Author: Yaki
Today I got to try "open claw" in a bit more depth. Here are a few thoughts…
1. Tools: The differences between tools aren't that significant
After using "Claude Code" for a long time recently, I feel that, essentially, the differences between tools aren't very significant. In terms of capabilities, there is almost no difference. By capability, I mean the "Agent" logic—the logic for completing a task in its entirety.
With traditional AI tools, such as Doubao and Zhipu, they only give you a method. You can follow it, and if you run into a problem, ask again to solve it. But as development tools like Trae have evolved, Agent programming has, in theory, already acquired a framework: it can write and execute code directly. The results are the same as with OpenClaw; both rely on large models to process context. The only difference is in the interface logic, which comes down entirely to personal preference.
As someone accustomed to using an editor, I feel somewhat uncomfortable with OpenClaw not having one. I miss the clear, at-a-glance view of files, as well as the clear display of execution steps—these are differences at the tool level!
2. Technology: Development capabilities are becoming democratized
Starting with Doubao, I began to get a vague sense that— "Technical and development capabilities are gradually becoming accessible to everyone." 。
Back in '23, I used GPT's step-by-step guidance to learn from error messages, and in just a few hours, I connected ChatGPT to a WeChat account. At the time, I also published a tutorial in the Way to agi Knowledge Base.
After that, I integrated Midjourney, then moved on to Coze workflows, Trae, Claude Code, and now OpenClaw. Essentially, the ability to “write code” has kept getting lighter and simpler.
An app, a mini-program, an official website—as long as you can describe your needs well and have some basic understanding of the general logic, or even if you don’t have any to begin with, you can build that understanding as you watch it execute. With this capability, development can be completed in an extremely short time.Just type and write up your requirements, and you can start designing your own product. This was something unimaginable before.
3. Knowledge dimension: A qualitative change in how knowledge is stored
Starting with Trae, when applications became able to read all the files on my desktop, the way knowledge was stored underwent a qualitative change.
Over the past roughly 3 years, the knowledge base content I’ve written myself amounts to about 1 million charactersor so, stored in Tencent Docs. Organizing, categorizing, and reordering content in Tencent Docs is very labor-intensive and inefficient, and the interaction between people and knowledge is extremely inefficient.
Since I started using Claude Code, I’ve converted almost all the information I produce into markdown format, including content generated by OpenClaw. I’ve also made it a hard requirement that it return content in markdown format and save it to a fixed directory, and I’ve equipped it with a knowledge graph skill. In theory, this can systematically organize the relevant knowledge and create links between pieces of knowledge—This was something that was completely impossible before!
I’m planning to move the content from Tencent Docs over soon, and I intend to give up Tencent Docs. Maybe in the future, the way we store knowledge won’t be paper books, but a markdown format that AI can understand!
4. Era dimension: How should we respond?
The evolution along this path has been incredibly fast—so fast that even keeping up with the changes is a struggle. Capabilities and applications are evolving almost by the day.
In November 2022, GPT3.5 took off. Then came Claude, Kimi, Doubao, image generation, video creation, development tools, workflows, and more—with progress getting faster and capabilities getting stronger. We were still talking about Seedance 2.0 over the New Year holiday, but as soon as the holiday ended, all the buzz was swept up by OpenClaw—In less than four years, everything has been upended countless times.
In this fast-paced world, how should we respond?
The first type: The early adopters
If you want to be among the first to try something new, you need to be able to gather and organize real-time information and explore its full potential. Behind every new capability is an upheaval of traditional workflows. That upheaval might mean a dramatic boost in efficiency, or it might mean rebuilding the entire process pipeline.
Things may not be perfect yet, but believe me: one day, they’ll arrive before us in a polished form. The process of evolution often brings many opportunities and challenges. If you can get involved at this point, you become part of the ecosystem, helping to change the world together!
It’s hard work, but also a lot of fun!
The second type: The users
This group is more focused on using these tools. As a capability is continually refined, optimized, and iterated on, it may reach an emergent breakthrough—a tipping point. When that fundamentally changes how we work and live, we may have no choice but to adapt, and that can be extremely painful.
When AI-generated information can nearly match what this group can produce, all that follows is anxiety and unease. The achievements they once took pride in are shattered in an instant. If they can’t adapt and master it fast enough,I believe there will be no mercy in the end!
In the end
There are no spectators in this race.
You can choose to ride the crest of the wave, or wait on the shore for it to reach you. But believe me—you can never choose “not to be hit by the wave.”
The only difference is this: when the wave comes, will you already be ready—or will it catch you off guard and knock you onto the sand?
Let’s encourage one another 🌊
I’m Yaki. Good night 🌙