“This article has participated in the call for good writing activities, click to view: the back end, the big front end double track submission, 20,000 yuan prize pool waiting for you to challenge!”
AI assisted coding to improve efficiency
GitHub Copilot was a huge hit last week. Thanks to the wave of AI+, GitHub helps developers intelligently generate code with the help of OpenAI’s Codex.
Aside from controversies like user privacy and code copyright protection, the features described by GitHub Copilot are really exciting. There is Tabnine AI Code Completion in front and Kite in the back. GitHub Copilot is only open preview version this time, and need to queue, just tried, still queuing. Whether it smells good or not has to be measured.
Efficiency, efficiency, efficiency
In the niches of development, front-end development is also one of the things AI is targeting, and AI + front-end has become politically correct. There is also a lot of exploration and practice in the field of front-end intelligence. Imgcook, who heard from Ali before, did a lot of Design to Code work.
Design to Code, as the name implies, is the Design draft into Code, currently mainly through AI training model to achieve this function. I’ve always been negative about Design to Code because in my business scenario, Design to Code is a bogus proposition. What we lack is more the link from product to design. Most of the middle and background products, because of the unified UI language, are often directly developed from demand without the participation of designers.
However, for some projects involving designers, it is more time-consuming to spend time in the link from product to design, because designers often have artistic background and have relatively difficult understanding of technology and business, which requires higher cost of communication and understanding. By contrast, design-to-development is not a bottleneck.
However, this idea gradually changed this year. This year, I began to take charge of the construction of technology culture and talent cultivation in the front end of science and technology, and my primary service is all the front end developers of the department. When I started this work, I made a rough statistics and found that the number of people at the front of the department was far beyond my imagination, which was really too few. Including this year’s recruitment, I deeply realized the rarity of talents. Yesterday yu Bo in the GMTC speech also mentioned this, is the industry is in urgent need of front-end talent.
What about the vacancy? On the one hand, continuous recruitment is more about how to make full use of existing talent. Efficiency improvement and talent training are the two methods we have at present. There is no need to say more about the cultivation of talents. There are many things that can be done to improve the efficiency. Due to the particularity of the traditional financial industry, our development is carried out in controlled virtual terminals. Improving the configuration of virtual terminals can help improve the development efficiency. From Intranet instant messaging tools and software and plug-ins that synchronize with the Internet, to the organization’s development mode and approval system for non-development related processes, they are constantly being optimized. Tools and platforms are also very important for improving performance.
Yu Bo mentioned that almost every business line in Ant has its own low code platform, which is very easy to understand. Low code does have different focuses for different vertical business field scenarios. We are also developing our own low-code platform. A few months ago, a “cancer of low code” article caused a lot of discussion and controversy in the community. Whether the opinion in the article is correct or not, some of the facts mentioned in the article do exist. Low code is not a new proposition, I entered the first project, the technology stack is the ancient EJB + JSP, also has low code/no code function, only configuration on the page, you can generate lists, forms, flows and a series of functions, isn’t it wonderful? But the real use, reality is cruel, enjoy the convenience of low code, but also suffered its difficult to extend, at the time for many business scenarios, is have a lot of special business logic is the low code function make uncertain, and these requirements normal coding actually it won’t take a long time, but in our system, will be very difficult to change. Accumulated more and more technical debt, and eventually had to spend a lot of effort to do incremental refactoring, which was the death of low code in the early years.
With the development of technology, especially the rapid development of the front end, many of the problems of low code have been solved now, and the low code solution for the vertical business domain has really helped us to free up some of the productivity from the repetitive work of CURD.
CodeFun, another Design to Code tool, opened beta last Friday, and the ability to intelligently generate production-grade Code is appealing.
Teacher Yang Fan, founder of CodeFun, also shared at the special session of front-end intelligence of GMTC, which was very popular.
After trying it out, the Sketch plugin is very easy to use, just select the project and artboard, and upload directly to the corresponding project in CodeFun.
Go to the project panel to directly view the code, you can set the data binding and interaction behavior, currently can export code platform applets and Vue.
It is a pity that GMTC was not able to participate in the event. Yesterday, CodeFun founder Mr. Yang Fan was invited to the company’s technology sharing brand “Minsheng Mini Tech Show”, which made up for the regret that he was not present to hear Mr. Yang’s share.
During the session, Professor Yang Fan gave us a wonderful presentation entitled “Design to Code: How to Generate Code that Programmers Really pay for”. It demonstrated how CodeFun helps develop auto-generated Code, including the challenges of implementing this feature, the related AI algorithms and so on. It was very rewarding.
Low code or front-end intelligence
Kids make choices, adults all make choices. Low code platform and front-end intelligence may not be a choice. At present, for us, back-end management system is more suitable to be built with low code platform. For projects requiring designers to participate in and requiring higher UI, front-end intelligence can help the development and efficiency through D2C. And at present, designer recruitment is relatively easier than development, so D2C has more space for imagination.
Every time low code or intelligent related technology is mentioned, there will be students worry that the job opportunities will be reduced. Productivity tools are always a developer’s friend. These tools and techniques can free us from the mire of repetitive work and allow us to focus more on creative work.
There is still much room for imagination in the field of front-end intelligence, such as I2C, Ideas to Code, or I2D2C, Ideas → Design → Code. In fact, this is the real pain point.
Another idea is another I2C, Image to Code; The ultimate ideal would be something like A2C, Anything to Code, like a Web page, typing in a URL, typing in an App, etc. If you can imagine a brain-computer interface that generates code directly, that would be the ultimate ideal.
conclusion
The shortage of front-end development talent has been a consensus in the industry, improving r&d efficiency can not do without tools and platforms. The AI + front end brings a lot of imagination to the future. It is a very useful attempt to improve both local and global efficiency. Expect more quality tools to emerge to help development teams improve their performance and ease the hiring challenge.