Machvix
Machvix is a real-time scoring engine that evaluates the speed and reliability of information retrieved from AI providers. It generates two independent scores for any query: MACH (how fast the answer was found) and VIX (how much the answer should be verified before relying on it). The scoring methodology is transparent and provider-agnostic, meaning it does not bias results based on whether a provider is free or paid. The engine supports multiple domains including cryptocurrency prices, weather data, recipes, and open-ended questions. Beyond basic scoring, Machvix includes a suite of reliability probes such as Sycophancy Probe (testing if a model caves under social pressure), Anchoring Bias Probe (checking sensitivity to planted numbers), Self-Consistency Check (detecting hallucination risk through repeated queries), and Verifiability (cross-checking answers against independent sources like Wikipedia). A Silent Model Drift Watch monitors answer consistency over time. The platform also features a trust leaderboard, shareable trust receipts, and a public corroboration scorecard comparing provider accuracy. Users can interact with Machvix via a live web demonstration or by running the engine locally using a Python script (`machvix_engine.py serve`), provided they set up API keys for Groq, OpenRouter, or Gemini. An API is available upon request for integration into other applications. The machvix.ai domain itself is listed for sale, but the product will continue under a new name if sold.
- Category
- Analytics
- Pricing
- Free
Problem solved
Machvix addresses the challenge of blindly trusting AI-generated answers by providing objective, real-time scores for answer speed and the need for verification. It helps users decide how much to rely on information from AI models, especially in high-stakes domains like finance and weather, where accuracy and timeliness are critical.
Use cases
- Verify real-time cryptocurrency prices from AI models
- Assess reliability of weather forecasts generated by AI
- Benchmark AI providers for speed and trustworthiness
- Detect hallucination risk in AI responses via self-consistency checks
- Monitor AI model drift over time with silent drift watch
- Generate shareable trust receipts for AI-generated answers
- Evaluate AI model resilience under adversarial prompts (sycophancy, anchoring)
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
- Business Users
Pros
- Real-time, provider-agnostic scoring
- Comprehensive reliability probes (sycophancy, anchoring, drift, etc.)
- Transparent methodology with public scorecard
- Free to use (demo and API access upon request)
- Multi-domain support (crypto, weather, general QA)
Cons
- Under active development with occasional connection timeouts
- Limited documentation beyond the live demo
- No clear open-source license for the engine
- Domain for sale creates uncertainty about long-term branding
- Requires third-party API keys for full local operation
Official website: https://machvix.ai/
Canonical LambdaBase page: https://www.lambdabase.com/tools/machvix