Search

4/1334 → 10/976; +0.72 percentage points of formation share.

The word "search" no longer means one thing in this group, and the split runs right down the middle: Phind and Lumona are consumer-facing engines that replace the query box itself, while Trieve, Reducto, and Onyx are infrastructure sold to companies that already have a search box but need better retrieval underneath it, whether that's documents, PDFs, or internal knowledge. That shift toward document-level and enterprise retrieval, seen in Reducto's document workflows and Glimmer's 1,000-page PDF problem, reflects what became practical once retrieval-augmented generation and long-context models matured through 2023, letting a small team build ranking and relevance tooling that used to require a search team and months of work, which is Trieve's explicit pitch. The category is accelerating at 3.42x, notably faster than the adjacent Web Data API formation's 2.89x, suggesting the harder-to-build retrieval layer is growing faster than the raw data-access layer feeding it. Yet an independent outcome evaluation tied to this formation complicates the consumer side of that story, finding that more personalized answers and explanations do not reliably produce durable learning, calibrated confidence, or transfer, and can leave users more dependent on the tool rather than less. That tension, infrastructure demand accelerating while the value of the answers it produces remains unproven for actual learning, is the real story sitting underneath a list of ten companies that otherwise all just say they do "search."

No comparable outside activity is connected yet.

These share a word with “Search” but are tracked as separate categories. The real overlap below is the actual share of Search’s companies also found in that category — not a guess.