Agentic coding
A form of AI-aided programming that reduces human intervention and lets agents work for longer on larger tasks. It differs from mindless vibe coding.
IDEs
- Archon
- Agent mode in Cursor
- Agent mode in GitHub Copilot
- Google Antigravity
- Agent mode in Replit
- Google Jules
Terminal-based
- Claude Code
- OpenAI Codex
- Gemini CLI
- Qwen Code
Record
- 2025-11-09 - Speculative Tree Search (thinking about giving AI more work while reducing the human review burden).
- 2025-08-19 - Retrospective of Agentic Coding Experiments
- 2025-08-03 - Agentic Coding Experiment 3. Specifications rather than a requirements document, plus stateless AI–human interaction.
- 2025-04-25 - Agentic Coding Experiment 2. Revised based on the thoughts in Software Engineering in the AI Era.
- 2025-04-12 - Agentic Coding Experiment 1. An attempt to imitate existing human workflows by defining planning, implementation, and refactoring agents.
- 2024-10-01 - Started using Cursor Composer (the former name for its agent feature) around this time.
- 2024-09-08 - Started using Cursor.
- 2023-10-21 - Tried Sweep AI and left a note.
- 2023-10-19 - Tried CodeRabbit and left a note.
- 2021-07-15 - Started using GitHub Copilot.
Articles
- 2025-06-12 - Agentic Coding Recommendations
- 2025-05-26 - Vibe Coding vs. Agentic Coding: Fundamentals and Practical Implications of Agentic AI
- 2025-04-02 - Best practices for using AI coding Agents
Negative(?) outlooks
- 2025-08-14 - Why LLMs Can’t Really Build Software
- 2025-08-05 - No, AI is not Making Engineers 10x as Productive