A bright, mobile-friendly placeholder for artbenchmark.org.
“Who decides what is art?” introduces the provisional art benchmark from Marsita the Ultra and Planetary Council. The page includes the poster, an expandable text transcript, and links to Planetary Council, Telegram, Twitter / X and Instagram.
Static HTML and CSS with no runtime dependencies. The production domains are attached
to the Cloudflare Worker artbenchmark. GitHub Pages is also enabled, but publishing
to GitHub Pages does not update that Worker. CNAME retains the GitHub Pages domain setting.
index.html— page copy, links, accessible poster transcript, and social metadata.styles.css— responsive cream, yellow and violet layout.assets/art-benchmark.jpg— full-size provisional poster, based on the supplied image.assets/art-benchmark*.webp— smaller, responsive versions for the page.assets/share-preview.svg/.png— editable source and 1200 × 630 social preview.assets/favicon.svg,favicon.ico, andassets/apple-touch-icon.png— site icons.robots.txt/sitemap.xml— crawler discovery.
Open index.html through a local static server to preview root-relative asset links.
When changing the public domain, update CNAME, the canonical URL, social image URLs,
robots.txt and sitemap.xml together. To update the share card, edit the SVG and
export it as a 1200 × 630 PNG. Social platforms may cache old previews.
Requires Node.js 22.12 or newer, npm, and access to the Cloudflare account that owns
artbenchmark.org. From a fresh checkout of this repository, run:
bash scripts/publish-cloudflare.shThe script builds dist/ from the public page and assets, then uses pinned Wrangler
4.162.0 to update the existing Worker artbenchmark. Wrangler uses your local login,
or opens a browser login if needed. Select the account that owns the domain if asked.
The configuration declares the two existing custom domains: artbenchmark.org and
www.artbenchmark.org. Cloudflare manages their locked Worker DNS records; no TXT
record or DNS import is needed. This deploy replaces the old website on that Worker.
The script then checks that both domains serve this checkout's HTML, stylesheet,
share image and favicon. If propagation takes longer, run node scripts/verify-live.mjs
again. The deploy may have succeeded even if the later verification times out.
Useful commands:
npm run build:site # prepare the public files only
npm run deploy:check # Wrangler dry run; no publication or login required
npm run deploy # publish, then verify the live domains
npm run verify:live # check the live domains againNo npm install is needed for these commands. Wrangler is downloaded by npx when
needed. If several Cloudflare accounts are available, you can select one by setting
CLOUDFLARE_ACCOUNT_ID in your terminal. Keep credentials in your local Wrangler
login or environment; never commit them.
For Cloudflare Workers Builds, use npm run build:site as the build command and
npx wrangler@4.162.0 deploy as the deploy command with main as the production branch.
The page keeps an empty WORKS array between the WORKS:BEGIN / WORKS:END markers
for compatibility with the existing syndicator. The placeholder poster is a page
asset, not a collection entry, so it will not be broadcast by that workflow.
The existing distribution tooling is retained for future use. Adding entries to the array makes them available to the syndicator; displaying a future collection would also require a new page layout.
Every new work is broadcast automatically by .github/workflows/syndicate.yml, which
runs syndicate.mjs. It reads the WORKS array (through the WORKS:BEGIN/WORKS:END
markers; keep them intact), checks syndicated.json for what has already gone out, and
announces works on six rails:
| rail | needs (repo secrets) | what happens |
|---|---|---|
| Internet Archive | nothing | Wayback snapshot of the page and the image |
| Nostr | NOSTR_NSEC |
kind-1 note signed with your key, 4 relays |
| Farcaster | NEYNAR_API_KEY, NEYNAR_SIGNER_UUID |
cast with image + site embeds, via Neynar |
| Telegram | TELEGRAM_BOT_TOKEN, TELEGRAM_CHAT_ID |
photo + placard caption to the channel |
META_PAGE_ID, META_PAGE_TOKEN |
photo post on your Page | |
META_IG_USER_ID, META_PAGE_TOKEN |
feed post with placard caption + hashtags |
Distribution is a discipline, not a dump. The workflow runs on every push to main and
three times a day (08:17, 14:17, 20:17 UTC). Each run broadcasts at most MAX_PER_RUN
works (default 1; set it under Settings → Secrets and variables → Actions →
Variables), newest first. New arrivals go out promptly and the back catalogue follows
at a steady pace. At 1 per run, 10,000 works is a nine-year broadcast. That is the point:
a feed people can live alongside, not one they mute.
- Rails without secrets are skipped, not failed. The Archive rail works from day one.
syndicated.jsonrecords which rails carried which work, so adding a key later back-fills only what that rail missed (still through the drip). - A failed rail (network, rate limit, expired token) stays unrecorded and retries next run. Rails that succeeded in the same run are still recorded and never post twice.
- A refused rail is one where retrying can't help. Instagram accepts only JPEGs under
8 MB with an aspect ratio between 4:5 and 1.91:1, so a panorama, a tall scroll or a PNG
is refused with the reason written to the ledger. Give that work an
img_socialfield (a cropped or padded JPEG) and the rail reopens on the next run.
npm run dry-run shows what the next run would fire; npm run syndicate fires it
(same env vars).
This is the fiddliest rail; it takes about 20 minutes, once.
- Switch the Instagram account to Professional (Business or Creator) and link it to your Facebook Page (Instagram → Settings → Account type / Linked accounts).
- At developers.facebook.com, create an app of type Business. Leave it in development mode: posting to Pages you admin doesn't need app review.
- In the Graph API Explorer, select the app and generate a User token with
pages_show_list,pages_read_engagement,pages_manage_posts,business_management,instagram_basicandinstagram_content_publish. - Run, locally:
It exchanges the token for a non-expiring Page token and prints
npm install META_APP_ID=… META_APP_SECRET=… node meta-token.mjs <the Explorer token>META_PAGE_ID,META_PAGE_TOKENandMETA_IG_USER_ID. - Paste those three into the repo's Actions secrets, then clear your terminal.
The Page token stops working if you change your Facebook password or remove the app. If that happens, repeat steps 3–5. Failed runs will tell you, and nothing is lost: the ledger retries.
Add all secrets at Settings → Secrets and variables → Actions. No credential ever appears in the repo; only signed posts leave.