Three steps to your first cascade. Then a weekly habit.

Chainge turns a structured exposure profile of your company into a weekly briefing of cross-domain cascades that touch your business. The profile itself runs in your own AI tool, so step one needs no install. You feed it your structural shape, Chainge pressure-tests it, and from then on the global signal corpus runs against you every week.

Below: how to use it, what the engine does behind the briefing, and where the data comes from.

Profile. Diagnostic. Cascades.

The user journey is three steps. The first runs in your own AI session. The second is a back-and-forth where Chainge challenges your profile. The third is the weekly briefing that lands.

research preview · for now there is no signup form. You reach Chainge directly, and the diagnostic and briefing arrive through a direct conversation. The flow below is what it looks like; once we open up to subscribers, the steps stay the same, the contact path changes.

Behind the briefing: four stages, and a loop.

What you read in the briefing is the output of an AI pipeline that reads public information, classifies it, surfaces cross-industry bridges, and verifies every claim against primary sources. Open methodology. Auditable cost. Every chain carries provenance.

Four stages run every week. The skeptic pass at the end is the part most pipelines skip; it kills more chains than it promotes, which is the point. Then the loop: every prediction gets a resolution marker, and the next week's briefing leads with the verdicts that landed.

Primary sources, not headlines about headlines.

Reasoning from structured primary sources is the credibility gap that separates Chainge from any wrapper around a chatbot. Below: every source the engine currently reads, plus what each one is doing in the chain.

• RSS news and policy feeds live

  Roughly ten editorial sources covering EU policy, central banks, climate and energy, technology, and academic economics: Politico EU, Euractiv, European Commission press, ECB, the IEA, Bruegel, the WEF, TechCrunch, plus arXiv for academic economics. Hourly polling. Surfaces the editorial signal layer of what is being talked about, and via arXiv the methodological shifts that move into policy and industry analysis with a lag.

• GDELT live

  The Global Database of Events, Language and Tone. Tracks news across one hundred-plus languages with fifteen-minute updates. The geopolitical pressure gauge: where in the world is escalating, where is de-escalating, where is the tone of coverage shifting before any single outlet writes the explainer. Read at scale, not headline by headline.

• FRED live

  Federal Reserve Economic Data. Currently fourteen macroeconomic time series in active rotation: nine macro indicators (rates, employment, inflation, industrial production) and five commodities (Brent, natural gas, copper, wheat, the dollar index). The macro spine the engine cross-checks every commodity and rates claim against.

• EUR-Lex CELLAR live since 2026-05-13

  The official EU legal database, queried directly via the publications.europa.eu REST API. Live access to every directive, regulation, implementing act, and Court of Justice ruling. When a chain claims "the CBAM implementing act applies to your imports," the chain links to the actual act, not a news summary of it.

• Polymarket live

  A prediction market: traders put money on the probability of specific future events. Used here as a continuous, crowd-priced probability layer for events the engine is tracking (an election outcome, a Fed cut, a Hormuz closure). When the engine says 24% odds, Polymarket usually has a market that says something nearby; the gap is interesting.

• Planned next: ACLED, World Bank, CME fed funds, encyclopedic context queued

  ACLED for conflict ground-truth, the World Bank development indicators for cross-country comparisons, CME fed funds futures for rate-path expectations, and a Wikipedia and Britannica context layer for entity and historical grounding. Each one closes a known gap. None is in production yet.

Every chain in the briefing carries a source chip naming which of these contributed the underlying claim. If you can't link back to a primary source, it doesn't ship as a fact, it ships as a model estimate.

Six questions readers ask first.

Six things readers ask before going further. Same answers I'd give in conversation.

• How is this different from McKinsey or any consultancy? Q01

  The qualitative quant; elite foresight, democratized. Consultancies charge millions for cross-domain synthesis built on largely public information. Chainge runs the same synthesis at $5-7 per weekly cycle, with every claim cited and every prediction tracked to a verdict. The methodology is auditable, the cost is auditable, the verdict is timestamped. McKinsey is a black box you trust because of brand. Chainge is a glass box you trust because of receipts.

• How is this different from ChatGPT, news, and my own analysis? Q02

  ChatGPT reads headlines about headlines. Chainge reads primary sources: EUR-Lex for regulations, FRED for macro time-series, GDELT for geopolitical events, Polymarket for crowd-priced probabilities, arXiv for emerging research. Every chain carries a source chip. Every prediction carries a verdict horizon. The reasoning is something you could do yourself. Doing it systematically, every week, against the global signal corpus, with a published verdict cycle, is what you'd otherwise pay a research desk for.

• Where's the track record? When do I see verdicts? Q03

  Every prediction carries a verdict horizon, capped at 90 days. Short-horizon predictions (a regulatory deadline this month, a commodity threshold this quarter) resolve faster. Long-horizon predictions resolve at the cap. First public verdicts land from W27 through W32 of 2026 onward, as horizons arrive. No retroactive editing. No narration after the fact. Until the first verdicts arrive, the chains, the methodology, and the data sources are all visible. Come back when the verdict cycle opens.

• What happens when a chain is wrong? Q04

  Every chain ships with resolution markers: regulatory deadlines, commodity price thresholds, prediction-market levels. When a marker resolves, a deterministic function returns right, partial, wrong, or unclear. Not judged by an AI. Wrong chains are published wrong. The next week's briefing leads with the verdicts that landed. Two consecutive wrong calls on the same domain trigger a methodology review, not a quiet edit. The failures stay visible. That's how the methodology compounds.

• What if a cascade is correct but I can't act on it in my market? Q05

  Chainge surfaces cross-domain links; it doesn't promise every link is actionable in every operator's market. Some chains resolve into exposures you can see but can't trade against. The briefing flags these conclusions explicitly as structurally unactionable. The value is still real: a known exposure you can't hedge is a known exposure you can communicate to your board, your customers, or your insurers.

• How much does it cost? When can I get it? Q06

  Chainge is in research preview. No signup form yet. Access is a direct conversation with Niels. The steady-state target is $5-7 per weekly run. That's a hard ceiling the engine is built around, not an aspiration. Open-tier pricing comes after the first verdict cycle resolves. Track record sets the price, not the pitch.

See it work, end to end.

Anne runs Brabant Advanced Materials, a specialty chemicals SME in Eindhoven. We took her through profile ingest, the diagnostic, and the five-hop cascade that came out, including the regulation the diagnostic caught that she hadn't listed and the German lobbying fight that decides her outcome.

→ see Anne's worked example

→ talk to Niels about a profile