<- all tokdocs
2026.10.14@jason_swet

Jason Murray's "taxonomy of visuals" is a real beta tool named Iris, and the biological ladder he borrows took 300 years to build

AI
Watch on TikTok

View on TikTok ->

The artifact is a 102.35-second vertical clip at 1080x1920 in an MP4 container, video encoded as HEVC Main profile in yuv420p at 2997/125 (23.976 fps) and 1,369,860 bps, audio as HE-AAC at 44,100 Hz stereo and 96,108 bps, for an overall bitrate of 1,470,934 bps and a file size of 18,819,207 bytes. The post timestamp is 1791999459 (2026-10-14 17:37:39 UTC). The uploader handle and channel nickname are both jason_swet; the audio is "original sound" by jason_swet. The description is 1,476 characters. It opens "I’m developing a taxonomy of visuals that connects images through the techniques and ideas behind them." and closes with "Visit visuals.modernartdirection.com to join the waitlist". At capture on 2026-10-19 01:54 UTC the metadata recorded 149,700 views, 19,400 likes, 191 comments, 1,641 reposts, and 11,138 saves. I read 14 of the 51 extracted frames (001, 004, 008, 012, 016, 020, 024, 028, 032, 036, 040, 044, 048, 051) and the 263-word, 17-segment transcript. Talking-head frames (012, 028, 051, and the lower halves of 001, 004, 008, 044, 048) show a bearded man in round glasses and a green "LOS YORK" trucker cap at a desk behind a podcast microphone, with shelving and a lamp behind him. Screen-recording frames (016, 020, 024, 032, 036, 040, and the upper halves of 001 and 004) show a node-and-line graph: circles labelled "Differentiate", "Group", "Sequence", and "Infer", boxes labelled "Feature", "Grouping", "Space", "Social", "Semantic", and "Cultural", a row of technique icons, and hundreds of thumbnails below. The card in frame 036 reads "01 OPERATION Differentiate", "02 DOMAIN Semantic", "03 TECHNIQUE Semantic contrast", and "04 EXECUTION WWF", a dash, "Before it's too late". Frame 008 shows a tablet and stylus over a grid of surreal portraits in an app whose header reads "Uncomfy", with a status bar reading "3:16 PM Wed Apr 8". Frames 044 and 048 overlay sepia animal engravings. Burned-in captions track the audio word for word. The transcript is one first-person pitch covering references versus ideas, an operation called "differentiate", and a biology analogy. The last segment ends at 102.12 seconds against 102.37 seconds of audio; nothing is hallucinated.

The tool has a name the video never says, and the waitlist page matches the screen recording level for level

The script calls this "a taxonomy of visuals" and never names it. The description points to visuals.modernartdirection.com. I fetched that page on 2026-10-19. Its title pairs "Iris" with "A Taxonomy of Visuals", its headline is "Explore how visual ideas connect.", and its footer reads "Iris™ by Modern Art Direction". The page describes the structure in one sentence: "The four levels ask: what are you trying to do (operation), what are you working with (domain), what move could help (technique), and how does it work in an example (execution)?" Those four labels are the row headers in frame 032 and the card in frame 036.

The page also confirms the description's beta claim: "Iris is in beta. Join the waitlist to hear when the full library opens." Access is gated: "The full library requires active member or invited course access," and "Joining the waitlist does not create an account or grant library access." His newsletter post "I'm building the art direction school I wish existed", which the newsletter index dated 17 days before my fetch (roughly October 2, 2026), says "I call it, Iris™" and "October Bootcamp students are going to be the first people to get access."

The audio names one operation, "differentiate", between 51.92s and 58.46s, and one technique, "semantic contrast", at 64.72s. The screen shows more. Frame 032 labels the operation row "Differentiate", "Group", and "Sequence", and frame 040 adds "Infer" as a fourth, selected in pink. Frame 032 repeats a domain row under each operation ("Feature", "Grouping", "Space", "Social", "Semantic"), and frame 040 adds "Cultural". The waitlist page publishes none of those names; everything beyond the four level labels comes from the screen recording.

The creator is Jason Murray, and "over a year" of teaching checks out against his own timeline, barely

The handle is jason_swet, and the TikTok profile I fetched on 2026-10-19 shows the bio "The Art Direction of Everything" and 107.4K followers. The video never gives a name. The Teachable course page for "NEXT Art Director: January 2026" describes the instructor as a Canadian creative director and art director based in Southern California with experience at Amazon's Brand Innovation Lab, BBDO, and Huge. Teachable's case study, bylined Chris Chan and dated May 21, 2026, names him as Jason Murray and says "When he quit his agency job in 2024, he did not build a course library."

At 5.90s he says, "I've been teaching art direction for over a year now." The case study says "Eight months after announcing the idea publicly, he had run three live bootcamp sessions, enrolled over 120 students," which places the public announcement around September 2025. It reports "He started his first cohort at $300 per person," then "raised the price to $800 for the January cohort," and gives cohort sizes of 42 and 35. His own newsletter post from early October 2026 says "A year ago taught my first bootcamp for art directors." A first bootcamp around October 2025 and a video dated October 14, 2026 supports "over a year" by weeks at most; "about a year" is the safer reading. The Modern Art Direction brand is older than the bootcamp. The Apple Podcasts listing for the "Modern Art Direction" podcast, hosted by Jason Murray, shows its earliest episode on May 9, 2025. None of these pages publish the current cohort's price.

"Decades of research" undersells the paper trail by about a century

At 38.88s he says, "From gestalt psychology to semiotics, there's already decades of research." The dates run longer. Max Wertheimer (1880 to 1943) published the phi phenomenon experiments in 1912 under the title "Experimental Studies on the Perception of Movement", which Wikipedia's gestalt entry treats as the movement's formal start, with Kurt Koffka and Wolfgang Köhler as co-founders. Wertheimer's grouping paper, "Untersuchungen zur Lehre von der Gestalt, II", appeared in 1923. That is 103 years before this video.

Semiotics is older still. Wikipedia's semiotics entry records that "the first use of the English term semiotics dates to the 1670s" and that John Locke proposed the term as the name for a new field in his 1690 "An Essay Concerning Human Understanding". The modern discipline's founding text, Saussure's "Course in General Linguistics", was published in 1916, compiled by Charles Bally and Albert Sechehaye from lectures given at the University of Geneva between 1906 and 1911, and proposed "a science that studies the life of signs within society." The technique vocabulary on screen is gestalt vocabulary nearly verbatim. Frame 040 lists "Closure"; frame 036 lists "Common region", "Figure-ground reversal", and "Selective focus"; frame 032 lists "Proximity clustering" and "Symmetry". Wikipedia's gestalt entry lists proximity, similarity, closure, continuity, common fate, symmetry, and figure-ground as the grouping principles, and its Principles of grouping entry credits Irvin Rock and Stephen Palmer with identifying additional principles beyond Wertheimer's set, without a date in the text I fetched. So "decades" understates the age of his own sources, and the technique layer of Iris is relabelled perceptual psychology, which fits his statement at 33.28s that "These principles have existed long before I ever started thinking about this."

The eight-rank ladder he recites was assembled by five people across 300 years

Between 83.40s and 96.44s he says the taxonomy is "Inspired by the way biologists classify the characteristics and evolutionary relationships of animals from domain kingdom phylum to class order family genus species," while frames 044 and 048 show sepia engravings of a primate skeleton and a plate of domestic and zoo animals. He credits "biologists" generically, which is accurate, and the history behind that word reframes his "very much a work in progress" line at 33.28s.

Linnaeus's "Systema Naturae", first published in 1735 with the tenth edition of 1758 serving as the starting point of zoological nomenclature, used five levels: "kingdom, class, order, genus, and species." The other three ranks arrived separately. Per Wikipedia's family entry, "familia" was first used by the French botanist Pierre Magnol in 1689, Linnaeus borrowed the word for broad plant groups in his 1751 "Philosophia Botanica", and the rank between order and genus in zoology was introduced by Pierre André Latreille in 1796. "Phylum" was coined by Ernst Haeckel in 1866. Domain is the newest. Wikipedia's taxonomic rank entry says domains were "so named by Carl Woese in 1990," and the three-domain system entry credits Carl Woese, Otto Kandler, and Mark Wheelis in a June 1990 PNAS paper. So the "domain to species" sequence he delivers in nine seconds spans Magnol in 1689 to Woese in 1990. The analogy holds better than he states it. A taxonomy that is "very much a work in progress" and built on principles that "have existed long before I ever started thinking about this" describes the biological one too.

The examples on screen are real campaigns with dates, and the reference app is Cosmos

The line "Designers collect things that look nice." at 10.76s plays over frame 008, a tablet showing a grid of surreal portraits. The app header reads "Uncomfy" with the search placeholder "Search in Uncomfy...", a "Menu" dropdown, and a faint "cosmos.so" at the top of the screen. Cosmos is listed on the App Store as "Cosmos: Search & Discover" by Cosmos Entity, subtitle "Your space for inspiration.", compatible with iPad, and the cosmos.so site calls its groupings "collections". "Uncomfy" reads as a collection inside Cosmos rather than an app. The tablet's status bar says "3:16 PM Wed Apr 8". April 8 fell on a Wednesday in 2026, so that footage is at least six months older than the October 14 post.

The "04 EXECUTION" example on the card in frame 036 is the WWF lungs image: an aerial photograph of forest shaped like a pair of human lungs with the tagline "Before it's too late". My Modern Met's April 11, 2009 post credits TBWA\Paris for WWF France; other search results date the campaign to 2008, and the adsspot and AdForum credit pages returned 403, so the year is 2008 or 2009 depending on source. Two more thumbnails in the same frame are datable. The magazine cover of a woman shaving her lathered face is Esquire, March 1965, art directed by George Lois with Virna Lisi as the subject, per the People's Graphic Design Archive entry, which quotes Lois: he "wanted to capture a woman being manly and still beautiful." The Rolling Stone cover of a nude man curled around a clothed woman is the January 22, 1981 issue, photographed by Annie Leibovitz on December 8, 1980, "just twelve hours before Lennon's death," per History.com. All three sit under "Semantic contrast", which is a defensible filing: each pairs two meanings that pull against each other.

Prior art the description does not name

The description says "That’s the part that I haven’t seen anyone do in this specific way." I could not verify a negative, and the specific operation, domain, technique, execution ladder is his. The closest published precedent I found is Phillips and McQuarrie's "Beyond Visual Metaphor: A New Typology of Visual Rhetoric in Advertising" in Marketing Theory, 2004, volume 4, pages 113 to 136, which, per the search result summaries, crosses "visual structure" (juxtaposition, fusion, replacement) with "meaning operation" (connection, comparison for similarity, comparison for opposition) into a nine-cell grid of ad figures. The full text returned 403 at Semantic Scholar, SAGE, and Taylor and Francis, so I am citing the listing and search summary rather than the paper. The overlap with Iris is "operation" as a top-level axis and "comparison for opposition" as a near cousin of "semantic contrast". Neither precedent is a graph of linked thumbnails, which is the part the video shows.

Key Takeaways

  • Verified: Artifact facts: 102.35 seconds, 1080x1920, HEVC Main at 23.976 fps, HE-AAC audio, 18,819,207 bytes, posted 2026-10-14 17:37:39 UTC, 149,700 views and 11,138 saves at capture on 2026-10-19.
  • Verified: The tool is named Iris, is "in beta", and visuals.modernartdirection.com describes the same four levels (operation, domain, technique, execution) that appear in frames 032 and 036. The video never says the name.
  • Verified: The creator is Jason Murray, a former art director at BBDO, Huge, and Amazon's Brand Innovation Lab who left agency work in 2024, per Teachable's May 21, 2026 case study and his course page.
  • Clarification: "Teaching art direction for over a year" matches his own newsletter ("A year ago taught my first bootcamp") within weeks; the Modern Art Direction podcast predates the bootcamp, with an episode dated May 9, 2025.
  • Partial correction: The video says "decades of research" for gestalt psychology and semiotics. Wertheimer's gestalt papers date to 1912 and 1923, Saussure's Course to 1916, and Locke's naming of semiotics to 1690, so the research runs a century or more.
  • Context: The eight ranks he recites were introduced by Magnol (1689), Linnaeus (1735 to 1758), Latreille (1796), Haeckel (1866), and Woese, Kandler, and Wheelis (1990). The biological ladder was itself a work in progress for three centuries.
  • Context: The screen shows four operations ("Differentiate", "Group", "Sequence", "Infer") and six domains; the audio names only "differentiate" and "semantic contrast".
  • Context: Frame 008 shows the Cosmos app with a collection titled "Uncomfy", timestamped "Wed Apr 8", six months before the post.
  • Unverified: "I haven’t seen anyone do in this specific way" (description). The nearest precedent, Phillips and McQuarrie's 2004 nine-cell typology of visual rhetoric, was reachable only as a search listing; its full text returned 403.

Resources

Published October 14, 2026. Writeup generated from a favorited TikTok.