Independent AI research labVOL. 001 / 2026 — KRYVETS
DOUBT.
TEST.
RESEARCH ≠ HYPE
Exploring where artificial intelligence actually delivers, where it fails, and what lies between the promise and the result.
Numbers without worship.
We do not worship models. We test outputs, disclose failures, and publish our methods.
01 / Original submissions
Open the complete research prompt for every rated model.
02 / Transparent scores
Penalty rules, category weights, and provisional status for each result.
03 / Inspectable methods
Every experiment has a benchmark task and explicit evaluation rules.