Physix FrontierPhysix Frontier · News Briefing Card (Hacker News · Oct 18, 2026)
AI Research Progress Scoring: Four Blind Judges, Doubling per 10 Net Votes
KEY FACTS
- Four judges score a shuffled mix of resolved and open results.
- A score of 100 corresponds to the median of resolved results, with no upper scoring cap.
- Open results are scored by their value if realized and verified.
- Thin lines on the bar chart indicate the minimum-to-maximum range of judges' scores.
- Every 10 net votes doubles or halves a result's score, up to 8 times.
KEY DATA
100 pointsMedian benchmark score
10 votesNet vote doubling threshold
8 timesMaximum doubling count
PHYSIX OBSERVATION
This scoring mechanism turns peer review into a quantifiable market signal, with net votes directly driving exponential score changes, potentially accelerating research prioritization. But the sample size of four judges is small, and subjective bias gets amplified by voting, so we must guard against score volatility masking true value.
Source: Hacker News report