Calibration

Calibration Dashboard

How well-calibrated is the Future Shock prediction engine? This page shows diagnostic metrics, calibration curves, and accuracy breakdowns.

Lower Brier score = better calibration. Perfect = 0.

Overall Metrics

Brier Score

0.121

56 scored predictions

Log Score

-0.629

Higher is better

Resolved

56 / 227

170 pending · 1 withdrawn

Accuracy

46.4%

26 correct · 24 wrong · 6 partial

Calibration Curve

0–10%
10–20%
20–30%
30–40%
40–50%
50–60%
60–70%
70–80%
80–90%
90–100%
Actual outcome ratePerfect calibration

Bars should align with the dots. Bars above = overconfident. Below = underconfident.

Signal Performance

SignalBrierResolvedSnapshotsCoverage
manifold0.0283321151%
PDM (Prerequisite Density)No snapshots yet

Accuracy by Category

capability35% · 14206~ / 40
economics40% · 230~ / 5
social100% · 400~ / 4
labor100% · 300~ / 3
regulatory100% · 200~ / 2
timeline50% · 110~ / 2
Correct Partial Wrong

Recent Resolutions

correct

AI agents will join the workforce in 2025

80% confidence · Resolved 2025-12-31

correct

AI smarter than humans in many domains

75% confidence · Resolved 2025-12-31

correct

The AI culture war begins in 2025

90% confidence · Resolved 2025-12-31

correct

No artificial general intelligence in 2025, despite claims by Elon Musk

95% confidence · Resolved 2025-12-31

correct

AI agents will be endlessly hyped but far from reliable in 2025, except in narrow use cases

85% confidence · Resolved 2025-12-31

correct

AI model profits remain modest or nonexistent in 2025

85% confidence · Resolved 2025-12-31

correct

US will have very little AI regulation protecting consumers in 2025

85% confidence · Resolved 2025-12-31

correct

Hallucinations continue to haunt generative AI in 2025

95% confidence · Resolved 2025-12-31

correct

Humanoid robotics: lots of hype, nothing close to Rosie the Robot in 2025

85% confidence · Resolved 2025-12-31

correct

Less than 10% of workforce replaced by AI in 2025

95% confidence · Resolved 2025-12-31

Free daily briefing on what actually matters in AI.