It has never been easier to write code quickly. We can now generate software by submitting a prompt to a coding assistant, and in seconds it can produce thousands of lines of code that might have taken several hours for a human to produce.
However, the quality of that code can vary significantly depending on many factors such as the model being used, the details of the prompt, the size of the context window, and more. As a result, the code produced by AI still needs to be reviewed carefully. Now the bottleneck has shifted from writing code to reviewing it.
Teams face a difficult trade-off, they either slow down development to carefully validate the code produced by AI, or they skip thorough reviews and risk shipping low-quality code (aka “code slop”).
The opportunity lies in finding new ways of shipping high quality code at the speed that AI can produce it. Instead of relying solely on human reviewers, other AI agents can be used to review code, flag issues, suggest improvements and produce better prompts. The exact process is something that is still being figured out. But avoiding the use of AI generated code is no longer an option as the technology matures.
How is your team working with coding agents?
