Full transparency on source selection, bias scoring, story clustering, framing analysis, blindspot detection, prediction market integration, and independent voice matching. Nothing hidden.
Last updated: March 2026
Sources are selected based on publication frequency, audience reach, and political diversity. The goal is to cover the full ideological spectrum with credible, consistently publishing outlets — not to maximize volume.
Each source is rated on a five-point bias scale: -2 (far-left), -1 (left), 0 (center), +1 (right), +2 (far-right). Ratings are informed by established third-party bias assessments (AllSides, Ad Fontes, Media Bias/Fact Check) and reviewed periodically. No source is rated by a single data point — ratings reflect consensus across multiple assessments where available.
Independent journalists are tracked separately from institutional outlets. They are not given a single bias rating — instead, their coverage is surfaced by topic area, allowing Signal/noise to identify when an independent voice breaks a story before mainstream outlets cover it.
Each refresh cycle ingests hundreds of new headlines across all monitored feeds. A semantic similarity model groups headlines that are covering the same underlying event — not just the same keyword, but the same occurrence.
This deduplication step is critical: without it, a single story would appear dozens of times under different framings. After clustering, each group represents one real-world event as covered by multiple outlets. The cluster preserves all source diversity — every outlet that covered it, regardless of bias tier.
Stories that appear across multiple bias tiers are prioritized in the briefing. A story that only appears in far-left outlets but nowhere else is flagged differently than a story with broad cross-spectrum coverage.
Once a story cluster is assembled, AI reads the actual article content — not just headlines — from sources across the bias spectrum and identifies how different outlets are positioning the same event.
Framing analysis looks for: what each outlet emphasizes in the first paragraph; what language choices (loaded vs. neutral terminology) differ between outlets; what context or background some outlets include and others omit; and what the "implied conclusion" of each outlet's coverage appears to be.
The output is not a political judgment — it is an empirical description of coverage differences. "Left-leaning outlets focused on X, right-leaning outlets focused on Y" describes what happened, not which framing is correct.
A blindspot is defined as a story that has significant coverage in one or more bias tiers but near-zero coverage in another. For example: a story covered by 12 left-leaning outlets and 0 right-leaning outlets, or vice versa.
Signal/noise calculates coverage distribution across all five bias tiers for each story. When the distribution is heavily skewed — above a configurable threshold — the story is flagged as a potential blindspot and surfaced with that label in the briefing.
Blindspot detection does not imply that the uncovered side was wrong to ignore the story. It surfaces the gap so you can decide whether it matters.
Signal/noise integrates live market data from Polymarket — a decentralized prediction market platform where participants stake real money on the outcome of real-world events.
For each major story in the briefing, Signal/noise attempts to match it to an active Polymarket contract. When a match is found, the current market probability is shown alongside the framing analysis — giving you a real-time measure of what informed (financially-incentivized) participants believe is most likely to happen.
Market matching is probabilistic and not always available. Prediction markets are most informative when they diverge significantly from the dominant media narrative — that divergence is surfaced explicitly when it occurs.
Signal/noise monitors 40+ independent journalists and Substack authors who are not affiliated with major institutional outlets. These voices are particularly valuable for stories that mainstream coverage underplays or where institutional incentives may shape editorial choices.
When an independent journalist publishes on a topic that matches a clustered story, their coverage is surfaced in the briefing — often hours before institutional outlets pick up the same angle. The independent voice is labeled and attributed so you can follow up directly.