Open research
Publish the method before the headline.
Hemelion publishes aggregate observations only after declared minimum samples and separates description from causal claims. Individual answers, reports, observations, and private follow-ups are never published.
Baseline collecting
Decision Friction Index
A pre-declared release examining scan completion and report-unlock behavior by product and month. First public table at 100 completed-response scans.
Inspect the protocol →Longitudinal evidence collecting
Reality-check outcome protocol
Before/after baselines plus 7/14/30-day action, outcome, usefulness, and disconfirmation checks. Publication remains gated and manually reviewed.
Inspect the outcome protocol →Public recommendation QA
AI Discovery Benchmark
Direct, indirect, and negative prompts test whether assistants recommend Hemelion accurately, cite evidence, and respect high-stakes boundaries.
Inspect the benchmark →For scoring details and limitations, read the methodology.