On research questions with a precise subject, we return what the people who work on that subject actually published and pointed at. Here is the test, and here is where we lose.
Not yet run: Exa, Ceramic, Brave, Linkup. Those columns are blank rather than filled in, and no overall verdict is drawn until they have been measured.
Every engine here can tell you what it returned. We also tell you how: what we understood the question to be, what each stage cost, what the ranking weighed, and which results we deliberately set apart. If the ordering is wrong you can say why, which is not true of a ranking nobody can see.
A weighted sum, not a product: an unscored author is missing data rather than a verdict. A page that mentions nothing you asked for keeps half its score, because a result that does not answer the question is not a result.
lecanemab or lekemab or donanemab or practiceposts_wide 1.8srecent_papers 3.1sposts 8.2spapers 8.2sgather 430msresolve 182msauthors 466msVerdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
amyloid hypothesis or amyloid beta or alzheimer's disease or wrongposts_wide 1.7srecent_papers 2.3spapers 2.6sposts 5.6sgather 744msauthors 317msresolve 579msVerdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
pfas or forever chemicals or pfas drinking water or evidenceposts_wide 1.9sposts 2.6sgather 4msrecent_papers 3.2spapers 8.1sresolve 0msauthors 303msVerdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
microplastics or plastic particles or microplastic pollutionrecent_papers 1.1spapers 1.3sposts_wide 1.6sposts 2.0sgather 390msauthors 257msresolve 408msVerdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
chain of thought or reasoning models or cognitive modelsposts 1.4sposts_wide 1.6sgather 0msrecent_papers 3.1spapers 4.8sresolve 0msauthors 219msVerdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
A one-off technical fault whose answer lives in a single thread. Knowing who the authorities in a field are does not help here, and our results show it: they are about the subject rather than about the error. This is a search engine's question, it is outside what this page claims, and we run it anyway because a table won on every row reads as rigged.
Verdict pending: read by hand, not scored by us.
Not yet run (no EXA_API_KEY).
Not yet run (no CERAMIC_API_KEY).
Not yet run (no BRAVE_API_KEY).
Not yet run (no LINKUP_API_KEY).
Long-tail lookups, local and navigational queries, and queries that are not in English. Those are a search engine's job and we say so on the about page as well: ask us first, and fall back on a search engine for the rest.
The query set, the harness and the raw per-query output are all published. Every run is date-stamped, because results move and a comparison without a date is a claim about a moment nobody can find again.
export AUTHORIX_EMAIL=you@example.com AUTHORIX_PASSWORD=...
export EXA_API_KEY=... BRAVE_API_KEY=... LINKUP_API_KEY=... CERAMIC_API_KEY=...
python3 tools/compare/run.py --out frontend/public/compare-runsAn engine without a key is skipped and recorded as not run, so any column can be refreshed on its own. The verdict per column is a human judgement and the harness never writes one.