The Weight of Words

Visualization: Line chart Average words per post by month. Captures shifts in posting verbosity — whether posts trend toward brevity or essay-length.

Lexical Archaeology

Visualization: Two-column layout

Left column: Cumulative vocabulary growth per quarter. Plots the number of unique words encountered for the first time, showing whether linguistic novelty accelerates or plateaus.

Right column: Word cloud of distinctive/rare vocabulary. Words sized by lexicographic rarity (inverse frequency). Surfaces the unusual, invented, and multilingual terms that distinguish this corpus.

Interactive element: “Did Wei Utter This Word?” — a text input oracle that searches the archive for any word and reports whether it appears, how many times, and when it was first used.

Daily Lexical Range

Visualization: Line chart (weekly average) Unique words per active day, smoothed to weekly average. Measures daily vocabulary diversity independent of posting volume.

The Compound Register

Visualization: Two charts Analysis of hyphenated compounds in the corpus. First chart: count of unique compounds over time. Second chart: categorization of compounds (e.g., adjective-noun, noun-noun, neologism). Reveals a distinctive stylistic pattern of compound coinage.

Grammatical Violence

Visualization: Chart Words that are bent across parts of speech — nouns used as verbs, adjectives used as nouns, etc. Surfaces deliberate morphological play and linguistic invention.

III. The Confession

“Technologies of the self and the examined life”

The Confessional Index

Visualization: Scatter/trend chart First-person pronoun density (I, me, my, myself) vs. emotional vocabulary density, both z-score normalized per month. Plots the intersection of self-reference and affect — months that are both highly personal and highly emotional appear in the upper-right quadrant.

The Profane Register

Visualization: Timeline chart Monthly profanity rate as a percentage of total words. Tracks profanity (from the PROFANITY word set) over time.

The Sentimental Instrument

Visualization: Timeline chart Positive vs. negative sentiment by month, using keyword-based sentiment (POSITIVE_WORDS and NEGATIVE_WORDS sets). Two overlapping area series showing emotional polarity.

The Linguistic Fingerprint

Visualization: Radar/spider chart (6 axes) Writing style analysis measured per quarter across six dimensions. Each quarter is plotted as a polygon on the radar, revealing how the stylistic signature shifts over time.

The Lexicon

Visualization: Horizontal bar chart Most frequent non-stop-word terms in the corpus. After filtering stop words and usernames, displays the top N terms by raw frequency.

IV. The Apparatus

“The machinery of temporal habit”

The Horologium

Visualization: Polar/radial chart (24 hours) Posting frequency by hour of day, rendered as a circular astrolabe. Separates original posts from quotes, showing when original thought vs. curation occurs. Named for the astronomical clock.

Hourly Distribution

Visualization: Heatmap Posting frequency by hour rendered as a heat grid. Shows concentration and gaps in the daily posting cycle.

The Weekly Rhythm

Visualization: Bar chart (7 bars) Day-of-week posting frequency. Reveals whether posting patterns differ between weekdays and weekends.

V. The Correlations

“The body, the silence, and the night”

The Night Scholar

Visualization: Scatter chart Sleep duration (from sleep.json) vs. posting output for correlated dates. Tests whether less sleep correlates with more posts — or whether the “night scholar” posts more when sleep-deprived.

The Research Pipeline

Visualization: Stacked horizontal bar Percentage of posts preceded by clipboard activity (copy events) within a 30-minute window. Tests whether posting is preceded by research behavior — copying URLs, quotes, or reference material before composing.

The Charging Confession

Visualization: Stacked horizontal bar Percentage of posts preceded by device charging within a 30-minute window. Tests the hypothesis that charging the phone triggers a posting session.

Cartography of Silence

Visualization: Timeline/gap analysis Maps silence gaps — periods with no posting — by duration and position. Reveals rhythms of absence: vacations, outages, deliberate withdrawal.

VI. The Reception (Conditional)

“How the discourse is received — partial metrics from the Threads API”

Only rendered if engagement data from the Threads API is available (requires raw posts with metrics field).

Stats Card Row

Six metric cards: posts with metrics, total views, average views, average likes, engagement rate, top post views.

Views vs. Likes

Visualization: Scatter plot Each post plotted by views (x-axis) vs. likes (y-axis). Dot size scaled by word count. Reveals whether longer or shorter posts perform better, and the views-to-likes conversion rate.

Most Viewed

Text display of the most-viewed post excerpt with its view and like counts.

VII. The Information

“Shannon’s children — entropy, surprise, and the architecture of unpredictability”

Stats Card Row

Eight information-theoretic metrics: tag entropy, Zipf alpha, Heaps’ beta, vocabulary size, mean surprise, word entropy, character entropy, conditional entropy.

The Taxonomy

Visualization: Squarified treemap The 20-tag LLM taxonomy with area proportional to post count. Each rectangle is colored by tag and labeled with name, count, and percentage.

The Surprise Distribution

Visualization: Density plot Per-post information surprise distribution. Surprise = -log2(P(tag)) — how unexpected each post’s primary tag is given the corpus distribution. Shows whether posts cluster around expected topics or spread into surprising territory.

The Character Strata

Visualization: Distribution chart Post length distribution across the 500-character Threads limit. Shows whether posts cluster near the limit, are typically short, or follow a bimodal pattern.

Mutual Information

Visualization: Heatmap Feature-to-topic mutual information matrix. Measures how much knowing one feature (length, time of day, media type) tells you about the topic tag. I(Tag; Length) dominates at 0.982 bits.

The Transition Matrix

Visualization: Matrix heatmap Topic-to-topic transition probabilities. Each cell shows the probability of posting about topic B immediately after posting about topic A. Reveals habitual topic sequences.

The Chaos Instruments

Visualization: Dashboard with 4 gauges Four unpredictability metrics displayed as instrument gauges: stay rate (tendency to repeat the same topic), burst posting frequency, hapax ratio (percentage of words used only once), and Heaps’ exponent (vocabulary growth rate).

Per-Category Vocabulary Entropy

Visualization: Horizontal bar chart Normalized vocabulary entropy by discourse tag. Shows which topics use the most diverse vocabulary and which are linguistically narrow.

Technical Notes

  • All visualizations are pure SVG generated at build time (no client-side charting library)
  • The page loads ~5 JSON data files and processes them in a single-pass loop over all 1,257+ days
  • Privacy filtering removes street addresses, “Country Village” strings, and location data
  • The 20-tag taxonomy comes from post-tags.json (LLM-classified via information-theory.mjs)
  • Engagement data (Section VI) requires raw Threads API exports which may not always be present
  • The interactive word oracle uses a lazy-loaded word index (word-index.json)