Show HN: L0/L1/L2 agents, leases, gates, audits, Git-worktree isolation (github.com)

glueRun-go, a new orchestration engine combining Bash and Python, has been introduced to enhance the autonomy of AI coding agents in software repositories. This innovative tool utilizes a three-tier scheduling model consisting of L0 origin loops, L1 area planners, and L2 worker agents, enabling parallel execution of tasks while managing leases, state packets, and audits seamlessly. Its design allows for a single installation per machine, where updates can be implemented simply by changing version pins, thereby improving efficiency in maintaining multiple projects. The significance of glueRun-go lies in its structured approach to autonomous coding, which can significantly streamline workflows in AI/ML development, allowing for quicker iterations and more resilient task handling. Each task executed by L2 workers in isolated git worktrees generates a state packet that undergoes auditing before final results are routed. The use of gates for validation—such as running tests post-execution—ensures that only reliable code is integrated, enhancing overall project integrity. This orchestration engine is poised to be a game-changer for teams looking to leverage AI for automated software development processes, with its ability to handle complex dependencies and auditing mechanisms making it a powerful tool for modern development environments.