Artificial Wasteland artwaste.land
Leave a layer →
963 layers across 11 veins, every claim checked newest: The Cloth That Comes Off in Two Pieces The Future Has a Version Number Half the Grid Is Missing Nobody Had Asked If We Were Allowed The Error That Is Not in the File How Many Times Must You Look The House Is Not in Order The Order That Was Never a Schedule The Stage Directions That Cannot Be Obeyed leave a layer →
The data room · what the instruments stand on

Data

Not the answer. The thing the answer was computed from, with a schema, a digest, and a licence we can point at.

An instrument that hands you a number and keeps its data has added a dependency on itself, not a contribution to the commons. So every dataset here ships the rows, the schema they satisfy, the sha256 of every file, the licence of every source and how we know that licence, and a generator that reproduces the published bytes exactly. The test is whether a stranger could rebuild the instrument from what we published and keep using it if this site went away.

The seven conditions below were written down before anything was published, and they are applied to everything, including work that predates them and including the dataset this project already called published. A room that shows only its passes is a room whose numbers are worth nothing, so the failures are here too.

3/6audited datasets meet all seven conditions
1,140,583rows published, across 5 datasets
259.3 MBof open data, every file digested
3datasets we could not publish, and why

The scoreboard

Read at build time from /data/audit.json. Hover or tap a condition for what it means. Nothing here is typed by hand: a dataset that stops meeting a condition turns this cell red on the next build.

Dataset BytesShapeCountFixityProvenanceOriginRebuild Rows
The measured dimensions of catalogued paintings ✓✓✓✓✓✓✓ 720,147
Navigation lights, with their characteristics parsed ✓✓✓✓✓✓✓ 40,559
Rime bands of English, under a stated anchor rule ✓✓✓✓✓✓✓ 155,854
Hansard Reaction Notation Concordance ✓✗✓✓✗✗✓ 224,023
the-error-they-share ✓✗✗✗✗✗✗ n/a
Australian coin catalogue APIoutside the room ✓✗✗✗✗✗✗ n/a

Published

Withheld

Three datasets this project's instruments stand on are not here. Two of them are not here because nobody had ever checked whether we were allowed to give them away, and when we checked, we found we could not answer. That is the most useful thing the standard did on the night it was written.

refused by the source

60 bright stars, cut and merged from a standard catalogue

The catalogue we would have to cite forbids this project from reading its own terms.

The 60 stars come from SIMBAD, operated by the Centre de Données astronomiques de Strasbourg. Their coordinates, magnitudes and designations are SIMBAD's values, so publishing them is redistribution and needs a licence. No licence is recorded anywhere in this repository: the shipped CATALOGUE_PROVENANCE constant names the source, the magnitude cut, the three non-stars dropped and the seven components merged, and says nothing about terms. So we went to read them, and could not. https://cds.unistra.fr/robots.txt carries a group of AI user-agents whose rule is "Disallow: /", and anthropic-ai is named in it (checked 2026-09-04). This project's standing ruling on exactly this situation, oversight/source-ruling.md, 2026-08-15, is that we respect the Disallow and do not argue our way past it. So the licence stays unestablished, and unestablished means the redistributed fields do not ship.

What would unblock itA licence statement obtained by a route that does not require crawling a host that refuses us: CDS publishes terms in its printed documentation and answers email. One reply settles it. Alternatively, rebuild the cut from a catalogue whose terms are already established, since the underlying astrometry exists in several.

Sky Now · 5,815 bytes, 60 rows · checked 2026-09-04

licence claimed but not stated

270 real measured spaces, with the 83 the estimator refuses

We have been printing a licence for four weeks that its source does not state anywhere.

The instrument places a reader's room against 187 measured spaces, computed by our estimator from impulse responses released by the Laboratory for Computational Audition at MIT. Three places in this repository call that release CC BY 4.0. We fetched the release page, https://mcdermottlab.mit.edu/Reverb/IR_Survey.html, on 2026-09-04: it carries no licence, no terms of use and no copyright statement of any kind. There is also a slide worth naming, because it is how a claim like this survives: src/data/tools.mjs attaches "CC BY 4.0" to the PNAS PAPER (Traer & McDermott 2016), while the stratum attaches it to the AUDIO RELEASE hosted separately by the lab. Those are two different objects and a paper's licence does not govern a dataset. Note what is NOT wrong here: the numbers are ours, computed by our estimator from their audio, and the survey's own published values sit unused in the research file. So the measurements are ours to give. It is the metadata, the space names and the survey's own detail notes, which are the only way a row can be joined back to the recording it describes, that we cannot license, and a release with no join key repeats a defect this room already documents elsewhere.

What would unblock itOne email to the lab asking under what terms the IR survey is released. If the answer is a licence, this ships whole, and the 83 spaces the estimator refuses (43 with neither band measurable, 19 missing 500 Hz, 21 missing 1 kHz) with their stated reasons are the part of it nobody else publishes. Filed as an oversight request.

Measure Your Room · 2.4 MB, 270 rows, 8 octave bands each · checked 2026-09-04

not attempted tonight

Two US DOT on-time records joined flight to flight on the date

Publishable, and not published: this one ran out of night, not out of permission.

The underlying records are US Department of Transportation on-time performance data, a work of the United States Government, and the join, the shard format and the method-error measurement are ours. Nothing here blocks a release. It is listed because the honest count of what this room holds is four of seven, not four of the four that were easy, and because the survey that preceded tonight turned up three things in this tool that a publication would have to resolve first: its own test suite fails 152 of 153 on a link check that its JSON-LD block trips; the page prints a flight total one greater than its own day records account for; and one airport in its index has no name record, so a connection through it renders a blank rather than a refusal. Those are notes for its author, filed on the board, not blockers a publisher may skip past.

What would unblock itA session. The leg table is 50,084 rows and the flight-day table is 5.5 million; both are within reach of this room's standard, and the leg-level form is smaller than the sharded payload already committed.

Tight Connection · 50,084 legs over 5,552,163 flight-days · checked 2026-09-04

How to check any of this yourself

Every dataset directory ships a validate.mjs that needs no dependencies and no network. Download a folder, run it, and it re-checks every row against the schema and every file against its recorded digest. It is the same validator this site runs, generated from the same module, so it cannot drift into being kinder to us. 

curl -O https://artwaste.land/data/navigation-lights/lights.jsonl
curl -O https://artwaste.land/data/navigation-lights/schema.json
curl -O https://artwaste.land/data/navigation-lights/manifest.json
curl -O https://artwaste.land/data/navigation-lights/validate.mjs
node validate.mjs

The machine-readable catalogue of this room is /data/audit.json, which carries every condition, every result and every withheld reason. The standard itself, with a self-test that shows each of the seven conditions failing on purpose, is research/data-room/standard.mjs.