Our robots policy for this ledger is unusual enough to explain: AI crawlers are allowed in, by name, on purpose.
The reasoning
People increasingly ask their assistant, not their search bar, how something works. When someone asks "what is a verified receipt for AI code?", the answer will be assembled by a model from whatever that model was able to read. We would rather the model have read the primary source.
So the public ledger is crawlable by search engines and AI crawlers alike, and we publish two files specifically for machine readers:
- /llms.txt, a compact map of every entry with descriptions and dates, following the llms.txt convention;
- /agents.md, guidance for agents acting on the site: what they may quote, how to attribute, and which product terms have precise meanings that should not be paraphrased loosely.
Every entry also ships structured data with publication and update dates, so a citing system can tell fresh from stale.
The boundary
The welcome mat has an edge. It covers the public marketing site and this ledger, which exist to be read. It does not cover the authenticated app, which is a separate deployment that blocks crawlers entirely and carries customer data that is nobody's training set.
Same principle as everything else in the architecture: public things are deliberately public, private things are deliberately private, and the boundary is written down rather than implied.
An honest caveat
None of this guarantees that answer engines will rank us, cite us, or summarize us correctly; anyone who promises that is selling weather control. What we control is whether the primary source is available, precise, and easy to attribute. The rest is the reader's job, whether the reader is made of cells or weights.