HEALTH A reasoning model hit 88.6% diagnostic accuracy on clinicopathological cases. In the same window, a five-hosp…+ HEALTH A reasoning model reached 88.6% exact or near-exact accuracy on clinicopathological cases.+ EDUCATION Small-scale district pilots report gains for students who previously had no outside tutoring access. POLICY Post-market monitoring standards for clinical AI tools remain unsettled as clearances accelerate. LABOR The labor market's two truths: large projected role creation and concentrated, measurable displacement. LABOR Employment for coders aged 22–25 has fallen roughly 20% against its late-2022 peak.+ LABOR New AI-adjacent skills already carry wage premiums in 1 of every 10 job postings in advanced economies. LABOR Current estimates vary 5x depending on methodology.
Sunday, July 12, 2026
TruvaceThe trace, not the pitch
The Index

What the evidence says. What the public feels.

The Sourced Indexranks AI’s measured gains and problems on verified evidence. The Public Pulse tracks what readers report, register, and debate. The Impact Ranking tracks which AI companies carry the largest visible public footprint.

Two instruments, never one number

The Sourced Index and the Public Pulse are separate products and stay that way. The Sourced Index answers: what does the strongest available evidence say AI is improving or worsening? The Public Pulse answers: what does the public feel are the biggest AI gains and harms right now? A public concern is not automatically evidence. A sourced finding is not automatically public sentiment. No view on this site collapses the two into a single truth score.

What counts as verified evidence

Peer-reviewed papers, government and regulatory reports, court and regulatory filings, academic datasets, audits from credible independent organizations, and independent journalism. Vendor self-reporting is never treated as verified evidence. Evidence strength weighs peer review and official statistics above journalism, and corroboration across independent sources above any single report.

Sourced Index scoring

Sourced Index Score = Impact 30% + Evidence Strength 25% + Scale 20% + Confidence 15% + Recency 10%. Every subscore is derived live from the stored record — the editorial severity tag and quantified magnitudes in the reading (impact), the types and count of stored sources (evidence strength), the measured reach in the sourced text (scale), corroboration, peer-reviewed share and editorial certification (confidence), and the age of the evidence (recency — recent evidence counts more, but landmark evidence never falls to zero). Every entry shows its component breakdown, not just the final number.

Public Pulse scoring

Public Pulse Score = Report Volume 30% + Sentiment Intensity 25% + Momentum 20% + Affected-Group Spread 15% + Corroboration Level 10%. The inputs are public signals on the record: registered readings, comments, and reader submissions. Corroboration measures whether public reports include links, documents, or institutional detail — it does not make a public claim verified. Because individual reading timestamps are private, momentum is derived from comment recency and publication age.

Evidence status on public items

Each Pulse entry carries one of: No sourced support yet, Partially supported, Strongly supported, Contradicted by sourced evidence, or Under investigation — set by comparing the public reading against the Sourced Index entry for the same topic.

What the scores are, and aren’t

Scores are directional instruments, not absolute truth. They update as evidence changes — the same recalibration discipline as the rest of the record, and revisions to certified records appear in each record’s public version history. Editors can override or annotate a reading only through the versioned revision process, with a stated reason. Public submissions can flag issues earlier than formal evidence; formal evidence can reveal underreported realities. The Alignment Matrix exists to show exactly that gap.

Editorial rules

Every sourced entry carries at least one stored source, a “what this doesn’t fix” limitation line, an evidence-strength reading derived from its actual sources, and a last-updated date. Every public entry carries its score breakdown, volume, momentum, corroboration level, evidence status, and explicit labeling that it reflects public feeling, not verified fact.

Recomputed live from the record · Jul 12, 2026, 12:12 PM