The output is a screen, not an answer.
Parse uses language to help build a transparent stock screen. It does not hide the criteria behind a recommendation list.
Describe the idea
Write the screen the way you would explain it to another person.
Parse translates it
Your words are mapped to explicit metrics, thresholds, and a ranking. The model builds the screen; it does not choose stocks directly.
Confirm before screening
Parse shows every interpreted filter before it screens the universe. Edit, add, or remove a criterion if the interpretation is not what you meant.
Run the verified screen
Only after you confirm the criteria does Parse execute the screen. Direct edits remain visible and intentional.
When wording is ambiguous
Ideas such as “safe,” “cheap,” and “quality” do not map to one universally correct metric. Parse either maps the wording to a supported criterion it can justify, uses one of the documented defaults below, or leaves the criterion out rather than silently substituting something else. You review the result before the screen runs.
What Parse means when you do not give a number
These defaults are used only for the qualitative meanings listed here when you have not already named a metric or threshold. Any default appears in the review step before screening and can be edited or removed.