THE SOURCE CoDE

The work and its context

THE SOURCE CoDE is Art Blocks project 187 on the contract at 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270, on Ethereum mainnet, credited to Ofir Liberman. The public index returns 180 invocations against a maximum of 180, a declared dependency of p5@1.0.0, an aspect ratio of 0.6153, a licence string of CC BY-NC 4.0 with a null licence link, and a display_static value of false ¹. The collection page describes it as an Art Blocks 500 release in the Factory category, a fixed edition of 180 unique artworks, released on 5 November 2021, and carries a View live control ².

The aspect ratio is the first thing that marks this project out. Almost every work in this batch is square; this one is a portrait rectangle at roughly 1 to 1.625, and it is a portrait rectangle because the work is a card. Each token is a tarot-like plate: a bordered field, a row of small glyph boxes along the top, a central figure with a round head and a triangular robed body, and a row of eight tiny marks along the bottom edge. The artist's own site says so directly: "It is like a digital tarot card, only instead of you picking a card from a fixed card set, you mint your unique card" ³.

The second thing that marks it out is a number that does not match. The mint description, printed in full on the collection page, states that the work is "a set of unique 1024 karma cards minted in the Genesis phase of THE SOURCE CoDE" ¹². The same page, a few lines above, states a fixed edition of 180 ². The index reports 180 invocations against a maximum of 180 and a complete flag of true ¹. The artist's own site repeats the 1024 figure ³. Whatever was planned, the edition on chain is 180 and it is closed; this package reports the 1024 as a stated intention and the 180 as the indexed fact, and does not try to reconcile them.

The third is that the published trait values do not come from the artist's program. That is not a guess. The program builds its own features object and logs it, and reading that object out of the running sketch shows it disagreeing with the published metadata in three separate ways on a single token: the key names are misspelled, one value is the literal string undefined, and the colour words are different ⁴. The construction section sets this out in detail, because it is the most consequential finding in this package and it bears directly on how much weight a catalogue can put on Art Blocks trait data.

Two structural facts to fix here. The work has no controls: there is no key, mouse or touch handler anywhere in the program, and keyPressed and mousePressed are undefined in the running instance ¹⁴. And it is static in what it shows while being anything but static in what it does. The program never calls noLoop, so its draw function runs at full frame rate indefinitely, rebuilding and re-masking two full-canvas image buffers on every frame; the picture it produces is identical every time. At the live renderer the sampled canvas was unchanged across five seconds while the frame counter advanced by 265 ⁴. There is no audio ⁴.

Conception and development

The artist's website of record resolves, which in this batch is not to be taken for granted, and it is a real project page rather than a studio index or a link hub ³. It is served over plain HTTP rather than HTTPS, and it still displays a drop countdown reading EXPIRED, five years after the drop.

What it says is a complete account of the system, and it maps onto the program with unusual precision. "At the base of the TSC system there are 10 symbols. Each symbol represents a dimension of reality and your life," the site explains; "the card holds two sets of symbols. The main, at the top of the card, which generates the root of the karma and the second set that holds the complete summary of the card's meanings" ³. In the program there is a list of exactly ten dimension names, and there are exactly two symbol groups drawn on the card: three boxed glyphs across the top and a row of eight small marks along the bottom ¹⁴. The bottom row is not decorative. It is the token's eight generated parameters, drawn one glyph each, in order. The card carries its own seed on its face.

"The Core," the site continues, is "the center of focus and main visual attribute of the card" and "holds the message of the card to the owner" ³. In the program a function of that name draws the central figure, and which of ten figures it draws is selected by the same parameter that names the published Action trait ¹. The ten Action values are Crystalize, Hold, Settlement, Heal, Teleport, Fight, Flow, Look, Visit and End, and each is a different arrangement of head, eyes, robe and horizon ⁵.

The framing is explicitly divinatory rather than decorative. The site describes the work as "a self discovery Techno Mysticism system that can help you find personal alignment with every aspect of your life", says every card "holds cryptic meaning for the card holder", and instructs the buyer to "mint with purpose, and the card will guide you" ³. The mint description on Art Blocks uses the same language: an Intent is generated when a card is minted, and it is "the most powerful tool for navigating the endless possibilities in life" ¹. A catalogue should represent that framing as the artist's, which it plainly is, without adopting it.

Two promises on that page were not kept in any form this session could find. The site tells holders to expect "a visit to the ORACLE where you can get a reading of a card meaning", and links a Discord ³. No oracle was located and the Discord was not followed. The indexed artist_interview field is null ¹.

The program itself shows signs of having been finished in a hurry. Its features object misspells two of its own key names as "Dmension"; its list of Action names misspells one entry as "Settelment"; a helper for choosing randomly from a list is defined and never used, and computes its index from ninety-nine per cent of the list length, which would have made the last item unreachable had it been called; and three console.log statements survive in the shipped program ¹⁴. None of these affects the image. All of them are permanent.

How the work was constructed

The script was retrieved as the script field of the projects_metadata row for 0xa7d8d9ef8d8ce8992df33d8b8cf4aebabd5bd270-187, from the Art Blocks public GraphQL index, on 2026-09-18 at 07:47:47Z ¹. It is 10,392 bytes and its SHA-256 digest, computed in the browser over the exact returned bytes using the Web Crypto API, is feec2a97e2f31acd738c977cd2068c4d49298954146fe7beaad473a7074a75c7 ⁴. The row reports script_count of 1 and the declared runtime is p5@1.0.0 ¹. No artist code is reproduced here.

The random generator is, for the fourth time in this batch, the same construction: a three-shift xorshift seeded from parseInt(tokenData.hash.slice(0, 16), 16), a slice that includes the 0x prefix and so parses fourteen hexadecimal digits, with an output step returning the absolute value of the state modulo 1,000 divided by 1,000 ¹. Amoeba, Blockbob Rorschach, Mental pathways and this project all carry it, by four different artists, across releases spanning nineteen days. The case for a shared snippet is now strong enough that a platform-wide survey would be worth someone's time.

Eight draws are made, in one statement, before anything is drawn: a point-of-view selector in the range 1 to 100, an action index, a base symbol index, a base density between 3 and 6, a vision symbol index, a vision density between 3 and 9, and two colour indices in the range 0 to 8. Those eight values become the six published traits and they are also the eight glyphs along the bottom of the card ¹.

Three adjustments follow. If the two colour indices collide, the second is redrawn, but from a narrower range of 0 to 5 rather than 0 to 8, so a collision quietly restricts the vision colour to the first six of the nine available ¹. If the token is invocation 0, every one of the eight values is overwritten with a hard-coded literal: point of view Normal, action End, base symbol 4, base density 6, vision symbol 0, vision density 3, base colour 3 and vision colour 1 ¹. That is an artist's proof whose appearance is fixed in the source and does not depend on its hash at all, and reading the running sketch for token 187000000 confirmed every one of those eight values in place despite a hash that had produced different draws ⁴.

Then the point of view reassigns colour roles. The nine-colour palette is fixed for the whole edition, and every card is drawn as two stacked layers, a base and a vision, each of which has a fill colour and a stroke colour. On a Normal, Chaotic or Fixed card the drawn colour indices fill the layers and the strokes are a near-black. On a Dream card the fills become that near-black and the drawn colours move to the strokes. On a Light card the fills become white and, again, the drawn colours move to the strokes. Chaotic additionally switches the base layer from one repeated glyph to a fresh random glyph in every cell, and Fixed forces the vision colour to equal the base colour ¹. The published Dimension traits name the colour the hash drew, which is correct in every case; what the point of view changes is whether that colour appears as a fill or as an outline.

The point-of-view selector has a boundary defect worth recording, because it is measurable. The selector is a chain of comparisons using strict inequalities on both sides of each band: below 60, above 60 and below 70, above 70 and below 80, above 80 and below 95, else the last category. The four exact values 60, 70, 80 and 95 satisfy none of the intermediate tests and fall through to the final branch. Across the edition seven of the 180 tokens drew one of those four values ⁵, so seven of the nineteen tokens labelled Light are there because of gaps between the bands rather than because they landed in the band intended for them. More than a third of that trait's population is an artefact of four missing equals signs.

An independent reimplementation of this draw order, the collision redraw and the invocation-0 override was run against all 180 token hashes. All six published traits reproduce on 180 of 180 tokens: PointOfView, Action, BaseDensity, VisionDensity, BaseDimension and VisionDimension ⁵.

That reproduction also settles where the traits come from, because it only works one way. Using the colour indices as they stand after the point-of-view reassignment reproduces BaseDimension on 145 of 180 and VisionDimension on 122 of 180; using them as drawn, before the reassignment, reproduces both on 180 of 180 ⁵. The program's own features object is built after the reassignment. So the published metadata cannot have been taken from it.

Reading the running sketch makes that conclusive, and it does so three different ways on three different tokens. On token 187000040, a Light card, the program's features object reads BaseDmension: "White undefined" and VisionDmension: "White unknown", while the published metadata reads BaseDimension: "Blue soil" and VisionDimension: "Yellow unknown" ⁴⁶. Three faults are visible at once. The keys are misspelled. The base dimension word is the literal string undefined, because the expression that builds it indexes a variable holding the number 0.6153 instead of the list of ten dimension names, a one-character mistake that the vision line does not make. And the colour words are the post-reassignment White rather than the drawn Blue and Yellow, which the program has in fact preserved in two other variables whose values read 6 and 3, being exactly Blue and Yellow ⁴.

On token 187000027, a Dream card, the same object reads "Black undefined" and "Black skin" where the published metadata reads "Grey water" and "White skin"; the preserved indices read 7 and 0, being Grey and White ⁴⁶. On token 187000015 the object reads Action: "Settelment" where the published metadata reads "Settlement" ⁴⁶. A misspelling in the program, corrected in the metadata, cannot be a coincidence of derivation.

The conclusion is that Art Blocks runs a separate feature-extraction program for this project which reimplements the parameter derivation correctly, spells the keys and the action name correctly, and reads the colour indices before the point-of-view reassignment. That program is not part of the project record and was not retrieved. For this project the published traits are better than the artist's own, and a cataloguer relying on either the on-chain features object or the console output would be propagating three errors.

The drawing itself is layered. Two off-screen buffers are filled with a tiled glyph, one at the base density and one at the vision density, each glyph chosen from ten forms. Two more buffers are filled with a silhouette, chosen by the action index from ten shapes, which is then converted into an alpha mask by moving the red channel into the alpha channel and zeroing the colour channels. Each glyph layer is masked by its silhouette and composited, so the robe and the ground read as fields of pattern cut to shape. The central figure, the border, the three top glyph boxes and the eight-glyph parameter row are drawn on top ¹.

That masking is also where the work becomes expensive. The mask is rebuilt from scratch inside the draw loop, twice per frame, each time copying and reading back a full canvas of pixels. Because there is no noLoop, this happens about sixty times a second forever, to produce an image that never changes. The browser emits a performance warning about repeated pixel readback on every frame; the renderer console filled with hundreds of them during a short observation ⁴. A gallery running several of these tokens on one machine should know that.

One further reading from the source should be recorded with its uncertainty. The routine that draws the silhouette shapes calls the shape-begin and shape-end methods on the off-screen buffer it was handed, but calls the vertex and curve methods without a receiver, which addresses the main canvas rather than the buffer. Whether that produces the intended mask in every case, or works by accident because of how p5 sequences those calls, was not determined here; it is reported as read, not as a demonstrated defect.

The mint

The index records activation at 2021-11-01T18:30:38Z, a public opening at 2021-11-04T19:00:00Z, and completion at 2024-03-10T20:31:11Z ¹. The collection page states 5 November 2021 ², which is the opening instant rendered in a timezone ahead of UTC.

For the fifth time in this batch the first mint precedes the recorded activation. Token 187000000 was minted at 2021-10-18T06:02:50Z ⁵, fourteen days before activation and seventeen before the public opening, by the address the index records as the artist's ¹⁵. It is the only token minted before the opening, and it is the hard-coded proof described above. Five projects out of ten now show the same pattern; a catalogue should treat activated_at as a listing or configuration event, not as a first-mint date.

The index records a price of 200000000000000000 wei, or 0.2 ETH ¹, and the description states no auction schedule, so this package does not assert a mechanism.

Taking the opening as the start endpoint and the final mint as the end endpoint, the public phase ran from 2021-11-04T19:00:00Z to 2024-03-10T20:31:11Z, a span of 857.06 days, and covered 179 mints ⁵. This is the slowest sale in the batch by a wide margin, and the shape is unusual too: it never had a rush. The first public mint came at 2021-11-04T19:34:02Z, 34.03 minutes after the opening. Anchored on the opening, 3 tokens were minted within the first hour, 7 within six hours, 17 within a day, 74 within a week, 102 within thirty days and 172 within a year, leaving 7 to arrive over the following fourteen months ⁵.

Those numbers are worth holding against the rest of the batch. Blockbob Rorschach sold 1,024 tokens in under ten hours. This project sold 17 in its first day and took two and a third years to place 180. The work is not less considered; if anything the framework behind it is more elaborate than anything else here. What it did not have was an audience that recognised it on sight.

The gap between the stated 1024 and the actual 180 belongs in this section as well as the first. A project whose description promises a Genesis phase of 1024 cards, and which closes at 180 after two and a third years, has a story in it that the record does not tell. Nothing consulted here says whether the maximum was reduced, whether the Genesis phase was always meant to be a subset, or whether the remainder was simply never minted. The index reports a maximum of 180 and a completion flag of true, which is all this package asserts ¹.

Across the 180 tokens the index reports 121 distinct owner addresses ⁵. The artist address holds two tokens.

The Art Blocks 500 wallet set was refreshed once this session from the platform's own profile index, by querying the ab5d user profile for its linked accounts. The query returned three linked addresses and all three were checked against every one of the 180 token rows ⁷⁵. Exactly one token in this edition sits in that set: token 187000040, invocation 40, held by 0xdd19308a9731cdecb67e0d80ca6823394208c045 and minted at 2021-11-06T20:02:45Z ⁶⁷, two days after the opening.

Copyright and licensing

The index records a licence string of CC BY-NC 4.0 for this project and a link_to_license of null ¹. The identifier is fully specified and resolves unambiguously to one deed, which permits sharing and adaptation in any medium subject to attribution, prohibits use of the material for commercial purposes, and prohibits applying additional legal or technological restrictions ⁸.

There is a conflict here that does not appear on the other projects in this batch. The artist's own project page, the website of record named in the index, carries the notice "all rights reserved Ofir Liberman" at its foot ³. All rights reserved is the opposite of a Creative Commons grant. The two statements cannot both be operative in the form each is written.

This package does not attempt to resolve which governs. It records what each source says, with its access time, and notes the shape of the problem: the licence identifier is attached to the project in the platform's index with no link and no accompanying terms, while the artist's own publication of the work asserts reserved rights. A catalogue that reproduces these images on the strength of the index string alone is relying on a claim the artist's own site appears to contradict. This is a question for the coordinator and, if it matters, for the artist.

Three rights domains should be kept apart, as elsewhere in this batch.

The first is the artwork, covered above, with the additional note that the record states no terms of sale and no terms for holders. The work is framed by its maker as a personal charm to be kept, gifted or sold ³, which is a statement about how to live with a token rather than a statement about copyright.

The second is the artist's code. The program is retrievable from the index, which establishes visibility and nothing else. No licence statement or copyright notice appears anywhere in the program text, and the all-rights-reserved notice on the artist's site, whatever it governs, does not read as a grant. Accordingly no artist code is reproduced in this package. What is returned instead is a description of the mechanism, the byte count, the digest, and reproductions of the published traits that can be checked against the index independently.

The third is the dependency. The work declares p5@1.0.0 and does not render without it. The library carries its own licence, which is not the project's licence and is not stated in the project record, and this session did not establish it.

A fourth domain is specific to this project and should be recorded even though it is unusual. The published trait values are demonstrably not the artist's; they are produced by a separate feature-extraction program that corrects three faults in the artist's own version. Those corrected strings are what every marketplace, wallet and catalogue displays as this work's attributes. Whose authorship they represent, and under what terms they may be reproduced, is not addressed anywhere in the project record.

Close reading of examples

Four tokens were examined at the live renderer at generator.artblocks.io, each identified first by a TokensByIds query against the index so that hash, owner, mint time and published traits came from a response rather than from memory ⁶, and each opened at its canonical Art Blocks token page, all four of which returned HTTP 200 on a real fetch from the artblocks.io origin ⁴. All four were observed with the browser pane displayed, and in each case the sketch's own global variables and its internal features object were read alongside the pixel sampling.

Token 187000040 is the Art Blocks 500 holding and was the first sustained observation. Its hash begins 0x612eeee4 and it was minted at 2021-11-06T20:02:45Z to 0xdd19308a9731cdecb67e0d80ca6823394208c045 ⁶. Its published traits are Point of View Light, Action Fight, Base dimension Blue soil, Base density 5, Vision dimension Yellow unknown, Vision density 8 ⁶.

On screen it is a cream card with a faint blue grid across the whole field, a figure with a circular head, two solid almond eyes and a small rune between them, and a triangular robe hatched in yellow, standing on a horizon line. Three boxed glyphs run along the top and eight small marks along the bottom edge.

The live reading is the centrepiece of this package. The canvas measured 1120 by 1820 device pixels, a ratio of 0.6154 against the indexed 0.6153. The point-of-view selector had drawn 97, which lands legitimately in the final band. The two colour indices as drawn were 6 and 3, which are Blue and Yellow, and the program had moved them into its stroke variables and set both fill indices to 0, which is White ⁴. That is exactly why the grid lines are blue and the robe hatching is yellow while everything they fill is cream: on a Light card the drawn colours survive as outlines. The published Dimension traits name those drawn colours and are therefore correct.

The program's own features object, read from the same running instance, read {Action: "Fight", BaseDensity: 5, BaseDmension: "White undefined", PointOfView: "Light", VisionDensity: 8, VisionDmension: "White unknown"} ⁴. Set beside the published metadata for the same token, it is wrong in three independent ways at once, and the published version is right in all three. The eight-mark row along the bottom of the card carries the same problem in visual form: it encodes the colour indices after the point-of-view reassignment, so the card's own summary row records White and White while its metadata records Blue and Yellow.

The frame counter advanced from 1030 to 1240 across the observation while the sampled canvas stayed identical, confirming a full-rate loop drawing an unchanging picture, and keyPressed and mousePressed were undefined ⁴.

Token 187000000 is the artist's proof and was the second sustained observation. Its hash begins 0xcd27763a and it was minted at 2021-10-18T06:02:50Z, seventeen days before the opening, and is still held by the artist address ⁶¹. Its published traits are Point of View Normal, Action End, Base dimension Yellow skin, Base density 6, Vision dimension Braun basalt, Vision density 3 ⁶.

On screen it is the most striking card of the four: a field of overlapping yellow arcs, a brown hooded figure filling the lower two thirds, and a cream skull-like head with crossed eyes and a straight mouth. The action value is End and the card looks like it.

Every one of its eight parameters read in the running sketch exactly matches the literals hard-coded in the source for invocation 0, even though the selector value drawn from its hash was 18 and would have produced a different result had it been used ⁴. This is a token whose appearance is not a function of its hash. Its features object again read "Yellow undefined" for the base dimension while reading "Braun basalt" correctly for the vision dimension, which localises that fault precisely to one line ⁴.

Token 187000027 is a Dream card, one of 17 in the edition. Its hash begins 0xc983b740 and it was minted at 2021-11-06T03:16:57Z to 0x665654f2d2a151be2d0f8e3697e6ce780f732af2 ⁶. Its published traits are Point of View Dream, Action End, Base dimension Grey water, Base density 3, Vision dimension White skin, Vision density 9 ⁶.

On screen it inverts to a night plate: a very dark blue ground, pale grey wave glyphs across the upper field, a robe of fine white arcs, and a large circular head with crossed eyes. The live reading shows both fill indices set to the near-black and the drawn colours moved to the strokes, with the preserved indices reading 7 and 0, which are Grey and White ⁴. Its features object read "Black undefined" and "Black skin" where the published metadata reads "Grey water" and "White skin", which is the same disagreement as on the Light card arriving by a different route ⁴.

Token 187000015 is a Chaotic card, one of only 9 in the edition and the rarest point of view. Its hash begins 0x64aa0cb7 and it was minted at 2021-11-05T16:32:12Z to 0x074fdc302f8d3c0e8b11c80f2a07bf2a3b8ca855 ⁶. Its published traits are Point of View Chaotic, Action Settlement, Base dimension Braun soil, Base density 5, Vision dimension Grey blood, Vision density 9 ⁶.

On screen the chaos is immediately legible. The upper brown band is filled not with one repeated glyph but with a different glyph in almost every cell, while the lavender lower field keeps a single regular motif; between them sits a plain cream dome with no face. That is exactly what the point of view does to the base layer and only the base layer. The live reading confirmed the chaotic flag set and the two layer densities at 5 and 9 ⁴. Its features object read Action: "Settelment", the misspelling from the program's own list, against a published value of "Settlement" ⁴⁶. The canvas was unchanged across five seconds while the frame counter advanced by 265, so the per-cell randomisation happens once at setup and is then redrawn identically forever ⁴.

Significance and limitations

This project is the clearest demonstration in the batch that Art Blocks trait metadata is a separate artefact from the artist's program, with its own authorship and its own reliability. The other nine projects here were examined for whether the published traits could be derived from the program. This one answers a sharper question: what happens when the program's own derivation is wrong. The answer is that the platform's version silently wins, and it is better. Three faults in the artist's features object, two misspellings and a one-character indexing mistake that yields the literal string undefined, are all absent from the published metadata, and the colour values are read at a different point in the program's execution than the artist's own code reads them. Anyone building a catalogue from on-chain artefacts alone, on the reasonable-sounding principle that the artist's code is the ground truth, would produce a worse record than the platform's.

The second contribution is a measurable defect with a rarity consequence. The point-of-view selector compares with strict inequalities on both sides of each band, so the four exact values at the band edges fall through to the last category. Seven of 180 tokens drew one of those four values, which means roughly a third of the tokens labelled Light are there because of four missing equals signs. Rarity in this edition is, to that extent, a typographical fact.

The third is the gap between what the work announced and what exists. The description promises a Genesis phase of 1024 cards and the edition closed at 180 after two and a third years, with 17 sold in the first day. The artist's site still runs an expired countdown and still promises an oracle. Whatever happened, this is a project whose published description describes a thing that did not come to be, and a catalogue that reprints the description without the edition size will mislead its readers about the scale of the work.

Against all that, the work itself is coherent and carefully made. The ten dimensions, the two symbol sets, the core figure and the eight-mark summary row are all exactly what the artist's site says they are, and the card carries its own parameters on its face in a way almost nothing else in this batch does. The intent to make a functioning divination system, rather than a series of pictures, is legible in the structure and not only in the copy.

The limitations of this package should be stated as plainly as the findings. The four visual observations cover four tokens out of 180, chosen to cover the extremes of the point-of-view trait and to include the hard-coded proof, so nothing here supports a claim about the typical card. The reading of the mask routine's mixed use of buffer and main-canvas drawing calls is reported as read from the source and was not demonstrated to produce a visible defect. No on-chain transaction values were read, so nothing here measures what buyers paid or establishes the sale mechanism. The oracle and Discord promised on the artist's site were not pursued. The collection page's detail accordions were not expanded. No contact was made with the artist, and the rights conflict between the indexed Creative Commons identifier and the artist's own all-rights-reserved notice is reported rather than resolved.

Several questions remain open, and this package records them as open rather than filling them.

What happened to the other 844 cards? The description and the artist's site both state a Genesis phase of 1024. The edition is 180 and complete. Nothing consulted says whether the maximum was reduced, whether Genesis was always a subset of a larger plan, or whether the rest was simply never minted.

Which rights statement governs? The index says CC BY-NC 4.0 with no link; the artist's own project page says all rights reserved. Both are published, neither is dated in a way that establishes precedence, and the difference decides whether these images can be reproduced at all.

Where is the feature-extraction program, and who wrote it? It corrects three faults in the artist's code and reads the colour indices at a different point in execution. Its output is what every marketplace shows. It is not in the project record.

Was the band-edge fall-through in the point-of-view selector noticed? Seven tokens sit in the Light category because of it. A single equals sign in four places would have moved them, and a third of that trait's rarity with them.

Does the mixed drawing context in the mask routine matter? The routine begins and ends shapes on an off-screen buffer while issuing the vertices to the main canvas. Establishing whether the masks are correct in all ten silhouette cases would need a per-case comparison that this session did not run.

Finally, is the shared random generator a template? This is the fourth project in this batch carrying the identical construction with the identical two defects, by a fourth artist. A survey of Art Blocks scripts from 2021 for the same sixteen-character seed slice and the same modulo-1000 output step would settle how far it spread and how many editions inherited its thousand-value quantisation.