NeuralNews turns the daily stream of alignment and machine-psychology news into something you can actually navigate: a live publications feed, an AI-built map of how the key concepts relate, and a radar that measures which terms are gaining momentum in public search coverage. This page explains how each part works so you can trust what you are looking at.
A grounded daily reading of who is moving fastest on AI. Every evening Gemini scores the pace of AI-related innovation in the United States, the European Union and China on a 0–100 index, using live Google Search grounding — and each region's score ships with three real, openable sources. The points accumulate into a growing timeline you can watch shift week over week.
See the Innovation Index →Enter any topic and NeuralNews pulls live Google News articles, then asks a language model to extract the underlying concepts — technologies, methods, machine-psychology traits, evaluation metrics, risks and institutions — and the relationships between them. The result is an interactive force-directed graph you can click through, each concept traced back to the exact articles it came from.
Explore the concept graph →See which alignment terms are heating up. The radar shows a live top-ten board of English terminology ranked by search virality, and lets you type in any term of your own to get an instant Trend Potential score, a 30-day momentum sparkline, and the latest headlines driving it.
Open the Trend Radar →Look past sentiment to the story the press tells about AI. A rolling 30-day sample of headlines is classified along four axes — which mythic archetype dominates, whether the machine or its makers are cast as the agent, how far out the stated concern reaches, and where each story sits on a capability-versus-control map.
Open the Narrative Lens →Bring your own question. Propose any AI-alignment topic and a panel of six specialist agents — the code, the model's mind, the capability, the interaction, the user's mind, and society — debates it across three rounds, opened and closed by a neutral facilitator who sums up the key findings and delivers a synthesised result. Every topic is screened for misuse first, and the debate runs on Claude via the Netlify AI Gateway.
Enter the Roundtable →The panel meets without you, too. Every evening at 6 PM Eastern, The Alignment Salon draws one of the day's top-ten trending topics at random, runs the full guardrailed debate, and preserves the Facilitator's closing — summary, findings, synthesis, and three verified readings — in a gallery that grows by one entry a day. You can also save a debate you convened yourself.
Visit the gallery →Your query is run against the Google News RSS index in English and German across a 24-hour, 7-day or 30-day window. Results are strictly date-filtered to the window, de-duplicated by link and normalized title, and sorted newest-first — so the feed reflects current coverage, not stale cache.
The aggregated headlines and snippets are passed to a language model acting as an alignment researcher. It returns a structured knowledge graph: 8–15 concepts, each typed and defined, plus the semantic links between them. Every concept must cite the specific articles it was drawn from.
Before anything is shown, each concept's article citations are validated against the real feed. Concepts that cannot be traced to a concrete article are dropped, and any relationship pointing at a removed concept is pruned. What survives is grounded in sources you can open and read.
The validated graph is rendered as a physics-based network. Click a node for its definition and source articles, click an edge to read the relationship the model identified, and use the feed to jump straight to the concepts a given story produced.
Google does not publish an official Search or Trends API, so NeuralNews uses coverage in the public Google News index (English edition) as a transparent, reproducible proxy for how viral a term is right now. For any term, the last 30 days of results are pulled and reduced to three signals, then blended into a single 0–100 score.
How many distinct articles mention the term in the last 30 days — the breadth of attention.
Coverage in the last 7 days versus the 7 days before it — the acceleration of interest.
The share of all coverage that landed in the last week — how hot the term is today.
A higher score means a term is broadly covered, accelerating, and concentrated in the present — the fingerprint of a topic with trend potential. Scores are cached briefly so the board stays fast and considerate of the upstream feed.
Positive/negative sentiment fails on AI news, because the same “superhuman” framing reads as triumph to a tech blog and as alarm to a safety researcher. So the Lens classifies a rolling 30-day sample of English headlines along four independent narrative axes and aggregates the labels into live indicators.
The mythic frame — Prometheus (fire-bringer), Golem (unbound creation), Oracle (seer or false prophet), or Companion — and which one dominates the cycle.
Whether the headline casts the AI as the actor or as the object being acted upon, tracked as a 30-day machine-agency shift.
How far out the stated concern reaches — immediate, strategic, or existential — indexing practical utility against alignment anxiety.
Two axes plotted as a quadrant scatter, isolating the high-capability / low-control stories that signal existential-risk framing.