W10-Post-OpenClaw Era: The Coding Singularity Has Arrived, but the New Production Relations Have Not

The lobster fever triggered by OpenClaw has become increasingly surreal, almost magical-realist at this point. It has already become art. I’m staying far away.

With the lobster frenzy, I have a feeling. The idea that “code is cheap” has now fully expanded beyond programmers. Many people have developed a real sense of how low the barrier is to building software (the barrier and the ceiling are two different things). The power to shape the conversation about its second-order effects is likely to shift, and the driving force for change may no longer be something that is only internally controlled.

Watching the new products that emerge after the post-lobster era.

1. Claude’s Cowork. In the past, Claude has already blocked two of my accounts, and since I wanted to try Cowork, I went through the trouble of registering another one. I completed a few small tasks, and the experience was decent, though it still needs to keep pulling my attention back through frequent authorization prompts. One task was to help me organize my Google Drive folders and files. Since it couldn’t directly write to Google Drive, Cowork opened a browser, went to the Google Script site, wrote a script to organize the files, and completed the task through web operations. I haven’t had a deep experience with it yet, so this week I’ll explore more complex workflows.

2. Perplexity’s Computer. It’s too expensive; access is limited to Max subscribers paying $200 a month. I mainly read this articleThe AI is the Computerand took away two insights. First, a judgment about the specialization of model development: no single model family can do the best job for you without other models. Second, a new insight into the nature of computers: the traditional moat of operating systems may weaken, and the essence of the computer will shift into an intelligent system that can orchestrate multiple AI capabilities and connect internet knowledge bases with personal data.

3. OpenAI has brought in Peter Steinberger, the author of OpenClaw. OpenClaw is now overseen by an independent foundation and has nothing to do with OpenAI. Steinberger’s work at OpenAI going forward will be developing the next generation of personal agents. I’m looking forward to it.

4. What product is Google holding back? I haven’t seen it yet.

Recently, several questions have been lingering in my mind, and I’ve been thinking about them continuously.

1. Why has AI coding progressed so rapidly, yet only a small number of people are deeply pursuing this alluring intelligence?

   I now think this question is not that important. On the surface, it just means everyone has different interests and levels of time invested, but eventually they can catch up—and in a short time.

2. Is coding fundamentals, which we have always focused on, still an important dimension? What will create the gap in the future, and how do we provide guidance that fits future development?

   People always say it’s taste, judgment, the ability to define products, and so on. These have always mattered, too. The key is how to cultivate them, and whether there are scalable methods suitable for the general public. It’s like always saying empathy is important—how do you cultivate empathy?

3. The productivity singularity in coding has arrived, but outside of coding there are obstacles everywhere, and nothing can move quickly. What is the breakthrough point?

   A new set of production relationships has not yet emerged. Code production capacity can improve by tens or even hundreds of times, but all the surrounding processes are still stuck at 1x speed: requirement definition, design collaboration, review mechanisms, result acceptance, and operations growth.