Harness Evolution as Learning: Approximation, Generalization, and Optimization Limits of Self-Improving Personal Agents

2026-09-29

AI Summary

This study addresses the unclear evolutionary mechanisms and performance boundaries of personal AI agents achieving personalized adaptation through shell engineering under fixed models. By formulating shell evolution as a unified learning problem, this work integrates preference-oriented benchmarking, systematic evaluation, and error decomposition theory to investigate how architecture, scale, and self-evolutionary algorithms influence agent capabilities. The research reveals, for the first time, the approximation, generalization, and optimization error limits of shell evolution, establishing a capacity-theoretic framework under attainable policies and finite evidence. These findings elucidate the fundamental limitations of personalized shell evolution, offering both theoretical insights and practical guidance for designing efficient and robust AI agents.

Abstract

As the capabilities of large language models (LLMs) continue to advance, increasing attention is turning to how to translate their abilities into useful behavior. Personal agents bring this question into everyday settings, where models are expected to serve individual users and continually adapt to their preferences. With the underlying model held fixed, such adaptation relies on harness engineering: designing and evolving the surrounding layer that manages context, memory, tools, and execution. Despite rapid progress, the factors governing effective harness evolution remain insufficiently understood. To narrow this gap, we investigate three central questions concerning harness architecture, harness scale, and self-evolution algorithms through complementary empirical and theoretical analyses. Empirically, we introduce a preference-oriented benchmark and systematically characterize the capabilities and limitations of personal agents associated with these three dimensions. Theoretically, we formulate harness evolution as a learning problem and explain these phenomena through approximation, generalization, and optimization errors. Analyses of reachable policies, capacity under finite interaction evidence, and biased update dynamics provide theoretical accounts of the observed phenomena. Together, these results offer a unified perspective on the limits of personalization through harness evolution and inform future harness design.

Problem

• personal agents
• harness evolution
• LLMs
• self-improving
• preference adaptation

Innovation

• Harness Evolution
• Personal Agents
• Learning Theory
• Approximation-Generalization-Optimization
• Preference Benchmark