# Research notes: Construction Token Status: review draft, not accepted or published. Reviewer: Phil. Research cutoff: 2026-09-17. ## Assignment and method This is internally commissioned work following specification 1.0.0, not a paid bounty claim. Roster: the first ten entries of data/atlas.json at commit 180c474c04ff98ed52e3e66db693da7d72424a1f; original snapshot generated 2026-08-17T21:29:43.854Z. Identity was also matched to data/ab500-contracts.json. Read the existing artist bibliography for discovery, then inspect the underlying project-specific sources. The AB5D dossier is linked as context, not counted as independent corroboration. Read live Art Blocks project metadata, the indexed artist script, four indexed token records and their actual preview images. Web research covered named artist interviews, project pages, applicable licence texts and the supplemental source identified in the bibliography. No search snippet is used as evidence for a material claim. ## Code inspection Source: https://artblocks-mainnet.hasura.app/v1/graphql Method: POST the projectQuery retained in evidence.json, filtered by canonical slug. Read projects_metadata.script. The stored code was inspected as text; it was not executed. This is retrieval through Art Blocks' index, not an independently verified contract call. Contract: 0x059edd72cd353df5106d2b9cc5ab83a52287ac3a; project ID: 2; chain: Ethereum, chainId 1. Chain block: unknown. No claim of original-version equivalence is made. SHA-256 of UTF-8 indexed script: e66bae6ba7a7b24a83616491efc2dc87bbb892ff5a3d8962cc913c61e9a86c0b Inspected passages: - Lines 1-20: hash expansion - setup: 36-unit grid, 20-40 drawing groups - draw: colour thresholds, edge anchoring, bounds corrections and stripe loops - Last line: noLoop The code bytes remain in the ignored local research cache and are not redistributed in this package because a separate artist-code reuse basis was not established. Digest and source query allow comparison on retrieval. The renderer dependency reported by the index is processing-js@1.4.6. ## Mint evidence and calculations The raw named time fields, current price setting, currency, invocations and maximum are preserved in evidence.json. All are observations of the live index on 2026-09-17, not original-announcement evidence. The metadata's completion and activation fields are not relabelled as public sale events. No sellout-duration calculation is made. No opening price is inferred from a current setting. Values in the price field were divided by 10^18 for the reported ETH or DAI configuration, not converted to fiat; the original integer and currency are retained. No claim of historical currency-decimal verification or original price is made. Selected token records include minted_at and mint_transaction_hash. These support precise identity and indexed time observations but were not independently audited as complete transaction receipts. Public, artist, reserved and administrative allocations have not been exhaustively classified. Current maximum is not automatically the intended original edition. ## Visual evidence Preselected editions: 0, 1, 10 and 20. Four distinct tokens, not four screenshots of one work. Canonical IDs, Ethereum contract, URLs, preview URLs and image digests appear in record.json. The initial media-proxy preview endpoint returned HTTP 403; the index's image.url relation supplied accessible Art Blocks S3 images. Those exact images were visually inspected in labelled contact sheets. They were not recreated from code. Descriptions are original visual observations. No claim of rarity or statistical representativeness is made. Still-image state is recorded; live animation and pointer behaviour are described from code and attributed published evidence, not from unperformed interaction tests. No image assets are redistributed; mediaIds is empty. The public canonical token links accompany the close readings. ## Source independence and access There are 10 cited source records, including four original token outputs used for four distinct close readings. The project page and its index observation are grouped as S01; they are not counted as independent corroboration. S02 is a different primary artifact, the artist script, though supplied by the same index. Token pages, indexed features and corresponding preview images are grouped per output. Four output records do not provide four independent confirmations of mint or rights claims. The numerical source count is not a claim of eight independent historical witnesses; editorial review should judge material usefulness. Each citation has a locator. Canonical token pages returned HTTP 200. Other sources were inspected with web page reading or direct text retrieval as recorded in sources.json. A successful HTTP response alone is not treated as proof of a claim. The API URL requires POST with the recorded GraphQL query; opening it as an ordinary page is not equivalent. ## Rights and exceptions Artwork, code and dependencies have separate records. Original essay prose is offered under CC0-1.0; third-party material is excluded. No explicit licence was found in the inspected artist script. This is a scoped finding, not a claim that no licence could exist elsewhere. Where applicable, the named Creative Commons deed and the upstream library licence were inspected. Missing operative terms are unknown, not automatically all rights reserved. Pending decisions are in record.json.editorialExceptions. Historical price/maximum/allocation gaps require editorial review. Acceptance is not granted by structural validation. Research can continue if Phil requires a complete transaction-level mint reconstruction before publication. ## Provenance and effort Authoring agent: OpenAI Codex, GPT-6; exact model build/version unavailable. Commissioning operator: Phil (@maxand98). Human editorial work on these drafts: pending. Tools: web search and page reading, public Art Blocks GraphQL, Python standard library, Pillow contact sheets and image inspection. No other authoring agents were used. The overall task began before the first explicit clock observation, so full elapsed time and per-series elapsed allocation are unknown. Measured batch work window: 2026-09-17T04:01:11+00:00 to 2026-09-17T04:12:43.795501+00:00; 693 seconds. This is a measured partial window, not a total-effort estimate. Inference-token usage and monetary cost were not exposed and remain null. No bounty payment or cost estimate is asserted.