The Impact of AI Coding on Programmers
Categories: Share
After using Claude Code for less than a month and spending $326 (actually $20 for Pro + $50 for top-up), the Cursor I had been using for a few months has become obsolete. Using AI well can easily reach the level of a P6+ engineer. As an engineer, I feel both surprised and scared.
The surprise is that the AI Coding ability is really strong. It has basically solved some of the difficult technical problems in non-front-end fields that I have not been able to solve well in recent years after continuous communication and debugging. It’s even like a friend who buys equipment in a game by spending money. I can’t help but give money to Anthropic because it surprised me so much. It’s more like making a great friend who is technically proficient and kind to people. In the future, the so-called single-person operation will really feel like having a very cost-effective team in the hands of students who know how to use tools, use their brains, and understand user needs.
The fear is that the traditional manual coding that I used to be confident in has become worthless in front of the current AI. A cruel but clear trend is that pure coding ability is no longer the moat of programmers. The current AI can easily replace programmers who are purely translating requirements. This is also the scary part. In addition, the Internet industry is basically in a quagmire of cost reduction and efficiency improvement, which will make this change faster.
I remember sharing two years ago when the environment was not good, The Breakthrough of the Next Generation of Engineers, should be to be a product engineer, that is, to know where the user has needs, and then independently use a good product solution to undertake it, while the product is very easy to use, and you are very good at operation and promotion, attracting more people to use it. It’s just that the ability of AI Coding was still very weak at that time. Today, it should be a product engineer who is good at using AI.
For the next generation of good engineers, coding ability only accounts for 30% of the advantage, 20% is in quickly discovering and understanding the essence of business needs and knowing why, 20% is in architectural design, like an architect telling AI what you need and the front-end and back-end architecture to ensure better implementation later, 10% is in communicating more clearly with AI to make her execution more in line with your wishes, and another 20% is in the final product quality control and operation and promotion control. Good wine is also afraid of deep alleys. No matter how powerful the AI ability is, it is also afraid of users who will not toss.
I feel that AI Coding brings more than just a doubling of work efficiency for engineers. In fact, this is not the key. The more critical thing is to be able to handle more complex product thinking and technical decisions faster, and to speed up the verification of business iteration ideas. It feels like changing from a code farmer to a digital product architect. Of course, aesthetics will be more important in the current software design engineering. Perhaps if we want to say that good engineers in the current era need to have a very good ability, it is product design and aesthetics. This is also why it is very convenient for smart designers to transform into engineers.
However, I don’t like the kind of promotion that teaches novices that they can do anything with AI without understanding the principles and technologies. For computer fundamentals, software architecture design, and interaction design capabilities, these are the foundations of engineers. It’s the same with or without AI. This is something that cannot be lost. What needs to be cultivated more is the ability to make products.
Perhaps the qualitative change had not yet arrived before. Claude Code made me feel that the qualitative change of AI Coding has arrived. Purely wrapping a skin and calling other people’s models to make an editor is actually not a long-term solution. It has slowly become a competition of model-as-a-product capabilities. In addition, I feel that for individuals, there is another way to enjoy the convenience of AI more, which is to invest in AI.