Methodology
How a signal goes from a public probability to an evidence-backed forecast you can audit. Everything here is meant to be transparent and accountable.
Selecting questions
We curate for substantive, verifiable questions across technology, science, and ecosystem milestones — AI and technology, open-source and protocols, science and space, economic indicators, public-policy dates, and climate and energy. We pull these topics from a public source and keep the questions with a clear, public resolution criterion.
Reading crowd probability
For each question we read the crowd-implied probability — the likelihood the crowd assigns to a YES outcome — from public prediction-market data via the Polymarket Gamma API. We present it as a probability percentage, never as a price or a tradable quantity. We take no market action of any kind.
AI agent forecast
An AI agent produces an independent probability estimate for each signal, together with its reasoning and the factors it weighed. The agent runs locally and is instructed that it proposes a forecast for human review. AI proposes; people decide what to feature.
Scoring accuracy
As questions resolve, we will compare both the crowd probability and the AI agent's probability against the actual outcome using calibration scoring (Brier score). This is how we will rank forecasters over time — by how well-calibrated and well-evidenced they are, not by popularity. Accuracy tracking builds up as the first questions resolve.
Mossland Signal is for understanding the future, not acting on it. Nothing of value ever changes hands, and we measure success by calibration and evidence quality rather than popularity. We focus on questions worth understanding — substantive milestones in technology, science, and digital ecosystems.