Attorney review packet: first-run funnel copy and /tools copy (card #4835)

Status: research-only. This document does NOT constitute legal advice. Before any held string ships or any funnel flag flips, consult a securities/investment-adviser regulatory attorney licensed in Pennsylvania (and federal practice) for the advice-posture and hold-list items, a consumer-protection/advertising attorney for the subscription-disclosure items, and a privacy attorney for the privacy-notice items. Last updated: 2026-10-31 UTC. Snapshot: origin/develop at 78b417d24. Sources as of that date; verify freshness. Delivery: NOT delivered. No counsel was contacted and no engagement was opened. Delivery goes to the engagement tracked in #3141 when counsel is available (new engagements are on hold).

Card: #4835 (epic #4818, funnel) and the /tools review queued by epic #4819.

Appendices (the inventories, one row per string). Each row carries an ID, a key (appendix A) or a file:line location (appendix B), the string and its guardrail tags; the source file, owner, route, flags and status of a row are stated once in the header of its group, so they are recorded at group granularity, not per row (the held-strings table in appendix A is the exception and carries a file per row):

• docs/legal/research/funnel-copy-inventory-2026-10-31.md (appendix A, 480 rendered or pre-staged rows plus 9 held strings; the held journal line also appears in group A7 as PAY-059, marked HELD and not counted)
• docs/legal/research/tools-copy-inventory-2026-10-31.md (appendix B, 592 rows plus pre-staged proposals for three pages)

TL;DR

Nothing is counsel-cleared: of the funnel strings on develop, 8 are held with the counsel-hold-4835 tag and one more retention line is held by the lapse audit, none of the 9 renders, and every other funnel string renders behind default-OFF flags while the /tools strings carry no flag and go live on the next promotion of develop to main. The decisions staged for counsel are the four hold categories (edge, "your plan", leak-to-rule, numeric goal), the subscription-disclosure copy, two open privacy-notice points, and the advice-adjacent wording on three calculators, as 34 numbered questions in section 9. The web trial and its reminder email (#4831), the personal-rule store (#4828), the watchlist (#4829) and three calculator pages (#4842, #4843, #4844) are not on develop, so their copy is pre-staged or absent here and must be inventoried again when each lands.

1. What is in this packet and what is missing

1.1 Covered (on develop at 78b417d24)

1.2 Not on develop (what is missing and what it changes)

The task for this card named #4842, #4843 and #4844 as dependencies that did not land. The check against develop found three further gaps that matter for counsel. Everything below is either pre-staged from an open PR or the design stage (clearly labelled in the appendices) or absent.

2. Context counsel needs

These statements are what the code and the decisions say today. Each has the file that backs it; none is a legal conclusion.

1. Personalization uses only the user's own inputs. The Plan screen is a pure function of the answers held in the visitor's browser tab (lib/plan/planModel.ts, header comment: "The Plan is an ECHO"). Every value shown is a value the visitor chose, or the registry default a skipped question stores, which is labelled "Default setting - change anytime." The five journal prompts are picked by the style answer id and nothing is generated (journalTemplateFor). A skipped benchmark stays empty; there is no default and no suggestion.
2. Tickers are never proposed. The quiz offers a fixed list of supported instrument classes (US stocks and ETFs, options) from lib/traderProfileRegistry.ts; no screen has a ticker or symbol field, and the privacy text says no ticker symbols and no free text are stored. /tools examples use the placeholder XYZ and scripts/lint-tools-copy.js fails the build on a list of real tickers. The one planned surface where a customer types symbols is the watchlist (#4829), which is not on develop.
3. Execution stays deterministic. No funnel screen reads, creates or changes a trade order (the paywall and confirmation call the billing endpoints, which handle subscriptions only). The epic decision is that the personal rule is an advisory prompt only and never blocks, delays or changes an order (#4828 acceptance criteria require a test that no order path reads it); that store is not on develop. The Wheel Structure Engine (#4286) is out of scope here and has its own counsel-gated language track.
4. No AI-generated content is shown. A word-boundary search of the funnel libraries, components and routes for OpenAI, Anthropic, LLM, GPT and generative returned nothing; the design rules say the same (docs/design/funnel-v2/README.md, "no AI text").
5. Audience and geography. The funnel is for public signups; the closed beta is unchanged. EU/EEA and Quebec visitors are geoblocked at signup, and on /tools the signup call to action is hidden and analytics are suppressed for those visitors and for an unknown region (lib/tools/region.ts).
6. The company. The footer line reads "MooseQuest LLC dba Raxx" (PRM-024). Raxx is positioned as a rules-and-record tool, not an adviser; the existing analysis is in docs/business/legal/2026-07-01-securities-attorney-question-list.md, docs/business/raxx-marketing-compliance-brief-2026-05-27.md and docs/legal/research/getraxx-differentiation-copy-review-2026-06-04.md. This packet does not restate it. The journal prompts on the Plan screen are the Shape 1 journal (docs/legal/research/shape-1-personal-sentiment-journal-compliance-2026-06-05.md).
7. Stage. Funnel copy ships behind default-OFF flags; /tools ships with no flag. The operator decision of 2026-10-29 is that /tools publishes without waiting for counsel and that review runs in parallel; the same decision keeps the four funnel hold categories held until counsel clears them.

3. Hold list

Status vocabulary: HELD (tagged in code, not rendered, blocks the flag); RENDERED-NEUTRAL (neutral replacement is on screen behind a default-OFF flag, not cleared); CLEARED (none: no sign-off is recorded for any string). Held strings carry the tag counsel-hold-4835; they are exported only for tests and no component imports them (verified by search at the snapshot; tests PromiseCopy4825, PlanCopy4827, PlanScreen4827, ConfirmationCopy4834 prove none renders).

3.1 Held strings (not rendered)

HELD-06 to HELD-08 are stand-ins that name their category; the code comment in lib/plan/planCopy.ts says the real held strings come from this packet and replace them. This packet does not invent held wording. Replacing the stand-ins is a code change for the card that records counsel's answer.

3.2 Rendered neutral wording that sits on a hold category

These are on screen (behind flags) in place of the held wording. They are listed so counsel clears or rejects them, not only the held text.

The quiz rows in the last two lines render on onboarding_quiz_v1 alone, not on onboarding_plan_v1; section 6.1 records what that means for questions A3 and A5.

4. Guardrail map

The codes are the "Touches" column of both appendices. The column is a mechanical pattern match, a triage aid and not a legal conclusion; a dash means no pattern matched, and counsel reviews every string regardless.

Design rules that are not copy but touch the same screens: hide unavailable features, never gray them; Pro is preselected but "Continue with Free" has equal weight (ADR-0152 section 2.5).

4.1 Mechanical scan of the shipped strings

A scan was run over every row of the two inventories, held rows excluded (the 9 held strings, including PAY-059, which is HELD-09 listed in its code position). Two banned-word lists exist, one per surface, and each was run over the surface it was written for. They differ on purpose: the funnel list bans "target", "win rate" and "must", which are input names and validation wording on /tools.

The funnel list in full (case-insensitive; the exact patterns are in check.py): will, edge, leak, leaks, biggest mistake, guarantee (so guaranteed too), optimize or optimise, win rate, recommended for you, best ticker or tickers, AI pick or picks, forecast, projected, projection, expected, predict (so prediction too), should, must, your plan, upgrade, target (so targets too); and the names Stripe, Alpaca, Postmark, PostHog, Apple, Schwab, Interactive Brokers, Tradier, Polygon, Massive, Intrinio, Cloudflare, Infisical, Sentry.

The /tools lint list in full is RULES in the script. In outline: about 40 broker, payment-processor and data-vendor names; advice, promise and prediction wording (recommended for you, guarantee and its forms, beat the market, you will, you'll, most traders, expected, predict and its forms, chance of, probability of, blow up, safe and its forms, best, optimal, good, bad, fear, greed, panic, forecast and its forms, annualized, should); tier and price names (Pro, Pro+, Founder or Founders, pricing); PDT and pattern day trader; and 14 real tickers. One phrase is allowed: the mandated H2 "How many shares should I buy?". The lint does not match "prediction" (its pattern is the whole word "predict" with an optional s or ed), "win rate", "target" or "must".

Counts at the snapshot (appendix A 480 rows, appendix B 592 rows):

What the funnel hits are:

• "your plan" (3 rows). EML-004 ("Review or cancel your plan"), EML-007 (the email button label) and BIL-002 ("There is nothing to cancel on your plan right now."). The funnel list bans the phrase because "your plan" framing is a hold category. These rows use "plan" as a billing noun, not in the trading sense, and all three are pre-staged from open PRs (groups A14 and A15). They belong to the G1p set that question A2 asks counsel to decide (listed in section 3.2); until counsel answers A2 they are neither cleared nor treated as held.
• "Apple" (8 rows found by the scan, group A12, plus a ninth string). Apple is named only in the platform-required subscription wording, the documented exception to the no-processor-name rule (ADR-0152 section 2.7). The scan finds 8 rows: IOS-035, IOS-036, IOS-052, IOS-053, IOS-056, IOS-057, IOS-059 and IOS-060. A ninth customer string also names Apple: IOS-061, the terms text for the no-trial case, which is the IOS-060 text minus the sentence "The free trial lasts 7 days." and so still contains "Apple Account" twice. Appendix A stores IOS-061 as a description of that difference, not as its full text, so a word scan over the rows cannot flag it. So "no vendor name" is true of every funnel string except these 9 iOS strings (8 rows the scan finds and IOS-061). Question B7 asks counsel to confirm the exception and lists all 9.
• "must" (1 row, IOS-068). The iOS purchase error "You must be signed in to your Raxx account to make a purchase." states a sign-in requirement and is not a trading instruction. The funnel list bans "must", so it is reported here; the packet does not rewrite it.

Why the /tools hits under the funnel list are acceptable on /tools (each is covered by a question where counsel should look):

• "target" (13 rows). "Target price", the label of a price the visitor types on the position size page, and the R multiple "at target" worked out from it (REG-002, PSZ-007, PSZ-013, PSZ-015, PSZ-021, PSZ-025, PSZ-051, PSZ-066, PSZ-072, PSZ-077, PSZS-035, PSZS-049, PSZS-056). It is an input name, not a goal or return target, which is why guardrail G1 is not used on /tools.
• "win rate" (22 rows). The name of an input the visitor types on the risk of ruin page, and the method text that echoes "the win rate you entered" (REG-014, REG-016, REG-017, ROR-005, ROR-012, ROR-013, ROR-021, ROR-028, ROR-030, ROR-033, ROR-038, ROR-039, RORS-007, RORS-016, RORS-022, RORS-026, RORS-032, RORS-036, RORS-046, RORS-054, RORS-062, RORS-064). It is the visitor's own number, not a statement about win rates in general.
• "must" (4 rows). The validation message for strike order on a spread, "For a bull call spread the long call strike must be below the short call strike." and its bear put twin (OPT-073, OPT-074, OPTS-050, OPTS-051). It states a condition the arithmetic needs, shown as an error on the visitor's own entry, not an instruction to trade.
• "should" (1 row, PSZ-010). The mandated H2 "How many shares should I buy?" (epic #4819), the one phrase the lint allows. Question E1.
• "predict" (3 rows, ROR-001, ROR-031, RORS-052). Each is a negation: "not a prediction of results", and the method paragraph "it is not a prediction of results". Question E2.

The scan does not replace review: it cannot see meaning, and several of the questions below concern strings that pass it.

5. Copy inventory summary

Row counts are in the appendices. Owner, flags and status are repeated in each group header there.

Gap for PM: neither the epic nor the code names a copy owner for /tools (the funnel names marketing-strategist in ADR-0152 section 2.7). A named owner will be needed to receive counsel's rewrites.

6. Flag gating (acceptance criterion 5)

At the snapshot no onboarding_* flag is in the prod or staging build lists, so every funnel route answers 404 for customers. This packet changes no flag.

6.1 The quiz flag and questions A3 and A5

onboarding_quiz_v1 alone controls the quiz screens. lib/funnelGate.ts gates /start/quiz/* and the signed-in quiz steps in /wizard on that flag (PUBLIC_SCREEN_FLAGS.quiz = [QUIZ_FLAG]; funnelLive). The Plan flag is not consulted for them. Those screens render:

• the habit question and its five options (QZ-008 to QZ-010, QZ-032 to QZ-037), the leak-to-rule category and the subject of question A3;
• the benchmark question, the number field and the units (QZ-011 to QZ-013, QZ-038 to QZ-046), the numeric-goal category and the subject of question A5 (a) and (b).

What the record says about gating them:

• The 2026-10-29 decision (epic #4818, card #4835) keeps strings in the four hold categories held until counsel clears them. Card #4835's last acceptance criterion names onboarding_plan_v1 and onboarding_paywall_v1 as the flags that stay OFF until sign-off is recorded. It does not name the quiz flag.
• The onboarding_quiz_v1 comment in backend_v2/api/feature_flags.yaml lists three prerequisites (#4837, #4824, public signup that can enrol an account) and notes that counsel decides the 30-day notice. It names no #4835 prerequisite. The onboarding_plan_v1 comment names #4835 and also requires the quiz flag's own prerequisites.
• The design register (docs/design/funnel-v2/README.md, "Copy register") says rendered copy is neutral and that only the held strings wait for #4835.

So, on the record as written, A3 and A5 do not gate the quiz flag: the held wording is not on the quiz screens and the neutral wording is. That is a reading of the card text. Nothing records that the decision weighed the habit and benchmark questions specifically, so this packet does not treat it as settled, and an operator who relies on section 6 alone should not flip onboarding_quiz_v1 believing A3 and A5 have been considered.

Open decision for Kristerpher, to be recorded on #4835 before onboarding_quiz_v1 is flipped; one of:

1. The quiz may ship ahead of A3 and A5. The neutral wording renders on the quiz flag alone and counsel's answers arrive later (A6 then covers any rewording, F2 any later review).
2. A3 and A5 gate the quiz flag as well. Then onboarding_quiz_v1 needs a #4835 prerequisite line in feature_flags.yaml (a separate code-side change, not made here) and the quiz row above adds A3 and A5 to counsel's conditions.

7. Claims to verify (what each notice says versus the code that backs it)

Counsel is asked to review the wording; engineering has the code. Where a row says "open PR", the claim cannot be confirmed from develop.

Other accuracy points counsel may want to know about (each describes a feature that is not on develop yet, and each sits behind a flag whose checklist already requires the feature first):

• Promise step 2 and step 3 describe the personal-rule reminder and a comparison with the rule and benchmark (PRM-020, PRM-023). The rule store is #4828; the comparison tile is a second PR on #4827.
• The Plan states "The comparison appears here once trades are recorded." (PLN-023), which promises the same tile.
• The paywall promises a Day-5 email and a billing-page cancel (PAY-023, PAY-029); both are in open PRs.

8. Facts with citations (legal framework counsel will apply)

This section lists the primary sources the questions rely on. It states what each says; it does not apply it to Raxx.

• Investment adviser definition. 15 U.S.C. 80b-2(a)(11): any person who, for compensation, engages in the business of advising others, directly or through publications or writings, as to the value of securities or the advisability of investing in, purchasing or selling securities. Exclusions include the publisher exclusion in (D). Text: Cornell LII, 15 U.S.C. 80b-2; controlling interpretation cited in the existing analysis: SEC Release IA-1092.
• Online negative option sales (ROSCA). 15 U.S.C. 8403: a seller must clearly and conspicuously disclose all material terms before obtaining billing information, obtain the consumer's express informed consent before charging, and provide simple mechanisms to stop recurring charges. Text: Cornell LII, 15 U.S.C. 8403; the FTC statute page is FTC, Restore Online Shoppers' Confidence Act (its summary covers the post-transaction third-party provision).
• FTC "click to cancel" rule. The FTC's rule was vacated by the Eighth Circuit on 2025-07-08 (Custom Communications, Inc. v. FTC, No. 24-3137), per secondary sources; the opinion itself was not retrieved: Covington, Eighth Circuit vacates FTC negative option rule. The FTC's own page records a 2026 advance notice of proposed rulemaking (published 2026-03-13, docket FTC-2026-0265) and the 2026-02-12 notice revising the rule: FTC, Negative Option Rule. Whether any rule is in force is a question for counsel; ROSCA and Section 5 apply regardless.
• California automatic renewal law. Cal. Bus. & Prof. Code 17602, as amended for contracts entered on or after 2025-07-01: offer terms presented clearly and conspicuously in visual proximity before the agreement is fulfilled, including the post-trial price; affirmative consent; a retainable acknowledgment with cancellation instructions; a pre-renewal notice 3 to 21 days before the end of a free trial or promotional price that lasts more than 31 days; and online cancellation through a prominently located direct link or button when sign-up was online. Text mirror: California Public Law, section 17602; the official source is leginfo.legislature.ca.gov, which refused an automated fetch, so verify against it. A 7-day trial is below the 31-day threshold. Other states have automatic renewal laws; the list that matters for nationwide public signups was not researched here (unsourced, confirm with a consumer-protection attorney).
• Deception standard. FTC Policy Statement on Deception (1983): a representation, omission or practice likely to mislead the consumer acting reasonably in the circumstances, and material. The FTC's staff report on deceptive design is Bringing Dark Patterns to Light (September 2022) (cited by title and date only; the report text was not summarized here).
• Endorsements and typical results. FTC Endorsement Guides, what people are asking: an ad featuring above-average results needs proof they are typical or a clear statement of generally expected results. There are no testimonials or results claims in the funnel or on /tools (no hits in the scan).
• Email. FTC CAN-SPAM compliance guide: a message is transactional or relationship only if its content falls in a listed category, which includes notices about a change in the terms or features of a subscription or account; such messages carry accurate routing information and are otherwise exempt from most provisions; commercial messages need a valid physical postal address and an opt-out. The category boundaries are read narrowly.
• Options legend. FINRA Rule 2220, paragraph (d)(2)(A)(vii): no member or associated person may use an options communication that fails to include a warning to the effect that options are not suitable for all investors. The rule binds members and associated persons; the existing analysis concludes Raxx is not a member (docs/legal/research/getraxx-differentiation-copy-review-2026-06-04.md, section 1.3). Cited as an analogue for the /tools disclaimer, not as a requirement.
• App Store subscriptions. App Store Review Guidelines, 3.1.2: before asking a customer to subscribe, clearly describe what the user gets for the price, and communicate the requirements in Schedule 2 of the Apple Developer Program License Agreement (not retrieved here); free trials are offered through App Store Connect.
• Raxx's own prior positions. docs/business/raxx-marketing-compliance-brief-2026-05-27.md (sections 1 to 3 and 5): the line Raxx cannot cross, risky phrasings, pre-approved candidate phrasings (all still subject to attorney review), and the footer disclaimer draft.

8.1 Jurisdiction flags

9. Questions for counsel

Priority: P1 gates a flag flip or a hold; P2 runs in parallel and does not block. Pro: SEC = securities/investment-adviser regulatory attorney; CONS = consumer-protection/advertising attorney (not the CPA, who is a separate pro); PRIV = privacy attorney.

A. Hold categories (gate onboarding_plan_v1; A3 and A5 also concern strings on the quiz flag, see section 6.1)

B. Subscription and billing disclosures (gate onboarding_paywall_v1)

C. Advice posture on funnel screens

D. Privacy notices (belong with the privacy and terms engagement, #3149, per the note on #3141)

E. /tools (published without waiting for counsel; review runs in parallel)

F. Process

Count: 34 numbered questions (A 7, B 7, C 3, D 6, E 9, F 2).

10. Counsel sign-off record (complete when counsel responds)

Record this on #4835 before any held string ships. The opinion or memo itself is stored in the private legal Drive folder, not in the repo (policy on #3141).

Delivery log (this card opens no engagement):

11. Staleness check and how to regenerate

Both inventories are snapshots of 78b417d24. The copy modules are the single source for the funnel (lib/promise, lib/quiz, lib/plan, lib/paywall, lib/confirmation); a few strings live inline and are listed in groups A10 to A15. To see whether a re-review is needed, run from a clean checkout:

git fetch origin
git diff --stat 78b417d24 origin/develop -- \
  frontend/raxx-next/lib/promise frontend/raxx-next/lib/quiz frontend/raxx-next/lib/plan \
  frontend/raxx-next/lib/paywall frontend/raxx-next/lib/confirmation \
  frontend/raxx-next/components/promise frontend/raxx-next/components/quiz \
  frontend/raxx-next/components/plan frontend/raxx-next/components/paywall \
  frontend/raxx-next/components/confirmation frontend/raxx-next/components/settings/SetupAnswersTab.tsx \
  frontend/raxx-next/components/wizard/steps frontend/raxx-next/components/wizard/WizardShell.tsx \
  frontend/raxx-next/components/onboarding 'frontend/raxx-next/app/(public)/start' \
  frontend/raxx-next/lib/tools frontend/raxx-next/components/tools frontend/raxx-next/app/tools \
  ios/Raxx/Features/Billing docs/legal/policies \
  backend_v2/api/templates/email/billing

Any output means a string may have changed since this packet. To rebuild the tables: the funnel modules are plain TypeScript exports with no framework imports, so they can be loaded under any TypeScript loader and walked (functions called with sample arguments); the /tools static text comes from extractCopy in frontend/raxx-next/scripts/lint-tools-copy.js (it returns JSX text with HTML entities undecoded, such as ', ' and →; decode them before listing, or the string is dropped or misread), and the run-time sentences come from calling evaluatePositionSize, computePayoff, describeResult, simulateRuin and resultSentence. The run-time sentences in appendix B are sampled (each page's golden examples and the edge cases that trigger a note, an error or a limit), not exhaustive; the appendix's method notes say which cases were added after review and why. The generators used for this snapshot were throwaway scripts and are not committed.

12. Timing

• No statutory deadline was found for any item. Dates that matter: /tools has no flag, so it goes live on the next promotion of develop to main (the last promotion, 70fe475, did not include it); the quiz flag may not flip before the notice question D2 is answered; California's amended renewal law applies to contracts entered on or after 2025-07-01; the FTC's 2026 rulemaking docket (FTC-2026-0265) is open and its comment window was not checked.
• Counsel availability is the constraint named in the card's risks. Held strings are isolated in the *_HELD_COPY exports so the cleared remainder can ship behind the flags.

13. Pre-delivery checklist (internal; remove before sending)

• Re-run the staleness check in section 11 and refresh the inventories if any listed path changed.
• Engineering confirms what consumes the stored risk answer (question C2) and what the hosted checkout does at card entry (question B2).
• Re-inventory the web trial copy, the reminder email and the billing notices from develop once PRs 4887, 4891 and 4892 merge; replace groups A14 and A15 (pre-staged) with the shipped text.
• Re-inventory /tools when #4842, #4843 and #4844 land (appendix B group B7 is a design-stage proposal only).
• PM assigns a copy owner for /tools (section 5).
• Send nothing until the operator lifts the hold on new engagements; delivery goes to the #3141 engagement and is logged in section 10.

Sources

Primary and repository sources used for facts:

• https://www.law.cornell.edu/uscode/text/15/80b-2 (Investment Advisers Act definition and exclusions)
• https://www.sec.gov/files/rules/interp/1987/ia-1092.pdf (SEC Release IA-1092, as cited in the existing analysis; not re-fetched)
• https://www.law.cornell.edu/uscode/text/15/8403 (ROSCA, negative option marketing on the Internet)
• https://www.ftc.gov/legal-library/browse/statutes/restore-online-shoppers-confidence-act (FTC page for ROSCA)
• https://www.ftc.gov/legal-library/browse/rules/negative-option-rule (FTC page: 2024 rule, 2026 notices, docket FTC-2026-0265)
• https://www.insideprivacy.com/united-states/federal-trade-commission/eighth-circuit-vacates-ftc-negative-option-rule/ (secondary: Eighth Circuit vacatur, 2025-07-08)
• https://california.public.law/codes/business_and_professions_code_section_17602 (text mirror of Cal. B&P 17602; official source leginfo.legislature.ca.gov)
• https://www.ftc.gov/legal-library/browse/ftc-policy-statement-deception (FTC Policy Statement on Deception)
• https://www.ftc.gov/reports/bringing-dark-patterns-light (FTC report, September 2022)
• https://www.ftc.gov/business-guidance/resources/ftcs-endorsement-guides-what-people-are-asking (Endorsement Guides)
• https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business (CAN-SPAM guide)
• https://www.finra.org/rules-guidance/rulebooks/finra-rules/2220 (FINRA Rule 2220)
• https://developer.apple.com/app-store/review/guidelines/ (App Store Review Guidelines 3.1.2)

Repository sources: docs/design/funnel-v2/README.md, docs/design/seo-tools/README.md, docs/architecture/adr/0152-funnel-pricing-and-trial-decisions.md, backend_v2/api/feature_flags.yaml, docs/legal/policies/privacy-policy-v1.md and privacy-policy-v1-history.md, the copy modules named in section 5, issues #4818, #4819, #4835, #3141 and PRs #4887, #4892.

Before you act on this, consult a securities/investment-adviser regulatory attorney (Pennsylvania and federal), a consumer-protection/advertising attorney for the subscription items, and a privacy attorney for the notice items. This packet is research to prepare that conversation; it contains no advice.