Record
Roles
- Co-founder & CEO, H3MPortals, CSFN, and creative systems for brands and studios.
- Creative Director, LaundryBrand films and AI-hybrid production.
- Creative Director, We Are RoyaleLaunches for Riot Games, Xbox, and Ubisoft.
- Creative Director, IndependentWith PopClutch, Troika, Transistor, Open Road, 72andSunny, and PARTNER — Hulu, Paramount+, BYUtv, The CW, Marriott Bonvoy, Vizio.
- CEO and Chief Creative Officer, TheoryIdentity, web, packaging, and animation — 7-Eleven Hawaiʻi, Four Seasons, Consumer Reports.
- Co-founder and Creative Director, ADAMAgency creative direction.
- Visual Manager, Levi Strauss & Co.Retail visual management.
Research
- Portals: Persistent, Editable 4D Spatial World Models on Edge DevicesCVPR 2026 Workshop on 4D World Models — Second author
- Intent-Based Persistent Spatial World Models on Edge DevicesU.S. provisional patent application (H3M) — Co-inventor · patent pending
- Compiled Semantic Field NetworkH3M research record — Concept originator
Awards
- Awwwards Honorable Mentionryanjbrant.com
- Clio Awards (3)Integrated marketing — Claws, Legion, Snowfall
- Cannes Lions (multiple)Riot Games — Naafiri, a single entry
- Clio GoldGoogle Cloud at the Las Vegas Sphere
- Telly GoldParamount’s Yellowstone at the Sphere
- NYX Gold (2)Virtual Wild — Husky tradeshow
Press
Teaching
- Adjunct Professor, University of HawaiʻiSoftware and emerging-technology courses.
Brands
Much of this work came through studios and agencies, including We Are Royale, PopClutch, Troika, and 72andSunny.
- Riot Gamesvia We Are Royale
- Xboxvia We Are Royale
- Ubisoftvia We Are Royale
- Huluvia PopClutch
- Paramount+via PopClutch
- CBS
- HBO
- FX
- FOX
- Warner Bros.
- Turner
- TBS
- Showtime
- The CWvia Troika
- BYUtvvia Troika
- MTV
- Nickelodeon
- Burger King
- McDonald’s
- Marriott Bonvoyvia 72andSunny
- St. Regis
- Four Seasonsvia Theory
- 7-Elevenvia Theory
- Vizio
- Lionsgate
- New York Mets
- The Jennifer Hudson Show
- Anivive
- Ben & Jerry’s
- Intel
- Logitech
- EA SPORTS
- Dell
- Alienware
- Cadillac
- Dr. Simi
- Dayforce
- Washington’s Lottery
- Bounce
- Foundation Building Materials
- Toyota
Studios
- We Are Royale
- Laundry
- PopClutch
- Troika
- Transistor
- Open Road
- 72andSunny
- PARTNER
Sources
Every number on this site points to one of these.
- Career record supplied by Ryan BrantSelf-reported. A dated CV is available on request.
- Résumé, January 2024Self-reported.
- Portals: Persistent, Editable 4D Spatial World Models on Edge Devices (opens in a new tab)Tunick, Brant, Pennock, Kasowski. Ryan is second author.
- Systems and Methods for Intent-Based Persistent Spatial World Models on Edge DevicesProvisional application under 35 U.S.C. §111(b); patent pending. Inventors: Tunick, Brant, Pennock, Kasowski.
- CSFN Phase 5 report: held-out reasoner benchmarkOne synthetic microworld. A zero-parameter baseline scores 31/37, so the learned margin is 31→36. Not evidence of general reasoning.
- Compiled Semantic Field Network: Current Architecture and EvidenceSelf-published research record.
- CSFN development record, 15–16 October 2026Work in development. The count is the whole automated suite at commit a523320b29 (5,142 passed, 2 expected failures, fixed seed). The real-text rounds are exploratory: no result from them is claimed here, and the earlier test-world result is unchanged.
- We Are Royale Unleashes Naafiri in New ‘League of Legends’ Character Spot (opens in a new tab)
- Faraway Downs — ryanjbrant.com (opens in a new tab)Self-reported.
- Display-room prompts 01–120 and the sentence router — shared/worldgen/roomPrompts.ts and roomRouter.tsRyan’s prompts, in four series of thirty (01–30, juxtaposition 31–60, sport 61–90 and franchise 91–120). Each line quoted on the page is a phrase from these prompts in his wording; three (rooms 03, 10 and 19) are trimmed from longer sentences. The file is unchanged since 2026-09-28. A sentence opens a prebuilt room only when it names one; otherwise it composes a world.
- Component catalogue — shared/worldgen/componentCatalog.json1,576 catalogue entries from all 120 display rooms (1,562 distinct component ids: a few are shared between rooms): 574 lighting, 358 architecture, 231 decor, 154 display, 83 backdrop, 77 sculpture, 68 furniture and 31 textile. Names, roles and sizes in metres are the catalogue’s. It is generated from the room files and grows as rooms are built: it held 593 from 48 rooms on 2026-09-28 and 1,567 from 120 on 2026-09-29.
- The art-direction formula — docs/art-direction/FORMULA.md and shared/worldgen/artDirection.ts45 rules distilled from 59 critic reports on the display rooms. 31 are gates over a set’s plan that answer in about 9 ms, without rendering. The rules are a floor drawn from what the critic rejected, not a guarantee of a passing score, and one reader labelled the reports.
- Looks: light and grade as data — shared/worldgen/looks.ts16 looks. 10 are transcribed from the display rooms’ own light rigs, each checked against the room file it cites; the Oxblood Supper Club’s is “low tinted fill, one flagged neutral key”. 6 are authored for spoken worlds and not yet rendered.
- Room words become pieces in any world — shared/worldgen/roomWords.ts and roomPieces.tsA room’s own words bring up to two of its signature pieces into a spoken world, never its lights, floors or shells. The Last Mile at Sunrise’s are its finish tape, street and gantry. Placement is still rough; placing pieces by the formula’s set grammar is next.
- Display-room review renders and the worlds showcase — scripts/showcase_shots.py and harness/world/showcase.htmlThe harness’s own renders, drawn in the browser: each room’s entrance view from its newest review sheet, and its components view, every component on a studio floor, measured against its declared size. The lettering in the rooms is the harness’s placeholder house name. The ten rooms shown are from the review sheets of 2026-09-29; by 2026-10-02 the harness held renders of all 120 rooms.
- The house name is a setting, and a room is separate items — shared/worldgen/room_products.ts and harness/world/rooms.htmlIn the harness the house name stamped on a room’s pieces is a setting (MAISON VALE by default, and any other name on request), and a room’s items are separate components, each listed in the catalogue with its size; in the app a world’s pieces can each be selected, moved, turned, resized and deleted. Ryan, 2026-11-03: the logo, like every other item, can be adjusted or swapped on its own.
- Spoken worlds: a sentence painted with the app’s own brushes, then edited in words — shared/worldgen/stage.ts, stageGarden.ts and src/features/studio-rebuild/worlds/worldPainter.tsText becomes a recipe of a few hundred bytes (403 after seven edits), the recipe a plan, and the plan a lit scene at any of five levels of detail; each edit’s report is a diff of two plans. Built and reviewed in the browser harness on 2026-09-27 and 28. The same plan now runs inside the app’s Composer (see the entry on Text to World in the app).
- Low-poly worlds from a sentence: the casual pack taken apart into library assets and recipes — commit 86667c53bb, and the sentence fixes after it100 premade miniature scenes, made by other artists and supplied to the project: they are not H3M’s work, and are shown here for illustration, with Ryan’s go-ahead, on a non-commercial site. Taken apart into 6,311 objects, 4,531 de-duplicated assets and 212 categories, each tagged, and kept as recipes; every placement was rebuilt from its source and checked, 6,311 of 6,311 within 0.5% of the object’s size. A low-poly sentence builds, remixes and mixes them; its hour, weather and named things reach the scene, rain and snow fall on it, and a thing added stands clear, dry and in front. The four pictures are the browser harness’s own renders of 2026-10-29, one seed, none edited. In the app the scenes load from a cloud bucket, each piece fetched the first time it is used (commit 396290dcbf); not yet released.
- Text-to-world priorities: the library, the set grammar, the scene queue and the composer — docs/PRIORITIES.mdDirection, not done: one library of room components, rebuilt library models, paint brushes and looks, searched library-first; pieces placed by the formula’s set grammar and checked by its gates, which the creator does not yet call; every one of the 120 scenes critiqued to 7 and on to 8–9, with levels of detail and at most 200K faces on mobile; and the creator measured on a phone, not only in the browser.
- Diorama slices: a spoken world as one small tile — docs/PRIORITIES.md, with the code imported in the Composer branchMore than twenty art-directed places (the radar counts 22; the code imported on 2026-10-13 lists 23 entries): fifteen Ryan named (among them a city park, a zen garden, a pond garden, Stonehenge, a desert and an ink-wash landscape) and seven more (a lake, Mars, a crystal base, a block world, a meadow, the moon and a forest trail). Each is one bounded tile with its own ground, water and light. Built and judged in the browser harness by 2026-10-10 (“never on a phone”, the radar says), then imported into the app’s Composer.
- Text to World in the Portals app — rail tool, prompt tray, 120-room strip and world nodeIn the app’s Composer from 2026-10-14: a Text to World tool, a prompt tray, a strip of the 120 rooms, and a world placed in AR as one item whose pieces can be selected, moved, turned, resized and deleted (commits 9acb52cfa7 and 552540d1f3). First tried on a phone on 2026-10-14. Not yet released: the 4.10.0 (80) build was archived on 2026-11-01 and waits to go to TestFlight, and the phone checks of placing and reopening a world are still open.
- Lab — h3m.studioCompany site copy; the awards are confirmed by Ryan, who won them for various studios. Ryan, 2026-11-03: the Naafiri entry won multiple Cannes Lions.
- Ryan Brant | Creative Director — Honorable Mention (opens in a new tab)
- Meet Ryan Brant (opens in a new tab)
- CLEAR Finance — Correctness & Integrity Core (design spec)A design document, draft for review. It records an audit of the earlier app and sets the rules the engine is built to; it is not an external audit. The 23 defects are a tally of the items it lists, not a number it states.
- CLEAR commit history, June 5–7Test counts as reported in commit messages; a static count of the suite agrees at 280. The suite was not re-run for this page.
- CLEAR agent tools and their testsThe proposal tool’s input schema has no status field, so a proposal cannot approve itself. Separate action tools act when the user asks.
- CLEAR repository, read for this page (branch correctness-integrity-core, June 7)A read-only review of the code, not an audit: the interface elements rebuilt here, where the eligibility gate is wired and where it is bypassed, what the review queue holds out, and which modules write audit rows.
- Latent Compute Runtime — research paper and test and evaluation planDesign documents and an early code scaffold. The primary test they define has not been run.
- Kalshi research log, RL-003: can LLMs price markets better than the market?Phase 2 calibration: a four-agent open-source LLM ensemble and a single-model baseline, scored by Brier score against the market price on 45 of 50 selected markets. The expectation was recorded after the test.
- Overnight lab: backtest verdicts on spot-crypto strategiesBacktests and paper trading; no live results are cited.
- Hyperfluid profit-fix plan, Phase 0: make the numbers trustworthyPaper trading. The audit replaced a dashboard alert’s funding profits with the net paper result the order log supports.