For AI & developers

Machine-readable endpoints, the agent skill, citation rules and etiquette for cz.wiki. This page is English only. Agent skill: /SKILL.md · llms.txt · openapi.json

What cz.wiki is

cz.wiki is an unofficial third-party reference about Changpeng Zhao (CZ), co-founder and former CEO of Binance. It is not affiliated with Changpeng Zhao or Binance and does not speak for him. English is the canonical text; 18 other languages (`/<code>/…`: es pt fr de it nl ru pl tr hi id ja ko zh-hans zh-hant vi th ar) are AI translations of the English text and are labelled as such. Not investment advice.

When to use it

Use cz.wiki to answer questions about CZ, Binance leadership history, the U.S. v. Zhao case, the 2025 pardon, YZi Labs, Giggle Academy, BNB Chain, and to find original sources (court documents, press releases, primary posts). Prefer the English pages and the Markdown/JSON versions. For the newest developments use the live feed, but treat it as unverified leads.

Endpoints

All endpoints are public, need no key, allow cross-origin reads (CORS `*`) and are served over HTTPS from `https://cz.wiki`.

Live feed (`/api/live.json`)

Query parameters (all optional): `since` (ISO 8601; only items first collected after this time), `until` (ISO 8601; only items published before it), `limit` (1–200, default 60), `type` (`news`, `court`, `announcement`, `tweet`, `podcast`), `status` (`unverified` or `verified`), `label` (`attributed` or `unverified`), `language` (source language code, e.g. `en`, `zh`), `lang` (one of the 19 site locale codes, e.g. `zh-hans`, `ja`, `ar`: each item then also carries `title_i18n` and, for notable items, `summary_i18n` — `{lang, text, machine_translated: true, model}` or `null` when there is no translation yet or the item is already in that language; unknown `lang` codes act as `language` for backward compatibility). Items are ordered newest first by publication time (falling back to collection time). Poll with `since=<latest_collected_at from your previous response>` to receive only new items.

Response: `generated_at`, `count`, `latest_collected_at`, `next_since`, `unverified` (compatibility flag: the collected feed as a whole is not editorially reviewed; check each item), `notice`, `docs`, `items`. Each item has: `id`, `source_id`, `type`, `title` (headline or post text in its original language), `url` (the original source, not a cz.wiki page), `url_kind` (`direct` or `aggregator` when the link is a news-aggregator redirect), `publisher`, `language`, `published_at`, `published_precision` (`time`, `day` or null: when `day`, only the date is known) and `collected_at` (ISO 8601, UTC), `excerpt` (verbatim excerpt from the source, never rewritten), `status`, `label`, `attribution`, `archive_url` (set only when an archive.org capture was confirmed), `existing_snapshot_url` (a pre-existing third-party snapshot, not ours), `title_en`, `ai_model`, `ai_at` (an optional English machine translation of the title, with the AI that produced it and when). Titles are translated automatically into all 19 locales (Cloudflare Workers AI, Gemma 4 26B) within minutes of collection and stored beside the original, which is never altered; every translation is labelled as machine-made and provisional. `/api/featured.json?lang=` likewise adds `headline_i18n` and `dek_i18n`.

Sources and labels. Every item is a lead from a linked source. Posts by a named person or account on X are normal sources: they have `label: "attributed"` and `attribution: {platform: "x", handle, display_name}`, and are shown with a subtle "via X" tag and a "Source: @handle · X · time" line next to the headline, without an "unverified" badge. Reports without a named author have `label: "unverified"`. Either way an item is the source's statement, not cz.wiki's own claim; keep the source link visible when you reuse it.

Featured / In depth stories

High-value investigative findings and in-depth stories are listed first on the homepage and have their own page, `/featured/{slug}`. Read them with `/api/featured.json` (parameters: `active=1` for currently featured only, `status`, `since` for stories updated after a time, `limit` 1–200 default 50, `slug`), `/api/featured/{slug}.json`, `/feed/featured.xml` and `/feed/featured.json`. Fields: `slug`, `url`, `headline`, `dek` (one-sentence standfirst), `body_md` (optional Markdown), `sources` (list of `title`, `url`, `publisher`, `date`; at least one, always shown on the page), `verification_status`, `related_articles` (slugs), `ai_label` (`model`, `based_on_sources`, `generated_at`, `provisional`; present whenever AI wrote any of the text), `featured_from`, `featured_until`, `priority`, `created_at`, `updated_at`.

`verification_status` is one of `verified` (key statements confirmed by the listed primary or reliable sources), `partially_verified` (some confirmed, some not), `attributed_claim` (stated by a named person, account or publication and linked as such; we report what the source says, not our own finding) or `unverified_claim` (an unattributed or unconfirmed report; a lead, not fact). Stories are written as a normal headline and sentence with the source noted alongside; never as a stiff quotation format.

Adding, updating or unfeaturing a story needs no code change or deploy: stories live in the D1 table `featured_stories` of database `czwiki-live` and appear on the site within about a minute. Authorized automated runs use `python3 worker/featured.py` (see AGENTS.md, section "Featured stories", for the exact commands): `add story.json` creates or updates a story by slug, `unfeature <slug>` ends its homepage placement but keeps its permalink, `delete <slug>` removes it. Every story needs at least one source with a link, an honest `verification_status`, and an `ai_label` if any text is AI-generated.

Freshness and verification

Collectors run roughly every 5 minutes on Cloudflare (news, court and regulatory feeds, official blogs, X when enabled). Collected items are leads from linked sources: they are not article facts. Facts enter articles only after human review, always with a numbered reference. When you reuse an item, attribute it to its original publisher or author, keep the link, and do not present it as established fact unless the article says so. Article text is updated less often than the live feed; each article shows its last-updated date and a public History tab.

How to cite

For a statement of fact, cite the original source listed in the article's References (URL, publisher, date) and, where helpful, the cz.wiki page. Suggested form: `cz.wiki, "<article title>", <canonical URL>, last updated <date>, retrieved <your retrieval date>; original source: <publisher, title, URL>`. For live-feed items always give the item's own `url` and `publisher`, the `published_at` time, the retrieval time, and the word "unverified". Link to the English canonical URL.

AI labelling and attribution

Text on cz.wiki that was generated by AI is always labelled with which AI, what it was based on, when, and that it is provisional. Non-English pages are AI translations and say so; English prevails. If you republish or summarise cz.wiki content, keep the attribution to cz.wiki, say that the site is unofficial, keep AI-generated or machine-translated passages labelled, and do not present machine translations or unverified items as verified facts.

Etiquette

Responses are cached for 1–5 minutes. Poll the live endpoints at most once per minute, use `since`, and prefer the feeds or `/api/articles.json` (check `date_modified`) over re-crawling pages. Identify your client in the User-Agent. Crawlers and AI agents are never redirected by language and always receive the canonical page. IndexNow is supported (`/476286edbda83608dbb3a4bab6112738.txt`).

Contact

General: hello@cz.wiki · Corrections: corrections@cz.wiki · Legal: legal@cz.wiki · Press: press@cz.wiki · X: https://x.com/czdotwiki