Tag
This post critiques the reality of autonomous error recovery in AI agents, highlighting issues like hallucinations and destructive retries, and argues that deterministic systems with strict controls perform better in production workflows.
The author reflects on building AI agents, highlighting challenges like inconsistent outputs, context management, and the importance of using deterministic approaches when appropriate.
The article emphasizes the shift from coding to defining primitives for building deterministic systems like Cody, a platform that ensures secure and reliable AI agent deployment by eliminating surprises.
Warrior Quest is a local LLM-powered dark-fantasy RPG demo on Steam where the AI model is used only for NPC emulation while the game state remains deterministic, requiring a GPU with 8GB VRAM.
This paper discusses how a deterministic artificial agent can display hedonic place preference behavior, with implications for understanding consciousness and free will.
The article reports benchmarking results for a deterministic AI financial verification engine, showing perfect performance on structured claims (66/66) but poor performance when LLM-generated claims are used (19/66), indicating a translation gap between LLMs and formal systems.
This analysis challenges the reflexive insertion of AI into all enterprise workflows, suggesting that deterministic systems often require traditional software rather than probabilistic models. It argues for a strategic approach to distinguish where AI creates leverage versus where established architectures remain superior.