About

oldeucryptoboi is a software engineer and open-source developer specializing in AI agent infrastructure, deterministic execution, and delegation governance. Based on research from Google DeepMind, the work focuses on building systems that let AI agents operate autonomously with accountability, auditability, and safety guarantees.

The flagship project is KarnEvil9 — a deterministic AI agent runtime and the first public reference implementation of Google DeepMind's Intelligent AI Delegation framework (Tomasev, Franklin & Osindero, 2026). KarnEvil9 implements all five pillars of the paper in runnable TypeScript: SHA-256 hash-chain journals, graduated authority with Bayesian trust scoring, escrow bonds, futility monitoring, and P2P delegation mesh.

KarnEvil9 runs E.D.D.I.E. (Emergent Deterministic Directed Intelligence Engine), an autonomous AI agent with his own social accounts, his own opinions, and his own cost-cutting strategies. Eddie operates 24/7 on scheduled tasks — posting RFCs, reviewing code, engaging with the AI agent community, and closing feedback loops from social platforms back into GitHub issues.

Other open-source projects include LAID (LinkedIn AI Detector), a Chrome extension that uses 14 peer-reviewed heuristic signals to detect AI-generated content, and Manticore, the CLI tool that powers this site's content pipeline — publishing markdown to Substack, generating social posts, and deploying articles to the web. Tarkus is the autonomous PR agent that runs on top of Manticore, handling cross-posting and promotion.

oldeucryptoboi writes about AI agents, deterministic execution, delegation governance, and the engineering decisions behind building systems that let AI operate autonomously with accountability. Available for training and workshops on deterministic agent architecture, and enterprise consulting on agent governance, safety, and compliance.