Harness Engineering for Self-Improvement
Lilian Weng published a post titled "Harness Engineering for Self-Improvement" examining how harness engineering relates to recursive self-improvement (RSI). The post traces RSI's origins to I. J. Good (1965), who defined an "ultraintelligent machine" as a system that can surpass humans in all intellectual activities and design better machines to improve itself, and to Yudkowsky (2008), who used the phrase for a feedback loop in which an AI uses its current intelligence to improve the cognitive machinery producing that intelligence.
Weng writes that in modern AI this feedback loop may mean a model rewriting its own weights directly, or, more broadly, improving the training pipeline and deployment system to enable a better successor model. She states that the layer between the raw model and real-world context appears as important as raw model intelligence, and defines a harness as the system surrounding a base model that orchestrates execution and decides how the model thinks and plans, calls tools and acts, perceives and manages context, stores artifacts, and evaluates results.
The post outlines three harness design patterns: workflow automation, the file system as persistent memory, and sub-agents with backend jobs. It describes coding agent harnesses such as Claude Code, Codex, OpenCode, and Cursor-style agents. Weng predicts that the near-term path of RSI is unlikely to begin with a model directly rewriting its weights, and that harness engineering will move toward a meta-methodology in which the harness itself becomes an optimization target.
For harness optimization, the post describes Agentic Context Engineering (ACE; Zhang et al. 2025), Meta Context Engineering (MCE; Ye et al. 2026), and Meta-Harness (Lee et al. 2026), which it characterizes as a harness for optimizing harnesses. On workflow design it covers The AI Scientist (Lu et al. 2026), ScientistOne (Meng et al. 2026), the Autodata agent (Kulikov et al. 2026), Automated Design of Agentic Systems (ADAS; Hu et al. 2025), and AFlow (Zhang et al. 2025).
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
The concept of recursive self-improvement (RSI) dates back to I. J. Good (1965) , where he defined an “ultraintelligent machine” as a system that can surpass humans in all intellectual activities and design better machines to improve itself. Yudkowsky (2008) used the phrase “recursive self-improvement” for a specific feedback loop: an AI uses its current intelligence to improve the cognitive machinery that produces its intelligence. This feedback loop in modern AI may indicate the model rewr