Current limits, as of August 2026.
What this product cannot do today, in the same register as the rest of the site: stated, dated, and checkable. When a limit falls, the changelog will say so.
Retrieval is top-k.
An answer reads the gated set. A relevant paper can be missed at retrieval, and the counts on each answer show what this one read.
Output tracks corpus depth.
Ask about a niche the library barely covers and you get thin-evidence flags or a bail. Coverage by field is on /corpus.
It runs on a frontier model.
A misread of a paper's claim is possible. Citations are checkable, which bounds the damage without removing it.
Search coverage is partial.
Search runs on summaries and embeddings: 13,944 of 15,139 papers today. The rest are reachable by browse and metadata filters.
Heavy tools are slow.
A strategy run with its adversarial review costs around 420 credits; a corpus question about 76 credits, shown exactly on every result. Deep runs take minutes, quick ones seconds.
It does not backtest.
Outsample tells you what to test and what is wrong with the idea on paper. Whether it survives data is your job, on your infrastructure.
What a limit looks like in the product.
“No other paper pairs in the corpus engage directly enough with cross-sectional momentum to constitute a genuine contradiction.”
retrieved 295. used 50. cited 2. $0.0407. Real capture, 2026-06-11.
Coverage figures measured 2026-08-24 against the shared library. Per-run costs are from the metered captures of 2026-08-04 (honest_numbers.md); the bail capture is from 2026-06-11.
How the pipeline works, with one real measured run: /methodology. What the corpus actually covers: /corpus.