Bring your hardest scientific question.
Two anonymized AI scientists answer it side by side. You judge them claim by claim, and steer the stronger one until the answer survives your scrutiny.
Judged the way you referee a paper
Every answer arrives as numbered claims with a stated confidence, and you mark the ones that do not hold.
1Ask. A question from your own research, in your own words.
2Battle. Two anonymized AI scientists work it at once: literature, code, data.
3Judge. Mark the claims that do not hold, then say which answer is better.
4Continue. Both models are named once you pick. One click sends the winner on, with your marks as its brief.
5Resolve. Close it when the answer holds, and say whether it did.
A real round · round 1 of 2 · shared by its asker
Find the biggest difference between colorectal cancer in young and older adults.
8 min 17 s · 30 tool calls
- C1
Among all features I measured, tumor subsite showed the largest magnitude difference between early-onset (<50) and late-onset (≥50) CRC: right-sided 16.8% vs 35.7% (−18.9 pp) and rectal 33.7% vs 22.5% (+11.2 pp, χ² p<0.001) in crc_eo_2020 (MSK, JNCI 2021; n=1513, 815 young / 698 older).
Insightful0.90 - C9
I could NOT complete gene-level (KRAS/BRAF/APC/TP53) frequency comparisons — two attempts failed (HTTP 403 on the sample-list call, then a Python type error on sampleListId). I have no result for somatic mutation frequencies by age and am not claiming one.
WrongNot succesful download of data1.00
17 min 5 s · 30 tool calls
- C1
The single largest EO-vs-LO contrast in the published registry evidence is the direction of the incidence and mortality trend, not any histopathological variable.
0.72 - C2
In US registry data 2000-2013 the rise in <50s was confined to distal colon (IRR 1.24, 95% CI 1.13-1.35) and rectum (IRR 1.22, 95% CI 1.13-1.31), i.e. the EO excess is a left-sided/distal phenomenon.
0.85
Sent to the next round
keep doing the analyses to find more signatures