GENEVA, SWITZERLAND – JULY 6: Ameka, a humanoid robot from British manufacturer Engineered Arts, interacts with visitors in Geneva, Switzerland on July 6, 2023. The two-day summit in Geneva, hosted by the United Nations, will bring together around 3,000 global experts from leading technology companies, educational institutions and international organisations to discuss the potential of artificial intelligence to empower humanity. (Photo by Johannes Simon/Getty Images)
Getty Images
Code is complicated. Software engineers typically spend four (sometimes five) years earning their degrees, which inevitably means countless late-night hours programming as a “hobby” and, as the real-world stereotype goes, usually fueled by soda and pizza.
But AI is on the rise, and its impact on software application development practices rarely strays from discussions focused on modern enterprise IT architecture. AI-based code assistants and automation engines will naturally be built by software engineers, so the fear mongering on the subject may be overblown. The question now is whether we are moving beyond the experimentation and prototyping phase into a realm where automation coexists alongside human engineers.
Shred Code Fat
Named either after a technique for reducing body fat or after the shredding process that involves splitting and fragmenting data, Shreds is one of the newer contenders in the code automation space. According to the company, their Shreds.AI product “regenerated” (completely rewrote) the entire codebase that runs the WordPress content management system. Originally written in the PHP programming language and developed over several years, Shreds completed the task in less than 24 hours, rewriting the popular CMS into modern Java (version 17) code. The resulting open source code repository is available here. The repository is based on a microservices architecture.
“These results highlight the power and potential of Shreds.AI to transform the software development landscape. As shown in this example of WordPress code regeneration, Shreds.AI not only saves companies time by generating code quickly, but also addresses the issues of software obsolescence and maintenance. It also increases return on investment by automatically transforming legacy code and architecture to modern standards,” said Soufiane Amar, founder and CEO of Shreds.
Amar is understandably optimistic about his team's work on the product, which he describes as “human-level engineering automation,” with Shreds.AI trained to perform tasks traditionally managed by software engineers, managing the complexity of large-scale software and making decisions about component integration.
Not just generative AI
Unlike generative AI models, which typically generate short code snippets for specific tasks, this tool can create code files with tens of thousands of lines of code with consistency and orchestration. Described as a meta-AI that leverages other AI engines, Shreds integrates with about 10 different AI services, ranking and assigning tasks to each service to ensure the highest quality output. Additionally, Shreds.AI incorporates code reviews, either internally or through the marketplace, and makes any necessary adjustments to ensure high-quality output.
Shreds itself is currently in beta development, but it should be an interesting start as the organization behind the technology is now hard at work strengthening its cloud infrastructure, hiring additional engineers, and expanding beta access to more businesses.
In more established enterprise circles (but in the realm of code assistants), Amazon Web Services, Inc. (AWS) is announcing the general availability of Amazon Q, an AI-powered assistant that accelerates software development by generating accurate code and providing testing, debugging, and multi-step planning and reasoning capabilities.
Amazon Questions and Answers
Amazon Q is built to connect to enterprise data repositories to logically summarize data, analyze trends, and interact with the data, enabling users to get answers to questions across business data such as corporate policies, product information, business results, code base, and employees.
“Amazon Q is the most capable generative AI-powered assistant available today, with industry-leading accuracy, advanced agent capabilities, and best-in-class security, helping developers be more productive and business users make faster decisions,” said Dr. Swami Sivasubramanian, vice president, Artificial Intelligence and Data, AWS. “Since we announced the service at re:Invent, we've been amazed at the productivity gains we've seen from developers and business users. Early indications show that Amazon Q is helping our customers' employees become more productive at work by over 80%, and we believe this will continue to grow with the new features we plan to introduce.”
Sivasubramanian and his team remind us that modern developers spend only 30% of their time (or less) coding, with the rest of their time spent performing tedious and repetitive tasks related to infrastructure provisioning, updating and addressing, and administrative tasks related to system maintenance.
Less filth, more structure
When (developers) switch projects, they have to spend time studying the existing codebase and understanding the programming logic. There's also all the work involved in testing and refactoring the code, upgrading the application, and debugging and optimizing. AWS wants developers to spend less time “in this coding quagmire” (their own words) and more time building applications that are actually useful.
“When we launched Amazon Q Business in April 2024, we also previewed Amazon Q Apps, a feature of Amazon Q Business that enables users to create generative AI-powered apps based on their organization's data. Users can build apps using natural language and securely publish them to their organization's app library for anyone to use,” Prasad Rao, principal partner solutions architect, AWS UK, wrote in the AWS News Blog.
Within four months of starting use, BT Group has generated over 100,000 lines of code with Amazon Q Developer, automating approximately 12% of the labor-intensive, repetitive, and time-consuming tasks performed by software engineers using the platform. The AWS service is providing BT Group with 15-20 code suggestions per active user per day, with a 37% acceptance rate by software engineers.
The rise of AI code assistants is undeniable. At this inflection point in the industry, we need to think about how far we can take these tools (even though a full CMS rewrite is a long way off), how much of their functionality is focused on shifting the burden of tedious infrastructure management to front-end software application development, and how thoughtfully we can combine computerized and humanized code in a symbiotic, well-coordinated, and harmonious way.
