Ghost Core

The company appears to be a solo project named Ghost Core, self-described as a personal AI system designed to build a user-approved cognitive core, validate deductions before saving them, and explain why each answer is truly personal. The author states that the system separates memory into inspectable records with provenance, lifecycle state, sensitivity, and user approval, allowing users to review, approve, edit, reject, or defer AI-generated deductions.

What changed: This project represents an extension of prior research into a "personal cognitive mirror" into a testable prototype focused on consent, transparency, and user control. It was built as part of the OpenAI 2026 hackathon submission.

The single most important open question: Is there any evidence that this system has been tested with real users beyond the author's own use, or whether it has moved beyond a proof-of-concept stage?

Analysis basis: This report is based solely on the self-reported project description provided by the author. No external verification, traction data, revenue figures, customer names, or third-party sources are available.

• The description states that Ghost Core is a Python and Streamlit application with local persistent storage.
• It uses GPT-5.6, Gemini, and Codex for its functionality.
• It implements a system where:
  • Memory is separated into inspectable records with provenance, lifecycle state, and sensitivity.
  • Users can intentionally save direct statements.
  • AI-generated deductions are reviewed before becoming active.
  • Users can approve, edit, reject, or defer those deductions.
  • Users can inspect, deactivate, reactivate, or permanently delete memories.
  • The system shows which records influenced an AI response.
  • Sensitive memories can be kept out of a request unless explicitly selected.

Inference: Based on the description, Ghost Core is a prototype that allows users to maintain control over how their personal data is used by AI systems. It is not a commercial product but a demonstration of a concept.

• The author states that most AI systems can remember information about a user, but memory is not identity.
• Ghost Core aims to be a transparent, user-controlled cognitive and relational mirror, where:
  • Memories, interpretations, and deductions remain visible, correctable, reversible, and owned by the person they describe.
  • No deduction becomes identity without an explicit user decision.

Claim: The system is positioned as a way to avoid AI silently deciding what a person is, instead giving users control over their own cognitive representation.

• Not evidenced. The description does not identify specific customer segments or personas.
• The author describes the system as for individuals who want to maintain control over how AI systems understand them, but no explicit target user group is named.

Absence of evidence: No indication of who the intended users are, beyond a general audience interested in responsible AI.

• Not evidenced. There is no mention of pricing, monetization strategy, or business model.
• The system appears to be a prototype built for demonstration purposes, not a commercial offering.

Absence of evidence: No indication of how the product would generate revenue or whether it is intended for sale.

• Built with:
  • Python
  • Streamlit
  • SQLite (for local persistent storage)
  • APIs: OpenAI, Codex, Google Gemini API
• The system supports:
  • Three independent and stateless Compare calls.
  • A Transparent Context Matrix showing what each response receives.
  • Separate Council adviser and synthesis stages.
  • Inspectable memory provenance and lifecycle controls.
  • Validation and safety checks around selected and sensitive records.

Inference: The architecture is designed to be auditable, testable, and transparent. It uses local storage and explicit governance rules to manage personal data.

• Not evidenced. There is no mention of:
  • Customers
  • Revenue
  • Usage metrics
  • Product adoption
  • Market testing or feedback

Absence of evidence: No signs of traction, product maturity, or real-world deployment beyond the prototype.

• Not evidenced. The description does not reference existing competitors or similar products.
• It is unclear whether other systems exist that attempt to offer user-controlled AI memory or personalization.

Absence of evidence: No competitive landscape or comparison with existing tools is provided.

• Solo project: Only one team member (Silvia Contarelli) is listed.
• Prototype only: The system is described as a prototype built for a hackathon, not a commercial product.
• No user testing: No evidence of real-world use or feedback from users beyond the author.
• Limited scope: The system appears to be a demonstration of principles rather than a scalable solution.
• Unverified claims: The description makes strong claims about control and transparency but lacks evidence of implementation effectiveness.

Inference: The project is in an early stage, with no clear path to commercialization or user adoption.

1. What specific user feedback has been gathered during development?
2. Has the system been tested with more than one user?
3. How does the system handle edge cases or ambiguous inputs?
4. Is there a plan for scaling beyond local storage and single-user use?
5. Are there any technical limitations in implementing full encryption or portability?
6. What are the long-term goals for this project beyond the hackathon?