Introduction: This article does not discuss the individuals and teams involved, but rather presents some viewpoints on Python in the AI era. It includes advice for both non-engineers and professional engineers.
01
Advice for Non-Engineers: Learn and Use Python WellFirst, the advice for non-engineers is to learn and use Python effectively.The reason is quite simple: currently, all AI applications and frameworks are just starting out and are far from maturity. Therefore, most of the “successful” AI applications and “mature” AI frameworks are more about success and maturity in terms of reputation. In fact, most AI applications have not passed PMF validation.In the industry, the excitement among major companies about AI stems from the high expectations for the future, ten or twenty years down the line, which will completely disrupt many things, rather than the currently popular applications and platforms. The current applications and platforms are too insignificant compared to what will be available in ten years.For these reasons, the current “successful applications” limit our imagination, and the “mature frameworks” constrain our creativity, which in turn affects our understanding of AI. Subtle cognitive biases ultimately lead to significant differences in direction and maturity among many teams, products, and companies.The best way to break this constraint is to directly use LLM APIs, allowing us to bypass all limitations.Python’s AI-friendliness (with many open-source projects) and its very low learning threshold make it suitable for anyone without a formal background to start learning. Moreover, with the assistance of AI, everyone can quickly learn Python programming independently and explore AI capabilities.In my nearly twenty years of professional career, I have seen countless cognitive biases permeate various areas, and these absurd issues arise from absorbing second-hand and third-hand information, echoing what others say after reading self-media articles. For example, in the “5G era,” everything is viewed through the lens of 1 millisecond latency, completely ignoring the constraints of the speed of light, with various media promoting that remote surgeries can achieve 1 millisecond latency over distances of 5000 kilometers; in the AI era, digging deep into many bizarre viewpoints reveals that many people are relying on outdated Chinese translation materials (many technologies change significantly within 1-2 months), and they never bother to check the most original first-hand information.The best way to obtain first-hand information in the AI era is through Python + LLM API calls, creating something useful for oneself. AI has already granted this capability to everyone, including product managers, operations personnel, and managers at all levels (including myself). Understanding the technical value without engaging with original information is akin to high school reading comprehension, where many standard answers and the original author’s intentions are completely misaligned.The current information landscape is too noisy, and I have a simple filtering method: default to ignoring information from these individuals:
- Relying on second-hand and third-hand information to understand technology, such as MCP, etc. (ignoring official websites and only looking at incomplete translation materials and unprofessional self-media articles, then outputting various viewpoints). I ask: Where did you get this information?
- Not having personally written any AI applications. I ask: Have you developed any?
- Developed AI applications that they themselves do not consistently use. Even for AI practitioners, this is a common phenomenon. They promote various AI products they have never actually used. I ask: How do you use it? How often do you use it?
Whenever I encounter various strange viewpoints, I often ask the above questions, and after asking, I understand the reasons.
02
Different Perspectives on Python for Professional Engineers
Finally, I have slightly different views on professional engineers choosing Python.
- Python is not suitable for large, multi-person collaborative, long-term maintenance systems; this applies to all dynamic languages;
- Despite its limitations, Python still has use cases for professional engineers:
- Prototype validation, write and throw away
- Dependency on mature open-source Python projects with no suitable alternatives