How Outsample works, and where it fails.
This page explains what runs when you ask a question, using one real run and its measured numbers. The limits we know about have their own page: /limits.
Every question runs three stages.
| stage | what happens | real run |
|---|---|---|
| Retrieve | Hybrid search over the corpus: dense embeddings plus keyword recall. This stage casts wide. | 293 papers |
| Gate | A relevance pass keeps what bears on the question. If too little survives, the answer is flagged as thin evidence. | 41 kept |
| Synthesize | The model writes the answer from the gated set and cites the papers it used. | 7 cited |
The real-run column is a disagreement query on cross-sectional momentum, 2026-08-26, run on the Max tier (Opus 4.8). $0.1353. Its refusal line and measured numbers are further down this page.
An answer reasons over the gated set, typically tens of papers. No retrieval system reads fourteen thousand papers per answer, and a product that implies it does is lying to you.
Answers are constrained to what retrieval returned.
Synthesis writes from the gated papers. It does not assert past what it found: thin retrieval is flagged as thin on the answer, and a question with no support in the corpus ends in a bail. Citations name the papers an answer used, and each links to its source.
Every answer carries its counts and its cost. From the run above:
retrieved 293. used 41. cited 7. $0.1353.
When the evidence is not there, it says so.
Asked where the literature genuinely disagrees on cross-sectional momentum, retrieval returned 293 candidates and the gate kept 41. The papers largely agree that raw momentum profits are real; the genuine splits are about what drives them. The answer mapped three real paper-vs-paper disputes, then drew the line at the axis the corpus cannot support:
“I did not manufacture conflicts on the decay-versus-persist axis itself because the corpus does not contain a paper claiming raw momentum profits fully disappear; the substantive splits are on mechanism (time-series vs cross-sectional, industry-level vs stock-level, factor-level vs stock-level).”
It did not pad the result with pairs that do not exist. A question the corpus cannot support ends in a statement of the limit and a list of what was actually found.
Two kinds of paper, marked in the citations.
Most of the corpus is full text. A smaller set, mostly classics that sit behind journal paywalls, enters as reference entries.
| Full text | Parsed body in the corpus. Synthesis reads the paper itself. | 13,454 (88.9%) |
|---|---|---|
| Reference entry | Metadata, abstract, and a link out to the publisher. | 1,685 (11.1%) |
Citations mark which kind they cite. A reference entry can ground an answer only as far as its abstract goes; the link out covers the rest.
Measured 2026-08-24 against the shared library. Full composition, sources, and field coverage: /corpus.
The limits we know about have their own page.
Six of them. Stated plainly. Dated. What retrieval can miss, where coverage runs thin, what this product does not do: /limits.
The Free tier includes two full synthesis runs, total. Run one on a topic you know cold and read the citations it returns. That test is worth more than this page.