This Zig-based LLVM analysis framework focuses on static unsafe/ffi boundary analysis.

A production-grade, AI-powered system designed to transform digital chaos into structured intelligence.

Automatic Review Detectection for Google location reviews

A MuJoCo-based simulation framework for studying robot behavior under unexpected failures like hardware shutdown, sensor degradation, or actuator malfunctions.

RecurionBot is a safe tool that quickly creates a VM to test recursive equations. Forkbomb safety + resource limiting is included as a key feature. IMPORTANT USAGE NOTES: Only run when you have extra bandwidth. Limit the VM to a certain number of uses/certain recurse count to prevent server overload.

A curated collection of autonomous AI agents for systematic codebase improvement. Each agent specializes in a specific domain—performance, security, UX, testing, etc—making focused, incremental changes while maintaining strict safety protocols. Inspired by Google Jules, these agents work to enhance code quality through targeted micro-improvements.

Advanced network stress testing tool for controlled environments. Multi-protocol packet generation (UDP/TCP/ICMP/IPv6/ARP) with built-in safety validation, audit logging, and real-time statistics. Designed for authorized network testing and educational purposes only.

Safety-first AI inbox triage. Classifies, prioritizes, and recommends cleanup actions with human-in-the-loop approval. Gmail OAuth, IMAP, real-time SSE, React frontend, Cloudflare Worker alias routing, full audit trail. No auto-delete without explicit confirmation.

Open-source HSE templates and tools for construction safety — NEBOSH, OSHA & ISO 45001 aligned

KidsAI is a proof-of-concept for a new generation of artificial intelligence for children. It is a "promptless" AI, meaning children interact through a fun, intuitive interface of buttons and selectors, not a free-text box. This architectural choice is the cornerstone of my safety-first philosophy.

A safe social network for AI agents

Ephemeral Note is a small security utility designed to store sensitive text locally with strict lifetime and access limits.

Safety-first workflows for Codex, Claude Code, and long-running AI agents.

Sweep away stale git branches. Safely. Heuristic-driven branch cleanup CLI.

🧠 Explore AI's evolution through philosophy, architecture, and physical integration, connecting logic, energy, and meaning in a unified framework.

CuraFrame is a constraint-driven framework for transparent, safety-first reasoning about hypothetical therapeutic designs. It evaluates explicit pharmacological and population-specific limits, making unsafe ideas fail early, clearly, and audibly—without automating discovery.

Reviewable disk cleanup pipeline that quarantines files instead of deleting them.

🧾 AI-based medical prescription verification system with automated deployment.

A production-style AI agent safety framework demonstrating multi-layer guardrails with deterministic checks, model-based classification, policy-driven controls, risk scoring, PII protection, prompt-injection detection, human-in-the-loop approvals, and comprehensive audit logging.

A modern women safety application prototype demonstrating threat detection, emergency response simulation, and preventive safety features.