Most people planning to move to another country are making one of the largest decisions of their lives on the worst information they will ever act on. The rules are real, public, and precise — and almost nobody encounters them that way. What they encounter instead is a consultant's brochure, a forum post from 2019, or a "free assessment" designed to produce a sales call. Ancieen 安心 — 安心 means "peace of mind" — is our attempt at the opposite: a mobility-intelligence engine that takes a person's real situation and a real goal, and returns ranked, cited, cross-border pathways with an official source and an as-of date on every claim. The problem with how migration information works today Three structural failures, and none of them are about effort: Advice is sold, not computed. The party telling you which route to take usually earns more from some routes than others. That is not necessarily dishonest, but it is not neutral either. Rules change violently and silently. Points thresholds, occupation lists, and investment minimums move several times a year. An article that was correct when written becomes actively harmful without changing a word. Free matchers cannot actually compute eligibility. A school-matching quiz can filter a list. It cannot evaluate whether you satisfy a points table, because it has no machine-readable rules to evaluate against. The engineering idea: a Mobility Graph The core insight is that immigration is a graph problem, and that modelling it as one makes the hardest feature free. Nodes are statuses a person can hold — Chinese citizen, Canadian study permit, post-graduation work permit, Portuguese golden visa, Grenadian citizen. Each node records what it grants: work rights, study rights, duration, renewability, and — critically — the passport mobility that comes with it. Edges are legal transitions between statuses via a specific named program. Each edge carries cost, processing time, an official source URL, and temporal validity. Conditions hang off each edge as machine-evaluable predicates, each with its own citation. One condition type changes everything: an edge can be gated on already holding a status or passport from a different country . The moment that exists, the graph stops being a set of per-country silos and becomes genuinely global. Why springboard routes fall out for free The feature we are proudest of is the one we never built. A 跳板 — a springboard route — reaches a destination by way of an easier intermediate country. The canonical example: US EB-5 investor residency runs to roughly $800,000 and years of waiting. But Grenada offers citizenship by investment, and Grenada holds an E-2 treaty with the United States. Chinese citizen → Grenadian citizen → US E-2 investor visa is dramatically cheaper and faster than the direct route. The map above shows the same shape in the product: a direct route from China plotted against a springboard through Portugal, where a Golden Visa opens onto EU residency. The engine draws both and lets you compare them on cost, time, and what each step actually grants. We did not hand-code either chain. They are simply multi-hop paths through a graph whose edges can be gated on another country's nationality or status. Every springboard route the engine surfaces is an ordinary shortest-path result over an unusual data structure. Add a country's rules to the graph and new chains appear on their own — including ones nobody on the team had thought of. This is the part that reads as creative work rather than engineering work. Choosing the right abstraction meant the valuable behaviour emerged instead of being enumerated. The guarantee: the model never invents a rule Immigration is regulated, and wrong information carries real consequences. So the language model is deliberately confined to the two jobs it is genuinely good at: Turning messy human input into structured facts. A paragraph about someone's life becomes typed fields. Narrating a result that has already been computed , with inline citations. Everything in between — every eligibility determination, every points calculation, every path search — runs in a deterministic rule engine against a versioned, cited knowledge base. A validator then drops any sentence not backed by a citation from that knowledge base. This matters more than it sounds. Ask a general-purpose chatbot whether you qualify for a visa and it will answer fluently, confidently, and without any mechanism that distinguishes a remembered rule from an invented one. Our engine can only state what it can cite, and it stamps every claim with the date the rule was verified. What it actually does for a person An honest three-way verdict. Every in-scope program returns likely eligible, almost, or not yet — including the ones you fail. Knowing what you do not qualify for is half the value. A gap list you can act on. Not "improve your English" but which specific threshold, on which specific program, unlocks which specific route. Budget-aware filtering. Enter your available funds and investment routes above it disappear. No fantasy planning around a golden visa you cannot fund. Springboard chains you would be unlikely to find manually, because finding them requires knowing two countries' rules and the treaty between them at once. Sponsor employer lookup across a register of 127,000+ licensed visa-sponsoring employers, scoped to the countries on your route. School and program matching across 200+ institutions, with tuition estimates and field-of-study rules. A rule radar that watches every source rule for change and flags the people whose plans a change affects. Depth of data The numbers we publish are deliberately unglamorous, because they are the ones that matter: Measure Current Countries and regions 19 Cited pathways 149 Sourced rules 283 Sponsor employers ingested 127,000+ Not testimonials, not user counts. Depth of verifiable data is the only thing that makes the product's central promise checkable. How it is built Three separate services, each with one job: Component Role Stack Collection pipeline Scrapes and parses official rules and registers into versioned, cited bundles with diffs between versions Python The engine Rule evaluation, multi-objective pathfinding, change radar, grounded narration Python / FastAPI The application Planner, accounts, credits, billing, and the B2B radar dashboard Next.js 14 / TypeScript The split is deliberate. Because rule data is versioned and diffed on the way in, "this rule changed on this date" is a property of the pipeline rather than something the application has to reconstruct. The radar is a consequence of the architecture, not a feature bolted onto it. Who it is for Students choosing between programs where the post-study work rules matter more than the course. Skilled workers who want to know which of their attributes is actually the binding constraint. Families weighing routes where dependants, schooling, and timelines interact. Agencies and advisers , who use the rule radar to learn a threshold moved before their clients do. What it deliberately is not Ancieen provides information, not advice . It reports what published rules say, with sources and dates, and computes which of them a set of facts satisfies. It does not tell anyone what to do, does not represent anyone, and is not a substitute for a licensed immigration professional. That boundary is a product decision as much as a legal one — the value is in showing your work, and an engine that shows its work cannot also be a black box that issues verdicts. Frequently asked questions How is this different from asking ChatGPT about visa eligibility? A general-purpose model generates a plausible answer from training data with no mechanism separating a real rule from an invented one, and no date attached. Ancieen computes eligibility in a deterministic engine against a versioned knowledge base, and every sentence that survives the validator carries a source and an as-of date. What is a springboard route? A path to a destination through an easier intermediate country, typically because the intermediate country's passport or status unlocks a program the original nationality cannot access. They arise naturally from multi-hop search across a cross-jurisdiction graph. How do you keep rules current? The collection pipeline re-fetches official sources and produces versioned bundles with diffs. Changes propagate to the engine and trigger the radar for affected users and subscribers. Is this immigration advice? No. It is information retrieval and rule computation with citations. Applications and legal representation remain the domain of licensed professionals. Which countries are covered? 19 countries and regions today, spanning traditional destinations and the smaller jurisdictions that make springboard chains work. Coverage expands as each jurisdiction's rules are sourced and cited to the same standard. Ancieen 安心 is an in-house applied-AI project; figures above reflect the knowledge base as of July 2026. Try the planner at ancieen.com/for-you , or see the agency rule radar at ancieen.com/for-agencies . For enquiries about the engineering behind it, get in touch through our consultation page.
