In May 2023, the Meituan Technical Committee issued a guide for the establishment and roles of the software development sequence, unifying the original four directions of front-end development, back-end development, system development, and data development into software development, and no longer setting up these four separately. sequence.
Why merge channels?
First of all, through the construction of technical infrastructure in recent years, the underlying technologies, tools, frameworks and solutions in various directions have basically become stable and perfect. For example, the Web front-end has basically converged on the two major schools of Vue or React, and the Web back-end technology stack It is almost dominated by Java's Spring Boot. There are many ready-made tutorials or open source codes for reference to implement commonly used functions. No matter which functional direction it is, as long as the basic knowledge and skills of programming are solid, it is easy to learn and apply it now.
Secondly, with the strengthening of AIGC, the technical threshold for different functions is getting lower, and the threshold for understanding business and code is also getting lower. The barriers between traditional functions (such as products, technology, front-end and back-end algorithms) and functions are tending to dissolve, and individuals and teams have the opportunity to become more cross-functional and cross-functional. At this time, the conditions for building more cross-functional and cross-functional teams are more mature, and closer collaboration and integration of business and technology are more possible.
Finally, according to the theory of seven stages of industrial development (starting period, growth period, maturity period, integration period, nirvana period, harvest period and transformation period), most Internet-related industries have entered a period of change, so we see Alibaba has carried out a major split this year, and Tencent and Byte have to overhaul their rank systems in order to adapt to the major changes that big model technology has brought to the Internet and society as a whole.
Impact on business
Technology is developing more and more rapidly and complexly, and we will find that doing many things often requires comprehensive capabilities.
From the perspective of the enterprise, building software development sequences can help eliminate the sense of boundaries between subdivided sequences, simplify the division of labor in the product development process, conduct flexible division of labor based on actual business needs, and improve collaboration and work efficiency in the operation chain.
At present, the wave of technological innovation triggered by breakthroughs in large model technology has, on the one hand, brought huge opportunities for change on the application side, and on the other hand, it will also profoundly affect every link and field in the software development process. With the continuous advancement of AI technology, the current professional division of labor barriers for software engineers will be further broken down. In the future, every engineer will be able to complete the tasks that previously required multiple engineers across professional technology stacks based on understanding the product business needs. Completed development work.
When a company's R&D does not limit itself to a front-end engineer, a back-end engineer or an algorithm engineer, it will be easier to understand both the front-end and the back-end, as well as the algorithm. Then the output of this R&D will definitely be better than that of the front-end. With the addition of background and algorithms, the output of the three students is even higher.
In this way, it can not only guide R&D students to pay more attention to value creation, but also reduce barriers to talent flow and improve the overall efficiency of corporate R&D.
Implications for software engineers
From the "Seven Liberal Arts" in the ancient West, to the "Books and Numbers of Ritual, Music, and Archery" in ancient China, to the modern "Liberal Education", general knowledge and skills have always been important to people's growth and development. Have a greater impact.
From the perspective of each engineer's growth, establishing a software development sequence can help everyone have more diversified development opportunities and promote the flow of talents. It allows everyone to continuously broaden their skill boundaries and try more possibilities on the basis of deep cultivation in this field, and become comprehensive talents with both technical depth and breadth.
On this basis, engineers can also establish a more comprehensive technical and business perspective. As technology gets closer to each business link, they are more likely to break through boundaries to solve problems, find opportunities to "invent and simplify", and create value for customers.
But this does not mean that we must master all skills to become a full-stack engineer. The specific skills we need to master should be determined by the needs and development stage of the business.
Of course, in actual work, mastering more knowledge and skills will help everyone better understand the entire technical system and promote efficient resolution of cross-domain problems.
Become a comb-type talent
All technologies are advancing at an unprecedented speed, many traditional industries will be completely subverted, and many professions are at risk of disappearing.
In recent years, most people should have heard of the following three types of talents:
-
T-shaped talents: "—" represents broad knowledge, which is the accumulation of underlying thinking and logical abilities; "|" represents professional depth, which is the accumulation of a certain skill or professional direction. T-shaped talents represent "broad knowledge, "Professional" with one specialty and multiple abilities.
-
π-shaped talents: Also called M-shaped talents , it is a talent concept originated from Singapore. π has one more vertical line than T, which refers to advanced composite talents who possess at least two professional skills and can cross-integrate multiple knowledges. Talent.
-
Comb-type talents: have in-depth professional knowledge in multiple majors, while maintaining a lifelong learning habit at the top level . This habit is the "—" of the comb, which represents strong underlying thinking and logical abilities. Its thickness determines your Whether it has the ability to transfer knowledge.
If we want to avoid being eliminated in the tide of scientific and technological progress, it is not safe to just be a T-shaped talent with one specialization and multiple abilities. We need to become a comb-shaped talent who is proficient in multiple fields.
In fact, becoming a generalist who specializes in multiple fields is not as difficult as you think. It is much easier to enter the top 25% in three or four fields at the same time than to enter the top 1% in one field, provided that you master efficient learning methods.
The current external market environment and social environment are changing too fast. If you only focus on a certain field, it will be difficult to become stronger. There is nothing wrong with accumulating skills, but in an era of rapid change, the skills we are accumulating may soon become outdated. .
some cases
Here I will give three examples.
Rich second generation experiencing life
Classmate A is a friend I only met in 2022. At that time, I moved to a new department and became colleagues with him. However, half a month after I arrived, he planned to resign. Since I was more enthusiastic, he often continued to work with me. keep in touch.
He must have graduated from a university in the United States with a bachelor's degree in computer science in 2015. He must have entered Microsoft as an intern before graduation. After less than a year, he returned to China and started his own business. He made 3D design software. The revenue should be over 10 million, and the team size is about 50 people.
I stopped starting my business in early 2021 and joined Byte. After working for more than 3 months, I joined Meituan. Then I worked for more than 7 months before leaving Meituan. After leaving Meituan in 2022, he took a three-month break and then started looking for a job. During this period, he often communicated with me about the interview situation. At that time, I was actively trying to arrange for him to return to Meituan, and also assisted him in interviewing other companies.
As a scarce talent in the field of rendering, it is easier for him to find a job. I suggested that he go to Wondershare Technology to take charge of a business unit, but in order to be able to balance his postgraduate studies at the University of Hong Kong, he finally went to Tencent to do cloud rendering.
However, cloud rendering has not taken off, and now he has switched to making large models. The direction is to generate 3D through AI. At the age of 30, he has an annual salary of nearly 2 million.
Today, he also told me that he plans to set up a stall selling cocktails part-time in Futian, Shenzhen. As the heir to the largest wooden floor manufacturer in South China, he knows how to experience life.
In addition to mastering rendering technology, this young man is also proficient in front-end and back-end, data processing and algorithm-related knowledge and skills involved in application development. In addition, he is in his prime age, so it is easy to get high-paying job opportunities.
The technical master who successfully converted the algorithm
Classmate B is a friend I met during the summer of 2008 when I performed a hip-hop dance program during military training. Not only were our majors the same as mine, both in measurement and control, but after graduation, we both went to work in the army for a few years, then left the army in the same year and joined the same foreign company.
He left the army half a year before me and joined a foreign company to develop operating systems for SIM cards. The system kernel was developed in C language and the application layer was in Java. He used VB for testing and later learned Android by himself.
I left a foreign company in 2015 and joined 360 to do Android development. Later I secretly learned deep learning without telling me, and published many papers at top AI conferences, such as CVPR. I transitioned to algorithm research around 2017 and joined Didi in 20. Didi was working on recommendation algorithms, and currently switched to Haomo Zhixing to develop vision-related algorithms.
I have known him for more than 10 years. He is relatively introverted and loves to mess around with technology. He built a very impressive high-voltage discharge device in college. At the same time, two of his school dance friends are great masters in the field of AI. After graduation, they both joined Google. One of them studied abroad in the United States, which ranks first in computer science, and earned an annual salary of more than one million yuan just after graduating with a PhD. The other only graduated with a master's degree, but after graduation, his annual salary was more than 2 million.
We often practice hip-hop together. During every break in dance practice, I see him reading papers. At the same time, he has a master to guide him, so it is easier to transform into algorithms.
I love tossing
Ten years ago, I would never have imagined that I would become a programmer. After leaving the army, I initially wanted to study for a postgraduate degree and then take the civil service exam. But later, because I came into contact with the legendary Internet celebrity temple, I gave up the re-examination and focused a lot of energy on exploring the issues of birth, aging, sickness and death that everyone needs to face.
Because I joined the information technology group of the temple at that time, many of the volunteers were coders. I also applied for the computer science major at Tsinghua University and had a certain programming foundation. In addition, classmate B also wrote code, so after his recommendation, I I joined a foreign company and started my programming career.
Before 2018, the technical directions of classmate B and I were quite different. I gradually mastered common technologies in application development, while he went deep into algorithm development.
Most of the public welfare projects in the temple are done by the front-end and back-end. Before I came to Meituan, I joined a startup company, started my own business, and was exposed to a large number of projects, such as back-end Java, Go, PHP, and Ruby. , C#, Nodejs, Python, front-end React, Vue, cross-end RN, flutter, and uni-app, all participated in development and maintenance.
After joining Meituan in 2020, I took on more of a project leader role and was responsible for material construction, engineering construction, inference engine construction, etc. I also participated in part-time PM training organized within the department, and also provided more than 200 employees in the department. A backend has done front-end development empowerment training.
During this period, the biggest challenge for me was that the construction of the inference engine required systematic learning of knowledge related to deep learning. Compared with the mutual transformation of system development, front-end development, back-end development, and client development, the transformation from application development to algorithm was more difficult. Much bigger.
Of course, it is more challenging to transform from development to product technology leader. This requires learning a lot of business and management knowledge. This is the direction I have recently changed.
Let’s make an advertisement first: follow me so you won’t miss it. I will share my follow-up posts on terminal intelligence, project management and business.
Summarize
The practical application of large language models is triggering a revolution in software engineering.
Software engineering is essentially a large-scale mental collaboration among humans. Due to its high complexity, different professional roles have been divided.
The adoption of LLM in software engineering will create breakthroughs and even subversion in many engineering fields. This also urges us to carefully examine the changes in professional capabilities and the definition of professional roles.
AIGC is giving birth to a new paradigm of software engineering. As a catalyst and innovation engine, LLM is opening up a doubling of software development and innovation efficiency, and even bringing about the possibility of exponential improvement.
In the face of paradigm shift, we cannot rely on speculation to infer the future. We can only get involved, identify opportunities, solve problems, and continue to trial and error and evolve in practice.