How content is made here

Facts, never advice. Every page reports observations with their reporting period and source. We do not recommend, rate, or rank securities or managers by attractiveness — lists are ordered by stated, measurable criteria ("most widely held", "largest reported value"), never by labels like "best".

News selection. Headlines are aggregated from named mainstream outlets. Press-release wires and buy/sell-rating content mills are excluded by a standing blocklist. Every item names its publisher, shows the publisher's own summary, and links to the original article — no bodies are copied. Inclusion is not endorsement.

Filing recaps are templated, not written. The "data desk" items are generated from parsed filings by fixed templates — every figure comes from the database. No language model writes news here.

Where AI is used, and its rules. The Research Desk uses a language model for one job: narrating the rows a structured query already computed. It is forbidden from inventing numbers, giving advice, or answering outside the returned data, and each answer shows the query, period and row count it narrates. Site translation into 15 languages is AI-assisted; legal pages remain English. The engraved section artwork is AI-generated original illustration (decorative, depicting no real place or event). The site-wide AI-provenance note appears in the footer.

Corrections. When the SEC record changes (amendments, restatements), pipelines recompute affected figures automatically and the filing history shows the sequence. If you find an error the pipelines missed, use Contact — or the original filing linked on the page, which always wins over our parse.

Independence. Data display is never influenced by advertisers or partners. If affiliate links are ever introduced, they will be disclosed and marked at the point of decision.