The Dawn of Liberation is Approaching
It is now 2026-01-05.
Recently, I started using vibe-kanban to experiment with project management at an engineering level, and I found the results quite promising.
It provides a Kanban board, allowing me to manage tasks and progress from a Team Leader's perspective. When using Claude Code before, I always felt like a Senior Developer mentoring a Junior Developer; I needed to focus intensely on a single task. Claude Code lacks the concept of work-tree isolation, so I couldn't manage multiple tasks simultaneously. After starting to use vibe-kanban, I can break down tasks into multiple cards and advance them in parallel. Each card has its own context and progress, and I can switch my focus at any time without worrying about losing information.
This approach makes me feel more like a "foreman" than a "coder." I can better coordinate various tasks, allocate resources, and monitor progress instead of getting bogged down in specific coding details.
I quickly realized that my bottleneck shifted to creating new tasks and reviewing tasks (yes, obviously, tasks still need review). The sense of unease didn't disappear just because I started using vibe-kanban. In fact, I feel it has become stronger because I now have more tasks to manage instead of focusing on just one.
In short, a moment of satisfaction, followed by dissatisfaction again.
But clearly, vibe-kanban is meaningful and useful. However, it made me realize a problem: there's a contradiction between my desire for control over details and my desire for rapid progress. This contradiction itself is caused by my bandwidth limitations. It existed before, but now it's just become more pronounced.
So I started thinking, why do I have such a strong desire for control? If I can't let go of my need to control details, I can't truly scale up. After all, I am a finite human being with limited bandwidth. Thinking about technical leads, they don't write code themselves but drive project progress through management and coordination. How do they dare to let go of their control over details? Simply hiring more people is obviously not enough; they need a system to ensure project quality and progress. Humans and AI, as the underlying workforce, are essentially not that different in terms of reliability. Humans make mistakes, and AI makes mistakes. The key is how to design a system that maximizes overall reliability and efficiency.
You can refer to this article, where I discuss this issue in detail and propose some solutions.
Additionally, I realized that vibe-kanban occupies a different niche than Claude Code. It operates at a higher level than Claude Code / CodeX and similar products. In other words, advancements in Claude Code and its ilk represent progress for vibe-kanban, not a threat, because they exist at different levels of abstraction. Claude Code and similar tools are the underlying workforce, while vibe-kanban is a management tool. This division of labor allows them to work together rather than compete.
However, vibe-kanban is still a team-leader level tool, not a CEO-level tool. A CEO needs to focus on strategy and direction, not on specific tasks and progress. Therefore, I still need to find a way to enhance my strategic thinking abilities, not just rely on task management tools. Perhaps there's already a product called vibe-okr? 😆
I discussed a recent Agent project with a colleague. I think the tool he previously built to facilitate human review of AI Plans was interesting, but ultimately, it's heading in the opposite direction of liberation. His tool ultimately reinforces the desire for control rather than letting go of it. I believe this is a misconception. True liberation is about learning to let go of the need to control details, not trying to enhance that control through tools. Their tool might make people feel more in control, but it doesn't truly liberate human bandwidth.
Company level (?) > Team level (vibe-kanban) > Individual level (claude code, ...etc).
In short, there's still a lot of work to do, but I feel I've found a viable path.
