Future Language
Archive · v1A preserved experiment in learning weather representations. Explore its six-hour forecasts, recorded outcomes and comparisons with simpler reference models.
Archived · 2026-10-10
Administrative retirement: this exploratory approach did not demonstrate a predictive advantage over its continuous-input baseline.
Collection and recurring research reviews have stopped. Data, frozen forecasts, reference models and verification tools are preserved.
Calls still marked pending are unresolved at administrative closure. Collection has stopped; no future settlement is promised. Historical statuses, forecasts and scores remain unchanged.
This was an exploratory allocation decision, not a registered statistical verdict. The broader OrangeKraken radio and weather services continue separately.
Forecasts you can check
20,807 observations collected · 401 forecasts · 388 resolved · 8 pending · 5 without an eligible outcome
Final retained snapshot: 2026-10-10T02:57:22.342404Z. Collection has stopped; pending calls remain unresolved at administrative closure.
Download public forecasts and outcomesModel versus baselines
Mean multiclass Brier score; lower is better. Each model is scored on the same resolved forecasts. These exploratory scores are descriptive, and overlapping observations are not independent trials.
Language model
0.3829
Continuous-input baseline
0.3522
Historical-rate baseline
0.6730
Persistence baseline
0.9621
Recent frozen predictions
Six-hour temperature change from the recorded origin: cooler by more than 2°C, within ±2°C, or warmer by more than 2°C. Probabilities are frozen before the outcome.
| Station | Target (UTC) | Cooler / steady / warmer | Outcome |
|---|---|---|---|
| Miami, Florida | 2026-10-10T08:40:00Z | 49.9% / 48.4% / 1.7% | pending |
| Houston, Texas | 2026-10-10T08:45:00Z | 49.9% / 48.4% / 1.7% | pending |
| Boulder, Colorado | 2026-10-10T08:15:00Z | 47.1% / 52.3% / 0.6% | pending |
| Great Falls, Montana | 2026-10-10T08:40:00Z | 49.9% / 48.4% / 1.7% | pending |
| Miami, Florida | 2026-10-10T04:35:00Z | 67.1% / 32.1% / 0.8% | pending |
| Houston, Texas | 2026-10-10T04:35:00Z | 67.1% / 32.1% / 0.8% | pending |
| Boulder, Colorado | 2026-10-10T04:15:00Z | 92.3% / 7.3% / 0.5% | pending |
| Great Falls, Montana | 2026-10-10T04:35:00Z | 62.3% / 36.8% / 0.9% | pending |
| Miami, Florida | 2026-10-10T00:10:00Z | 85.6% / 13.9% / 0.5% | cooler |
| Houston, Texas | 2026-10-10T00:15:00Z | 46.2% / 45.4% / 8.5% | cooler |
| Boulder, Colorado | 2026-10-09T23:55:00Z | 46.2% / 45.4% / 8.5% | steady |
| Great Falls, Montana | 2026-10-10T00:15:00Z | 46.2% / 45.4% / 8.5% | steady |
| Miami, Florida | 2026-10-09T16:30:00Z | 2.3% / 14.8% / 82.8% | warmer |
| Houston, Texas | 2026-10-09T16:35:00Z | 3.2% / 17.8% / 79.0% | warmer |
| Boulder, Colorado | 2026-10-09T16:15:00Z | 3.2% / 17.8% / 79.0% | warmer |
| Great Falls, Montana | 2026-10-09T16:30:00Z | 3.2% / 17.8% / 79.0% | warmer |
| Miami, Florida | 2026-10-09T07:05:00Z | 49.9% / 48.7% / 1.4% | steady |
| Houston, Texas | 2026-10-09T06:55:00Z | 46.3% / 53.0% / 0.7% | cooler |
| Boulder, Colorado | 2026-10-09T06:35:00Z | 66.9% / 32.3% / 0.8% | cooler |
| Great Falls, Montana | 2026-10-09T07:00:00Z | 66.9% / 32.3% / 0.8% | cooler |
| Miami, Florida | 2026-10-09T00:40:00Z | 92.5% / 7.0% / 0.5% | cooler |
| Houston, Texas | 2026-10-09T00:40:00Z | 44.5% / 46.5% / 9.1% | cooler |
| Boulder, Colorado | 2026-10-09T00:15:00Z | 44.5% / 46.5% / 9.1% | cooler |
| Great Falls, Montana | 2026-10-09T00:40:00Z | 44.5% / 46.5% / 9.1% | steady |
| Miami, Florida | 2026-10-08T16:35:00Z | 2.5% / 15.6% / 81.9% | warmer |
| Houston, Texas | 2026-10-08T16:35:00Z | 3.4% / 18.7% / 77.9% | warmer |
| Boulder, Colorado | 2026-10-08T16:15:00Z | 3.4% / 18.7% / 77.9% | warmer |
| Great Falls, Montana | 2026-10-08T16:35:00Z | 3.4% / 18.7% / 77.9% | warmer |
| Boulder, Colorado | 2026-10-08T05:55:00Z | 66.3% / 32.8% / 0.9% | cooler |
| Great Falls, Montana | 2026-10-08T06:25:00Z | 66.3% / 32.8% / 0.9% | cooler |
| Miami, Florida | 2026-10-08T02:15:00Z | 58.5% / 40.3% / 1.2% | cooler |
| Houston, Texas | 2026-10-08T02:25:00Z | 92.4% / 7.1% / 0.5% | cooler |
Public results, private research
The public record includes frozen probabilities, observed outcomes, baseline scores, coverage and version commitments. Training datasets, learned representations, model weights and detailed research protocols are retained privately.
Downloads let you check forecast scores. They do not reproduce the private model or independently prove when a forecast was issued. This experiment has not established forecasting superiority.
Evidence commitments
Method: d1ee8844e213743f4f60e334cccbc53ac89145f13f3c69d322ad4423f158452a
Private evidence: 351aa2d8dbcc05137cea187adf70a19410d0422c8eb3bac232598c6a4f8b68d5
Public ledger: 09ae8ddbe3674e10149ac6fd92cd4b31e8bf73aa855fdc2634b3e3f385ca0aaa