HN Showcase
Back to browse
GitHub Repository

Deterministic agent runtime with explicit plans, typed tools, permissions, and replay. Implements Google DeepMind's Intelligent AI Delegation framework. Named after ELP's Karn Evil 9.

12 starsTypeScript

KarnEvil9, a deterministic AI agent runtime

by oldeucryptoboi·Mar 5, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerWizardryBig BrainZero to One

First public implementation of DeepMind delegation paper. Tested on Zork with governance that blocked 'attack'.

Strengths
  • •Direct translation of Tomasev et al. (2026) into working code—novel architecture, not just API wrapping.
  • •Hash-chain journal + Bayesian trust + escrow bonds create genuine accountability; demonstrated on 3-agent Zork swarm.
  • •Domain-agnostic governance layer (same framework for code refactor, game, financial workflow) is genuinely clever.
Weaknesses
  • •Early-stage GitHub project (3 stars); unclear production readiness or real-world testing beyond Zork demo.
  • •Paper is dated 2026 (future date suggests preprint or internal paper); legitimacy unverified.
Target Audience

AI engineers building multi-agent systems, financial services (HFT, trading), aerospace/safety-critical applications requiring auditability

Similar To

LangChain · Crew.ai · Autogen

Post Description

Built this over the past few months because I kept hitting the same wall with agent frameworks. You run something, it does... stuff, and then you're left trying to figure out what actually happened and why.

KarnEvil9 is a TypeScript runtime that implements the DeepMind delegation paper from earlier this year (Tomasev et al.). The core idea is pretty simple: every action goes into a SHA-256 hash-chain journal, agents earn trust through a Bayesian scoring model, and there are actual economic stakes via escrow bonds. If an agent screws up, it loses its bond. If it keeps failing, the futility monitor kills the loop.

The fun part was testing it on Zork I. I set up three agents in a swarm: one plans moves, one executes them against a Z-machine, one independently verifies game state. The governance layer immediately blocked the agent from attacking the troll because it classified "attack" as high-risk. Took me a while to realize the fix wasn't to whitelist attack commands, it was to make the system trust-aware so an agent with a good track record can take riskier actions.

The other thing I didn't expect: when Eddie (the autonomous agent that runs 24/7 on this) hit the Anthropic API credit wall, the futility monitor halted everything, and Eddie's next plan included switching to cheaper models for routine code reviews. Nobody told it to optimize costs. That came out of the delegation framework's cost-awareness primitives.

Happy to answer questions. https://oldeucryptoboi.com

Similar Projects