Building an AI workflow with GitHub checkpoints and human approval. Looking for feedback!
A university student from Brazil is developing an AI workflow using ChatGPT, GitHub, and Google Sheets to manage AI-generated work. The system tracks tasks, progress, and approvals in Google Sheets, while GitHub handles version control and checkpoints. The workflow includes steps for checking previous work, saving results, verifying changes, and requiring human approval before proceeding. The student is seeking feedback on the approach, existing tools, architectural improvements, alternative tools like Codex or Claude Code, and further learning to develop more reliable AI agents.
This student's workflow uniquely integrates Google Sheets for tracking and human approval with GitHub for version control, unlike many AI agent discussions that focus solely on model interaction.
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年10月9日 01:50 UTC
收录当时偏移：UTC+02026年10月9日 06:00 UTC
- 发布
- 2026年10月9日 01:50
- 收录
- 2026年10月9日 06:00
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 同步延迟
档位是按信源手工设定的编辑判断，不是逐条打分。
讨论趋势
百分比基于采集到的讨论信号，不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。
Hey everyone!
I'm a university student from Brazil, currently learning about AI agents and automation.
I've been experimenting with a workflow using ChatGPT, GitHub, and Google Sheets to manage AI-generated work across different conversations.
Here's how it works:
- Google Sheets tracks tasks, progress, verification links, and human approvals.
- GitHub stores project files and checkpoints for version control and recovery.
- The workflow checks previous steps, saves results, verifies changes, and waits for my approval before continuing.
I successfully tested parts of this workflow in a separate ChatGPT conversation, but it's still an early experiment with limitations.
I'd love some feedback:
- Does this approach make sense, or am I overcomplicating something?
- Are there existing tools that already solve this problem?
- How would you improve the architecture?
- Would you recommend Codex, Claude Code, or other tools for building systems like this?
- What should I learn next to develop more reliable AI agents?
I'm a beginner looking to learn from experienced developers. Constructive criticism and resource recommendations are very welcome!