Your engineers spend most of their time understanding, not building
Field studies and developer surveys agree. The bottleneck is not writing code, it is finding out why the code is the way it is.
Ask a team what slows them down and you will hear about meetings, tooling and tech debt. Measure what they actually do and a different picture appears.
The measurements
- A field study that tracked 78 professional developers over 3,148 working hours found they spent about 58% of their time on program comprehension: reading and understanding existing code (Xia et al., IEEE Transactions on Software Engineering, 2018).
- Atlassian’s 2025 developer experience research, with 3,500 developers and managers, found that half of developers lose 10 or more hours a week to inefficiencies, and that finding information (services, docs, APIs) is the biggest single cause.
- In Stack Overflow’s 2022 survey of over 70,000 respondents, more than two-thirds said they hit a knowledge silo at least once a week. In 2024, more than 60% said they spend 30 minutes or more a day searching for answers.
- Stripe’s Developer Coefficient estimated that developers spend 17.3 hours of a 41-hour week dealing with technical debt and bad code.
Understanding the code is only half of it
Much of that comprehension time is recoverable from the code itself: a good reader, or a good AI assistant, can work out what a function does. The part that is not recoverable is intent. Why 30 days? Why does this region skip the check? Is this branch dead, or protecting something?
That question has only two answers: a person who remembers, or a record someone kept. When neither exists, the engineer guesses, asks around for a day, or leaves the code alone forever. All three are expensive.
AI makes this more urgent, not less
AI coding assistants have made it much cheaper to change code. They have not made it any cheaper to know whether a change is safe for the business. An assistant asked to extend refunds to 45 days will make the edit correctly and has no way to know that the finance reconciliation assumes 35.
The fix is to give people and assistants the reason, with its owner, before they change the rule. See AI coding tools.
Sources
- Xia, Bao, Lo, Xing, Hassan, Li, Measuring Program Comprehension: A Large-Scale Field Study with Professionals, IEEE TSE, 2018
- Atlassian, State of Developer Experience Report 2025
- Stack Overflow Developer Survey 2022, reported by CIO Dive; Stack Overflow, insights from the 2024 Developer Survey
- Stripe, The Developer Coefficient, summarised by Lightspeed