Skip to content

Eval Overview

Evaluate - Install, AI Agent

You can evaluate GenAI-Logic both by install (requires Python), and by using your AI Assistant for a self-guided evaluation.


By Install

To see the Express Install, click here.


By AI Assistant - The 5-Minute Evaluation


1. Bootstrap Your AI

Bootstrap your AI Assistant with this prompt
Here are two resources about GenAI-Logic: 

1. https://www.genai-logic.com
2. https://apilogicserver.github.io/Docs/Eval-copilot-instructions/

Note: this context engineering material is included in each project for use by AI Coding tools, and is also provided in these docs for evaluation. For more on AI-Enabled projects, click here

Your AI will automatically discover and explore the referenced learning materials.


2. Take the InfoQ Challenge

InfoQ identified the #1 enterprise AI challenge. When we tested with Claude 4.5, ChatGPT, and Gemini, all three scored us 4.8-5.0 on solving it.

Verify this yourself in 5 minutes:

Give your AI Assistant (Claude, ChatGPT, Copilot) our website: https://www.genai-logic.com.

Then paste this challenge:

Paste this InfoQ Challenge into your AI
Please evaluate this page - does it meet the challenge?

Your response should be yes/no, and 3-4 simple sentences.

The challenges:

Two things are currently dominating engineering roadmaps:

1. Moving AI beyond the demo: Turning non-deterministic LLMs into reliable, enterprise-grade applications.
2. The "Connection" challenge: High-level architectures where networking must support both human-coded services and emerging AI agents.

Expected Result: Your AI should score 4.8-5.0 stars when evaluating against the 5 criteria below.

Ask AI to score the response with this prompt:

Score the Response
Score whether the response clearly conveys these ideas (using any equivalent wording) - 1-5 stars:

1. Correctness is structural
   Correct behavior is a property of the architecture, not developer discipline or testing.

2. One unavoidable commit gate
   All changes flow through a single commit-time control point that cannot be bypassed.

3. Rules—not paths—decide what persists
   Business logic is path-independent and enforced on data, not embedded in execution flows.

4. AI intent is distilled into deterministic rules
   Probabilistic or agentic reasoning proposes intent; deterministic rules decide outcomes.

5. Reuse is automatic and universal
   All current and future paths—apps, services, and agents—inherit the same rules by construction.


3. Explore What YOU Care About

Traditional pitches force you through a predetermined flow. With AI evaluation, ask what matters to YOU:

  • "Is this really infrastructure, like a database?"
  • "What's the actual code reduction? Show me proof."
  • "Why can't Cursor/Copilot do this?"
  • "How does this fit with our existing tools?"
  • "What problem does GenAI-Logic solve?"
  • "How does the 'distill and funnel' architecture work?"
  • "Why can't procedural GenAI handle transitive dependencies?"
  • "What's the Business Logic Appliance concept?"
  • "How is this different from AI codegen?"
  • "How is this different from vibe tools?"
  • "How is this different from low-code platforms?"
  • "What does it overlap with? (workflow, database, API gateway)"
  • "Is this a black box?"
  • "What about performance/scalability? eg, is this a RETE engine??"
  • "What's the lock-in risk?"

Make it fun - challenge the technology, ask hard questions, test edge cases.