The cognitive effort behind understanding code

May 28, 2025

"I still think that one of the finest tests of programming ability is to hand the programmer about 30 pages of code and see how quickly he can read through and understand it."

- Bill Gates

Various research suggests that reading, editing, and navigating code encompasses significant cognitive effort beyond simply writing new code. And contrary to popular belief, most coding is maintenance coding, so comprehension is a significant overhead cost in the software development process.

"All programming is maintenance programming, because you are rarely writing original code."

- Dave Thomas

We’ve found a variety of research focused on comprehension. While the details may differ, the significance of the effort in software development doesn’t.

Early Research Foundations

1986-1989

IBM started the Program Understanding Project at IBM's Research Division in 1986. Research by Thomas Corbi in 1989 found that more than half of the effort in accomplishing a task for the programmer is towards understanding the system.

50%+

of programmer effort goes to understanding systems

1979

Fjeldstad and Hamlen conducted research on software maintenance tasks and found that a significant portion of time, 47% for enhancement tasks and 62% for correction, is dedicated to understanding the existing code and documentation.

47%

enhancement task time

62%

correction task time

Fast forward a few decades, and the numbers hold steady or even increase:

Modern Research Validation

Recent Studies

Fast forward a few decades, and the numbers hold steady or even increase. A research on program development concluded that programmers spent 70% of their time in program comprehension.

Specifically, it has been estimated that reading and understanding code consumes nearly 60% of a software professional’s time.

70%

of developer time in program comprehension

60%

of a software professional's time is consumed by reading and understanding code

And we haven’t even addressed the time a developer may spend mentoring new developers, answering questions from business analysts, participating in knowledge-sharing meetings, or engaging in the myriad activities that rely on an understanding of what a company’s code does.

What would your developers do with the time they save by using COBOL Colleague?

Tags: productivity, software development, software maintenance

How a Novel Approach to AI Mitigates the Need for Comments in Code

October 25, 2022

Code comments are often difficult to understand, incomplete, out of date and untrustworthy to many developers, resulting in significant additional work and unintended business risks. Incorrect documentation results in time and money lost. Transitioning away from relying on developers to add and update comments in code and related documentation requires new methods and tools.

Steve Brothers, President of Phase Change Software, recently addressed this challenge in his article: "How a Novel Approach to AI Mitigates the Need for Comments in Code." He explains how new AI technology can exponentially improve software development productivity by assisting new developers with identifying code behavior and locating the exact place in the code where changes are needed.

Stephen Tullos is an Analyst with Phase Change Software. You can reach him at stullos@phasechange.ai.

Tags: AI, code comments, code documentation, software development, software engineering

How COBOL Code Can Benefit from Machine Learning Insight

October 24, 2022

Most dev tools are not yet capable of identifying the specific lines of code that need to be changed, and unearthing that information is hard cognitive work. While some tools can help improve productivity by suggesting what code to write, software developers still have to use their brains to add new features, fix bugs, implement changes to meet regulatory requirements, address security needs and solve challenging engineering problems. This can drastically affect productivity and increase the risk of application crashes.

Phase Change President Steve Brothers recently shared his thoughts on how COBOL Colleague offers an elegant solution that uses AI to automate the identification of specific lines of code that require attention to this problem in an article tiled: "How COBOL Code Can Benefit from Machine Learning Insight."

Read the entire article here.

Tags: AI, cobol, cobol colleague, machine learning, software developer application knowledge gap, software development

An AI alternative to code search tools

October 4, 2022

Would you believe that the average software developer spends roughly 75% of their time just searching through and understanding code to make necessary changes? When software engineers have to spend so much of their time just finding and understanding legacy code, before any real work gets done, they have less time to create new solutions to move an organization forward.

Phase Change President Steve Brothers recently penned an article for the Infoworld New Tech Forum titled, "An AI alternative to code search tools," about how AI tools are becoming available to close the application knowledge gap for developers, promising to exponentially improve developer productivity across applications. Specifically, Brothers wrote about Phase Change's COBOL Colleague, an AI-driven tool that helps developers quickly gain a mental model of a COBOL codebase, and zero in on the exact code they need to change.

Read the entire article here.

Tags: AI, artificial intelligence, cobol colleague, software developer, software development, software maintenance, steve brothers

Solving the issues with current documentation practices

September 1, 2022

Software development is typically a team endeavor. Developers may work on separate projects but many times their work intersects with modules others are building. Even individuals creating their own applications must refer back to prior work to track source-code changes and limit vulnerabilities. Creating proper documentation for teamwork and legacy code should be a top priority for all developers.

The consequences of missing or inadequate documentation impede application updates and new feature additions, or worse, affect end users by delivering buggy products or missed delivery deadlines.

Phase Change President Steve Brothers was recently interviewed for an article published by SD Times titled, "Solving the issues with current documentation practices," about how software development and maintenance documentation remains an issue. In the interview, Brothers said many times documentation is not a priority because of time constraints – developers feel they are paid and assessed on the code they create, not on documenting the process. And when they do provide comments, once again, project constraints can lead to inaccurate information. This failure to transfer knowledge leads to "slower and sloppier development."

Brothers also talked about coming AI tools that will automatically capture the knowledge developers put into the code, thus creating its own documentation, which never leaves the organization, even when the developers depart. Phase Change's AI tool, COBOL Colleague, will also help automate the process of searching for relevant code and data, which minimizes the need for extensive documentation.

Read the entire article here.

Tags: AI, artificial intelligence, developer productivity, documentation, productivity, sd times, software developer, software development, steve brothers

Combining developer knowledge with artificial intelligence to improve software maintenance

May 19, 2022

Enterprise software systems are complex and require specialized abilities and unique knowledge to update, add new features, and generally solve problems. They necessitate ongoing systems maintenance to grow and evolve, which costs your organization a significant amount of money – generally about three-quarters of your IT software budget. Unfortunately, because of the global software developer shortage, the typically brief developer average tenure at one job, and today’s inadequate source-code search tools, linters, and static and dynamic analysis tools, organizations across industries are struggling to maintain their software systems effectively.

Phase Change President Steve Brothers recently wrote an article for The Next Tech about how a novel approach to artificial intelligence (AI) software tools can help enterprises save a significant amount of time and money while minimizing the risks associated with making changes in complex software systems. The article, "Combining developer knowledge with artificial intelligence to improve software maintenance," discusses how AI and cognitive automation can automate the identification of the specific lines of code that require attention — no matter how entwined throughout the system that code might be – at machine speed. The tools also comprehend and reveal all of the upstream and downstream changes that will occur due to code modifications so developers can be confident when updating source code to add new features, fix bugs, meet regulatory requirements, and address information security concerns.

Read the entire article here.

Tags: AI, application maintenance, artificial intelligence, developer productivity, developer-engineer, enterprise software, enterprise software maintenance, industry turnover, software developer, software development, software maintenance

Improving developer productivity on the mainframe with artificial intelligence

March 3, 2022

Mainframes are the central data repository in an organization’s data processing center. They support thousands of applications and input/output devices while simultaneously serving thousands of users. Most corporate data still lives on the mainframe, and these systems offer advanced capabilities, flexibility, security, and resilience to downtime. Unfortunately, mainframe management and modernization can be costly, risky, and can damage an organization’s reputation by crashing internal and customer-facing applications if developers don't know the system.

Phase Change President Steve Brothers recently authored an article for Techslang.com titled, "Improving Developer Productivity on the Mainframe with Artificial Intelligence," which discusses the roles mainframes play in multiple industries including finance, healthcare, and government, and the difficulties reliant organizations face maintaining and integrating them with modern tools.

To maintain and improve critical mainframe applications, software teams rely on seasoned developers who have developed an intimate understanding of their systems. Unfortunately, many of these experienced programmers are aging out of the workforce or opting for other opportunities – creating a loss of knowledge about those organizations' mainframe applications.

In the article, Brothers explains how AI can automate the process of precisely and accurately identifying code that requires attention — no matter how dispersed throughout the system it might be. By guiding these AI tools through describing the application behavior that needs to change, developers don’t have to search through and develop an intimate understanding of, massive source code bases to reveal the specific lines implementing that behavior. They can now collaborate with an artificially intelligent coworker to augment their own intelligence and be guided exactly to the code that matters.

Read the entire article here.

Tags: AI, artificial intelligence, assistive AI, developer productivity, developer-engineer, legacy systems, mainframes, maintaining legacy systems, productivity, software developer, software development, software maintenance, steve brothers

You can use artificial intelligence to fix your broken code

February 23, 2022

Mainframe systems are used across industries and around the globe, with over 10,000 currently in worldwide use. They are relied on by some of our most important institutions, including 96 of the world’s 100 largest banks, nine out of 10 of the world's biggest insurance companies, 23 of the 25 largest U.S. retailers, and 71 percent of Fortune 500 companies. Unfortunately, often because of a lack of detailed understanding of these mainframe systems, making source-code changes can be costly, risky, and can tarnish the organizations' reputations.

Phase Change President Steve Brothers recently wrote an article for BuiltIn.com titled, "You Can Use Artificial Intelligence to Fix Your Broken Code," which explains how artificial intelligence (AI) can help developers better understand the codebase, and help them find code responsible for application behavior at machine speed. Developers will no longer have to pore over millions of lines of code to unearth the intent of previous developers and find the source code that requires change.

Read the entire article here.

Tags: AI, application maintenance, artificial intelligence, assistive AI, business agility, developer productivity, developer-engineer, legacy systems, mainframes, maintaining legacy systems, software developer, software development, software maintenance, steve brothers, system knowledge

How banks should leverage the power of automation

February 16, 2022

Mainframes are widely considered the backbones of many global financial services firms because they deliver unparalleled security, stability, and processing power. From credit card payments and ATM transactions to loans and mortgages, mainframes are relied on by 44 of the top 50 banks to host core applications that deliver secure experiences based on real-time data analytics.

Phase Change President Steve Brothers recently penned an article for TechBullion.com titled, "Banking automation: How banks should leverage the power of automation," in which he examines how these critical mainframes systems also present modernization challenges.

Mainframe systems are complicated and require meticulous processes to continue providing core operational value. While they are fully capable of running newer applications and systems to create new products and revenue streams, their ongoing support and modernization are challenging.

Brothers believes automation and artificial intelligence (AI) could greatly assist banking firms in maintaining and enhancing their mainframes because the key to sustaining these systems is precisely identifying the functionality created by the source code that is intertwined throughout the system — and changing that behavior without unintended consequences. Using a new AI approach that's designed to sift through large quantities of code in the same way humans do, AI-powered tools can aid developers in their frequent search through the deluge of code to rapidly identify where they need to make a change.

Read the entire article here.

Tags: AI, application maintenance, artificial intelligence, brain drain, business agility, developer productivity, developer-engineer, domain knowledge, enterprise software, enterprise software maintenance, industry turnover, legacy systems, mainframe legacy software, mainframes, maintaining legacy systems, productivity, software developer, software development, software maintenance

AI rises to the challenge with COBOL

June 3, 2021

A May 28 article published by TechRadar pro, and written by Phase Change President Steve Brothers, explains how the well-reported "COBOL skills shortage" is not really a fundamental problem for enterprises that rely on mainframe systems. The real challenge is application knowledge. Developers can learn COBOL in less than 6 months. What they can't learn quickly is specific application knowledge because that knowledge comes from experience.

Steve also describes how AI tools that assist developers in identifying and locating code responsible for specific behavior will help them reveal the application's intent and expose code that requires change. The developers will learn the application through task completion while remaining productive for the organization.

Click here to read the full article on TechRadar pro.

Tags: AI, application knowledge gap, artificial intelligence, cobol, cobol applications, knowledge gap, software developer, software developer application knowledge gap, software development, software maintenance

Early Research Foundations

Modern Research Validation

Contact