Raxx · internal docs

internal · gated

Project Memory: Sentiment Roadmap — Shapes 1, 2, and 3

Status: operator-locked roadmap 2026-06-05. This file is a living project-memory reference for the three-shape sentiment feature architecture. Updates are appended with date stamps.

Shape definitions

Shape 1 — Personal Sentiment Journal (in design, 2026-06-05) - User logs their own emotion at structured points: - Pre-trade label: Bullish / Bearish / Neutral / High-Uncertainty - Post-trade outcome label: Disciplined / Confident / Panicked / Surprised / etc. (final taxonomy: data-scientist decision) - Backtest queryable across those labels — user can see "how did my iron condors perform when I marked Disciplined vs Panicked?" - All data is user-generated, user-specific, surfaced only to the user who entered it. - No aggregation across users. - No scoring or recommendation.

Shape 2 — Aggregated / Anonymized Cohort Patterns (future) - Sentiment data aggregated across users to surface population-level patterns. - Not in design as of 2026-06-05. - Higher legal exposure than Shape 1 (aggregation across users raises different Advisers Act and privacy-law questions).

Shape 3 — Scoring or Recommendation (future) - Raxx uses sentiment data to score users or recommend strategies/labels. - Not in design as of 2026-06-05. - Highest legal exposure of the three shapes.

Research completed 2026-06-05 by business-legal-researcher agent. Full research: docs/legal/research/shape-1-personal-sentiment-journal-compliance-2026-06-05.md

The condition is that the user-asserts-first principle holds. If the platform auto- classifies or suggests labels in a way that pre-selects answers, exposure rises toward Shapes 2 and 3.

Advisers Act §202(a)(11): Shape 1 does not satisfy Prong A (advice concerning securities) when the user asserts their own label and Raxx returns only filtered historical data. No recommendation is made. The "inanimate tool" framing from the P6 attorney question applies. Formal written opinion recommended before launch.

S7-18-22 (PDA): Withdrawn June 12, 2025. Not a current obligation. Residual risk is Advisers Act §206 anti-fraud if any label-suggestion UI nudges behavior.

FINRA 2210 / 2360: Raxx is not a FINRA member. These rules do not apply directly.

Reg BI: N/A. Raxx is not a broker-dealer.

FTC §5 / PA UTPCPL: Exposure arises from framing, not from the feature itself. "Trading psychology" or "helps you trade better" copy creates a substantiation requirement Raxx cannot currently meet. Neutral self-reflection framing ("see how your trades went when you felt X") avoids the substantiation issue.

CCPA / CPRA: Emotion labels are treated conservatively as potentially Sensitive Personal Information (psychological trends / inferences category, §1798.140(ae)). Requires in-flow notice before first use, right-to-limit-use disclosure, and retention/ deletion pathway. Privacy-policy update required before Shape 1 ships.

Claim 2 verdict: "Backtest results must be auth-gated, never on marketing pages." — TRUE

Sep. 2023 SEC enforcement sweep (nine investment advisers) confirmed that hypothetical/ backtested performance on public websites violates Marketing Rule 206(4)-1. Although Marketing Rule does not directly bind Raxx (not a registered IA), the same prohibition applies via FTC §5 on public pages. Auth-gate is load-bearing for both the legal posture and the brand trust promise ("your data, for your eyes").

User-own-data results are substantively different from general hypothetical performance, but not safe on public marketing pages — a reasonable consumer visiting a marketing page cannot distinguish "this user's actual results" from "this is what the product would do." FTC §5 substantiation applies regardless.

Claim 3 verdict: "User always asserts first, Raxx never auto-classifies." — LOAD-BEARING

Design erosion scenarios and their legal consequences:

Design choice Risk level Consequence
Blank field, user types or selects Clean No inference; user-asserted
Default = user's own modal prior label Low Slight bias; still user-assertion path
Default = population modal label ("most users pick Confident") Medium Platform inference applied to individual
"Suggested label" from prior behavior High Inference + potential §206 nudging exposure
Auto-tag from P/L sign or time-in-position Critical Clearly Sensitive PI inference; §206 exposure

Required disclaimer set (summary)

  1. Pre-trade label: "Your self-assessment — Raxx does not suggest or interpret your emotional state. For your personal review only."

  2. Post-trade label: "Your reflection — not Raxx's evaluation. Not used to score, rank, or recommend."

  3. Backtest-filter results page (top of page, non-dismissible): "Your historical results, filtered by your labels. Raxx did not classify these trades. Past outcomes in your own history do not predict future performance. This is not investment advice."

  4. Privacy in-flow notice (first use, one-time, acknowledgment required): Labels are stored privately, linked to individual trades, visible only to you, subject to deletion at privacy@raxx.app, not used to compare you to other users or generate recommendations.

Privacy-policy delta required before Shape 1 ships

Attorney questions added (Section Q)

Nine questions added to docs/business/questions-for-attorney.md, Section Q: - Q1: §202(a)(11) Shape 1 scope confirmation - Q2: auto-classification line-drawing - Q3: Marketing Rule inapplicability + FTC §5 standard - Q4: FINRA non-applicability confirmation - Q5: §206 anti-fraud label-suggestion design - Q6: Endorsement Guides / testimonial threshold for screenshots - Q7 (privacy): CPRA SPI classification — user-asserted labels - Q8 (privacy): in-flow consent adequacy - Q9 (privacy): GDPR Art. 17 relocating user

Estimated marginal cost — Shape 1 opinion

Research not yet conducted. Both are higher-exposure shapes.

Shape 2 (aggregated cohort patterns) will require Advisers Act §202(a)(11) analysis on the theory that showing "traders who feel Disciplined outperform" is market commentary, and CCPA §1798.100 et seq. analysis on using inferences from user data to surface population-level claims. Flag as needing BLR research before Shape 2 design begins.

Shape 3 (scoring / recommendation) likely triggers §202(a)(11) registration analysis at a minimum. Do not design Shape 3 without a formal opinion in hand.

Open decisions

Last updated: 2026-06-05 by business-legal-researcher agent. Research basis: docs/legal/research/shape-1-personal-sentiment-journal-compliance-2026-06-05.md