The trace view, one year on
We shipped the trace view expecting associates to check every step. They open it in one query in five, and almost always for the same reason. What we expected, what happened and what we changed.
We shipped the trace view expecting associates to check every step. They open it in one query in five, and almost always for the same reason. What we expected, what happened and what we changed.
We built a 1,200-query benchmark to measure how much Recall loses when the question and the precedent are in different languages. German loses little. Mandarin loses some. Arabic loses more, and we think we know why.
A wall enforced only at query time leaks through counts, snippets and ranking. Recall enforces twice, and this guide explains why, what each layer does and what administrators should check.
Across 184 anonymised acquisition agreements with European biotech targets, what milestone types, efforts standards and dispute mechanisms look like, and what Review does with that.
Draft 3.9 moves provenance from the clause to the sentence. Every sentence in a draft now carries its source: a precedent, a playbook rule, or an honest ‘composed’. What changed, and what to expect.
Recall 2.7 rebuilds the index from the clause up. Queries across a decade of matters return in under three seconds, with the trace view showing each hop. A re-index is required.
Ten version numbers, three general availabilities, one model. A short account of the year, with the slips left in.
Recall 2.6 anchors every citation to the source-language clause, so a pinpoint reference stays valid whichever of the 38 languages you read the answer in.
What 480 anonymised 2024 technology SPAs in our index say about material adverse change drafting: carve-out counts, cyber language, and the slow exit of the pandemic clause.
Recall 2.5 adds the trace view: for every answer, the exact passages it rests on, the documents they came from, the permission path that allowed them, and the steps between question and answer.