About Commodity Origins
Commodity Origins answers one question for every commodity on earth: where does it come from, who sells it, what does it cost, and why? It is a reference work, written in plain English for the person who typed "where does cocoa come from" or "why is coffee so expensive" into a search box or an assistant, and for students, journalists, small importers, procurement and supply-chain staff who need a citable answer fast.
What makes it different
- Every number on the site comes out of an open data pipeline with its source, unit and year attached. If a source is unavailable the page shows the last good snapshot and says so; it never shows a guess.
- The prose is written for this site from the data, with AI assistance (see how AI is used). Nothing is scraped or paraphrased from encyclopedias.
- Produced, exported, processed and consumed are always kept apart, because they answer different questions and point at different countries.
- It is built to be quoted: answers come first, facts carry their year and unit in the same sentence, and every page has a JSON and Markdown twin.
What it is not
It is not a trading terminal, a forecasting service or investment advice. Prices are monthly benchmark averages that lag the market. For bilateral trade flows use the Observatory of Economic Complexity; for daily prices use the exchanges.
How AI is used on this site
Commodity Origins is made with AI, and this section says how, so you can judge the pages for yourself.
- What AI does. Anthropic's Claude drafts and edits the written text on the site: commodity pages, articles, explainers, glossary entries and the analysis paragraphs. It also helps write the code that fetches the data, computes the tables and builds the pages.
- What AI does not do. It is not a source of numbers. Each figure is either computed by a script from a named source and inserted into the text with its unit and year, or cited to the document it is taken from, such as a tariff notice or a company report. The charts are drawn by code from those tables. There are no AI-generated photographs or illustrations, and no invented people: the only author is the real publisher named below.
- How it is checked. Most figures are checked by construction, because a script computes each one from its source; the rest carry the citation of the document they come from. Automated checks stop the build on broken links, duplicate titles and descriptions, missing sources and leaked secrets, and titles, descriptions and structured data are generated from the same data. The analysis section on each commodity page was also checked, claim by claim, against the published data by a separate AI reviewer before release. Not every page has had a second read. Errors found after publication are fixed and listed, with what the page said before, on the corrections page.
- What it has not had. No lawyer, financial adviser, geologist or other licensed professional has reviewed the pages, and they are not advice.
- Who is responsible. Alykhan Virani directs the work, releases each version of the site and answers for it. Write to [email protected] to report an error.
Pages that are mostly a table with little added text, such as glossary definitions, side-by-side comparisons and country-by-commodity pages under 250 words, are kept out of search engines' indexes and the sitemap. They stay available to readers.
Who
Commodity Origins is published by Alykhan Virani, Director at SimbaSystems, alongside sister references on critical minerals and the LNG chain. Corrections and questions: [email protected] or the contact page.
Funding
The site is free to read. Google AdSense ads can run, but only after a visitor chooses Accept all in the consent dialog; with Accept required no ads are loaded, and the cookies page says what is set. Affiliate links, if ever used, are disclosed on the page where they appear.