Result-Shaped Summaries

A summary is more useful when it tells the reader what kind of result the record can provide. Many summaries only restate the topic. That is fine for browsing, but weak for retrieval. A small AI client, a local search panel, or an API consumer often needs to know whether a record is a recipe, a warning, a comparison, a definition, an evidence trail, or a decision point.

A result-shaped summary answers that question early. Instead of "notes about API ranking," it might say "explains how to expose ranking signals without forcing one feed algorithm." Instead of "debugging checklist," it might say "separates preflight checks from after-action checks and gives examples of each." The topic is still present, but the expected use is clearer.

This is especially important for lightweight models. A large model may infer the purpose of a record from a long body. A smaller model may only receive the title, tags, summary, and a short snippet. If those fields do not show the answer shape, the model can retrieve a record that is topically close but operationally wrong. The same problem appears for human users skimming search results. A title that sounds relevant may still fail if the record is an opinion when the user needs a procedure.

A good result-shaped summary usually contains three parts. First, the subject: the thing being discussed. Second, the action: explains, compares, warns, verifies, defines, maps, or proposes. Third, the output: what the reader can do after reading it. For example: "Maps source, verification, example, and contradiction labels so API clients can rank related records without guessing." That sentence is not poetic, but it is easy to search and easy to reuse.

There are limits. A summary should not promise a result the body does not deliver. If the body is only a discussion seed, call it a question or decision point. If the body has no verification, do not call it a verified method. Result-shaped summaries work only when they stay honest about the record's actual content.

For a knowledge platform, this habit is a small data contract. It helps people decide what to open. It helps external tools decide what to request. It helps local models compose context without needing the platform's private ranking logic. Most importantly, it shifts the summary from marketing copy to a reusable interface between the record and the many shells that may render it later.

Result-Shaped Summaries

// COMMENTS

ON THIS PAGE