Predictor calibration
Per-predictor Brier scores and the κ (predictor discount factor) used to weight their evidence in the Bayesian intake update. κ is computed once per Brier recompute via scripts/ops/compute_brier_scores.py and applied at intake time. Global Brier (climatology baseline): 0.0211.
Predictors
104
37 with ≥1 resolved
Mean κ
0.561
population avg
At κ floor (≤0.11)
0
effectively silenced
Top predictor
Peter Diamandis
κ=0.875 N=15
Methodology
Brier = (1/N) Σ (prior_prob_i − outcome_i)², outcome ∈ {hit:1, partial:0.5, miss:0, killed:0}
quality_raw = clamp((0.30 − Brier) / 0.20, 0, 1)
# Brier 0.10 → quality 1.00
# Brier 0.20 → quality 0.50
# Brier 0.30 → quality 0.00 (and below)
κ (kappa) = clamp((quality_raw × N + 0.5 × N_prior) / (N + N_prior), 0.10, 1.00), N_prior = 5
Applied LLR = κ × LLR(polarity, evidence_strength)
posterior_logit = prior_logit + Σ adjusted_LLRPredictors (104)
| Predictor | Total | Resolved | Hits | Misses | Hit rate | Brier | κ ↓ | Calibration plot |
|---|---|---|---|---|---|---|---|---|
| Peter Diamandis | 160 | 15 | 10 | 0 | 66.7% | 0.0367 | 0.875 | |
| Codex Research Pack | 20 | 0 | 0 | 0 | — | — | 0.850 | — |
| Alex Wissner-Gross | 231 | 11 | 6 | 1 | 54.5% | 0.0341 | 0.844 | |
| Dave Blundin | 165 | 9 | 3 | 2 | 33.3% | 0.0491 | 0.821 | |
| Jensen Huang | 27 | 8 | 6 | 0 | 75.0% | 0.0128 | 0.808 | |
| Brett Adcock | 51 | 6 | 5 | 0 | 83.3% | 0.0040 | 0.773 | |
| Emad Mostaque | 14 | 4 | 3 | 0 | 75.0% | 0.0073 | 0.722 | |
| Jack Dorsey | 6 | 4 | 3 | 0 | 75.0% | 0.0109 | 0.722 | |
| Ben Lamm | 47 | 3 | 2 | 0 | 66.7% | 0.0043 | 0.688 | |
| Dario Amodei | 17 | 3 | 1 | 0 | 33.3% | 0.0363 | 0.688 | |
| Andrej Karpathy | 10 | 3 | 3 | 0 | 100.0% | 0.0067 | 0.688 | |
| Elon Musk | 69 | 3 | 1 | 0 | 33.3% | 0.0142 | 0.688 | |
| Andrew Yang | 35 | 3 | 0 | 0 | 0.0% | 0.0178 | 0.688 | |
| Kevin Weil | 9 | 3 | 2 | 0 | 66.7% | 0.0200 | 0.688 | |
| SpaceX | 3 | 3 | 3 | 0 | 100.0% | 0.0011 | 0.688 | |
| Dara Khosrowshahi | 89 | 3 | 2 | 0 | 66.7% | 0.0105 | 0.688 | |
| Leopold Aschenbrenner | 23 | 3 | 2 | 0 | 66.7% | 0.0417 | 0.688 | |
| Eric Schmidt | 68 | 3 | 3 | 0 | 100.0% | 0.0064 | 0.688 | |
| Superforecaster Community | 5 | 2 | 2 | 0 | 100.0% | 0.0000 | 0.643 | |
| Salim Ismail | 49 | 2 | 1 | 0 | 50.0% | 0.0144 | 0.643 | |
| Jimmy Ba | 5 | 2 | 2 | 0 | 100.0% | 0.0122 | 0.643 | |
| Alex Finn | 28 | 2 | 1 | 0 | 50.0% | 0.0122 | 0.643 | |
| David Holz | 5 | 2 | 2 | 0 | 100.0% | 0.0163 | 0.643 | |
| Morgan Stanley | 42 | 2 | 1 | 0 | 50.0% | 0.0442 | 0.633 | |
| Nvidia | 1 | 1 | 1 | 0 | 100.0% | 0.0064 | 0.583 | |
| Morgan Stanley / Georgia Tech / Intel | 1 | 1 | 1 | 0 | 100.0% | 0.0064 | 0.583 | |
| OpenAI (Sam Altman-led) | 1 | 1 | 1 | 0 | 100.0% | 0.0025 | 0.583 | |
| Micron | 1 | 1 | 1 | 0 | 100.0% | 0.0225 | 0.583 | |
| Demis Hassabis | 15 | 1 | 1 | 0 | 100.0% | 0.0064 | 0.583 | |
| Anthropic | 1 | 1 | 1 | 0 | 100.0% | 0.0144 | 0.583 | |
| Joseph Moore | 6 | 1 | 1 | 0 | 100.0% | 0.0001 | 0.583 | |
| Joe Liemandt | 16 | 1 | 1 | 0 | 100.0% | 0.0064 | 0.583 | |
| Video Narration (SpaceX) | 1 | 1 | 1 | 0 | 100.0% | 0.0064 | 0.583 | |
| Peter Diamandis / Salim Ismail / Andrew Yang | 1 | 1 | 0 | 0 | 0.0% | 0.0025 | 0.583 | |
| Michael Saylor | 10 | 1 | 0 | 0 | 0.0% | 0.0025 | 0.583 | |
| Meta / a16z (Andreessen, Horowitz) | 1 | 1 | 1 | 0 | 100.0% | 0.0025 | 0.583 | |
| New Market Pitch | 1 | 1 | 1 | 0 | 100.0% | 0.0025 | 0.583 | |
| Sam Altman | 20 | 1 | 0 | 0 | 0.0% | 0.0625 | 0.583 | |
| Ben Horowitz | 22 | 0 | 0 | 0 | — | — | 0.500 | — |
| Brent Bornick | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| CATL | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Daniella Amodei | 2 | 0 | 0 | 0 | — | — | 0.500 | — |
| Dario Amodei / Anthropic | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Amy Webb | 3 | 0 | 0 | 0 | — | — | 0.500 | — |
| Chamath Palihapitiya | 5 | 0 | 0 | 0 | — | — | 0.500 | — |
| China (government) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| David Friedberg | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Dr. Don Mucalem | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| EY (Ernst & Young) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Goldman Sachs | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Brett Adcock / Elon Musk / Vinod Khosla | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| 1 | 0 | 0 | 0 | — | — | 0.500 | — | |
| BIS Research | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| April Rinne | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Equinix | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Alphabet | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Fabric8Labs (NEA / Intel Capital funded) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Gerd Leonhard | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Gwynne Shotwell | 5 | 0 | 0 | 0 | — | — | 0.500 | — |
| Gwynne Shotwell / xAI | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| IEA | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Ian Bremmer | 3 | 0 | 0 | 0 | — | — | 0.500 | — |
| Industry analysts (synthesis) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Isaiah Taylor | 2 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jared Isaacman | 3 | 0 | 0 | 0 | — | — | 0.500 | — |
| Mike Wilson | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Multi-Forecaster Synthesis | 3 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jared Isaacman (NASA administrator) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jason Calacanis | 5 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jason Calacanis / David Sacks | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Seattle Met / Washington State regulators | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jennifer Li | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Jensen Huang / Morgan Stanley | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Joe Liemandt / MacKenzie Price | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Lyten | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| MacKenzie Price | 4 | 0 | 0 | 0 | — | — | 0.500 | — |
| Marc Andreessen | 8 | 0 | 0 | 0 | — | — | 0.500 | — |
| Marc Andreessen / Ben Horowitz | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Mark Cuban | 7 | 0 | 0 | 0 | — | — | 0.500 | — |
| Mark Pack Donovan | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Meta | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| NVIDIA | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Nick Bostrom | 6 | 0 | 0 | 0 | — | — | 0.500 | — |
| Nvidia (All-In Podcast analysis) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| OpenAI Codex Lead | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Pete (audience, data center builder) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Peter Dannenberg | 5 | 0 | 0 | 0 | — | — | 0.500 | — |
| Peter Zeihan | 4 | 0 | 0 | 0 | — | — | 0.500 | — |
| PolyMarket | 2 | 0 | 0 | 0 | — | — | 0.500 | — |
| Prediction markets | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Ralph Losey | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Ramez Naam | 5 | 0 | 0 | 0 | — | — | 0.500 | — |
| Ray Kurzweil | 8 | 0 | 0 | 0 | — | — | 0.500 | — |
| SMIC (All-In Podcast analysis) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Samsung (All-In Podcast analysis) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Samsung executives | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| San Francisco AI community | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| SoftBank | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| TSMC (All-In Podcast analysis) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Unknown | 2 | 0 | 0 | 0 | — | — | 0.500 | — |
| Unnamed friend (accountant manager) | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Unnamed frontier lab mid-level executive | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Unnamed tech CEO | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
| Zipline | 1 | 0 | 0 | 0 | — | — | 0.500 | — |
Calibration plot legend: x-axis = stated probability (mean of bin), y-axis = observed hit rate. Diagonal = perfect calibration. Green dot = within 5pp of identity. Red = overprediction (predicted higher than actual). Yellow = underprediction. Dot size scales with bin sample size.