All (118)Personal Introduction (1)AI Summary (8)Project Management (5)Investment Strategy (1)Financial Market Analysis (5)AI Software Engineering (9)AI Learning Methodology (1)AI Interview Design (1)AI Research (8)Content Management (9)Capital Protracted War (18)Cognitive Development (1)Philosophical Reflection (3)Entrepreneurship (1)AI Tools Development (9)Product Development (3)Troubleshooting (2)AI Cost Analysis (1)System Architecture (1)Technical Log (3)Quantitative Finance (6)AI Social Systems (3)Personal Reflection (1)AI Scheduling (2)
Agentic Engineering Development Practices and AIM Scheduling System Concept
AI Scheduling
👤 AI developers, software engineers, technical managers, and individuals interested in Agentic Engineering and AI-driven development
This article documents the author's practical experiences in 2026 using Agentic Engineering to develop auth-mini and SOAR projects, noting that current AI development has reached Senior SDE level, capable of independently completing the full development cycle from creating a worktree to merging PRs, but lacking task management and planning capabilities. The author analyzes the limitations of current AI development's serial task execution and proposes developing an AIM (Agentic Intelligence Manager) scheduling system to improve concurrency. This system will manage the lifecycle of multiple OpenCode Sessions, using a SQLite database for state coordination, aiming to increase daily Token consumption and PR output to billions and tens of units, respectively.
- ✨ Agentic Engineering can independently complete the development cycle from worktree creation to PR merging, surpassing traditional development efficiency
- ✨ Current AI development lacks task management and planning capabilities, preventing it from reaching Team Leader level
- ✨ Proposes developing an AIM scheduling system to manage multiple OpenCode Sessions, enhancing concurrency and development efficiency
- ✨ AIM will use a SQLite database for state management, eliminating the need for API design, allowing AI to operate directly via SQL
- ✨ Expected that AIM implementation will significantly increase daily Token consumption and PR output
📅 2026-04-15 · 728 words · ~4 min read
Reflections on AI Scheduling and Task Management Optimization
AI Scheduling
👤 AI developers, project managers, technical decision-makers, and those interested in AI scheduling and task optimization
This article discusses how the AI scheduling project Agent Harness has reached a performance sweet spot, with the main issue shifting from improving accuracy to optimizing iteration speed. The author proposes breaking tasks into basic units, establishing dependency management, and constructing a task graph, allowing AI to execute tasks in parallel based on dependencies. Through the OpenCode session mechanism, AI can autonomously determine task blocking status, split tasks, and modify the environment to achieve efficient scheduling. The article emphasizes that with low AI Token costs, we should let AI grow naturally, reduce manual review, and accelerate project progress.
- ✨ Agent Harness has reached a performance sweet spot, with iteration speed becoming the main issue
- ✨ Proposes a parallel execution strategy based on task splitting and dependency management
- ✨ AI can autonomously determine task dependencies, split tasks, and modify the execution environment
- ✨ Suggests reducing manual review and letting AI grow naturally to enhance efficiency
- ✨ With low AI Token costs, economic cost is not currently a concern
📅 2026-04-11 · 685 words · ~4 min read