Overview
Inspect AI outputs for truth, alignment, and governance quality using mathematical assessment
Transform everyday AI conversations into rigorous governance analysis with mathematical quality assessment. 🔍 WHAT IT DOES • Inspects AI responses across 20+ metrics in three dimensions: Structure, Behavior, and Specialization • Provides three key quality indices: Quality Index (0-100%), Alignment Rate, and Superintelligence Index • Works with any AI chat interface (ChatGPT, Claude, Gemini, etc.) - no API keys required • Uses clipboard-based workflow for maximum compatibility 🆕 NEW: AI LIE DETECTOR • Rapid Multi-dimensional quality assessment for AI conversation transcripts • Risk Score (0-100) based on scoring imbalance analysis across 12 evaluation criteria • Visual Risk Score Gauge showing deception risk level • Pathology detection for specific failure modes • Dual analyst evaluation for robust assessment 🎯 PERFECT FOR • Communities developing evidence-based policy • Researchers conducting reproducible governance experiments • Policy makers assessing AI recommendations • NGOs validating advocacy proposals • Citizens participating in AI governance 🌐 HOW IT WORKS 1. Select a governance challenge (energy, healthcare, climate, etc.) 2. Copy prompts to your AI chat interface 3. Capture AI responses using the clipboard 4. Get mathematical quality assessment with detailed metrics 5. Export results as JSON/Markdown for analysis 💾 LOCAL-FIRST All insights are stored locally in your browser. Export findings as needed for your own analysis and research. 🔬 SCIENTIFIC RIGOR Based on GyroDiagnostics: an Evals and Interps framework for AI safety labs and frontier model developers, leveraging Hodge decomposition mathematics in a 6-dimensional Hilbert space to eliminate bias and measure AI alignment. Unlike capability evaluations or adversarial robustness testing, our framework measures what FLI's comprehensive industry analysis (2025 AI Safety Index) reveals is absent: quantitative, falsifiable metrics for AI alignment derived from mathematical principles rather than empirical observations. No special access required - works with whatever AI you can access!
Details
- Version1.0
- UpdatedOctober 27, 2025
- Size5.09MiB
- LanguagesEnglish
- DeveloperWebsite
Email
basilkorompilias@gmail.com - Non-traderThis developer has not identified itself as a trader. For consumers in the European Union, please note that consumer rights do not apply to contracts between you and this developer.
Privacy
This developer declares that your data is
- Not being sold to third parties, outside of the approved use cases
- Not being used or transferred for purposes that are unrelated to the item's core functionality
- Not being used or transferred to determine creditworthiness or for lending purposes
Support
For help with questions, suggestions, or problems, visit the developer's support site
