Data
0.0B
0
Records
Models
0.0B
0
Parameters
Platforms
0.0M
0
Requests
Tools
0
AI tools
Browse our archive of the most common polling questions asked over the last 50 years. Compare question compositions and construction across time and create a survey with established historic cross tabs per question.
Ask in plain words and get answers from the scores, scripts and shorts on record — never an invented number. Try five questions free before you sign in.
Browse all 209 partner videos in one library, each with its review thread and, where tested, its measured lift. Pick what to run from evidence instead of memory.
Post short statements and vote them agree, disagree or pass. Watch the opinion groups form in the votes and see where your coalition agrees — and splits — before you take a position.
Compare any result against 1,583 effects from campaigns' own randomized experiments (APSR 2024). See the ordinary range and tell a strong result from an average one before you scale it.
Show a random panel one of your videos or a placebo, then ask how they would vote. Measure the video's effect alone — before you buy the airtime.
Work from four desks in one shared library: Voter Freedom, Yes on 3, Unite America and PAX. Check what another desk already made or tested before you pay to shoot.
See every poll on open primaries side by side: thirteen public polls and seven official ballot results, 2014 through the 2026 Massachusetts tracking — so one favorable number can't stand in for the record.
Read a finished test as a printable document: every video's lift against its own placebo, by subgroup, with what the result can and cannot establish. The Massachusetts study's two waves are there now.
Draft and revise scripts beside the clips they become, and start the next one from the script that tested well. Not built yet.
The workspace that replaces the bare terminal for working with coding agents.
Point at anything on a page and hand its exact location to an AI.
A 0.6B bge-m3 encoder fine-tuned on Nuclear Science References — papers, nuclides and reactions — for semantic search by meaning.
A 0.1B ModernBERT reranker that scores candidate nuclear-science passages for precision retrieval on top of the encoder.
A 0.6B bge-m3 encoder fine-tuned on biomedical literature for semantic search across health and pharmaceutical datasets.
A 0.1B ModernBERT reranker for biomedical passage ranking, pairing with the Biomed Encoder in a two-stage retrieval pipeline.
A 0.1B ModernBERT reranker trained on the 44B corpus — bills, votes, filings and law — for cross-domain legislative retrieval.
A 22.7M MiniLM reranker — fast and lightweight — for low-latency legislative re-ranking where the full reranker is too slow.