Public Scorecard
Every prediction, published before its outcome, Brier-scored.
Every prediction is published before its outcome, with a confidence and a falsification criterion. Resolved predictions are scored with the Brier score — (outcome − confidence)², 0 is perfect, 0.25 is a coin flip — using the confidence at authoring. Backtests are method calibration on frozen historical vantage points; they are labeled, scored separately, and never marketed as predictions. Checkpoints are fast-resolving, pre-registered probes of a live prediction's mechanism; they are scored in their own lane, separately from the headline predictions.
Live predictions
| # | Prediction | Horizon | Confidence | Status |
|---|---|---|---|---|
| 1 | The model layer doesn't keep the take-rate | 2027-12-31 | 79% | open |
| 2 | The rails tax becomes a rule: a network mandate lands by the Apr-2027 cycle | 2027-06-30 | 60% | open |
| 3 | The walled garden captures the margin instead — expected MISS | 2027-12-31 | 37% | open |
| 4 | The app layer bifurcates: integrators expand margin, wrappers compress | 2027-12-31 | 70% | open |
| 5 | Frontier labs retain premium-SKU pricing power | 2027-12-31 | 60% | open |
| 6 | Model-API pure-plays suffer broad gross-margin collapse | 2027-12-31 | 38% | open |
| 7 | The carrier JV will not take control of AST SpaceMobile | 2027-06-30 | 70% | open |
| 8 | AST consolidates the neutral-host position | 2027-06-06 | 55% | open |
| 9 | The carrier JV stays multi-vendor: no satellite-operator exclusivity | 2027-06-06 | 72% | open |
| 10 | Being the structural winner does not make ASTS the trade — expected MISS | 2027-06-30 | 42% | open |
| 11 | The device layer does not capture the carrier D2D channel | 2027-06-30 | 60% | open |
| 12 | The integrated platforms do not bypass the carrier-wholesale channel | 2027-06-30 | 62% | open |
| 13 | The resilience carve-out is a recurring price floor, not a one-off | 2027-12-31 | 67% | open |
| 14 | The rent lands at the non-China cell/inverter complement, not at module assembly | 2027-12-31 | 60% | open |
| 15 | EU module manufacturing reshores and captures the rent — expected MISS | 2027-12-31 | 22% | open |
| 16 | The electrode reprices toward its value (throughput rent re-rating) | 2027-12-31 | 68% | open |
| 17 | "Value over volume" discipline holds (the producer keeps its self-imposed constraint) | 2027-12-31 | 63% | open |
| 18 | The operating inflection does not reward equity — expected MISS | 2027-12-31 | 40% | open |
| 19 | Electrical-equipment scarcity rent persists | 2027-12-31 | 78% | open |
| 20 | No clean GPU→power baton-pass by horizon | 2027-12-31 | 62% | open |
| 21 | IPPs deliver further outsized constraint-driven re-rating | 2027-12-31 | 38% | open |
Checkpoints (mechanism calibration)
| # | Prediction | Horizon | Confidence | Status |
|---|---|---|---|---|
| 1 | Oct-2026 rules cycle: first agentic compliance language appears | 2026-11-30 | 35% | open |
| 2 | Frontier flagship list prices hold through Q3'26 | 2026-09-30 | 72% | open |
| 3 | AST Q2'26 print: guidance held, no control-changing event | 2026-08-31 | 78% | open |
| 4 | GrafTech Q2'26 print: "value over volume" discipline holds | 2026-08-31 | 80% | open |
| 5 | GrafTech Q3'26 print: the H2 repricing actually shows up | 2026-11-30 | 65% | open |
| 6 | GE Vernova Q2'26 print: gas-turbine backlog holds | 2026-08-15 | 85% | open |
| 7 | Nvidia FQ2'27 print: no GPU margin compression | 2026-09-15 | 90% | open |
Backtest (method calibration)
| # | Prediction | Horizon | Confidence | Status |
|---|---|---|---|---|
| 1 | Public cloud captures net-new enterprise compute, modularizing the on-prem stack | 2016-12-31 | 66% | ✓ hit |
| 2 | The integrated hardware-IT incumbents lose independence and margin | 2016-12-31 | 60% | ✓ hit |
| 3 | The modular operator (AWS) becomes the disproportionate profit pool | 2016-12-31 | 55% | ✓ hit |
| 4 | Control: the integrated private-cloud / OpenStack alternative fails to win | 2016-12-31 | 58% | ✓ hit |
| 5 | Intel fails to build a profitable mobile-SoC business despite the CEO mandate | 2020-12-31 | 68% | ✓ hit |
| 6 | Intel loses its process-technology leadership to TSMC | 2020-12-31 | 52% | ✓ hit |
| 7 | Value migrates to the modular layer: foundry + architecture + fabless overtake Intel | 2020-12-31 | 58% | ✓ hit |
| 8 | Control: the datacenter franchise holds, where Intel's model is aligned | 2020-12-31 | 70% | ✓ hit |
| 9 | Open weights reach frontier parity and take the volume tier | 2025-12-31 | 60% | ✓ hit |
| 10 | The proprietary labs flee upmarket rather than defend the commodity tier | 2025-12-31 | 66% | ✓ hit |
| 11 | Meta remains the durable anchor of the open-weight frontier | 2025-12-31 | 55% | ✗ miss |
| 12 | Control: low-end disruption damages the proprietary incumbents | 2025-12-31 | 30% | ✗ miss |
| 13 | A hyperscaler (Amazon) secures control of MSS spectrum | 2026-09-09 | 62% | ✓ hit |
| 14 | Globalstar is acquired at a substantial premium | 2027-03-09 | 66% | ✓ hit |
| 15 | Carriers stay tenants: no MNO acquires MSS spectrum | 2026-09-09 | 73% | open |
| 16 | Control: the re-rating extends to all MSS holders (Iridium) | 2026-09-09 | 40% | open |
Resolved: 14 · mean Brier 0.156