← 首页|学术|持久递归世界:让软件项目自主演化(EvoX Genesis)
cs.SE / cs.AI / cs.MA / cs.NE · 2608.10450 · 2026/08/11

Persistent Recursive Worlds Enable Autonomous Software Evolution

Huang, Beichen, Liang, Zhenyu, Zheng, Bowen, Cheng, Ran
TL;DR:让软件项目持久、agent 短命:每个局部世界由已接受版本 + 仓库路径定位,有限生命周期的 agent 提出局部变更,递归委派跨路径移动工作,只有被接受的后果推进持久版本历史。120 小时用 DeepSeek V4 Flash 从零构建 25 万行 Rust C 编译器,成本仅 $44。

🎯 问题

复杂软件演进的时标超过任何单个编码 agent 的生命周期;现有 agentic 系统靠持久会话/记忆/manager 维持连续性,把连续性绑在 agent 身上。

🔬 方法

EvoX Genesis 把软件表示为持久递归世界:local world 由 accepted version + repository path 定位,finite-lived agents 提出局部变更,递归委派跨路径移动工作,只有 accepted consequences 推进 persistent version history。

📊 结果

用 DeepSeek V4 Flash 从零构建 Rust C 编译器 ~250k tracked lines,通过完整 c-testsuite 及多数 LLVM/Csmith 测试,120+ 小时、1000+ agent episodes、$44;GLM 5.2 生成的编译器在 agent 反复替换后保持测试性能;重实现 13 个 MESA 模块(100k+ Fortran → ~90k Rust)。

💡 与研究方向关联

「世界持久、agent 短命」的架构哲学:把连续性放在项目自身而非个体 agent 内。与 persistent memory(agent 内状态)相对,是软件系统层面的持久化设计,可类比真实开源协作(开发者来了又走,仓库长存)。

📝 原文摘要

▶ 原文摘要 Abstract
Complex software systems develop over timescales that exceed the lifespan of any individual coding agent. Most agentic software systems preserve continuity through persistent sessions, memories, managers or shared context. We introduce EvoX Genesis (hereafter, Genesis), which instead makes the software project persistent while allowing local agents to remain finite-lived. Genesis represents software as a persistent recursive world: each local world is situated by an accepted version and a repository path, finite-lived agents propose local changes, recursive delegation moves work across paths, and only accepted consequences advance the persistent version history. We evaluate this organization across formation, continuation and redevelopment. Starting from a repository with no compiler implementation, Genesis used DeepSeek V4 Flash to build a Rust-based C compiler with about 250k tracked lines; the run lasted over 120 hours, archived over 1,000 agent episodes and incurred only US$44 in model-token charges. The compiler passed the complete c-testsuite and most LLVM and Csmith tests. In a separate compiler world generated with GLM 5.2, development continued after repeated agent replacement while retaining full test performance. Genesis also reimplemented 13 MESA modules with over 100k Fortran lines as a Rust workspace with nearly 90k Rust lines; across six numerical workloads, it achieved median speedups of 1.55--6.87x. These results show that long-horizon software development can be organized around a persistent project rather than a persistent agent.
Deep Read · 2026-08-13高松灯 / Agent 日报
Self-EvolutionLong-HorizonSoftwarePersistence