How do Vietnamese programmers manage when AI reduces their income?

Minh Tu, 24 years old, a programmer living in Ho Chi Minh City, shared his feeling of “ambiguity” before choosing a career. Graduating with honors from a school with a reputation for information technology training, he was quickly accepted into a software outsourcing company in the area with a “living” salary.

However, at the end of last year, the business used the reason of deploying AI agents in some stages, as well as a decrease in orders, to adjust the salaries of some positions. He is in the affected group because he is new to the profession and does not have enough seniority. “I just started running a motorbike taxi, 3-4 hours every night. I don’t plan to do it long term, but it gives me time to consider how I should change my career,” Mr. Tu said.

Quoc Khanh, 28 years old, takes advantage of technology knowledge to create content and make money as an affiliate (affiliate marketing). Formerly a programmer for a software company in Da Nang, he switched to making technology equipment review videos on TikTok and YouTube.

Initially, he shared tips on using AI to program faster, such as how to use AI to support debugging or creating application interfaces, then gave further instructions on assembling and programming hardware. Instead of appearing directly, he uses photo and video creation tools to produce content posted on social networks. Taking advantage of his followers, he attached affiliate links to products such as GPUs, PCs, hard drives and online AI courses. Each month, income from this source ranges from 15-20 million VND.

“AI is changing the software industry very quickly. If you understand technology deeply enough, programmers can not only write code but also make money from sharing knowledge,” Khanh said.

The Hoang, working for a software corporation in Da Nang, also feels difficult before the wave of AI Agents. Once frequently honored for his high productivity, he is now repeatedly “surpassed” by his colleagues. “Hard work is what holds me back, when I appear slow compared to my colleagues who know how to use AI in certain stages to achieve higher work efficiency,” he said.

To improve himself, he spent millions of dong registering for short-term courses on machine learning, LLM and creating AI Agents on online platforms, accepting “learning from scratch” some skills in the new era. Every night, he spends 3-4 hours studying and tinkering with personal projects such as internal chatbots for businesses and AI data analysis systems.

Similarly, despite having more than 10 years of experience, Dinh Long, a programmer in Hue City, is taking the time to learn deeply about AI. He invested more than 100 million VND to buy a powerful computer to run LLM at home to increase data control, especially with things that belong to business secrets such as code, documents, ideas, and customer data.

“The goal is not just to use AI, but to understand how it works,” he said, adding that he took on a consulting contract for a small business that wanted to deploy AI internally.

Investing in machines running LLM, or Local AI, is becoming popular among programmers. On social networks, many programmers’ groups have posts related to sharing experiences of running AI at home. Minh Vu, an AI expert in Ho Chi Minh City, said that instead of using OpenAI’s API, which Anthropic has to send to a third-party server for processing, Local AI helps everything be processed right on the personal computer, so the data does not leave the device.

“This is very important in cases such as programming with internal source code, analyzing corporate documents, researching sensitive data or working in high-security environments,” this expert explains. “Even businesses are applying this approach.”

Mr. Pham Quang Khang, Consulting Director of Rikkei Japan, assessed that AI agents are causing major changes in the programming field, forcing programmers to shift their role from “only using tools” to “mastering capabilities”.

“Learning commands, workflows, or even running Local AI models shows that they care about pragmatism and data security,” he assessed. “However, not everyone who can run a model on Local AI can make a good product. What makes a difference is the ability to set the right problem, control output quality and put AI into a disciplined process.”

According to Dr. Le Duy Tan, lecturer at the Department of Information Technology, International University – Vietnam National University, Ho Chi Minh City, compared to previous waves of technology, the impact of AI Agent on the programming profession is much greater. Cloud or mobile computing trends mainly change deployment platforms and application types, while AI Agent directly impacts the software thinking and production process.

“Simply put, new technology previously made us change our ‘workplace’, but AI Agent makes us change our ‘way of working’,” Mr. Tan said.

According to GitHub research published late last year, with some tasks, programmers using Copilot can complete 55% faster. However, data from the Stackoverflow platform shows that only 43% of programmers believe in the output accuracy of AI, meaning AI helps speed up but does not replace human testing.

According to Mr. Tan, having to constantly learn AI and test new models makes many programmers fall into a state of “not keeping up with technology”. He assessed this as a “real feeling”, when AI changes much faster than previous technology cycles. Therefore, many people feel like just learning a tool today will have a new tool next week.

However, this does not mean “trying everything”. He believes that professionals need to separate short-term trends from core competencies for long-term investment. For example, “hot” models can change very quickly, but thinking about data, testing, system architecture, security and output evaluation is always valuable.

With Local AI, according to Mr. Tan, programmers have been very “proactive and dynamic” in the face of new trends. Building a personal AI infrastructure at home helps control data, test quickly and is not completely dependent on external services. “But we need to look at the reality that not everyone needs to go in that direction,” he emphasized. “Most software development needs today are more suitable for API or cloud models because they are more optimal in terms of cost, power, maintenance and scalability.”

This expert predicts that personalization of AI infrastructure may become a trend in the future, but not for everyone, but more suitable for research teams, AI engineers or cases where high data security is needed. In the context of AI Agents being able to write their own code, Mr. Tan believes that programmers should focus on the sustainable advantages of humans: understanding users’ problems, designing architecture, making technical decisions in the context of many constraints as well as verifying whether the solution is correct, safe, and maintainable or not.

“The future does not belong to those who only know how to code quickly, but to those who make the right technical decisions,” Mr. Tan added.

Agreeing with the opinion, The Hoang said he has applied new knowledge to speed up work progress. “AI code is very fast, but needs people who understand the system to check, optimize, and ensure safety,” Hoang said. “I don’t think AI will replace programmers, but those who don’t use AI will definitely fall behind.”