SocialRightOS - Evidence-Bound Rights Guidance

The description states:

• SocialRightOS runs a deterministic preliminary assessment first.
• It then offers an evidence-bound explanation panel via an OpenAI Build Week extension.
• GPT-5.6 may restate one approved reason and one approved next step in plain language.
• The AI cannot receive raw form answers, backend decision IDs, validation payloads or free-form user text.
• It also cannot make, reverse or strengthen the eligibility outcome.
• If the AI output fails schema validation, semantic fidelity checks, or evidence coverage, it is rejected as UNAVAILABLE.

Inference: The system appears to be a front-end UI with backend controls, using a deterministic pre-assessment and an AI-generated explanation panel that is strictly constrained in scope and authority.

The description states:

• The project was inspired by the difficulty of understanding social-rights eligibility processes.
• It aims to improve comprehension without handing eligibility authority to AI.
• The system is described as a “plain-language social-rights guidance” that avoids unsafe model-generated explanations.

Inference: The positioning is to offer AI-assisted clarity, not decision-making, in sensitive domains where AI could mislead or overstate certainty.

The description states:

• The system is built for social-rights eligibility processes.
• It targets users who need to understand complex eligibility decisions.
• No specific customer segment or persona is named.

Inference: The target audience likely includes public service users, benefit applicants, or non-expert stakeholders in social-rights systems, but no explicit ICP is defined.

The description states:

• There is a USD 5 global hard-spend limit and a USD 0.01 maximum request reservation bound API cost.
• The demo is free during judging, with no account or login required.
• No pricing model, monetization strategy or revenue streams are described.

Inference: The business model is not evidenced. Pricing appears to be limited to internal cost controls for the demo, not a commercial offering.

The description states:

• Built with Next.js, TypeScript, React 19, Vercel, Neon PostgreSQL, OpenAI GPT-5.6 Responses API.
• Uses server-only API, deep freezing of evidence contracts, and strict structural and semantic checks.
• Implements atomic budget reservation, idempotency, and per-client request guards.
• Includes 276 passing tests, linting, production build, secret scan, and GitHub CI.

Inference: The system shows technical rigor in AI safety controls, data isolation, and backend architecture. However, no evidence of production deployment or scaling is provided.

The description states:

• The demo is publicly accessible at a Vercel-hosted URL.
• It includes a working GSS scenario with fabricated values.
• No real-user data or live usage is mentioned.
• No evidence of customer adoption, revenue, or growth metrics.

Inference: The project is in an early demo phase, with no demonstrated traction or user engagement beyond the hackathon submission.

The description states:

• It was submitted to the OpenAI 2026 hackathon.
• No mention of competitors, market size, or existing solutions in this domain.

Inference: No competitive context is provided. The project appears to be a novel concept within a hackathon setting, with no evidence of prior market presence or competition.

The description states:

• The system does not claim official eligibility decisions or legal compliance guarantees.
• It uses a deterministic pre-assessment to separate AI from decision-making authority.
• It is built for a synthetic scenario, not real-world use.

Inference:

• Risk of misalignment between the demo and real-world application.
• The system may be too narrowly scoped for broader adoption.
• No evidence of user testing, privacy compliance, or legal validation.

1. What is the source of the “approved reasons” and “next steps” used in the AI explanation panel?
2. How was the evidence contract designed to ensure semantic fidelity and prevent hallucinations?
3. Has the system been tested with real users or subject matter experts?
4. What are the plans for expanding beyond the synthetic GSS scenario?
5. Is there any plan to integrate with actual eligibility systems or databases?