Loosely-Structured Software: Engineering Context, Structure, and Evolution Entropy in Runtime-Rewired Multi-Agent Systems Conference: arXiv Preprint; 2026;

Weihao Zhang, Yitong Zhou, Huanyu Qu, Hongyi Li · 2026

Read Loosely-Structured Software: Engineering Context, Structure, and Evolution Entropy in Runtime-Rewired Multi-Agent Systems Conference: arXiv Preprint; 2026; by Weihao Zhang, Yitong Zhou, Huanyu Qu, Hongyi Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15690.txt.

Introduction

Zhang and co-authors frame a new software class for runtime-rewired agent systems. As agent teams grow, free-form chats drive behavior and drift grows. Prompt tuning alone cannot hold such systems steady, they argue. Their Loosely-Structured Software shifts focus to runtime entropy control. Three layers manage views and context, dynamic structure, and evolution. Design patterns act as semantic control blocks inside the flow. Early tests support the framework's core mechanisms.

Catalog record

CREATORWeihao Zhang, Yitong Zhou, Huanyu Qu, Hongyi Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE83288
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.15690

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Loosely-Structured Software: Engineering Context, Structure, and Evolution Entropy in Runtime-Rewired Multi-Agent Systems Conference: arXiv Preprint; 2026; Weihao Zhang, Yitong Zhou, Huanyu Qu, Hongyi Li Abstract . As LLM-based multi-agent systems (MAS) become more autonomous, their free-form interactions increasingly dominate system behavior. However, scaling the number of agents often amplifies context pressure, coordination errors, and system drift. It is well known that building robust MAS requires more than prompt tuning or increased model intelligence. It necessitates engineering discipline focused on architecture to manage complexity under uncertainty. We characterize agentic software by a core property: runtime generation and evolution under uncertainty . Drawing upon and extending software engineering experience, especially object-oriented programming, this paper introduces Loosely-Structured Software (LSS) , a new class of software systems that shifts the engineering focus from constructing deterministic logic to managing the runtime entropy generated by View-constructed programming, semantic-driven self-organization, and endogenous evolution. To make this entropy governable, we introduce design principles under a three-layer engineering framework: View/Context Engineering to manage the execution environment and maintain task-relevant Views, Structure Engineering to organize dynamic binding over artifacts and agents, and Evolution Engineering to govern the lifecycle of self-rewriting artifacts. Building on this framework, we develop LSS design patterns as semantic control blocks that stabilize fluid, inference-mediated interactions while preserving agent adaptability. Together, these abstractions improve the designability , scalability , and evolvability of agentic infrastructure. We provide basic experimental validation of key mechanisms, demonstrating the effectiveness of LSS. 1. Introduction Large language model (LLM) multi-agent systems (MAS) are experiencing a “Cambrian explosion” ( 47 ; 4 ; 54 ; 51 ; 13 ; 33 ; 29 ) . Across academia and industry, teams are assembling systems that search, plan, code, and iterate, aiming to turn probabilistic models into reliable and productive infrastructure. Powered by semantic understanding and intelligent reasoning, LLM agents can synthesize information, propose actions, and coordinate across tools. Yet progress in MAS has also exposed a hard limit: capability does not scale linearly with the number of agents ( 18 ) . In practice, larger teams often amplify errors and coordination overhead, and some claims argue that simpler agent architectures can better align with user intent ( 31 ) . A critical gap persists: while prototypes dazzle with emergent capabilities, production systems often hit a ceiling of complexity where additional agents add more failures than utility. Figure 1. The engineering focus shifts from designing specific logical code to designing runtime generation and evolution. Much of the current MAS literature focuses on building “more intelligent” orchestration frameworks ( 56 ; 43 ; 55 ; 16 ) . Increasingly, however, the dominant failures are not merely a matter of prompt tuning or base model intelligence, but an engineering problem of architecture (recently emerging OpenClaw-/NanoClaw-like architectures ( 33 ; 29 ) also reflect this trend): the discipline required to manage complexity, interfaces, and drift under uncertainty. Accordingly, this paper treats MAS as a new kind of software system and develops an engineering perspective that draws from, adapts, and extends conventional software engineering principles and patterns. MAS need not be treated as an opaque “intelligence-on-intelligence” black box; rather, they can be approached as software systems that can be designed, governed, tested, and evolved. Engineering this new kind of software requires confronting its core property: runtime generation and evolution under uncertainty . Classic software architecture assumes build-time modular decomposition and slow-changing boundaries. Agentic software violates these assumptions through three coupled “physics”: • View-Constructed Programming : Unlike modular software where logic is encapsulated in code, an agent’s effective program is determined by a step-specific View constructed from global Artifact files (e.g., system prompts, skills, plans, tools, memories) and projected into its context window at runtime ( 27 ; 34 ; 21 ) . • Runtime-Generated Structure, Abstraction, and Semantic Binding : Unlike code-defined linking, component connections are formed dynamically via semantic understanding. The system’s connectivity and abstraction can be generated on the fly via agent/tool selection, temporary team organization, and interfaces negotiated in natural language rather than fixed function signatures ( 51 ; 13 ; 38 ; 15 ; 42 ) . • Endogenous Evolution : The system’s executable substrate (Artifact files that mediate behavior in-context) is itself rewritable by the system, and thus forms the foundation of adaptation and self-improvement ( 43 ; 24 ; 46 ) . This paper defines Loosely-Structured Software (LSS) as a software paradigm characterized by these properties. While many existing frameworks provide valuable structure through deterministic pipelines to ensure reliability (e.g., LangChain, AutoGPT) ( 4 ; 54 ) , LSS explores an alternative path by treating runtime variability as a first-class design dimension (Figure 1 ). For engineers trained in traditional software development, this gap signals a deeper shift. Instead of writing deterministic logic behind stable interfaces, one must engineer the conditions under which agents construct Views, bind to capabilities, and revise their own Artifacts. A recurring challenge is deciding what to constrain and what to leave flexible. Traditional software design workflows favor stable abstractions (protocols, intermediate representations) and stable topologies (e.g., classic Object-Oriented Programming, OOP patterns). In LSS, however, abstractions and architecture are often unstable at design time and can be generated and revised at runtime. Moreover, because agents interpret the same Artifact through task-driven perspectives, the abstraction can be viewpoint-dependent rather than objective. Addressing these issues cannot rely solely on adding more higher-level wrappers. Figure 2. Loosely-Structured Software: a software system with unstable, runtime-generated structure & abstraction and its dominant engineering focuses (View, Binding, Evolution). Beyond these design differences, practical MAS characteristics further complicate engineering: (i) View-boundedness (context budgets make context pollution a main failure source) ( 21 ; 27 ; 10 ) , (ii) collaboration cognitive gaps (agents often lack explicit supply–demand contracts, so coordination devolves into semantic probing) ( 51 ; 38 ; 1 ) , and (iii) persistent-memory and Artifact maintenance (large, evolving stores of skills, traces, and long-term memories accumulate operational debt and require coherent management) ( 34 ; 58 ; 53 ) . Together, these characteristics drive the poor scalability observed in practice: as more agents are added, the system bears greater “context pressure” and a growing communication tax without proportional gains in task success. Figure 2 summarizes the key differences between LSS and traditional software in both behavior and system features. These differences shift the engineering objective from “constructing deterministic logic” to “managing runtime entropy distribution”: guiding the growth of the system itself. We introduce a three-layer framework that systematizes this entropy management: Layer 1 (View/Context Engineering) manages the execution environment and attention span (taming Context Entropy); Layer 2 (Structure Engineering) manages the organization of artifacts and agents to enable dynamic capability discovery and binding (taming Self-Organization Entropy); and Layer 3 (Evolution Engineering) manages the lifecycle of self-rewriting artifacts (taming Evolutionary Entropy). Building on this framework, we also adapt reusable software design wisdom—especially from OOP ( 9 ) —and intellectualized it as dynamic engineering mechanisms for regulating information flow and distribution. Unlike OOP design patterns that typically map to static class hierarchies and compile-time relationships, LSS Design Patterns (e.g., Semantic Router , Mediator ) operate as semantic control blocks that stabilize fluid, inference-mediated interactions. These patterns transform structural uncertainty into a feature that increases designability —ensuring that an

Read the complete text