# Pathfinders research

Next essay after Archetype, now published and verified as essay 25.

## Sources inspected

- https://www.artblocks.io/collection/pathfinders-by-luxpris : Playground, 1,000 works, public release 2 March 2021; start-screen click erases and starts drawing, next click pauses, viewer may save image. Keep page-derived prose under 200 words.
- https://www.artblocks.io/articles/in-conversation-with-josh-katzenmeyer-luxpris : full interview read in search result, Jeff Davis, 2 March 2021. Artist describes Pathfinders as personal reflection on generative practice, deterministic palette/brush identity with new paths, and viewer curating a preferred image. Earlier musical practice, IBNIZ and web-development learning provide context. Keep all derived prose under 200 words; source does not prove complete output determinism.
- Full 13,744-character indexed script read, not executed for analysis. Cached script.js and four token records under ignored series-expansion/pathfinders-by-luxpris.

## Metadata

Project25, Ethereum 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270, p5.js1.0.0, 1000/1000, CC BY-NC4.0 with null terms URL. Current price setting0.1ETH. Firstmint2021-02-28T17:45:13; activation2021-03-02T01:13:32; second/start18:01:07; completion18:18:25. No public allocation or original-price proof.

## Code findings requiring careful wording

Setup creates min-window square, installs canvas mouseClicked controls, scales against2400, draws circuit grid then noLoop. Initial state off. First click clears background and starts loop, changes to on. Second click noLoop/paused. Third click clears, starts loop, sets shadow; does NOT reset t, noise or clock. Starting screen therefore is not the moving work. User autoplay requirement needs existing approved wrapper pattern, not static preview or ordinary iframe that stays on start screen.

Hash pairs choose palette, brushes, avatar and start-grid parameters. Custom xorshift drives starting grid. p5 noise drives movement; no noiseSeed call found. millis drives changing brush dimensions, and t advances each draw by a hash-derived .001-.005 increment. Thus hash fixes important vocabulary but does not establish repeatable pixel trajectory across sessions, window sizes or observation times. Avatar randomSeed is set each frame separately.

Palette choices specify background, stroke and avatar fill/stroke. drawImage overlays low-alpha circles, lines, triangles and squares through10 primary and4 secondary routines (last secondary none). Some endpoint selections require highest byte because floor(mapRange(...0,9/3)); no audited rarity claim. Branch setGridSize for nominal4x4 uses l<69 && l<=84 after prior l<=69, appears unreachable. Can state inspected conditional issue in notes, do not convert to observed feature distribution. One line call has six args; do not invent intended geometry beyond actual runtime evidence.

Draw accumulates rather than clears every frame; secondary-colour heavy border and lower-right avatar medallion redrawn. Shadow blur enabled early and on restart then reverted around avatar. Viewer pause time materially selects result. Distinguish pause from restart: paused click clears prior drawing, continues state progression. Need test actual controls and capture successive frames for four outputs.

## Remaining

Four official live outputs not yet visually inspected. Need browser unlock, inspect initial and moving states, pause/restart, fresh AB5D-held token and autoplay approach, tagged p5 library and CC deed source records, nine-source manuscript/package, presentation checks and publication.

## Additional verified evidence

Fresh owner observation for held #401: token25000401, owner0xdd19308a9731cdecb67e0d80ca6823394208c045, mint2021-03-02T18:04:52. Bubblegum Palette, primary brush0, secondary1, small start grid. Cached held.json. No claim of visual inspection yet.

Tagged p5.js source https://raw.githubusercontent.com/processing/p5.js/1.0.0/src/math/noise.js read in full. Lazy table initialization uses Math.random; noiseSeed replaces table via seeded generator. This substantiates the script-level finding that absent noiseSeed, paths are not fully hash-determined. Keep discussion derived from this source under200words. Tagged p5 licence and CC deed also opened/read.

Prepared ignored autoplay-candidate.html using the existing Gen2 wrapper pattern. Fetches canonical renderer at viewing time; calls the artist's controls() once only when state is off and initial frame exists. No artist drawing code copied. This candidate is UNTESTED and not in the publication tree. Need actual autoplay/pause/restart verification before promotion.

Prepared construction, mint and rights prose in construction-draft.md, a research fragment explicitly excluded from publication. Full essay still requires four actual live inspections and source-led close readings.

## Live visual inspection resumed

Canonical generator #0 inspected in Chrome at 1689x796. Initial light-grey field, thin black diagonals/verticals and hollow circles, thick black border, lower-right black avatar medallion. First click clears that grid: large faint concentric circles and a small wedge appear. Continued drawing accumulates broad dark sweeping arcs, tight filament meshes and angular fans; far denser than initial state. Second click stops drawing; later screenshot unchanged. Third click clears accumulation and restarts sparse rings in a different place. Not a reproducibility test across reloads.

Canonical #1 inspected initial and two active states. Slate-blue field, pale grey-green border and avatar; initial pale circuit-like lines and circles. First click produces faint linked rings and thin pale threads. Later state has a luminous folded triangular web at upper centre/right with large loops and chains of small circles toward lower left. Unlike #0's broad dark arcs, active marks read as pale suspended threads against dark ground. Session-specific states, not immutable token composition.

#10, #20, held #401 and autoplay candidate still require live inspection.

## Remaining comparison works inspected live

#10 initial violet ground, muted grey border and avatar, thin pale circuit-like lines/circles. Firstclick clears grid and introduces extremely faint broad ring and several small drifting marks. Continued drawing: broad nearly concentric arcs spanning field, pale looping ribbon cluster toward centre/lower area, small beadlike curls to left. Lower contrast than #20; violet remains visually dominant despite accumulation.

#20 initial black ground, white border/avatar, much denser and finer grid than #0/#1/#10. Firstclick clears to one bright ring, tapering spoke and small pale marks. Continued drawing produces several overlapping luminous large rings/oval bands, internal wedge/fan forms and strings of small linked loops at left/right. Strong black/white separation makes transparency and overlapping edge brightness legible. Full token25000020. Both initial and twoactive states actually inspected in Chrome. Timing was not controlled as an equal-duration experiment.

AB5D #401 autoplay candidate loaded successfully on local server without anyclick, then displayed moving dark navy/purple circular accumulations and chains of small loops on pink ground, navy border and small cream robotic avatar. This is actual visual inspection of sandboxed canonical renderer fetched atviewtime, not a cachedpreview. First click inside iframe was performed, but the test tab disappeared before a delayed comparison. Do not claim embedded pause stability or restart verification yet; reopened test tab for completion.

Fresh autoplay test completed in replacement Chrome tab1423996842: started automatically on load; click paused at overlapping navy circles and chain forms, delayed screenshot unchanged; next click cleared to sparse points/circles on pink ground. Thus autoplay, pause and restart are all observed in wrapper. Full article embedding and mobile checks remain.