Pre-Registered Forward Forecasts (May 2026)
Frozen predictions on Pew AI Trends and Pew Top Problems benchmarks before ground truth is published.
Retrospective benchmarks can always be questioned after the fact. Forward forecasts remove that ambiguity: we publish predictions before official results land.
In May 2026 we froze forecasts for two upcoming Pew releases: AI Trends (15 items) and Top Problems (13 items). Predictions were generated with the same production stack customers receive, without post-hoc tuning on those items.
Why forward benchmarks matter
Results will be updated on lewsearch.com/predictions when ground truth is available. We report misses as plainly as hits.
- Ground truth did not exist in our training or calibration pool when forecasts were locked
- Public adjudication when Pew publishes official marginals
- Clear pass/fail criteria for directional and item-level error
Product validation
Customer-facing benchmark tables and live predictions are on lewsearch.com/methodology. Due diligence materials available on request.