1. Claude Opus and GPT Sol
Claude Opus and GPT Sol provide bounded factual review where required and available. The deterministic Engine ranks, sizes and applies risk rules. AI reviewers may flag supported source or sizing defects, jointly block permitted hard-risk fresh entries, or select an offered verified source statement. They cannot choose replacement stocks, invent weights or override portfolio rules. Independent deterministic checks can establish and correct supported sizing errors.
The diagram describes roles, not a promise that both providers freshly review every output. Review scope, timing and availability vary. Missing evidence or incomplete checks can restrict publication. Neither AI review nor a completed software check establishes that every financial fact or conclusion is correct.
2. Separate research features
The current connected-account review is deterministic: no AI-provider request or extra AI charge is made. Confirming it saves a review, not an order. Research Chat uses the approved Claude model shown in the app. Company explanations use a separate Claude research workflow and may reuse a dated explanation based on shared public company evidence. A cached explanation is not a fresh analysis or current-data guarantee.
3. What reaches providers
Enabled Engine reviewers receive bounded cases, approved rules and retained public-source evidence. Research Chat sends the question, recent conversation and relevant published model, ticker or tool context to Anthropic. Watchlist and Screener requests can include selected symbols. Connected-broker holdings are not automatically attached to chat; sensitive information entered in the conversation may still be sent. Do not enter account credentials or other confidential information. See the Privacy Policy.
Anthropic and OpenAI APIs process requests under their applicable arrangements, potentially outside Taiwan. Retention settings do not establish that every provider retains nothing. Shared company explanations and private conversations have separate purposes and storage.
4. Charges and control
Included chat replies are used first. Extra paid replies require opted-in prepaid AI credit and settle at actual metered provider cost without platform markup. The selected model, estimate and user-set spending limits appear in the product. A temporary hold is not the final charge; uncertain outcomes may remain pending reconciliation. Undelivered replies are not charged. No automatic AI-credit card top-up occurs. Subscription renewal is separate. See billing terms.
5. Limitations and questions
AI output is general research information, not investment advice. AI can produce false, incomplete, stale or biased statements, unsupported citations or unsuitable suggestions. Read the evidence, dates and deterministic model result. AI cannot place brokerage orders, and its output is not a personal suitability assessment or a promise of return. Applicable consumer and financial-services obligations are not waived by calling an output AI-generated.
Equity Quant, a sole proprietorship in Taiwan, provides this service. Business address: No. 68, Zhonghua Street, Taipei, Taiwan. Contact support@equityquantai.com for AI, privacy or support questions. See the Risk Disclosure.