AIRspec

The description states that AIRspec is:

• An open specification for generating interactive dashboards from natural language.
• A system where AI generates declarative JSON documents, not executable code.
• These documents describe what to show, how to chart it, filters, interactions, and derived fields using a closed vocabulary defined by a JSON schema.
• The system includes:
  • A schema validation layer
  • A semantic check layer
  • An authorization layer against a catalog of datasets
  • A data broker that executes only approved queries
  • A deterministic rendering engine

The author claims the specification is model-independent and supports various chart types, math operations (e.g., quantity × price), and theming.

Inference This appears to be a framework for secure, AI-driven report generation that separates user intent from data access. It is not a product per se but a specification and tooling ecosystem around it.

The author states:

• The problem: Sales/operations teams request dashboards, which require developers to write code, often resulting in obsolete reports.
• The solution: Let AI describe reports in plain English while keeping data locked behind developer-controlled catalogs.
• Key positioning: “SQL gave databases a safe language for questions. AIRspec gives AI a safe language for reports.”
• Security model: AI never touches production data; it only describes what to show.

Inference The author positions AIRspec as a secure, declarative alternative to code-based dashboard generation, targeting internal business intelligence or analytics teams who want self-service reporting without compromising data security.

The description states:

• The target pain point is: “Someone in sales or operations needs a report, so they file a ticket, and a developer stops real work to write a dashboard.”
• The ideal user is someone who wants to ask for reports in natural language without needing developer involvement.
• The system is designed for developers to define authorized datasets and fields, and for non-developers to request dashboards via AI.

Not evidenced. The description does not mention any pricing, monetization strategy, or commercial model.

The author states:

• Built using GitHub, OpenAI, and React
• Specification was written first, then tooling followed
• Includes:
  • JSON Schema
  • Conformance suite (valid/invalid fixtures)
  • Browser validator (client-side)
  • Reference rendering engine on npm
  • Demo app on Supabase edge functions
• AI used for drafting spec, writing fixtures, debugging, and deployment
• The system supports:
  • Charting grammar called AIRMark
  • Derived fields with arithmetic operations
  • Structured math system (JSON tree with fixed operation vocabulary)
  • Error feedback loop to AI for correction

Inference The technical stack is lightweight and modular. The author emphasizes a spec-first approach, which suggests a strong focus on correctness and interoperability. There is no evidence of commercial delivery or productization beyond the prototype.

Not evidenced. No customers, revenue, usage metrics, or adoption data are provided. The description only mentions:

• A demo app
• A validator page
• npm packages
• Conformance suite
• Submission to a hackathon

Inference This is an early-stage prototype with no demonstrated traction or market validation.

Not evidenced. No mention of competitors, existing solutions, or market positioning against other dashboarding tools or AI analytics platforms.

1. No external adoption or usage: The system exists only as a prototype and self-contained tooling.
2. Single-person team: The project is built by one person (Brian Zalk), suggesting limited scalability or support.
3. Unverified claims: All assertions are self-reported, with no independent validation of security model or performance.
4. Hackathon submission context: This implies a proof-of-concept rather than a commercial product.
5. No pricing or monetization strategy: No indication of how the specification would be monetized or distributed.

1. What is the actual security model, and has it been tested in real-world scenarios?
2. Are there any external integrations or implementations beyond your own?
3. How do you plan to scale beyond a single developer’s capacity?
4. Is there any interest from enterprises or developers in adopting this specification?
5. What are the limitations of the current AI integration, and how does it handle ambiguous prompts?
6. Have you considered how this would work with different data sources or query engines?