What is Relocation Atlas?

Moving countries is not one decision. A visa can let someone enter without permitting their real work, supporting their family, or leading to the future they want.

Relocation Atlas turns those trade-offs into a plan a household can inspect. It starts with the outcome — temporary stay, renewable residence, or a documented citizenship path — then keeps legal blockers separate from preferences. Budget, work, family, cities, confidence, source dates, and unknowns stay visible all the way from the first calculation to a portable My Plan.

The current beta compares 40 routes across 35 countries and is optimized for Russian passport holders, whose consular access, work authorization, and route availability can differ from generic immigration advice. Profiles stay in the browser. Atlas has no account and sends no questionnaire answers to a server.

Inspiration

Relocation advice often ends where the difficult part begins: “Here are five countries you might like.”

For a family, the real questions are more concrete. Can we legally do our work? What happens to a spouse or child? Does this status lead anywhere? What would make the answer change? Which fact is official, which is a planning estimate, and which is still unknown?

We built Atlas because uncertainty should be part of the interface, not hidden behind a recommendation.

What it does

  1. Builds a local household profile around outcome, budget, work, family, geography, and time horizon.
  2. Applies hard legal constraints before ranking softer preferences.
  3. Explains one leader and three distinct trajectories instead of returning an opaque top-three list.
  4. Lets the user compare routes, test one changed scenario, and inspect official sources, confidence, checked dates, work rights, family effects, and unknowns.
  5. Turns 2–3 user-selected finalists into My Plan, with anchor/fallback roles, city choices, separate readiness gates, a verification board, and a local JSON export.

The dataset currently contains 40 routes, 35 countries, 50 city planning profiles, 63 route-source links, 11 official-source work profiles, and an initial child-education layer for the United Kingdom, Germany, and Portugal.

The final product layer

The July 20–21 product work turned Atlas from a route comparison into a complete decision workflow:

  • the result now explains one leader and three different trajectory roles while preserving hard blockers and strict-presence constraints;
  • My Plan carries the profile, selected comparison, blockers, unknowns, checked dates, official URLs, and user verification progress;
  • What changes the result? tests one alternative budget, presence, work, or family assumption without mutating the real profile;
  • the child layer separates dependent status, a child's own education route, the accompanying parent, and access to nursery or school;
  • users assign anchor, fallback, or research roles to their own finalists, select dated-dataset cities, and review five independent readiness gates;
  • a reversible English presentation and anonymous judge demo make the complete path testable without personal data;
  • v0.12.0 covers the entire English product surface — including what-if, all 40 routes, Visas, Countries, Data, drawers, errors, and recovery states;
  • a 1,168-string locale validator blocks publication if untranslated user-facing copy returns;
  • stable sticky/anchor geometry, low-cost transform motion, reduced-motion support, and keyboard focus make the long experience calmer on desktop and mobile.

How we built it

Codex with GPT-5.6 was the primary product, engineering, research, and QA environment throughout the build window.

Codex provided the agentic workspace: it inspected the repository, ran research and validation workflows, edited code, coordinated GitHub Issues, tested the browser, diagnosed failures, and verified releases. GPT-5.6 powered the long-horizon reasoning and code generation inside the primary build thread.

Together they helped us:

  • turn official-source research into a normalized route/country/city/work/family schema;
  • separate visa family from legal outcome and blockers from comparative fit;
  • implement a bilingual, local-first application in dependency-free HTML, CSS, and JavaScript;
  • design profile migrations, explicit stale-result behavior, what-if isolation, finalist roles, and portable plan export;
  • build deterministic schema and English-locale validators plus a privacy gate that publishes only the interface and dated derived dataset;
  • test desktop/mobile behavior, keyboard focus, route drawers, import/export, cache safety, and exact production artifacts.

The public app intentionally has no runtime AI dependency. GPT-5.6's meaningful role is the auditable build workflow that created, challenged, and verified the decision system — not a decorative chatbot added to the final screen.

Key decisions

Outcome before destination. Atlas starts with temporary stay, residence, or citizenship path, not a country quiz.

Visa type is not legal outcome. Nomad, talent, education, family, investment, and work are separate from what a status can ultimately unlock.

A score cannot erase a stop sign. Unauthorized work, insufficient resources, nationality restrictions, physical-presence conflicts, and unverified facts stay visible.

The user makes the final choice. Atlas can narrow the field, but only the user assigns an anchor or fallback. Those roles never rewrite the legal data or ranking.

Privacy is architectural. The profile and exported plan remain local; the public build excludes personal intake, evidence, internal status, and research notes.

Challenges

The hardest problem was preventing a ranking from creating false confidence. A single average cannot safely combine legal eligibility, family consequences, work rights, costs, and incomplete evidence.

We also had to make a dense decision journey coherent on mobile, preserve causality when a profile changes, migrate saved data across schema versions, and stop GitHub Pages or browser caches from mixing incompatible releases.

Accomplishments we are proud of

  • A complete consumer journey from anonymous profile to an inspectable, portable plan.
  • A non-trivial local decision engine with hard blockers and no backend or account.
  • Visible source dates, confidence, unknowns, and planning exclusions instead of false precision.
  • A judge-safe public artifact that is independently runnable while the private evidence room stays private.
  • Deterministic validation, repeatable publication, complete bilingual presentation, reduced-motion support, and accessible keyboard behavior.

What we learned

High-stakes products need an information architecture for uncertainty. A source link is not enough; a user also needs to know when it was checked, which passport it applies to, what outcome it unlocks, and what remains unresolved.

We also learned that local-first software can still feel personal. Removing the backend reduced privacy risk and made the judging experience faster and easier to audit.

What is next

  • Research complete nationality layers for Ukrainian and Belarusian passport holders.
  • Deepen dedicated employment and self-employment routes.
  • Replace broad city planning ranges with household-specific cost baskets for finalists.
  • Add a source-update workflow that preserves review dates and change history.
  • Test Atlas with relocation communities and accessibility users before expanding coverage.

Try it

Open the live demo and select Load judge demo. No account, API key, or personal data is required. The anonymous scenario calculates automatically; judges can inspect the questionnaire, explainable result, what-if controls, comparison, route sources, and My Plan export.

Disclosure

Relocation Atlas was created during the OpenAI Build Week submission period with Codex and GPT-5.6. Immigration facts are dated and linked where available. Atlas is an informational comparison and planning tool, not legal advice or a probability of approval.

Built With

Share this project:

Updates