Skip to main content
Prism - Programming with Confidence

Programming with Confidence in an Uncertain World

Get StartedView on GitHub
// AI responses automatically include confidence
analysis = llm("Analyze this code for vulnerabilities")
// Returns: "No critical issues found" with 85% confidence

// Make decisions based on the confidence level
uncertain if (analysis) {
  high { deploy_to_production() }    // confidence ≥ 0.7
  medium { request_human_review() }  // 0.5 ≤ confidence < 0.7
  low { block_deployment() }         // confidence < 0.5
}
🎯

First-Class Uncertainty

Attach confidence levels to any value and propagate uncertainty through calculations automatically.

🤖

AI-Ready Design

Built for the age of AI with explicit confidence tracking and the @prism-lang/confidence library for extraction.

🌊

Confidence Flow Control

Make decisions based on confidence thresholds with intuitive uncertain if statements.

📊

Confidence Extraction

Extract confidence from LLMs and other sources with @prism-lang/confidence - consistency checks, response analysis, and more.

🔧

TypeScript First

Full TypeScript support with type safety, IntelliSense, and modern tooling integration.

📦

Modular Design

Pick only what you need with our modular package ecosystem: core, confidence, and LLM modules.

Built for the Age of AI

Handle uncertainty in AI responses, sensor data, and complex computations with confidence

AI Applications

Handle LLM responses with explicit confidence tracking and decision thresholds.

IoT & Sensors

Manage noisy sensor data with uncertainty bounds and confidence propagation.

Risk Analysis

Make risk-aware decisions with confidence budgets and differential analysis.

Scientific Computing

Propagate measurement errors and uncertainty through complex calculations.