Every load-bearing claim on Suit Up traces to a source captured at the time of citation. URLs are not used as the primary reference because they rot. Instead, sources are archived locally inside the project alongside the structured frontmatter the wiki layer reads.

Archival mechanics

New entries enter the pipeline through one of three paths: a curated RSS feed list, a Playwright-based scraper for sources without reliable RSS (most APAC bank press rooms, regulator publications), or a manual drop via the Telegram bot. Each entry is written into the project repository as a markdown file with structured frontmatter (title, source, URL, date, jurisdiction, named institutions and projects, asset class). The original HTML or PDF is preserved where the source permits redistribution; otherwise the canonical metadata is preserved alongside the citation.

Citation

Wiki entries cite their underlying source inline at entry-level. Where a source is paywalled, the citation points to the underlying document or filing rather than the paywalled summary. Where a claim is sourced from a primary regulator filing, the citation links to the filing on the regulator's site. Where a claim is sourced from trade press, the citation links to the trade-press article and notes the underlying source the article references.

Relevance and importance scoring

The ingestion pipeline scores each new entry on two axes. Relevance answers "is this on-topic for institutional tokenisation?" Importance answers "does this matter operationally, or is this background noise?" Both scores are produced by Claude Haiku 4.5 against a calibrated rubric and written into the entry's frontmatter. The homepage news feed prefers high-importance items by default; the broader wiki layer surfaces entries by their structural relationships rather than by score.

Corrections

Corrections are dated, attributed, and preserved in the wiki entry's revision history. The corrections log aggregates them in reverse chronological order. Where a correction changes the editorial conclusion of a wiki entry, the original conclusion is preserved in a struck-through form so the change is auditable.

What we do not do

We do not republish full articles from third-party sources. We do not embed paywalled material. We do not use AI-generated content as a primary source; AI is used inside the pipeline to score, classify, and route, never to write the editorial copy that appears in wiki entries or the weekly briefing. We do not accept paid placement, sponsored content, or sponsored research. Editorial integrity is the only product.