Recently, Amazon CEO Andy Jassy announced that the company saved 4,500 developer years of work by using Amazon Q. “Yes, the number is crazy, but it’s real,” he posted on X.
With Amazon Q, the company has significantly reduced the time it takes to update Java applications. “The average time to upgrade an application to Java 17 dropped from what was typically 50 developer days to just a few hours,” he said.
He added that in less than six months, the company was able to upgrade more than 50% of its Java production systems to modernized Java versions with a fraction of the usual time and effort. “Our developers shipped 79% of the automatically generated code reviews without additional changes.”
“The benefits go beyond the effort we saved developers. The upgrades have improved security and reduced infrastructure costs, resulting in estimated annual efficiency gains of $260 million,” he claimed.
In fact, generative AI has made programming a breeze. Tools like GitHub Copilot, Devin, and Amazon Q simplify the development process, make application creation easier, and help developers and companies reduce technical debt.
Technical debt occurs when a company rushes to complete a product to meet deadlines without properly checking code quality and bug fixing. Incomplete documentation and inadequate testing can lead to bugs and inefficiencies that increase debt.
Converting the legacy codebase
Amazon isn’t the only company using AI to reduce its technical debt. San Francisco-based Databricks uses generative AI to quickly analyze and understand its legacy code base – which its CIO says has reduced the burden on engineers.
Seventy-five-year-old payroll company ADP is also using generative AI to convert COBOL to Java. “A big problem we and other established companies face is that we have COBOL running in our systems,” said Amin Venjara, ADP’s chief data officer. He added that very few programmers today are familiar with COBOL.
The Roseland, New Jersey-based company is exploring the use of generative artificial intelligence to convert its mainframe code from COBOL – a language developed in the 1950s and still widely used in banking and financial services – to Java, a programming language that has been around since 1995.
Wayfair, the online furniture retailer, uses generative AI-based coding tools to update old code. Although Wayfair, founded two decades ago, doesn’t use COBOL, it has “legacy code” in languages like PHP and outdated database code in SQL.
In addition, there is code written by developers who are no longer with the company.
GenAI helps with tedious tasks
Generative AI acts as an intelligent assistant that automates tedious tasks, suggests improvements, and improves code quality. Armand Ruiz, product VP at IBM, says his favorite use case for generative AI is software development.
According to Ruiz, GenAI has several use cases in software development. It can convert simple English instructions into code in the preferred programming language.
Code translation tools convert languages such as COBOL to Java and facilitate code modernization and migration. Debug tools identify and recommend fixes for code errors, improving code reliability.
In particular, IBM recently announced the IBM Watsonx Code Assistant for Enterprise Java applications, which is expected to be generally available later this year.
Generative AI also plays an important role in streamlining code maintenance through refactoring. Generative AI automates refactoring by suggesting or implementing code transformations. It can identify common antipatterns and suggest more efficient alternatives to ensure that refactoring follows coding standards and best practices.
Recently, the AI coding tool Cursor AI has been making waves on social media.
OpenAI co-founder Andrej Karpathy praised the AI-integrated code editor, saying, “Programming is changing so fast… I’m trying VS Code Cursor + Sonnet 3.5 again instead of GitHub Copilot and I think that’s a net win now.”
“Empirically speaking, most of my ‘programming’ over the last few days now consists of writing English (entering the generated differences and then checking and editing them) and doing a bit of ‘semi-coding’, where you write the first bit of code you want, maybe add a few comments to it so the LLM knows what the plan is, and then tab, tab, tab through the completions.”
In the past, software developers needed a long time to ship a product, but now that time has drastically reduced. GitHub recently launched Models, a platform that gives developers access to leading LLMs, including Llama 3.1, GPT-4o, GPT-4o Mini, Phi 3, and Mistral Large 2.
“With GitHub Models, developers can now explore these models on GitHub, integrate them into their development environment in Codespaces and VS Code, and use them in actions during CI/CD – all with just their GitHub account and free permissions,” explained Github CEO Thomas Dohmke.
Then there are AI software developers like Genie and Devin. Genie is designed to emulate the cognitive processes of human engineers, allowing it to solve complex problems with remarkable accuracy and efficiency. “We believe that if you want a model to behave like a software developer, you need to show it how a human software developer works,” said Alistair Pullen, co-founder of Cosine.
“Software development companies will start hiring programmers who can demonstrate in interviews that they have mastered AI-powered programming tools like Copilot or Cursor,” said Andriy Burkov, head of machine learning at TalentNeuron.
He added that the claim that LLMs cannot write reliable code is ill-conceived. “Most junior and mid-level programmers cannot write reliable code either. This is why software development has spent decades equipping itself with automated and semi-automated tools to ensure that code is production-ready.”
He further explained that LLMs are already being optimized specifically for coding. Companies are investing hundreds of millions of dollars to improve the coding capabilities of these LLMs because, if done correctly, coding is the only use case where they can charge enterprise customers $10 or even $100 per million tokens.
One problem developers face is the assumption that AI-generated code might contain bugs. Many AI startups are forming to address these concerns. One such startup is YC-backed CodeAnt AI.
CodeAnt’s AI-driven code review system can significantly reduce vulnerabilities by automating the review process and identifying potential issues before they reach the customer.
Thanks to advances in AI-driven coding tools and platforms like IBM Watsonx and GitHub Models, developers are undoubtedly better able to deal with legacy code, optimize code maintenance, and increase productivity.
