What changed—and why.

September 6, 2026

Optional support now tells the whole data story.

The launch policy still allows only one reviewed hosted payment provider and one-time contributions—never a recurring subscription. Before checkout, Role by Role names that provider and links its privacy policy, terms, and the Role by Role support terms. Provider links remain unavailable until a valid provider is actually connected.

Role by Role does not receive card numbers, but a payment provider may expose a receipt record and payer details in its merchant dashboard for reconciliation, refunds, disputes, and legally required records. Checkout carries no résumé, study history, score, job match, role-proof story, or Role by Role account identifier. The external handoff also suppresses opener access and referral data.

Public production remains blocked until the operator publishes reviewed refund, dispute, contact, and transaction-retention terms. Cash App and crypto remain deferred, and the first iPhone candidate contains no support or payment path.

Hosting metadata is not called “no data”

The Privacy page now distinguishes absent product analytics from the ordinary technical request metadata Cloudflare processes to deliver and protect the preview. It also names the disconnected Cloudflare account stack, publishes its implemented cleanup defaults as unapproved technical defaults, and identifies email-delivery records, operational logs, recovery copies, and legal holds that are outside the Worker’s D1 cleanup. Production still requires a verified processor inventory, final retention, operator access controls, rights process, and qualified legal review.

Application links remain a separate choice

The job privacy boundary now follows the learner through the external handoff: opening a verified listing is optional, suppresses Role by Role referral data and opener access, and subjects any later information the learner submits to the destination’s own terms. Role by Role does not apply or accept those terms for the learner.

Stale job results now fail closed

Job search and its freshness status now enforce the same 36-hour boundary. Search stays unavailable before a completed refresh, after a failed latest run, or when the last successful refresh is stale. Returned results are also limited by their own last-seen time and carry the freshness state used for the decision. Partial source coverage is disclosed rather than presented as a complete market view.

Stronger search-result identity

Every indexable route now has a unique page title and description, a canonical-matching social URL, and complete social-image metadata. The public-build audit rejects drift across those signals and rejects static content pages without exactly one primary heading. These checks improve clarity for people and crawlers without keyword stuffing or invented authority.

Mortgage scope stated accurately

The methodology and Terms now distinguish the open, source-backed mortgage-loan-originator guide and self-recorded Living Career Map from a test-prep product. Role by Role still offers no SAFE MLO question bank, approved course, or MLO curriculum. Insurance and cybersecurity remain research previews only.

Résumé workflow repaired with browser evidence

A production-like browser test found two defects that isolated analyzer tests had missed: the deferred scoring module used a duplicated asset path, and PDF.js could fall back incorrectly when started inside the disposable parser Worker. Both root causes were fixed. The same synthetic, non-personal PDF now completes extraction, editable-profile review, qualified scoring, separate revision consent, item-by-item decisions, download, and deletion on the deployed preview.

The browser made no request when the file was selected and consent was still unchecked. After consent, it fetched only the three same-origin static analyzer resources with body-free GET requests. Common contact details did not enter the editable profile or browser storage, and the original file reference and filename were released after extraction. This is controlled Chromium evidence—not a claim about physical devices, Safari, assistive technology, or every real résumé.

Larger mobile controls and stronger reflow checks

The mobile menu, track selector, target-date control, theme control, Job Copilot fields, and secondary actions now meet Role by Role’s 44-pixel product target in the measured mobile flow. A new exact-artifact check also covers 320-pixel reflow, landscape orientation, drawer completion, dialog focus containment and return, theme persistence, and a service-worker-controlled offline reload. Physical-device and screen-reader review remain launch requirements.

Published September 6, 2026 · Product, discovery, privacy, and support-policy record

September 5, 2026

Five securities tracks in the iPhone source.

The native SwiftUI candidate now lets a learner select SIE, Series 7, Series 63, Series 65, or Series 66. Every track includes ten evenly distributed recall questions and its complete original appointment-length form with an exam-specific clock. The generated native set contains the same 525 questions and exact content fingerprint as the audited web banks.

This remains source-preview evidence, not a shipped App Store claim. The independent securities-expert review is pending, and the app still requires a matching full-Xcode build and unit-test run, real-device and VoiceOver QA, signing, screenshots, TestFlight evidence, and final App Store review. The first native candidate remains offline and contains no account, analytics, jobs, résumé processing, payment, donation, or MLO test-prep path.

Readable details on smaller screens

A rendered mobile and tablet review found that several source notes, safety disclosures, résumé explanations, and job-freshness labels inherited dense 8–10 pixel desktop microtype. Consequential disclosures now retain a 12-pixel floor at every viewport, while explanatory copy and controls receive a larger mobile scale. The release audit enforces those floors; 390-, 768-, and 1440-pixel browser checks show no horizontal overflow or console warnings.

A clearer first decision

The dashboard now starts with the broader Role by Role promise instead of assuming every visitor is only there to get licensed. Three reversible choices—practice the selected credential, map a career, or strengthen an application—state what happens next and disclose the practice time before the learner commits. The page explicitly says that switching paths remains available, and an automated conversion-path check rejects missing routes, hidden effort, manipulative urgency, or a return to license-only framing.

The map becomes the mark

The identity's familiar yellow step-route now carries three meaningful map states: mint for an acknowledged start or completed prior step, blue for skills or evidence being built, and yellow for the learner-controlled next direction. The same first-party vector system generates the favicon, PWA icons, App Store icon, native icon, and social card, with an automated audit for exact exports, dimensions, and opacity. The states describe a path; they do not claim an exam pass, license, or job outcome.

Published September 5, 2026 · Product and native-source record

September 4, 2026

Clearer evidence, safer boundaries.

A mistake-review guide that teaches the decision

A new Field Note turns missed SIE questions into a seven-step repair: reconstruct the choice, classify the error, verify the rule, build a contrast, retrieve without choices, and return after a delay. It treats correct guesses as unresolved, links directly to the local mistake queue, and distinguishes research-informed advice from a validated SIE intervention or pass prediction.

A second, source-backed career path

The mortgage loan originator guide now separates covered-institution federal registration from state licensing, explains the federal education and test floor, links to current NMLS state checklists, and carries the current 120-item, 190-minute SAFE MLO test structure. It explicitly says Role by Role does not yet offer an MLO question bank, course, or approval predictor. Its job-copilot link prefills only the target-role field; résumé processing still requires a separate file choice and affirmative consent.

A usable mortgage pathway and terminology guide

The in-app MLO Living Career Map now branches by employer type into a state-licensed or covered-institution registration route. Every milestone is self-recorded, state approval is distinguished from merely applying, quiz activity cannot complete a credential, and optional account sync accepts only the route and allowlisted milestone booleans. A new source-backed Field Note answers the narrower “NMLS license versus registration” question and links readers into the map without claiming that Role by Role offers MLO test prep.

Dark mode across the whole public site

The learner’s light or dark appearance choice now follows them across the app, career maps, SIE page, Field Notes, community principles, methodology, privacy, terms, and updates. A shared tested policy preserves the rest of the local progress record, recovers safely from malformed or legacy state, synchronizes open tabs, and exposes the current choice through an accessible pressed-state control. Versioned styles prevent an older service-worker cache from masking the new theme on first navigation.

One route through the working experience

A compact Choose, Learn, Practice, Prove, and Apply navigator now carries the Living Career Map language across the working views. It identifies the current stage in text and assistive-technology state, keeps prior and future stages visually distinct without claiming that navigation completed a credential, works by keyboard, respects reduced-motion preferences, and reflows without horizontal scrolling on narrow screens.

Methodology scope correction

The methodology page and related career and study-guide copy previously understated the practice-form coverage after appointment-length forms had shipped. They now accurately list one original appointment-length practice form for each supported exam while preserving the important boundary that these are not actual exam questions, psychometrically validated predictions, a comprehensive course, or a source of multiple independent full-length forms.

A direct answer to Series 7 sponsorship

A new source-backed Field Note answers whether the public can self-enroll for Series 7: no, FINRA requires broker-dealer association for qualification exams such as Series 7. The guide separates that rule from the independently available SIE, explains Form U4 and the firm enrollment sequence, presents FINRA FIND as an optional directory rather than a placement promise, and gives candidates practical steps that do not overstate exam or license status.

Role-specific licensing routes

The Investment Specialist map now preserves three listing-dependent pathways found in recent SoFi job-listing snapshots: Series 7 + 66, Series 7 + 65, and Series 7 + 63 + 65. The page labels these as location- and date-dependent and directs readers to verify the current opening and official regulators.

Résumé review stays local

PDF and DOCX extraction, the readiness rubric, revision suggestions, and job-term matching now run in the active browser tab. The learner must review the extracted profile, authorize revisions separately, and approve or reject every proposed line. The score is explicitly not an employer score or offer probability. Match explanations now separate shared evidence from requirements not found in the reviewed profile, treat an unspecified location as neutral, and explain that extraction or wording may account for a missing term. Automated safeguards also confirm that changing only names, ages, or universities does not change ranking. The active workspace drops the original filename after extraction, and parser failures do not send raw details to the console.

Job-source provenance

New job feeds now fail closed. A source cannot be enabled without documented authorization, its canonical HTTPS URL, a terms-review date, and a removal contact. Public results carry the source URL and review date so a listing can be checked at its origin.

Optional support guardrails

Support remains optional and cannot change access, scores, matching, rankings, response speed, or service quality. The web launch now accepts exactly one reviewed Stripe or PayPal provider per deployment, names it at the decision point, and links its official privacy policy, terms, and Role by Role's support terms before checkout. Mixed providers, Cash App, crypto, lookalike URLs, and every iOS support-payment destination fail closed. A service-worker version update prevents older disclosure code from surviving an otherwise fresh release.

Mobile navigation and motion

The mobile drawer now removes off-screen controls from the keyboard order, moves focus into the active destination when opened, returns focus after Escape, and places focus on the destination heading after navigation. View transitions keep content visible from their first frame and still respect reduced-motion preferences.

Faster, verifiable releases

The main application code now loads as a deferred, versioned asset, reducing the decoded HTML document from about 203 KB to roughly 123 KB while keeping the complete initial decoded set under 273 KB. Every public build also carries a deterministic artifact fingerprint, and each production Worker can report the exact opaque release ID running behind its health check.

Planning dates are not appointments

The SIE flow now labels dates as either a private planning target or an official appointment the learner says was already scheduled with Prometric. It no longer implies that saving a date enrolls or books the learner, and it explains that FINRA's 120-calendar-day window begins after enrollment approval.

A source-backed SIE appointment decision

A new Field Note answers when to schedule the SIE without turning one practice percentage into a prediction. Its four-part decision checks enrollment, knowledge, timing, and real-life logistics; a seven-day observation plan links the baseline to mistake repair. Current FINRA and Prometric sources support the dated $100 fee, 120-day approved-enrollment window, online or test-center delivery, and cancellation boundary. A dedicated release audit rejects missing sources, drifted facts, unsupported readiness claims, or broken discovery links.

SIE appointment-length practice and corrections

The SIE form now contains 80 original items: the first 75 follow FINRA’s exact 12 / 33 / 23 / 7 scored-section allocation and the final five are transparent mixed-section practice standing in for the exam’s unidentified pretest items. Every added item maps to the current October 2025 FINRA outline plus at least one claim-level regulator or first-party source. The review also narrowed a closed-end-fund question to publicly traded funds, made private-securities approval explicitly dependent on selling compensation, and updated the Regulation S-P explanation for the current amendments. Independent securities-expert approval remains pending.

A direct SIE discovery-to-practice route

A new crawlable SIE page explains the 80-item appointment, scored blueprint, practice modes, scheduling boundary, primary sources, and independent-review status before inviting someone into the app. Its call to action opens the SIE practice track directly—even when another exam was previously selected—and its full page and styles are available in the offline shell.

A substantive offline iPhone practice path

The native iOS source now uses the same exact-fingerprint 80-question SIE bank as the web product. It offers a mixed Quick 10 and a Full 80 with a 105-minute clock that accounts for time spent in the background. A deterministic generator rejects fingerprint drift before Swift verification, while the native release gate independently checks all 80 items. The app remains offline, contains no web view or payment path, and still requires independent subject-matter review, a full-Xcode build, real-device accessibility testing, signing, and TestFlight evidence.

Full-length Series 7 practice

Series 7 now includes one original 130-question practice form with the official 225-minute limit. Its first 125 questions follow FINRA’s 9 / 11 / 91 / 14 scored-function blueprint exactly; the final five are clearly treated as mixed-function practice because the real exam’s pretest items are unidentified. Every added question includes a rationale, outline reference, the current FINRA outline, and a distinct claim-level primary-authority mapping. Answer positions are balanced 33 / 33 / 32 / 32. The source audit also excludes the superseded pattern-day-trader count and $25,000 minimum-equity rules after FINRA’s June 4, 2026 margin amendments. Independent securities-expert approval remains pending.

A crawlable Series 7 practice route

A dedicated Series 7 page now explains the 130-item appointment, 225-minute limit, sponsor requirement, SIE corequisite, four scored job functions, current reviewed fee, and the product’s one-form limitation before opening the exact Series 7 track. Four visible and schema-linked FINRA sources support the page. A dedicated release audit rejects format, blueprint, source, sponsorship, outcome-claim, or discovery-link drift.

Full-length Series 63 practice

Series 63 now includes one original 65-question practice form. Its first 60 questions follow NASAA’s current eight-domain scored blueprint exactly, while the final five are clearly treated as mixed-domain practice because candidates cannot know which questions NASAA designates for pretesting. The answer-key positions are balanced, every added question includes a rationale and outline reference, and each maps to the NASAA guide plus claim-level primary authority. Independent securities-expert approval remains pending.

Full-length Series 65 practice

Series 65 now includes one original 140-question practice form with the official 180-minute limit. Its first 130 questions follow NASAA’s 20 / 32 / 39 / 39 scored-domain blueprint exactly; the final ten are clearly treated as mixed-domain practice because the real exam’s pretest items are unidentified. Every added question includes a rationale, outline reference, the NASAA guide, and a claim-level primary-authority mapping. Answer positions are exactly balanced at 35 each, and independent securities-expert approval remains pending.

Full-length Series 66 practice

Series 66 now includes one original 110-question practice form with the official 150-minute limit. Its first 100 questions follow NASAA’s 8 / 17 / 30 / 45 scored-domain blueprint exactly; the final ten are clearly treated as mixed-domain practice because the real exam’s pretest items are unidentified. Every added question includes a rationale, outline reference, the NASAA guide, and a claim-level primary-authority mapping. Answer positions are balanced within one question, and independent securities-expert approval remains pending.

Crawlable Series 63, 65, and 66 practice routes

Three dedicated pages now connect each shipped state-law practice form to its distinct search intent. They state the current appointment composition, time, passing threshold, reviewed fee, exact scored blueprint, enrollment path, and registration limit using visible NASAA and FINRA sources. Series 66 explicitly preserves its Series 7 corequisite, while Series 65 avoids inventing one. Exact FAQ/schema parity, direct practice links, the single-form limitation, sitemap inclusion, and offline access are enforced by a dedicated release audit.

Role proof stays truthful and private

The Living Career Map now continues beyond licensing into three locally stored Situation–Action–Result stories: customer clarity, sound judgment, and ownership/recovery. A story becomes “ready to rehearse” only after useful detail and an explicit truth-and-confidentiality review. Stories are excluded from account sync and are never presented as employer verification, a hiring score, or an outcome prediction.

Published September 4, 2026 · Product and source-review record

What belongs here.

Role by Role records changes that affect a learner's understanding, privacy, safety, or ability to act. Each entry gets a real date and plain-language explanation. Cosmetic edits and routine maintenance stay out of this record.

Study articles are available in the Field Notes library and its Atom feed. Methods, limitations, and primary references live on the methodology page.