Cumulative Web Inc logo CWI BLOGThe label's news wire
CWI Blog

The CWI blog is now machine-readable

The CWI blog is now machine-readable

This blog is built for two audiences: people, and machines.

Every post on the CWI blog ships as three things at once:

  1. An HTML page — the article you're reading, with full SEO meta, Open Graph tags, and a canonical URL.
  2. A JSON twin — a machine-readable node with the slug, title, excerpt, body, date, tags, author, canonical URL, and the full JSON-LD record. Grab any post's JSON by swapping `.html` for `.json` in its URL.
  3. JSON-LD BlogPosting schema — embedded in every page, so search engines and AI agents parse the headline, date, author, and publisher without guessing.

The whole feed is one file: cumulativeweb.com/blog/index.json — newest first, every post, every tag. That feed is the ingestion surface for CWI's own machine-data layer, which can pick up new posts as content the moment they publish.

Pages also report back: each post fires exposure and click beacons to the machine-data layer, so the team can see what gets read and what gets clicked — the same measurement pattern as the Radio 365 track pages.

The machine-readable contract for the label starts at cumulativeweb.com/llms.txt. This blog is the news wire plugged into it.

Machine-readable

Every CWI blog post ships with a JSON twin. Copying fires an equip beacon so the machine learns what gets equipped.

curl https://cumulativeweb.com/blog/posts/cwi-blog-machine-readable.json
Copied — equip beacon fired.

One story, two readers

Every CWI blog post ships in two forms: the prose you’re reading, and a JSON twin carrying the same facts as structured fields. The prose serves humans — narrative, context, voice. The JSON serves machines — the label’s own agents, the CWI Discovery Engine MCP server (listed on mcprush), and any AI that queries the catalog. Same claims, two encodings, zero drift between them.

This isn’t a gimmick; it’s the label’s signal doctrine. "Think signals at all times" — every digital asset is built to be visible to and ingestible by other machines and agents. A blog post that only humans can read is half a post. The JSON twin makes the half whole: dates as dates, placements with scan dates, stream counts with observation dates, credits as entities.

What the twin carries

The structured twin mirrors the post’s verifiable claims: the headline, the entities (tracks, artists, producers), the numbers (with their as-of dates), the platform links (verified), and the schema type (BlogPosting, with the same JSON-LD the page carries). When the label’s agents answer questions about the catalog, they read the twin — which is why the numbers in chat match the numbers on the page: one source of truth, two readers.

The pattern extends across the site: press releases with NewsArticle schema, category pages with CollectionPage and MusicRecording data, video pages with VideoObject blocks. The blog is where the doctrine is most visible, because it’s where the label thinks out loud — and now the machines can listen in.

Frequently asked

Where do I find the JSON twin?

Next to the post — the "JSON" link beside "Read" on the blog index. Same facts, structured fields.

Can my agent read the CWI catalog?

Yes — through the CWI Discovery Engine MCP server: search_catalog, get_track, get_artist, and more, all free on free channels. The blog twins are the training set; the MCP server is the query layer.

Do the twins ever disagree with the prose?

They’re generated from the same source, and the label’s proofread discipline covers both: content, logic, metadata, and presentation verified before anything ships. If you ever catch drift, the contact page takes corrections.

For builders

If you’re building agents that touch music: start with the JSON twins for the fact patterns, then the CWI Discovery Engine MCP server for the query layer (search_catalog, get_track, get_artist, and more — free on free channels). The schema vocabulary is schema.org throughout, so your parser already speaks it. And if the twins ever drift from the prose, the contact page takes corrections — the doctrine only works if it’s enforced.

The standard

Machine-readable isn’t a feature — it’s the standard every CWI surface is held to. Press releases carry NewsArticle schema, category pages carry CollectionPage and MusicRecording data, video pages carry VideoObject blocks, and the blog carries JSON twins. One doctrine, every page: built for readers and for AI, verified before it ships.

Start building

Read a post, then read its JSON twin — compare the claims field by field. That’s the whole lesson: one source of truth, two readers, zero drift. Then point your agent at the MCP server and ask it something the catalog knows.