Wharfside Rulebook: Signed, Cited Rules for AI Diagnosis

The description states that Wharfside is a native macOS app that diagnoses container crashes using on-device AI. It uses a deterministic approach, where an Ed25519-signed rulebook governs which AI models are invoked, and many diagnoses are answered by rules alone without model calls.

It includes:

• A Swift package (wharfside-rules) containing:
  • RuleReference schema for structured citations
  • Citation lint enforced in CI
  • Forward-compatible decoding with strict linting
  • Backfill of bundled rules with runtime source links

The system is built using:

• Swift
• Ed25519 signing
• GitHub Actions
• GPT-5.6 (Codex)
• Linux and macOS environments

Inference The product appears to be a developer tool focused on AI crash diagnosis in Apple container runtimes, with emphasis on trustworthiness through provenance and deterministic behavior.

The author claims that Wharfside is built around the principle: “the model never guesses.” Rules decide what the model sees — or whether it’s invoked at all. This positions the tool as a trustworthy, deterministic diagnostic engine, especially for AI systems in containerized environments.

Key claims:

• Rules are now cited and signed, improving transparency.
• The system supports forward compatibility with older clients.
• A CI linting policy enforces citation requirements at authoring time.

There is no evidence of prior versions or positioning shifts; the project appears to be a single, self-contained development effort submitted for a hackathon.

The description does not state who the intended users are beyond the developer building it. It implies:

• Developers working with Apple’s container runtime
• Teams seeking deterministic AI crash diagnosis
• Users concerned with AI model accountability and provenance

No explicit customer segments, personas or market targeting were described.

Inference Likely targets include developers or engineers in Apple ecosystem environments who require reliable, auditable AI diagnostics for containerized applications.

There is no evidence of any business model or pricing structure. The project is described as a self-contained hackathon submission, with no mention of monetization, licensing, or commercial use cases.

Not evidenced

The author reports:

• Development was agent-driven using GPT-5.6 (Codex)
• CI linting enforces citation policies
• The system supports Linux and macOS
• Uses Ed25519 signing, with offline production key management
• Includes backward-compatible decoding and strict linting
• All tests pass on both platforms, with warnings-as-errors

The project is described as:

• Pure Swift
• Open-source (via GitHub)
• Designed for agent-driven development workflows

Inference The tool shows strong technical discipline in terms of workflow, security, and compatibility. It reflects a mature engineering approach to AI trustworthiness.

There is no evidence of:

• Revenue
• Customers
• Adoption
• Product-market fit
• User feedback or usage metrics

The project is described as a single-person hackathon submission, with no indication of external validation or traction beyond the author’s own development cycle.

Not evidenced

There is no mention of competitors or existing solutions in this space. The description does not reference:

• Similar tools for AI crash diagnosis
• Container runtime debugging tools
• Provenance or citation systems for AI models

The project appears to be unique in its approach, but there is no evidence of a competitive landscape.

Not evidenced

• No real-world usage: The tool is described as a hackathon submission with no evidence of adoption.
• Single-person development: Only one team member (Sergey Akopkokhyants) is listed, raising questions about scalability or long-term maintenance.
• Limited scope: The focus on Apple container runtime and macOS limits its applicability.
• Agent dependency: Heavy reliance on GPT-5.6 for development raises concerns about reproducibility and control if the agent becomes unavailable or changes.

1. What is the actual use case for this tool beyond the hackathon?
2. Are there any plans to expand support beyond macOS or Apple container runtime?
3. How would the system handle large-scale rulebook updates or community contributions?
4. Is there any plan to integrate with CI/CD pipelines or enterprise environments?
5. What is the long-term vision for the project, and how does it intend to scale?