VIEW CONTROLS
PHOTOSENSITIVITY The diagrams move continuously, and a few pulse or cut between light and dark; each is small, and none flashes more than three times a second. Rows on strobe describe works that flash much faster. Freeze motion, at the top of View controls, holds every diagram on its telling frame. If your device is set to reduce motion, the page starts frozen, and Start motion, in the same place, sets it moving.
MOTION AND ANIMATION
GAMES AND INTERFACES
CHARTS AND INFORMATION DESIGN
COLOR
COMICS AND FILM
TEACHING
For educators: ways of reading a scene shot by shot, which the table’s rows can supply names and drawings for.
EFFECTS AND MAKEUP
TOOLS FOR MAKING
This group was suggested by the reader Robertonawhim (Instagram).
BUILDING AND KEEPING THE SITE
IN PRINT, AT A LIBRARY
Every term's motion diagram beside its glyph, in table order, with only its place (Part, container, section) and its name. Terms repeated in Part II from Part I appear once, under Part I. Each theme folds out; tap a pair to go to its term.
Tap a Part to see its sections. Tap a section name to go there; tap ↓ beside it to list its terms. opens Histories, Works cited or the Colophon.
| Thumbnail | Glyph | Term | Kind | What it does | Seen in |
|---|
By maker, by formula, and A to Z.
Each maker, a person, studio or team, with the motion diagrams of their rows side by side, so one hand’s devices read together. Until then, makers are listed in the Index, under Names.
ForthcomingRows that follow a mathematical rule, sortable by family, level and rule size. Partial: not every device reduces to a formula.
ForthcomingEvery term once, in alphabetical order, with no duplicate entries. Each entry links back to every place the term appears in Parts I to V, and a renamed term points to its current name.
ForthcomingThe same tiles as the Table of Visual Elements at the top of the page, arranged differently. At the top, each theme is one column read simple to complex, with no dates. Here the themes run across as columns and time runs down as rows, centuries and then decades, oldest at the top, never named eras; within a cell, oldest first. Tiles move by default, with the same motion diagrams as the Table of Visual Elements; the Glyphs switch shows the still marks instead. Borders show how each tile is dated. A row with two contested dates (HH) appears at both. Below the grid, detached strips hold rows without a date yet, and rows without a theme, like the periodic table's separate radioactive rows.
One numbered entry per marked row or dialect family. An entry is placed by when the tradition became irreducible to a parent culture (field 1, usually a range). Rights a group claims and the recognition others grant or deny are kept as separate, dated fields, each naming who acted, so placement never stands in for status and status is never dropped. Open fields are marked [U] and are filled by later passes, starting with Turkey. Provenance: [R] retrieved · [Rz] reasoned by Claude · [U] unverified · [S] stated by R.
Opening all entries...
Every source cited in the library doc, plus those retrieved for each new section, grouped by section in table order.
NOTES · THE PROJECT AND THE NAME
Readers who have sent in items, searches or fixes. Each contribution is also credited where it sits in the table. To contribute, see Open Submissions.
Names, works, places and terms, each linked to every row that mentions it. A term’s heading opens its own row; the links after a heading are the rows that name it. Works are in italics; a family name in small capitals is the filing name.
Anthropic publishes no energy, water or emissions figures for Claude, so past our own counts everything here is an estimate Rz. The numbers move with every drop.
| Electricity | 9.5–28.4 kWh |
| Carbon, at the 2025 world-average grid of 458 g CO₂e per kWh (Ember) R | 4.3–13.0 kg CO₂e |
| Water, from 0.5 L per kWh (a large operator's on-site cooling) to 9.4 L (the US average on site, 1.8 L, plus 7.6 L to generate the power) (datacentres.com, Data Center Knowledge on LBNL) R | 5–267 L |
Block unused connectors, check deploys with the lightest call, count tool calls per thread, and hand threads off before their context grows heavy. Sources: Works cited, under Energy.
About 1,113 turns and 76.9–81.9 working hours so far.
Each turn is taken as 10–30 Wh, scaled down from Zeke Hausfather's logs of his own Claude Code use: about 150 Wh an agent prompt, 96% of it re-reading context (Futurism, Blogspan) R. It may be lower, since Watershed finds lab benchmarks overstate production inference 4–20 times (Sustainability Outlook), or higher, since training is not counted U.
Anthropic has published no per-query energy, water or emissions figures for Claude, and no audited emissions report or climate target (Ethisphere; azvai, Oct 2026). Per Bloomberg (12 Aug 2026) it is measuring its footprint with Watershed; in June 2026 it joined the Frontier Climate carbon-removal coalition (azvai).
California’s SB 253 has companies over $1B in revenue file Scope 1 and 2 emissions by 10 Nov 2026 (Greenberg Traurig). Cloud electricity likely falls under Scope 3, due from 2027, so a first filing may not show Claude’s own use Rz. Recheck after 10 Nov.
Per the IEA, data-center electricity demand rose 17% in 2025 (AI-focused centers 50%), heading for about 950 TWh in 2030, about 3% of world demand (First Trust, on the IEA). Its 2025 base case put 2030 at 945 TWh (TaiyangNews) R. Hausfather put his own heavy use at about 1.1 MWh a year (Blogspan) R.
Readers so far. Netlify Web Analytics, server-side (screenshots, 11 Oct 2026): 1,030 unique visitors and 3,033 pageviews from 11 Sep to 11 Oct, 948 visitors in the last seven days R.
What would move these readings. How much the page is read and cited, and how it compares with the directories not yet found.
Reasoned estimates, not an appraisal Rz. What the work could be worth in money, by route; adoption itself is usually free.
| Route | Range |
|---|---|
| A. Free adoption: citations, reputation, commissions | $0 direct |
| A2. Grants for open humanities tools: NEH Digital Humanities Advancement, up to $75,000 (Level I) or $150,000 (Level II) (NEH); Schmidt Sciences’ Humanities and AI Virtual Institute, up to $800,000 in 2026 (CNI) R. Both usually fund institutions; one listing says NEH is not taking applications for fiscal 2026 U. | up to about $150,000 for a prototype phase; up to $800,000 at the scale of a funded team |
| B. One-time license as an expert dataset to an AI lab or tool maker: 2,000 to 4,900 hours of expert work at $85 to $150 an hour Rz | about $170,000 to $735,000, if exclusive |
| C. Subscription reference over 3 to 5 years: a smaller niche than Mobbin at $50,000 to $500,000 a year, times 2 to 7 Rz | about $100,000 to $3.5M |
| D. Platform: tools, a file format and a marketplace around the vocabulary; needs a team, funding and years Rz | $10M or more, a long-tail outcome, unlikely |
| E. The work itself, as authored art | no comparable; only a buyer sets it U |
Nearest funded precedent: the Distant Viewing Lab (Taylor Arnold and Lauren Tilton, University of Richmond), maker of DV DataReel, with about $1.9 million in reported awards: NEH 2018 and 2022 (Richmond, Richmond), Mellon 2022 and 2025 (Mellon, Richmond) RRz, plus a share of a 2025 Schmidt Sciences award (23 teams, $11 million; Schmidt Sciences) R. Not all film work; none earmarked for DataReel.
Mobbin, a curated UI reference: $1.6M a year in revenue, 10,000 customers, 2024 (EveryDev). LottieFiles, a motion-design platform: about $47M raised, 3.1 million users (TechCrunch). Expert AI-training work: $85 an hour or more (Sacra). Small software businesses sell for about 2 to 7 times yearly revenue (Livmo, FE International).
The catch, revised 6 Oct. The page is public and free to read, so anyone can collect it.
Which holds is unknown until the grassroots search is done U; the ranges are ceilings under the first account, estimates under the second.
| Cost | Amount |
|---|---|
| Claude, on the human designers’ plan, counted as plan share from meter screenshots R: on 11 Oct 2026, 18 points of one session in about 70 minutes; 8% of the week. | 18 points of one session in about 70 minutes; 8% of a week |
| Context, not what the designers pay: Anthropic’s individual plans list at $20, $100 or $200 a month (Dupple, checked against Anthropic’s pages on 10 Oct 2026) R. | $20 to $200 a month, list |
| Unpaid hours, the human designers: 76.9–81.9 working hours, from the build line. Priced three ways below, because the price depends on the question asked. | $0 to about $10,410 |
| Domain, unit-table.com: registration $72.85 plus full domain protection $41.85, three years, no tax, from the designers’ receipt S. Renews in 2029 for $107.85 plus $22.95, $130.80 the three years S. | $114.70 for three years, about $38 a year (about $44 a year at renewal) |
| Hosting, Netlify Pro (deploy record) R: a flat $20 a month list since 2026 (PricingSaaS), 15 credits per deploy of 3,000 a month (MakerKit) R. Whether it covers other sites: not recorded U. | $20 a month, list |
No one is paid for these hours. A price depends on the question asked, so all answers stand side by side Rz, each on the build line’s 76.9–81.9 hours, recomputed at every drop.
| Method | Result |
|---|---|
| 1. Not priced. What did the makers give? Hours, not money: the work was done for open use, and a price would imply a sale that did not happen Rz. Unpaid is not unvalued: the human designers hold that work like this should be paid S. | 76.9–81.9 hours; $0 |
| 2. Market rate. What would commissioning it cost? Route B’s rate, $85 to $150 an hour (Sacra, on Mercor) R, fewest hours at the low rate to most at the high Rz. | about $5,470–10,410 |
| 3. Floor rate. What is an hour worth at the legal minimum? That depends on place, so no single figure Rz; see the map below. | depends on place |
| 4. Grant rate. What would a funder budget? NEH Level I: up to $75,000 over up to 24 months R. Prices the phase, not the hours, and claims no award Rz. | up to $75,000 for a prototype phase |
Monthly statutory minimum wages by country, from the ILO’s ILOSTAT (updated 11 October 2026) R. No country singled out; tap one, or pick from the list, for its local amount, year and kind of minimum. One five-step scale for both views Rz.
Selected
Tap a country, a small-state chip, or choose from the list: its figures appear here.
Not drawn at this scale R:
How to read it. One rate per country: national, or else the capital’s or widest sector’s R. Dollars at market rates, or PPP (2021 international dollars) for local prices R. Hatched: no statutory minimum; gray: no record; lined: no conversion; details on tap R. Three apparent unit slips (Ecuador, Uruguay, Cambodia) are corrected from national sources, originals on tap RRz. Natural Earth 1:110m, Equal Earth projection.
What would decide between them. Nothing: they answer different questions, so none is the true price Rz. The spread is the finding: the same hours run from $0 to market rate.
Left out. M.’s hours, not yet counted separately U; hours away from the chat, counted only where noted; a real wage’s on-costs and taxes; Claude, counted as plan share. None of these prices the work itself; routes A to E do Rz.
The card’s questions as a tree, each answer a bar on one log dollar scale ($0 on a stub; dashed bars are ceilings); the hours bar recomputes at every drop Rz. Compare Decision tree and Forest plot.
Inputs in their own units (hours, plan share, dollars), drawn at one width; outputs not yet measured, dashed. The table between them is a stock that reading does not use up Rz. Compare Flow band.
Is two people’s unpaid work doing what funded teams are paid for? Of Schmidt Sciences’ 23 teams in 2025, one overlaps: Bamman’s, with Arnold and Tilton, measuring close-ups, camera movement, storylines and sight against sound in television R. They measure devices across thousands of films; this page names and draws them.
Where the money goes. Award funds carry at most 10% indirect costs (Schmidt Sciences) R, so they pay for people, students, GPU compute, film collections, workshops and software upkeep; itemized budgets are not published Rz.
What this page stands in for. The research-assistant, developer and illustrator layers, the costs AI compresses most; not film-scale compute, collection rights, peer review, permanence or training students Rz.
Two readings. Comparable at a smaller scale: an NEH Level I or II prototype, done unfunded. A cost partly hidden: unpaid hours, M.’s uncounted, compute inside a subscription, so the cost shown is what makers and provider absorb, not what the work costs Rz. The human designers hold that it should be paid S.
What would decide. A comparable itemized budget, the full hours, and an outside accuracy check, which is what peer review buys Rz.
The human designers search for precedents before adding anything, to avoid redundancy S. A novelty pass on 12 Oct (13 searches) tested the page feature by feature and found these, each with its own audience and funding, noted per entry R:
What survives the pass Rz: no single feature is new. Not found: a vocabulary of visual devices that accounts for its own making in those devices, as the Energy card does.
Not yet searched U: grassroots directories (Are.na, Tumblr, Behance, personal sites), non-English sources, per-claim provenance marks, animated comparisons in glossaries. Any could overturn the narrow claim; sources welcome at editors@raisonne.ai.
About eleven close precedents, most personal or reader-funded, each serving its own audience; this page is offered in the same spirit, as a public service S.
Sources are listed in Works cited under Economics.
What changed on the page each working day, newest first, grouped by week for this month and by month before that.
One section renamed.
Walkers that keep striding where a loop starts over, two walkers turned to side view, two resources, a tile that enlarges, and an Index that answers at once.
Twenty-four more rows in the culture sections redrawn as scenes, so each reads without its definition; 64 of the 122 that draw people are done. Later: eight weak redraws drawn again, walkers in profile, beards that hold on a turned head, and four resources from NIST. In the evening: walkers that keep their build, legs that lift toward the camera, elbows that follow through, beards round the jaw, and the palette page level with palpik.ink.
Faces III joins the portraits; skin runs from light to dark along one violet; faces take a matte gleam and a soft shadow in their creases; rows of people take seeded faces. Later: a second set of head coverings, hands at the face in three-quarter and profile, eyes on measured proportions, and coverings on main-table tiles where their source documents them. Later still: forty rows in the culture sections redrawn as scenes, so each reads without its definition.
Palette changes repaint at once; 62 rows redrawn to read at table size; tiles on the light palettes sit on grounds of their palette’s own colors; Hours replaces Peachy on the Views page.
Build folds further in; a hundred and fifty-two rows source-checked; Matter loses its sub-groups; the top table is renamed; fold pills at one size.
Six rows on seeing; seventeen dark grounds and eighteen near-empty tiles redrawn as places and scenes; sharper sheet pictures; a loading sun.
An eye for the table, and rows tightened from proofs.
Measured values replace constructed ones, and the pear joins the objects.
The standard figure and the measured subjects reach the rows: one shared deer, owl and rabbit; the table figure set at its drawn canon; the first flora and object standards.
The figure standard opens from Build on its own page: eighteen model sheets of the table’s standard figure, smoothed where it was faceted, with hands for age and arches added. Model sheets for animals are under way and not yet on the table.
A second source check of 50 terms, with 10 from the first rechecked: three corrections, two terms folded away, and sharper wording and sources on eight more.
Closed-circuit loop given its own drawing; Resources folding in place again; llms.txt term counts corrected; Part II width fixed; dead links taken off; privacy note; wider Parts on desktop; archived links repaired; corrections from a source check of 50 terms; an Audits card (12 entries); 145 redrawn motion diagrams from the 31 Oct–2 Nov review rounds put on the table, and four walking rows redrawn as whole scenes.
83 motion diagrams set to close on the 24 s loop, pop-outs brought above the Contents menu, the menu’s view controls able to unfold again, Freeze frame moved under Access, the pop-out button moved into the header, one button standard, the Legend folding in place, 29 doubled tiles removed, card pills back to flat cross-colors, Peachy, Sky and Forest given their own card colors, a Redraws card in Build, and two rulings on figures and motion recorded (15 entries).
Two terms added, nine dated, 37 shared into Part I, steadier running text, a new On-screen text section, matched Tab pills, a style cleanup tried and undone and 23 more terms dated (16 entries).
Two fixes: no more “NaNs” row in the Temporal table, and long section titles kept on one line on the narrowest phones (3 entries).
A sturdier layout: the Temporal table no longer covered by the cards below it, Part II no longer widening the page on desktop, section titles resized when the type setting changes, and card spacing set by the stylesheet rather than measured after loading (6 entries).
Index filing and two names corrected; Klein light on dark; pills one color; Pop out on every dense card; a Resources card in the cheat sheet’s place; On deck removed; the third-person camera from behind, the Index in plain text and four Software rows.
Then the Wait cursor traced to the Lisa, the periodic table opening on its condensed preview, the bottom cards regrouped between two flowers, and the Index opening straight into the reader panel.
Then Full tree and Legend as their own cards, the Table of contents and the Contents menu notes cut, and Klein’s Part colors spread; a double rule above the periodic table; pop-outs closing from a corner badge; Resources popping straight out, Tab buttons on the two tables, and Updates listed by date.
Then the Body figure resized to measured proportions, the runtime scripts gathered, the Contents menu’s View in its own capsule with smaller headings, and Riso’s fold buttons lightened; then, from desktop notes, opened rows that squeezed their neighbors, header rows that could not be tapped, the Part cards at the periodic table’s width, the periodic table opening with its rows and the Condensed table as its own card.
Then one 700 px column on desktop and fold pills that had vanished into their ground.
Then the Tab arrow and the card titles under the periodic table, and a page that no longer jumps while it loads.
Then a new section, Anonymity + censorship, section names with +, shorter Part II names and new subsections, every header on one line, two rows to a line on desktop, rows moved into the subsections, Czechia and Slovakia, the veiled face in the Siyer-i Nebi, mysterious light and steam, Part II rows shown in full, five Iranian rows, toned-down rails, a clearer selected cross-link, the veiled-face row revised, and paragraph breaks in long definitions and Histories (99 entries).
Sources, the ECG row and the craft pass; search, Part maps and view controls.
Then one figure across the table, proportion rows and new terms.
Then six redraws, new Software and Games rows, contributors and the Body compass.
Then two-line rows, one-line headers, rows on smoke projection and the phantasmagoria, four from Pikmin, two from the farm games, the Sources cards and fold controls tidied, search, headings and Back adjusted, a turn count in Details, row padding and Part II headers, two type sizes matched, two Final Fantasy rows, the Contents colors, the menu brought to standard, a hamburger-only Contents button, tighter section spacing, Before and after, and colored Part pills.
Then four rows by makers and Part IV By maker, a summon animation, Rakuyo moonlight accents, a looping link preview, rebuilt Works cited back-links and smaller fixes.
Then session reviews, a second pass on Part IV, a Forest palette and the plumbob.
Then a general index, Corporeal under Matter, five rows on the gun barrel, split screen, the third-person camera, grid levels and need meters, and fixes to Sources and the Body compass (93 entries).
Layout, prior art, For builders, Economics, Details, text, figures, fixes.
Then motion, periodic table, glyphs, navigation, tile caption and views.
Then software, order, layout, view controls, education and fixes.
Then software, order, imprint, accessibility, full tree, view controls, pop-outs, color-safe palettes, defaults, arrows, motion and fixes.
Then ECG rows, software, maker palettes, motion loops, full tree and view controls.
Then palettes and fixes.
Then education rows, periodic table, fold pills and fixes.
Then the footer.
Then glyphs, figures and fold pills.
Then Details, Education and figures.
Then physics tags, Histories, page weight and the colophon.
Then the craft pass: Basics, Korea and the sky (69 entries).
Observance, Listening, Energy, Economics, Figures, Layout, Submit work, Text, Fixes (17 entries).
Body, Layout, Fixes (4 entries).
Lenses, Views, On deck, New rows, Layout, Naming, Propaganda, Rows, Fixes, Body (43 entries).
New section, New rows, Siblings, Propaganda, Prior art, Submit work, Layout, On deck, Lenses (25 entries).
Specimens (1 entry).
Loading, layout, On deck (3 entries).
Loading, views page (2 entries).
Decision and uncertainty diagrams, siblings and families, prior art, Economics diagrams, a world minimum-wage map, Energy drawings, On deck, layout, temporal table motion, symbol cheat sheet, Beta, loading, acknowledgement (22 entries).
Translation section, new rows, Energy card, Economics (7 entries).
Table layout and Seen-in examples (3 entries; includes late 6–7 Oct).
Layout, Parts V and VI, renamed rows (4 entries; from the working notes).
The first periodic table, mirror strips, a Math line (4 entries; from the working notes).
Five new Part I sections, a dates audit, culture containers (10 entries; from the working notes).
Entries FORTHCOMING, filled in from the working notes.
needs tendingsomesteadylittle
Data and model standards, by analogy Rz.
Marks: met, partly, not yet.
In CRediT terms (ANSI/NISO Z39.104-2022), with its degrees, as confirmed by R. and M. on 22 Oct S.
R. Conceptualization (lead), Methodology (lead), Supervision, Project administration, Visualization (lead), Writing – review & editing (lead), Writing – original draft (supporting)
M. Writing – review & editing (supporting)
Claude Opus 5.5, by analogy: Writing – original draft (lead), Investigation, Data curation, Software, Visualization (supporting)
In the four sections of the AI Usage Cards generator (Wahle and colleagues, 2023) R.
Project details. Unit Table, raisonne.ai lab; human designers R. and M.; model Claude Opus 5.5 (Anthropic), used in the Claude mobile app with web search, a code sandbox and Dropbox. Earlier sessions may have used earlier Claude versions; they are not recorded here U.
Methodology and experiments. Claude searched the web for each term’s sources and dates, drafted rows, sorted and checked them, and wrote the page’s code, its glyphs and its motion diagrams. R. chose the terms, the order and the look; M. reviewed. Every claim carries its provenance: R retrieved, Rz reasoned, U unverified, S stated by the designers.
Writing and presentation. Claude drafted the text under R.’s direction, and R. and M. edited it; the carousel slides were drawn by Claude from the page’s own tiles.
Ethics and limitations. A model can be wrong with confidence; the provenance marks show which claims were checked against a source. The human designers are accountable for what is published, and corrections are welcome at editors@raisonne.ai. The making is counted in Energy and Economics, as estimates with ranges.
For a small project, these five lines cover most of it Rz:
Every working session ends with a handoff note, called a Lantern after the save-point lanterns in Bloodborne, that carries the task to the next session and closes with a short review.
Four questions. What we set out to do, what happened, why, and what changes next. This was reached here on its own, then found to be the shape of the After Action Review, which the US Army developed in the 1970s (Morrison and Meliza, 1999) R.
Scores, both ways. R. is scored on clarity of requests, prioritizing and how specific the feedback is; Claude on accuracy, completeness, sourcing and efficiency. Each runs 1 to 5, with an average, the change from the last session, one or two things to improve and any flags carried forward. Tying each score point to described behavior follows behaviorally anchored rating scales (Smith and Kendall, 1963) R, in a much lighter form: the anchors are short and written by the two parties themselves Rz.
Mistakes log. Each note carries the mistakes made so far, with the check that now catches each one, so none is made twice; the Works cited back-link fault found on 24 October is the latest.
Limits. The scores are the two parties rating each other, not independent raters, so they show a trend within this project and nothing beyond it Rz.
Redraws. Counted from the notes kept inside each drawing’s code and from the review sheets that save R.’s marks; rounds marked only in conversation are added as they are found, so the counts run low.
Every drawing on the table, counted by how many times it has been redrawn.
Each check of the table, what it found and what changed, newest first.
A source check takes 50 terms twice a week, plus about 10 already checked as controls. Every claim is checked against its sources, for anything invented, contrary to its sources, missing a competing or earlier date, making several claims without a citation, or unconfirmed. From 5 Nov each term is also read for weakness, after the weeding criteria libraries use (CREW): a term another one already does better, one too slight to teach anything, or one that adds nothing beyond its link. Weak terms are only flagged; each is kept, folded into another or removed by R.’s decision. Links and drawings get their own checks. Once a week a simplification check reads the page itself rather than its terms: what could be cut, merged or made plainer (a layout, a control, a label said twice), with what each change would gain and lose; each is kept or changed by R.’s decision. The same week, the sections that measure the table itself (Energy, Economics, the footer counts) are brought up to date. A drift check runs weekly and before each release: the build against the locked palette record, each drawing against its last passed version, and the rulings in force against the record of rulings.
Filled: done. Hollow, in the accent color: due and not yet done. Above the source checks, the share of sampled terms found wrong (the first sample random, the later ones not).
| Date | Audit | Checked | Found | Changed |
|---|---|---|---|---|
| 17 Nov | Drift | every drawing, the palette record | 13 drawings changed since the 15:21 release, all of them recorded redraws or restores; the palette record behind the build | palette record brought up to date; nothing reverted |
| 17 Nov | Links | every outside link (3,521) | 2,761 loaded; 32 failed, each opened a second way: 1 broken, 2 moved; 728 refused automated checks or timed out | 1 repaired, 2 updated to their new address; 16 refusing hosts sampled, all load; the rest left for the next check |
| 16 Nov | Source check, third session | 50 terms + 10 controls (4 rechecked) | 3 contrary to sources; 7 claims too broad or unfound; 2 with sources in conflict | corrected; competing accounts added; 3 terms left for the next check |
| 16 Nov | Source check, second session | 52 terms + 10 controls | 1 date contrary; 1 wrong title; 9 claims too broad; 3 with sources in conflict | corrected; competing accounts added; 1 term left for the next check |
| 16 Nov | Source check, with a weakness reading | 50 terms + 10 controls | 4 dates contrary to sources; 8 claims too broad; 2 weak terms | dates and wording corrected; 2 terms left for the next check |
| 5 Nov | Source check, with the first weakness reading | 50 terms + 10 controls | 3 contrary to sources; 2 weak terms | corrected; 4 narrowed, 2 cited, 2 folded |
| 3 Nov | Source check | 50 terms, at random | 2 contrary; 4 missing a date; 6 uncited; 9 unconfirmed | corrected the same day; 5 still being checked |
| 3 Nov | Links | every source link | 104 no longer loaded | 27 moved, 46 archived, 31 citation only; 30 archive links repaired |
| 2 Nov | Duplicates and chaff | every drawing and term | 8 shared drawings; 9 not devices | 8 resolved; 8 removed, 1 folded |
| 2 Nov | Loops | every motion diagram | loops not closing | all closed, timing devices excepted |
| 9 and 12 Nov | Source check | missed; the 16 Nov check took the queue on | ||
| 19 Nov | Source check | next on the twice-weekly schedule | ||
| 19 Nov | Simplification; self-measures | first weekly pass due | ||
Notes on each check, in full:
Every drawing on the page was compared with the release of 17 Nov 15:21 and with the live page. Thirteen had changed, and all thirteen are in the record: the redraws of that afternoon and the two restored after the ruling on death and blood. No drawing reverted and no passed redraw was lost. The locked palette record lagged the build: the grounds re-chosen for the afternoon’s redraws had not been written into it. It now holds them; the palette values themselves had not changed.
Every outside address in the page, its scripts and llms-full.txt was requested: 3,521 in all, against the 3 Nov check of source links alone. 2,761 loaded. Thirty-two failed, and each was then opened a second way; 31 load, most from sites that turn away automated requests. One was broken: The Nation’s “Imperfect Cinemas” (Emily Witt, 2010) now links its plain address. Two Peachpit articles now link Adobe Press, where they moved. A further 728 refused automated requests or timed out. The 16 hosts with the most of them were sampled a second way and all loaded, among them Britannica, MoMA, the Met, GitHub, the AFI Catalog and PubMed; the 56 archive copies could not be reached from the checker at all. The rest are left for the next links check, best run from an ordinary browser.
The first queue is used up; this one was rebuilt from unchecked rows with no source attached that make more than one claim. Fifty terms and 10 controls, of which 4 were rechecked. Nothing invented. Three rows were contrary to their sources: the Fox and the Hound delay (about six months, one account a year), the Over the Garden Wall background guide (glazing named, then skipped to save time) and Albers’s folding exercise (no cutting where possible, glue only later). Seven were narrowed or had an unfound detail removed, among them the three arches in Ave Maria (now credited to a fan wiki), the Space Jam animator count and site credits, and the glass in Moholy-Nagy’s balance studies (glass-plate negatives). Two Science SARU rows now give both accounts of its founding. Three basic terms were not reached. Errors of the first kind: 3 of 50, a non-random sample.
The next 52 terms from the queue, including the two carried over, and 10 controls. Nothing invented. One date was contrary to its source: Zarrinkelk set up Iran’s ASIFA branch while studying in Antwerp, not in 1977. One title was wrong: Princess of Rome. Nine claims were narrowed, among them the 1934 firsts (two films, not one), the lamps in the mirror palace and the shooting place of Tehran Taboo. Three rows have sources in conflict and now show both: the date of the term item number, the length of Ahmadieh’s Mollah Nasreddin, and how Mayabazar’s moonlight was made. Two details were searched and not found and stay as stated; one term was not reached. The Opera Jawa count is settled at seven by the director’s own account. Errors of the first kind: 2 of 52, a non-random sample. The controls held; two carry the same unconfirmed details as before.
The next 50 terms from the same queue, every claim checked against a source, and 10 terms from earlier checks rechecked as controls. Nothing invented. Four dates were contrary to their sources: Search and find (Where’s Wally?, 1987), Puppy-dog eyes (Shrek 2, 2004), Awooga eyes (the copypasta, 2018) and National style (the film taken for Soviet work was from 1955). Eight claims were narrowed to what the sources say. Two terms were not reached and lead the next check; a few named details are still unconfirmed and stay marked. Weakness reading: Grain-loop still repeats Grain loop, and Korean, North and South is a scope note rather than a device; both flagged for R.’s decision. Errors of the first kind: 8% of a non-random sample, not a rate for the whole table. The controls all held.
50 terms from the queue of those making several claims without a citation, every claim checked against a source, and 10 terms from the 3 Nov check rechecked as controls. Nothing invented; 3 contrary to their sources (a date, a study’s viewers, a credit); 2 claims given citations; 4 narrowed to what the sources say; 2 terms folded away by the first weakness reading, one repeating others on weak sources and one now a name under Dolly. Errors of the first kind: 6% of the sample (95% range 2–16%); the sample was not random, so this is not a rate for the whole table. The controls all held. All corrections made the same day.
50 terms drawn at random, every claim checked against its cited sources and a second source where needed. Nothing invented; 2 contrary to their sources; 4 missing a competing or earlier date; 6 making several claims without a citation; 9 with a detail unconfirmed. The errors were corrected and the missing citations added the same day; 5 details are still being checked. Errors of the first two kinds: 12% of the sample (95% range 6–24%). From here: 50 terms twice a week until every term has been checked, and every new term checked before it is added.
Every source link tested: 104 no longer loaded; 27 now go to the same page at its new address, 46 to an archived copy, and 31 keep their citation without a link. Thirty archived links that pointed to an archive of an archive repaired.
Eight pairs of different terms found sharing one drawing, each resolved with its own drawing or by folding one term into the other; in a separate pass, eight terms that were objects rather than devices, or lacked a documented case, removed, and one folded into another as an also-name.
Every motion diagram checked against the 24-second loop; all now close on it except the devices whose ratios or timing are the point.
The table’s diagrams began as one-off drawings, so the same person or animal could come out a different size, build or pace from one row to the next. The model sheets draw each subject once, from measured sources, and every row draws from that one sheet: proportions from measured body data, joints held inside measured ranges of motion, timing from measured gaits and wingbeats, one engine so a correction lands everywhere, and every value marked with where it came from.
FIGURE
FAUNA
FLORA
OBJECTS
PALETTE
SHOWCASE
Unit Table © 2026 R. and M., raisonne.ai lab.
Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0): the written content (definitions, notes, Histories, section and Part notes), the selection and arrangement of the vocabulary, and the glyphs and motion diagrams as artworks.
Credit as: Unit Table (unit-table.com), raisonne.ai lab, R. and M., with a link to the license.
Commercial use is not covered by this license; commercial licenses are available on request at editors@raisonne.ai.
Code: the page's program code is under the MIT License (full text below).
Not covered: third-party images and stills, credited and used for criticism and review; their rights stay with their makers, who may ask for removal at editors@raisonne.ai. Quotations and titles of cited works are not covered.
The motion diagrams and glyphs are Claude’s own drawings of each technique, not reproductions of the makers’ works. Full entries, sources and provenance live in the library doc and on the library page.
Note on drafted text: much of the written content was drafted by Claude (Anthropic) at R.'s direction, from R.'s selections, recordings and screenshots, and reviewed by R. and M. None of it would exist without that direction. The license covers the rights R. and M. hold in it.
Fingerprint: unit-table b02aa5d5-9c42-479b-ad2d-c0162d972bb0.
MIT License Copyright (c) 2026 R. and M., raisonne.ai lab Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
The table teaches as some early-learning schools do: one quality at a time, made visible and handled Rz.
In class. Ways a teacher can use it, at any age Rz:
Lesson plans: FORTHCOMING.
Lineage. Froebel opened his first kindergarten in 1837, with play gifts (balls, blocks, parquetry tiles) given in order of complexity to teach shape, structure and perception (99% Invisible); (Publishers Weekly) R. Norman Brosterman’s Inventing Kindergarten (1997) argues the gifts prepared modern abstraction and architecture; Johannes Itten, who taught color at the Bauhaus, trained as a kindergarten teacher (Icon) R. Publishers Weekly calls parts of it informed speculation R.
Yoram and Sandra Gross made The First Animated Step (1975), a survey of the history of animation, both as a book and as a 24-minute film (ACMI; AustLit) R; the couple later set up a company to distribute their own films to Australian schools (Australian Screen) R. It was a row of its own until 11 Nov 2026.
Albers taught color with colored paper, practice before theory: Relativity of color R.
Montessori materials isolate one quality (size in the Pink Tower, color in the Color Tablets); paint sample cards can stand in for the tablets (Montessori Mom); (Numa School, weak) R. The same move as One variable changed Rz.
Waldorf form drawing (repeated patterns, reflections, interlaces) is often walked or traced in the air before it is drawn (Escuela Caracol, after Child and Man, 1987) R. Steiner set it out for the first Waldorf school, Stuttgart, 1919 (SteinerBooks, Schuberth and Embry-Stine, Form Drawing) R.
Rows. Froebel’s gifts and Form drawing and Wet-on-wet watercolor (first lesson still sought; Waldorf teachers welcome to review U) sit in Art school exercises; the Pink Tower and Color Tablets in One variable changed Rz.
Every color and typeface the view settings use, so the page can be rebuilt or matched elsewhere. Fonts load from Google Fonts; each type direction falls back to system fonts if they do not.
To: FromSoftware
Every term as plain text, for reading without scripts, is on its own page: unit-table.com/plain.html. The whole vocabulary as one Markdown file is at llms-full.txt, with a summary at llms.txt.