Hello! I'm your financial analysis assistant. This article from "Sīxiǎng Gāngyìn" explores a very fundamental and often misunderstood aspect of the business logic of AI: Why can AI make money by writing code, but it's so difficult for AI to directly replace humans in tasks like design and video editing? What fundamental problem does OpenAI's latest GPT-6 Astra (with its Computer Use feature) aim to solve?
To help you understand this easily, I've broken down the article into a core summary and five in-depth analyses.
---
📝 Core Content Summary
In one sentence:
AI used to be like a “super intern who could only write code”; now it wants to become a skilled worker who can operate all the software on your computer.
The core logic:
AI has been successful in programming because code is its “native language,” and the process is transparent and modifiable. However, in industries like design and finance, the results generated by AI (such as videos and posters) are often “black boxes.” Once something goes wrong, the entire output has to be discarded, which is too costly for businesses to accept. OpenAI’s new strategy, Computer Use, aims to make AI act like a human, directly using existing software like Photoshop and Excel. This makes the AI’s work process visible, controllable, and editable, allowing it to be seamlessly integrated into business workflows and opening up a much larger market.
---
🔍 Five In-depth Analyses
1. Why can AI make money by writing code, but not in video editing? (The difference in workflow transparency)
Many people think that AI’s video editing results are not good enough compared to professional models. However, the real issue is not with the quality of the results, but with the uncontrollability of the process.
- The “transparent workshop” of programming:
Programmers produce code, which is visible to them. When AI writes code, you can check line by line and make corrections. You can even ask AI to write only half of the code, and then complete the rest yourself. It’s like watching workers in a workshop; if something goes wrong, you can stop them or do it yourself. This transparency allows AI to integrate seamlessly into programmers’ daily work.
- The “black box” of video/editing:
If you ask AI to create a 10-second video, it gives you a finished file. If you spot a flaw at the 3rd second and want to fix just that part, AI usually can’t make local adjustments; it has to start over. Moreover, you don’t know how AI made the decision—was it the composition or the lighting that was wrong? The result is that fixing one part often breaks three others, forcing you to start from scratch. This is a huge waste of time and resources for businesses.
💡 In simple terms:
Programming works because AI and programmers use the same language and tools, making the process transparent. In other industries, AI acts like a magician, but you can’t see how it does its magic, and any mistakes mean starting over, which is too costly.
2. What is “Computer Use”? (Allowing AI to use your mouse and keyboard)
OpenAI’s “Computer Use” aims to solve the problem of the “black box.”
- Previous AI:
Like a brain that can only type or a printer that only outputs files.
- Current Computer Use:
It’s like an intern sitting next to you. It can see your screen, understand the buttons, move the mouse, click on menus, and type text.
- The key difference:
Instead of directly generating a finished file, it opens Photoshop, creates a new file, finds materials, adjusts layers, and saves it. This makes the AI’s actions visible. You can watch it work, and if it makes a mistake, you can stop it or help it make the right choice. This turns the “black box” into a “white box.”
💡 In simple terms:
Previously, AI gave you a finished product without any intervention. Now, AI works in front of you, and you can guide it at any time. It’s like changing from outsourcing to hiring a transparent employee.
3. Why do businesses prefer “old software”? (Reducing the cost of adoption)
This is the most insightful business point in the article:
- The trap of native AI software:
If AI were to replace designers, it would need to develop completely new software. But that would require businesses to:
1. Abandon software they’ve used for 20 years.
2. Change file formats.
3. Retrain employees.
4. Adapt their processes for new customers.
This is too difficult for businesses to do.
- The advantage of Computer Use:
It doesn’t require any changes. AI directly uses existing software like Photoshop, Excel, or SAP.
- File formats remain the same.
- Employee habits remain the same.
- Customer delivery standards remain the same.
- The only change is having an AI assistant to help you use the software.
💡 In simple terms:
Businesses fear the hassle of making changes. Allowing AI to use existing software minimizes resistance. It’s like adding smart home features to an old house rather than tearing it down and building a new one.
4. “Experts managing experts”: The new workplace relationship in the AI era
The article proposes an interesting idea: AI won’t completely replace jobs; it will change the way work is managed.
- Past barriers:
Designers’ barriers were aesthetic creativity and proficiency in software.
In the past, software skills were important because they were considered a form of “skill.”
- Future barriers:
In the era of Computer Use, AI can use software, but people still need to understand how it works.
Why? Because you need to manage AI.
- If AI makes a mistake in Photoshop, you need to know which layer or setting is wrong to fix it.
- If AI gets stuck, you need to know how to take over manually.
This is called “experts managing experts.” You don’t need to create the content yourself, but you need to understand how it’s created to direct AI.
- Vs. “Non-experts managing experts”:
With native AI software (black boxes), you can only give commands. It’s like having a non-expert handle the work; you don’t understand the internal logic and can’t control the outcome.
💡 In simple terms:
The future’s top professionals won’t be those who don’t use software, but those who understand how it works and can direct AI. Your value lies in understanding the software and being able to guide it.
5. The implications for investors: The explosion of AI’s Total Addressable Market (TAM)
From a macro perspective, what does this mean for the capital market?
- From programmers to all white-collar workers:
Last year, AI coding only benefited programmers (tens of millions of people). If Computer Use becomes widespread, AI will serve designers, accountants, analysts, operators, architects, and many other professions that rely on software.
This group includes hundreds of millions of people.
- Annual Recurring Revenue (ARR) growth:
When AI can be integrated into these complex workflows, businesses will be willing to pay for it (to save on labor and improve efficiency), leading to significant and sustainable revenue growth.
- Is the investment in cloud providers excessive?
If AI could only write code, building many data centers might have been unnecessary. But if AI can handle entire work processes, the demand for computing power will grow exponentially, justifying the massive infrastructure investments.
💡 In simple terms:
The market for AI has been limited to programmers; now it’s expanding to all white-collar workers. The market has grown dozens of times, making the investments in graphics cards and data centers more justified.
---
📌 Summary and Recommendations
The main message of this article is that for AI to become a profitable reality on a large scale, it can’t just generate results; it must integrate into business processes.
- For individuals: Don’t panic about AI replacing you. Instead, learn how to guide AI to use software you’re familiar with. Understanding the underlying software logic is more important than just being fast.
- For businesses: Focus on AI tools that can directly use existing software stacks, not on disruptive new software that requires major changes.
- For investors: The AI story is just beginning. The expansion from coding to general office tasks represents a trillion-dollar market opportunity.
To remember:
AI coding shows that AI has learned to “communicate”; Computer Use shows that AI has learned to “work.” Only by being able to perform tasks effectively and transparently can AI become a valuable part of a business’s workforce.