Beichen Huang, Zhenyu Liang, Bowen Zheng, Ran Cheng
Date
Context
EvoX Genesis puts continuity in the software project so development can outlast any individual coding AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition →. Agents work from an accepted project version and repository location, propose changes, and advance the persistent history only through accepted results. Research abstract.
The authors report a compiler-building run lasting more than 120 hours and spanning over 1,000 agent episodes. In a separate compiler experiment, development continued through repeated agent replacement while retaining test performance. These results support organizing long-running work around a durable project history that successive agents can continue.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
articleMindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program SynthesisYihao Chen, Shi Chang, Khaled Chawa, Feng Lin, Boyuan Chen, Shaowei Wang, Ahmed E. Hassan
articleOurArkRoy Zhao (Paul G. Allen School of Computer Science & Engineering, University of Washington), Zhenyu Zhao (Independent Researcher)
articleOuroboros: A Self-Developing Frontier Coding Agent with Reviewed Core EvolutionAnton Razzhigaev, Andrei Gritsaev, Andrei Kaznacheev, Nikita Dragunov, Roman Yampolskiy, Andrei Kuznetsov