Developers embrace AI Tools but face ‘Big Code’ challenges, survey finds

Artificial intelligence is both the best and worst thing to happen to developers, engineering leaders and companies as a whole, according to a new report from Sourcegraph, a code-intelligence platform that helps developers navigate and understand large and complex codebases.

The report is based on a survey of more than 500 software developers and engineers across various industries and regions. It reveals, among other surprising findings, that 95% of developers surveyed are already using AI tools to write code, such as GitHub Copilot, ChatGPT and Cody, an AI coding assistant launched by Sourcegraph last month.

While these tools can boost productivity and creativity, they also pose significant challenges for managing and securing the vast amounts of code being generated and modified every day. This problem is known as “Big Code,” and it has been a headache for years; but the report warns that it will hit crisis mode if companies don’t get a handle on how their developers use AI at work.

“Big Code has gotten worse over the last 10 years,” Sourcegraph founder and CEO Quinn Slack said in an interview with VentureBeat. “77% of developers, according to the study, say that their codebase grew five times in the last three years. And as we look into the future, AI is about to make that much worse.”

Big Code refers to the situation where codebases are made up of millions (sometimes billions) of lines of code written by thousands of developers over the last two or three decades. The consequences of Big Code include things like struggling to fix critical vulnerabilities and delayed productivity.

The report shows that only 65% of companies have a plan for Big Code and even fewer have any idea of how to approach using AI for software engineering. AI is clearly the best thing to happen to development teams in unlocking a new level of productivity, but if not done right could be the worst in terms of a “toothpaste out of the tube” situation for codebases and subsequent tech debt and security implications, the report says.

“If you’re an enterprise developer at one of our customers, and you need to make a change [to your code], that change is incredibly difficult to make because it could have a ripple effect on a hundred or a thousand other systems. Everywhere you look, there’s incredible complexity, because you have thousands of software engineers that are constantly changing the system.”

The report includes input from hundreds of Sourcegraph customers. The company provides code intelligence — such as a code search function — for engineering teams at four out of five FAANG companies (Facebook, Amazon, Apple, Netflix and Google), top tech companies like Canva and Uber, four of the top 10 U.S. banks, companies launching satellites into space and the team working on the Large Hadron Collider.

The report highlights several concerns developers at these organization have about AI’s impact on Big Code:

• 61% are concerned about AI’s contribution to tech debt.
• 67% are worried about code sprawl due to AI’s rapid growth.
• 76% fear the amount of new code created that will need to be managed.

Developers recognize the threat Big Code and AI pose to their companies’ ability to innovate and compete, with 72% seeing it as a real risk. They’ve identified several key areas where they need help:

• 95% want assistance in quickly getting up to speed on their codebase.
• 91% want more efficient ways to identify and resolve code issues.
• 91% would save significant time if their codebase were fully searchable across all sources and repositories.
• 88% desire tools that would allow them to achieve greater output with fewer resources.

Developers currently spend only 20% of their time in the codebase writing new code for core products, with that percentage dropping to 14% when accounting for non-code activities like meetings and documentation. This has led to developer dissatisfaction:

• 73% of developers experience more frequent blockages due to the size of their codebase.
• 85% struggle to maintain efficiency in their daily work.
• 82% wish they could spend less time searching for information or old code and more time coding.

In a world where AI is transforming developer tools and productivity, the report serves as a stark reminder that companies must address the challenges posed by Big Code head on. By providing developers with the right support and taking a proactive approach to managing AI-generated code, businesses can unlock the full potential of AI without succumbing to its potential pitfalls.