- From AI Coding to AI Development
- The real Real AI Engineer is the “human body.”
- For everyone to write the code, AI is the bridge to “Technology Inclusion”
- If AI can finish all the code writing, can everyone “program”?
- Where does the programmer go when everyone can program?
** “** On 11 June, at the 2025 Spring Volcanic Engine FORCE Basic Dynamics Conference, the byte beat Technology Vice-President Hong Ding-Qun revealed the latest user data for the AI IDE tool Trae: the overall monthly effort is over 1 million and over 80 per cent of the byte engineers are using AI-assisted development.”** When you see the data above, you can see that the AI programming tool has been one of the most hot topics in the technology world over the past year. The AI programming tool has entered the Developer’s Tool Chain from the Laboratory’s Toys. Google’s CEO Sundar Pichai says: “According to the AI support code, the productivity of the Google engineer has increased by about 10% (data from internal engineering hours analysis) As of earlier this year, more than 30 per cent of new codes were generated by AI (25 per cent last year) and all AI-generated codes require manual clearance. Pichai claims that he is also actually using tools such as Cursor, Replit, to construct small network applications through the “vibe code” (i.e. natural language interactive code) ** To summarize the current status of AI programming in one sentence, AI Coding has been upgraded from “a few lines that can be completed” to “a job that can be done with you” ** which really changes the flow of developers. But it comes with the idea that programmers are “released” and “replaced”. Even the CEO Dario Amodei of Anthropic (Claude) recently said: “In the next three to six months, AI will prepare 90% of the code, and in 12 months almost all of the code will be generated by AI.”
#penAI CEO Sam Altman also publicly stated on several occasions: “Our first reasoning model ranks about 1 million in the global programmer competition. o1 (September 2024) ranking = 9800 +3 (December 2024) ranking = 175 places Now, according to our internal assessment, our internal model, currently ranked 50th, is expected to be ranked first this year, and there is no sign of stopping.” The numbers, the words, are powerful enough for many to panic: So what’s the future of the programmer? What’s the future encoded tool like? What else can a man do?
From AI Coding to AI Development
To answer the above questions, I think we must first crack a core error — that is, to think that the value of the programmer is “writing code”. Dario Amodei’s assertion that “90 per cent of the code will be prepared by AI” is frightening because it uses the word “code line” as the core indicator for measuring the work. But in reality, the code is only the final product, and it’s more like the physicalization of building drawings.** The real value of programmers has never been brick-bricking, but designing drawings, planning structures, selecting materials, and ultimately taking responsibility for the quality, safety and preservation of the entire building.** What we produce is a robust system that solves particular problems, operates steadily in complex environments, and is sustainable**. This process is full of business understanding, trade-off trade-offs, and prejudging future changes.
-
Why do you use PostgreSQL instead of MongoDB?
-
Should this API be designed as synchronous or asynchronous?
-
How should micro-services be split in order to balance business cohesion and team independence?
-
How should the system ‘ s cache, downgrade, and flow-limit strategy be deployed in response to the two-nifteen flood peaks?
The real Real AI Engineer is the “human body.”
Although AI Coding is becoming stronger, Hong DingQin has repeatedly stressed in his speech that the true “Real AI Engineeringer” must be the product of close collaboration between AI and others. While 85% of the code was generated by AI, he was involved in architecture design, process control, logical review, and so on. Even when AI debug failed, he had to go back to manual intervention to solve the problem. He said very directly: “If we just throw our needs at AI without being involved in the process, then the systems that are created are often difficult to maintain, let alone optimize performance”. Hong Dingkun identified his identity as a “drivers” rather than as a “standers”: he wrote prompt, reviewed logic, and took over manual improvements at any time; and the ultimate location of the large sections of the code produced by AI was still dependent on a person’s judgement of the boundaries of demand, performance bottlenecks, security risks. In other words,** the role of the programmer evolved from a “coder” to “System adjudicator plus process designer.”** Will be eliminated, depending on how many lines you knock, and whether you can:
-
A precise and complete description of needs in one sentence;
-
I can see where the AI-generated program will fall into the pit;
-
Ties operations, tools, models into a smooth flow line. When product managers, designers, and even operations can “talk to code”, developers become “symphonic commands” linking business to AI – know technology and know business rhythm.
One of the words mentioned several times in the speech was: “AI can do a lot, but it’s still people who drive the process.” The current AI coding tool, although able to write the full code, is still very difficult for ordinary people, so I think that a phase of AI Coding should be a closer “human body”, and that humans would only need creative guidance, simple aids, process control, security clearance, and that the rest would be done by AI. Then we’ll be at the next stage of AI programming after the transition to human coordination.
For everyone to write the code, AI is the bridge to “Technology Inclusion”
In addition to professional programmers, HUN DingQin shared a particular case: an 11-year-old child, with the help of his father (a byte R & D engineer), developed a web site using Trae to capture and practice a number of topics. Behind this story is that Trae is trying to lower the “development threshold” by allowing non-professionals to participate in the construction of the digital world. Historically, every “democratization of programming tools” has led to a wave of expansion in the size of developers:
-
Compiling language to C language and opening up the first system-level developers;
-
C to Java/Python to promote the rise of Internet product developers;
-
Front-end visualization, low-code platforms, which enable non-engineers to build functions;
-
In the future, as AI’s programming capacity increases, it may be possible for “people who don’t have the code to build the product.”
In Trae’s case, the 11-year-old child completed a project that included not only reptiles logic, interface design, but also interactive and curator management with data interfaces, a task that in the past had to be completed with complete CS training. Now, with AI’s help, he has step by step to lead the system not only “workable”, but also “run”. This marks a turning point: AI is not just a tool to improve the efficiency of development, but a key to breaking technical barriers.
If AI can finish all the code writing, can everyone “program”?
At first sight, this question sounds like a technical idealism, but it is actually a reality that is happening: Today, “programming” is moving from a skill to a capability interface. Like using a search engine and making slides,** you don’t need to know the mechanism behind it, but you know how to solve the problem.** If we admit that the code is the language of communication between people and computers, then AI’s intervention is tantamount to creating a ** “real-time translation” . This translator understands your intentions and how the machine should be executed. You tell it, “I need a form to upload the operation and automatically score.” It finishes the code, adjusts the style, deploys the line, and you don’t need to care about POST requests or stride backs, not to write the proper text, not to rely on the import. So can we still say this guy can’t program? This may be the most worthwhile part of future programming to redefine. The future development ecology may be re-segregated. If we look at the next 10 years, AI programming penetration may result in the following structural changes: ** This is not the disappearance of programmers, but the conversion of programming from a skill to a common language of an “organizational digital system”. Developers will no longer be able to prove themselves by simply writing the correct syntax function, but will be able to organize a person’s intent into a system structure in a more efficient and structured way. You will use the AI tool to complete a whole function design, realization and upline, which may be more important than writing a sorting algorithm.
The future of “programming for all” is not a fantasy, but a necessity of ecological evolution.
The fall in technological thresholds is essentially a process of productivity release. Industrial revolutions allow everyone to open machines, printing and publishing, and AI programming tools give everyone the possibility of “building digital capabilities”. It means:
-
In education, 10-year-olds can also use AI for web applications;
-
At the working level, market operations can independently produce tools to analyse data;
-
At the start-up level, independent creators have the ability to build a business model for MVP prototypes.
As Hong Dingkun has said, the greatest significance of AI Coding is perhaps not to make professional programmers faster, but to allow “non-programmers to do so”. From “engineer explosion” to “programmable for all”, AI programming tools bring not skills migration but redistribution of technical powers.
Where does the programmer go when everyone can program?
AI is reprogramming the definition at an eye-to-eye rate and seems to be unstoppable! When 11-year-olds can build websites in natural languages, when non-engineers can automatically generate interactive pages in Trae, and when large models can generate, debug and go online business modules, “programming” is no longer the exclusive skill of a few, but is becoming a universal capacity for universal participation. So back to our original question: ** Does programmers matter when everyone can program?** The answer is — ** More important than ever and less important, but their roles must change radically.** Slowly, programmers are no longer “the people who write the code,” but ** problem molders, AI dispatchers, system architects, product responsible. They no longer hold IDE one line of output code, but rather stand at a higher level of the system to guide AI in building complex logic, assessing programme quality, and managing technological risks. And the future does not belong to the developers of “managering a language”, but to those who can transcend technology and business, understand human needs and have infinite creativity and organize the system. And the significance of this change is much greater than the replacement of a career path. When AI programming tools give “building power” to every ordinary person, it is effectively redistributing the right to participate in the digital society. The programmer is no longer just a builder, but also an enabler, which is the bottom of the real system of the digital age for all.** So the programmer’s direction was never “replaced” or “margined”, but ** from the person who wrote the code to the person who made the world write the code.** This is not only a leap in technology, but also a profound relaxation of the social fabric. It’s the programmers who rewrite these things, and they end up hiding their credit!