A public history, without pretending it is complete
This record begins on 10 October 2026. Earlier commits were not reconstructed, so it does not claim to cover every change since each tool was created. Going forward, updates that can change results or materially change the stated limitations belong here.
Final accessibility, performance and security pass
Improved focus movement in the step-by-step Calibration flow, narrowed page-specific CSS loading and added conservative response security headers without blocking the intentionally embeddable tools.
Result impact: Presentation, keyboard navigation and delivery hardening only; calculator formulas, research thresholds and saved plans did not change.
Privacy-thresholded research insights published
Added bilingual public status pages and machine-readable JSON for real, one-dimensional Calibration aggregates. The page never substitutes demo data and suppresses a whole dimension unless at least 100 eligible snapshots exist overall and every exhaustive cell has at least 20.
Result impact: Expired raw contributions can now be converted atomically into all-time category counts without retaining contribution IDs, dates, deletion credentials, filters or cross-tabs. No calculator or Calibration signal logic changed.
Consent-based research infrastructure prepared
Added a fail-closed submission, deletion and 30-day retention path for narrow categorical Calibration contributions. Collection activates only after its database, firewall rate limit, deletion secret and scheduled cleanup all pass readiness checks.
Result impact: Local plan saving still sends nothing. Optional sharing remains unavailable on deployments without the complete protected backend, and the interface must never report a contribution as received unless the server returns a deletion receipt.
Local plans separated from optional research sharing
Published the proposed research data dictionary, one-time consent boundary, 30-day raw-data limit and deletion constraints. Local saving still sends nothing, and research collection remains closed.
Result impact: Privacy and storage behavior were clarified. Clearing a Five Paths baseline now preserves local Calibration drafts and snapshots; no research data is collected.
Calibration method card and learning hub published
Documented the four-signal reflection method, data boundaries, implementation checks and primary references, then connected two in-depth bilingual decision guides.
Result impact: Documentation and navigation only; saved plans, signal logic and calculator results did not change.
FIRE Calibration published
Replaced the score-led Calm FIRE Planner with a bilingual, local-first reflection across numbers, capacity, meaning and resilience.
Result impact: The new tool does not produce a readiness score or forecast. It records separate planning signals so users can keep, adjust or revisit a plan without treating change as failure.
Stable citations and public registries
Added a versioned citation block to every tool plus public methodology and change-history pages.
Result impact: Presentation and traceability only; calculator results did not change.
Netherlands and Belgium policy snapshot corrected
Changed the Belgian basic gain allowance default from €20,000 combined to €10,000 per taxpayer, clarified the proposed Dutch 2028 regime and added primary government and parliamentary sources.
Result impact: The Belgian scenario can show higher modelled tax than the prior default. The tool remains a simplified scenario, not a personal tax calculation.
Belgium preset source and omissions clarified
Replaced a secondary policy link with the Belgian Chamber law file and made the omitted allowance explicit in the preset description.
Result impact: Source and limitation wording changed; the broad preset formula did not.
Model cards and data lineage published
Published purpose, inputs, transformations, exclusions, update policy and source lineage for six higher-complexity analytical tools.
Result impact: Documentation only; calculator results did not change.
Evidence registry introduced
Assigned every public tool a model version, review date, evidence class, freshness label and decision-sensitivity label.
Result impact: Documentation only; this is the beginning of the public history and earlier changes were not reconstructed.
Found a problem?
Use the issue-report link on the relevant tool page and include the inputs, expected result, actual result and reproducible URL. Errors that can change a decision receive priority and are recorded here after correction.